An ai intelligent analysis and structured generation method and system for vehicle test defects

By combining AI Agent and large language model with a comprehensive knowledge base for vehicle testing, the system automatically identifies vehicle testing scenarios and generates structured defect work orders. This solves the problems of low efficiency in vehicle testing defect processing and insufficient semantic understanding in existing technologies, and enables efficient and standardized defect entry and rapid fault diagnosis.

CN123068293APending Publication Date: 2026-11-10WUHAN KOTEI INFORMATICS
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611311370.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-27
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing methods for handling defects in vehicle testing suffer from problems such as low efficiency of manual data entry, non-standard descriptions, lack of semantic understanding capabilities for vehicle-related business by general AI, and inability to provide end-to-end troubleshooting assistance.

Method used

By collecting raw, unstructured data during vehicle testing, using an AI Agent for semantic parsing and scene recognition, and combining a large language model and a pre-built comprehensive knowledge base for vehicle testing for semantic reasoning, structured defect work orders are generated to match the troubleshooting logic of developers.

Benefits of technology

It improves testing efficiency, enhances defect clarification, possesses professional semantic understanding capabilities in the field of automotive testing, realizes a closed-loop value chain, and reduces the understanding cost and troubleshooting time for developers.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123068293A_ABST
    Figure CN123068293A_ABST
Patent Text Reader

Abstract

This invention relates to an AI-powered intelligent analysis and structured generation method and system for vehicle-mounted testing defects. The method includes: collecting raw unstructured data from vehicle-mounted testing; identifying testing business scenarios and extracting core defect elements through semantic parsing using an AI Agent; relying on a comprehensive vehicle-mounted testing knowledge base and calling a large language model for semantic reasoning to obtain standardized defect analysis results; structurally reorganizing the defect content according to the developer's troubleshooting logic, adaptively matching standard templates for business scenarios and filling in the content, and generating standardized defect work orders. This invention relies on a collaborative architecture of AI Agent and a large language model, possessing professional reasoning capabilities in the vehicle-mounted testing domain. It replaces manual compilation, improves testing efficiency, provides clear and standardized defect work orders, shortens the troubleshooting cycle, can be generalized and adapted to various vehicle-mounted testing businesses, and supports iterative optimization, achieving a closed-loop value chain for vehicle-mounted defect handling.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of vehicle testing and AI intelligent analysis technology, and in particular to an AI intelligent analysis and structured generation method and system for vehicle testing defects. Background Technology

[0002] With the rapid development of the intelligent connected vehicle industry, the vehicle's electronic architecture is constantly being upgraded and iterated, and the number of functional modules carried by the in-vehicle software is increasing, with the frequency of version iterations accelerating. Against this backdrop, the coverage and execution frequency of core R&D testing links such as vehicle integration testing and domain control function testing are increasing simultaneously. The number of defects generated during the testing process is showing a significant upward trend, placing higher demands on the efficiency and standardization of handling in-vehicle testing defects.

[0003] Currently, defect handling in the automotive testing field mainly relies on two mainstream methods: manual compilation and traditional template-based AI generation. In the manual compilation mode, testers manually compile and record defect-related content such as defect phenomena, reproduction steps, and hardware / software environment information, based on the actual conditions of the testing site. Traditional template-based AI generation solutions, on the other hand, primarily utilize the basic capabilities of general-purpose large models to generate text-based defect content based on preset fixed templates, thereby assisting in the defect entry process.

[0004] The existing methods for handling defects in vehicle-mounted testing mainly have the following technical problems: First, while manual compilation offers high flexibility, it involves a large amount of repetitive work and low efficiency in defect entry. In scenarios such as version regression testing and batch feature iteration testing, the manpower and time costs of manual compilation increase significantly. Furthermore, manually entered defect information is generally characterized by casual and colloquial language, easily leading to missing key information, non-standard field entries, and poor clarification of defect reports. After receiving defect reports, developers cannot quickly understand the fault scenario and anomaly characteristics, requiring repeated cross-functional communication for confirmation, directly lengthening the troubleshooting and problem closure cycle.

[0005] Secondly, traditional template-based AI generation solutions also have many shortcomings. For example, existing general-purpose large models can only complete text fill-in-the-blank generation based on fixed templates, lacking the professional business semantic understanding capabilities of the automotive testing field. They cannot automatically decompose test business scenarios, intelligently extract key defect elements, and the generated content does not conform to general automotive testing specifications, failing to assist developers in quickly locating the root cause of failures. Furthermore, existing technologies have not built a dedicated business knowledge base, defect standard specification library, and multi-scenario business template library for automotive testing, making it impossible to adapt to the differentiated generation logic of different vehicle models, domain controllers, and test scenarios. The solutions themselves also lack the autonomous reasoning capabilities for summarizing abnormal elements, identifying related modules, and inferring suspected root causes. In addition, the design perspective of existing solutions is relatively singular, focusing only on improving the efficiency of test-side order entry, without architecturally designing the entire chain from test discovery and defect flow to development and debugging. This fails to fundamentally reduce the understanding cost and troubleshooting time for developers, limiting the overall practical value of the technology. Summary of the Invention

[0006] This invention provides an AI-powered intelligent analysis and structured generation method and system for vehicle-mounted test defects, which solves the technical problems of inefficient manual input of vehicle-mounted defects, non-standard descriptions, lack of vehicle-mounted business semantics in general AI, and inability to assist in troubleshooting across the entire process in the existing technology.

[0007] In a first aspect of the present invention, an AI-powered intelligent analysis and structured generation method for vehicle-mounted test defects is provided, comprising: Collect raw, unstructured data during vehicle-mounted testing; The AI ​​Agent performs semantic parsing on the raw unstructured data to identify the business scenario to which the current test belongs and extract the core elements of the defect. Based on the business scenario and the core elements of the defect, the large language model is invoked to associate with the pre-built vehicle test comprehensive knowledge base for semantic reasoning, and the normalized defect analysis results are obtained. Based on the defect analysis results, the system is restructured and layered according to the developer's troubleshooting logic. Based on the business scenario, the corresponding standard template is adaptively matched, and the structured and layered defect content is filled into the standard template to generate a standardized defect work order.

[0008] Preferably, before collecting the raw unstructured data during the vehicle-mounted testing process, the method further includes: Construct a comprehensive knowledge base for vehicle-mounted testing; the comprehensive knowledge base for vehicle-mounted testing includes a general vehicle-mounted testing business knowledge base, a defect standard field specification library, and a multi-business scenario template library; Standard defect fixed fields are uniformly defined in the defect standard field specification library. These standard defect fixed fields include problem phenomenon, preconditions, operation reproduction steps, probability of anomaly recurrence, hardware and software environment configuration, associated vehicle module, and suspected root cause of the fault.

[0009] Preferably, the raw unstructured data collected during the vehicle-mounted testing process includes: Collect test operation logs, abnormal screens of the vehicle interface, system error messages and function interaction records during the vehicle testing process as raw unstructured data.

[0010] Preferably, the step of using an AI Agent to perform semantic parsing on the original unstructured data, identifying the business scenario to which the current test belongs, and extracting the core elements of the defect includes: The AI ​​Agent's perception and understanding module performs semantic parsing on the raw unstructured data to automatically identify the business scenario to which the current test belongs; In the aforementioned business scenario, the core elements of the defect are intelligently extracted. These core elements include abnormal behavior, complete operation path, software and hardware version, operating environment, and associated vehicle function modules.

[0011] Preferably, based on the business scenario and the core elements of the defect, a large language model is invoked to perform semantic reasoning in association with a pre-built comprehensive knowledge base for vehicle testing, resulting in a standardized defect analysis result, specifically including: By calling the large language model and associating it with the general vehicle testing business knowledge base, semantic reasoning is performed on the core elements of the defect, automatically filling in missing fields, unifying industry professional terms, and eliminating colloquial expressions to obtain a standardized defect description. Based on the core elements of the defect and the regularized defect description, a large language model is invoked to deduce the suspected root causes of the fault and the scope of functional impact, thereby obtaining the defect analysis results.

[0012] Preferably, based on the defect analysis results, a structured, layered reorganization is performed according to the developer's troubleshooting logic, including: Based on the conventional thinking sequence of developers troubleshooting vehicle malfunctions, the defect content in the defect analysis results is reordered and hierarchically organized so that the defect content is presented in the order of abnormal phenomenon, reproduction operation steps, software and hardware environment, related modules, and suspected root cause.

[0013] Preferably, after generating a standardized defect work order, the method further includes: The generated standardized defect work orders are archived and stored in the database. Collect usage feedback from both the testing and development teams; Based on the user feedback, the feature extraction rules of the AI ​​Agent, the Prompt generation rules, and the configuration of the in-vehicle test comprehensive knowledge base are iteratively optimized.

[0014] In a second aspect of the present invention, an AI-powered intelligent analysis and structured generation system for vehicle-mounted test defects is provided, comprising: The test data acquisition module is used to collect raw, unstructured data during the vehicle-mounted testing process; The scenario element extraction module is used to perform semantic parsing on the original unstructured data through an AI Agent, identify the business scenario to which the current test belongs, and extract the core elements of the defect. The defect semantic reasoning module is used to perform semantic reasoning by calling a large language model and associating it with a pre-built vehicle test comprehensive knowledge base based on the business scenario and the core elements of the defect, so as to obtain the normalized defect analysis results. The defect layering and reorganization module is used to perform structured layering and reorganization based on the defect analysis results and the developer's troubleshooting logic. The work order template adaptation module adaptively matches the corresponding standard template according to the business scenario, fills the structured and layered defect content into the standard template, and generates a standardized defect work order. The defect closed-loop iteration module is used to uniformly archive the generated standardized defect work orders into the database, collect usage feedback from the testing and development ends, and iteratively optimize the element extraction rules, prompt generation rules, and configuration of the vehicle-mounted test comprehensive knowledge base based on the usage feedback.

[0015] In a third aspect of the present invention, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable by the processor, wherein the processor executes the computer program to implement the steps of the method as described in the first aspect of the present invention.

[0016] In a fourth aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method as described in the first aspect of the present invention.

[0017] The AI-powered intelligent analysis and structured generation method and system for vehicle-mounted test defects provided by this invention have the following advantages compared with existing technologies: 1. Improve testing efficiency. This invention uses an AI Agent to automatically collect raw unstructured data and complete scene recognition and extraction of core defect elements, replacing the large amount of repetitive and mechanical data entry work in the traditional manual compilation method. This reduces the workload and time cost of manual compilation for testers and is suitable for high-frequency regression, multi-version iteration and other large-scale testing scenarios.

[0018] 2. Enhanced Defect Clarability. This invention utilizes a large language model to perform semantic reasoning on the core elements of defects, automatically completing missing fields, standardizing industry-specific terminology, eliminating colloquial expressions, and structurally reorganizing them according to the developer's troubleshooting logic. This results in standardized defect work orders with a standardized structure, complete information, and clear logic, allowing developers to quickly understand the fault scenario and abnormal characteristics without repeated cross-functional communication for confirmation, effectively shortening the debugging and root cause localization cycle.

[0019] 3. Possesses professional semantic understanding and autonomous reasoning capabilities in the field of vehicle testing. This invention constructs a comprehensive vehicle testing knowledge base that includes a general vehicle testing business knowledge base, a defect standard field specification library, and a multi-business scenario template library. Using an AI Agent as the core engine, it enhances the large language model, enabling the system to possess professional semantic understanding capabilities in the field of vehicle testing. This allows it to automatically decompose testing business scenarios, intelligently extract core defect elements, and further deduce suspected root causes of faults and the scope of functional impact, overcoming the limitations of traditional general-purpose large models that are merely text-fill-in-the-blank generation based on fixed templates.

[0020] 4. General and Scalable Architecture. This invention is not dependent on specific vehicle models or brands, and can be generalized and adapted to various vehicle and domain control testing businesses. The comprehensive vehicle testing knowledge base and multi-business scenario template library support continuous expansion, allowing for rapid integration of new business scenarios. It boasts strong reusability and low maintenance costs. Through scenario-adaptive template matching and differentiated weight adjustment mechanisms, core defect information can be highlighted for different testing businesses, improving the relevance and readability of defect content.

[0021] 5. Achieving a closed-loop value chain. This invention achieves standardized accumulation and traceable management of defect assets by uniformly archiving standardized defect work orders into a database. By collecting usage feedback from the testing and development ends and iteratively optimizing Agent element extraction rules, Prompt rules, and knowledge base configuration accordingly, the system's defect analysis accuracy and work order generation quality can be continuously improved. This realizes a closed-loop value chain from efficient test work order entry and rapid development troubleshooting to standardized accumulation of defect assets, forming a core technological barrier for AI Agents in the automotive testing vertical. Attached Figure Description

[0022] Figure 1Flowchart of the AI ​​intelligent analysis and structured generation method for vehicle-mounted test defects provided by the present invention; Figure 2 The structural block diagram of the AI ​​intelligent analysis and structured generation system for vehicle-mounted test defects provided by the present invention; Figure 3 This is a structural block diagram of the electronic device provided by the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Figure 1 This is a flowchart illustrating the AI-powered intelligent analysis and structured generation method for vehicle-mounted test defects provided by this invention. (Refer to...) Figure 1 The method includes the following steps: S1 collects raw, unstructured data during the vehicle-mounted testing process.

[0025] Raw unstructured data refers to multi-source test-related data generated during vehicle-mounted testing that has not been standardized or processed. This data may include test operation logs, abnormal vehicle interface screens, system error messages, and function interaction records. Specifically, relevant data generated throughout the entire testing process is collected synchronously during vehicle-mounted testing and used as input data sources for subsequent intelligent defect analysis.

[0026] S2, the AI ​​Agent performs semantic parsing on the original unstructured data, identifies the business scenario to which the current test belongs, and extracts the core elements of the defect.

[0027] Here, AI Agent (Artificial Intelligence Agent) refers to an artificial intelligence entity capable of acquiring environmental information through a perception module, making autonomous decisions based on a knowledge base and reasoning model, and executing specific tasks to achieve preset goals. In this embodiment of the invention, the AI ​​Agent serves as the core control engine, used for semantic parsing and intelligent analysis of raw unstructured data.

[0028] Specifically, the collected raw unstructured data is input into the AI ​​Agent for semantic analysis. The AI ​​Agent, through its built-in perception and understanding module, performs semantic analysis on the raw unstructured data, automatically determining the business scenario category corresponding to the current test, such as in-vehicle central control interaction test scenario or domain control function test scenario. After identifying the business scenario, the AI ​​Agent further extracts core information elements related to the defect from the raw unstructured data, namely the core defect elements. These core defect elements are the key information items necessary to describe a defect.

[0029] S3. Based on the business scenario and the core elements of the defect, the large language model is invoked to associate with the pre-built vehicle test comprehensive knowledge base for semantic reasoning to obtain the normalized defect analysis results.

[0030] Among them, the Large Language Model (LLM) is a large-scale artificial intelligence model with natural language understanding and generation capabilities, pre-trained on massive amounts of data. The Vehicle Testing Comprehensive Knowledge Base is a pre-built, dedicated knowledge base storing professional business knowledge and standards in the field of vehicle testing.

[0031] Specifically, after obtaining the business scenario and core defect elements in step S2, the system invokes the large language model, using the business scenario and core defect elements as input. This allows the large language model to perform semantic reasoning in conjunction with the vehicle testing comprehensive knowledge base. Here, semantic reasoning refers to the process by which the large language model, based on its semantic understanding of the input information and combined with professional knowledge from the knowledge base, performs logical inference and content generation. Through semantic reasoning, the core defect elements can be standardized using professional terminology and missing fields can be filled in, resulting in standardized defect analysis results.

[0032] S4. Based on the defect analysis results, perform a structured, layered reorganization according to the developer's troubleshooting logic.

[0033] Specifically, based on the conventional thinking sequence of developers troubleshooting vehicle malfunctions, the various types of information in the defect analysis results are reordered and hierarchically organized to ensure the logical coherence and completeness of the defect information.

[0034] S5. Based on the business scenario, adaptively match the corresponding standard template, fill the structured and layered defect content into the standard template, and generate a standardized defect work order.

[0035] Specifically, after identifying the business scenario in step S2, a standard template corresponding to the business scenario is adaptively matched from the vehicle testing comprehensive knowledge base, and the field weights and expression logic are adjusted accordingly. The standard template is a predefined, structured format template used to standardize the presentation of defective content.

[0036] After matching the corresponding standard template, the system fills the defect content, which has been structured and hierarchically reorganized in step S4, into the corresponding field position of the standard template, generating a standardized defect work order that conforms to the general specifications for vehicle project management. This work order can be submitted by testers to the defect management platform or used directly by developers for troubleshooting.

[0037] The technical solution provided in this invention automatically collects raw unstructured data and completes scene recognition and element extraction through an AI Agent, replacing repetitive manual compilation, significantly reducing test data entry workload and improving testing efficiency. By calling a large language model for semantic reasoning, defect content is professionally standardized and structurally reorganized according to the developer's troubleshooting logic, making defect ticket information complete and logically clear, effectively shortening the debugging and root cause localization cycle. Through the collaborative architecture of AI Agent and large language model, the solution possesses professional semantic understanding and autonomous reasoning capabilities in the automotive testing field, overcoming the limitations of traditional template filling and generalizing to adapt to various vehicle models and domain control testing businesses.

[0038] Based on the above embodiments, as a preferred implementation, before collecting the raw unstructured data during the vehicle testing process in step S1, the AI ​​intelligent analysis and structure generation method for vehicle testing defects provided in this embodiment of the invention further includes: A comprehensive knowledge base for vehicle-mounted testing is constructed, which includes three sub-bases: a general vehicle-mounted testing business knowledge base, a defect standard field specification library, and a multi-business scenario template library.

[0039] Specifically, the general automotive testing business knowledge base stores industry knowledge such as professional terminology, business logic, functional module division, test specifications, and fault association rules in the automotive testing field, providing a professional knowledge foundation for semantic reasoning of large language models. The defect standard field specification library stores the field definitions and filling specifications for standard defects, uniformly constraining the field setting standards of defect work orders and ensuring the consistency of output formats. The multi-business scenario template library stores defect description templates corresponding to different testing business scenarios. The field weights and expression logic of each template are configured differently according to the characteristics of different business scenarios, adapting to the defect generation needs of multiple scenarios.

[0040] Specifically, standard defect fixed fields are uniformly defined in the defect standard field specification library. These standard defect fixed fields include problem phenomenon, preconditions, operation reproduction steps, probability of anomaly recurrence, hardware and software environment configuration, associated vehicle module, and suspected root cause of failure. The generation of all defect content follows the above field specifications.

[0041] This embodiment pre-constructs a comprehensive vehicle testing knowledge base that includes a business knowledge base, a field specification base, and a multi-scenario template base. This provides professional knowledge support for subsequent AI Agent scene recognition and element extraction, semantic reasoning of large language models, and professional terminology standardization, enabling the defect analysis results to have professional semantic understanding capabilities in the field of vehicle testing.

[0042] Based on the above embodiments, as a preferred implementation, in step S2, the step of performing semantic parsing on the original unstructured data using an AI Agent to identify the business scenario to which the current test belongs and extracting the core elements of the defect specifically includes: After the collected raw unstructured data is input into the AI ​​Agent, the AI ​​Agent calls its built-in perception and understanding module to perform semantic-level parsing processing on the raw unstructured data. The perception and understanding module automatically determines the business scenario category to which the current test belongs by identifying key semantic information in the raw unstructured data. Optionally, the business scenario may include in-vehicle central control interaction test scenarios, domain control function test scenarios, etc. This embodiment of the invention does not specifically limit the granularity of business scenario division and can be flexibly adjusted according to actual test business needs.

[0043] After determining the business scenario to which the current test belongs, the AI ​​Agent intelligently extracts the core elements of the defect based on the business logic of the corresponding scenario. These core elements include the abnormal behavior, complete operation path, software and hardware version, operating environment, and associated vehicle-mounted functional modules. Specifically, the abnormal behavior describes the specific external manifestation of the defect; the complete operation path records the complete sequence of operation steps that triggered the defect; the software and hardware version identifies the software and hardware version information at the time the defect occurred; the operating environment records the system's operating status and environment configuration at the time the defect occurred; and the associated vehicle-mounted functional modules identify the vehicle-mounted functional modules involved in the defect.

[0044] This embodiment uses the perception and understanding module of the AI ​​Agent to automatically identify the scene and extract the defect elements, realizing the automated transformation of unstructured test data into structured defect information. This avoids problems such as arbitrary expression, missing information, and non-standard fields caused by manual processing, and ensures the integrity and standardization of the core defect information.

[0045] Based on the above embodiments, as a preferred implementation, in step S3, based on the business scenario and the core elements of the defect, semantic reasoning is performed by invoking a large language model associated with a pre-built vehicle testing comprehensive knowledge base to obtain the normalized defect analysis results, specifically including: S31, invoke the large language model to associate with the general vehicle testing business knowledge base in the vehicle testing comprehensive knowledge base, perform semantic reasoning on the core elements of the defect, automatically fill in missing fields, unify industry professional terms, and remove colloquial expressions to obtain a standardized defect description.

[0046] Specifically, the business scenario and core defect elements obtained in step S2 are input into the large language model, enabling the large language model to perform reasoning in conjunction with the general automotive testing business knowledge base. This general automotive testing business knowledge base stores industry knowledge such as professional terminology, business logic, functional module division, and fault association rules in the automotive testing field. Automatic field completion refers to automatically identifying and supplementing necessary information items missing from the defect description based on existing defect elements and business rules; standardizing industry terminology refers to converting non-standard and inconsistent expressions into standard terms commonly used in the automotive testing field; and eliminating colloquial expressions refers to removing informal and vague expressions from the defect description, ensuring the content conforms to professional written language standards. Through the above processing, a standardized defect description is obtained.

[0047] S32, based on the core elements of the defect and the regularized defect description, the large language model is invoked to deduce the suspected root causes of the fault and the scope of functional impact, and the defect analysis results are obtained.

[0048] Specifically, after obtaining a standardized defect description, in-depth logical deduction is conducted based on the core elements of the defect and the standardized defect description to analyze the possible causes of the fault, i.e., the suspected root causes of the fault, and the potential impact of the fault on other functional modules of the vehicle system, i.e., the scope of functional impact. Based on this, the suspected root causes and scope of functional impact obtained through deduction are added to the defect description as analytical conclusions, generating a complete defect analysis result that conforms to the general specifications for vehicle defect transfer.

[0049] This invention uses semantic reasoning based on a large language model to complete fields, standardize terminology, and eliminate colloquial expressions. Combined with development troubleshooting logic, it structures and layers defective content to generate standardized work orders that are complete in information and logically clear. This reduces the cost of cross-departmental communication in development and shortens the cycle of troubleshooting and root cause location.

[0050] Based on the above embodiments, as a preferred implementation, step S4, which involves performing structured hierarchical reorganization according to the developer's troubleshooting logic based on the defect analysis results, specifically includes: Specifically, after obtaining the defect analysis results, the information in the results is structured and hierarchically reorganized based on the conventional thinking sequence of developers troubleshooting vehicle malfunctions. Here, the developer's troubleshooting logic refers to the thought process and reading habits followed by developers when locating problems and performing root cause analysis after receiving a defect report. When troubleshooting, developers typically first focus on the specific manifestations of the defect, then trace the operational process that triggered it, then examine the software and hardware operating environment at the time the defect occurred, then identify the functional modules that the defect might involve, and finally, based on the above information, locate the cause of the fault. Following this thinking habit, the defect content is presented in the following order: abnormal phenomena, reproduction steps, software and hardware environment, related modules, and suspected root causes. Among these, abnormal phenomena describe the specific external manifestations of the defect, reproduction steps record the complete sequence of operations that triggered the defect, the software and hardware environment identifies the software version and hardware configuration status at the time the defect occurred, related modules identify the vehicle functional modules involved in the defect, and suspected root causes indicate the preliminary deduced possible causes of the fault. By reordering and hierarchically organizing the defective content in the above order, the generated defective content has a closed business logic loop, and the key information is presented completely and unambiguously.

[0051] This embodiment restructures and hierarchically reorganizes defect content according to the developer's troubleshooting logic, ensuring that the layout of defect work orders aligns with the developer's natural troubleshooting thought process from phenomenon to cause and from operation to environment. This guarantees the logical coherence and completeness of defect information, enabling developers to quickly understand the fault scenario and abnormal characteristics after receiving the information. This effectively shortens the defect troubleshooting and root cause location cycle and reduces cross-functional communication costs.

[0052] Based on the above embodiments, as a preferred implementation, in step S5, according to the business scenario, the corresponding standard template is adaptively matched, and the structured and layered reorganized defect content is filled into the standard template to generate a standardized defect work order, specifically including: S51, based on the identified business scenario, adaptively match the corresponding standard template from the multi-business scenario template library, and adjust the field weights and expression logic.

[0053] After identifying the business scenario of the current test in step S2, the system adaptively matches a standard template corresponding to the business scenario from the multi-business scenario template library in the pre-built vehicle testing comprehensive knowledge base. The weights of fields and the expression logic are adjusted based on the scenario characteristics. For example, the vehicle central control interaction test scenario emphasizes the operation path and interface response presentation, with a higher weight for the "reproduce operation steps" field and an emphasis on the completeness of the operation sequence. The domain control function test scenario emphasizes the presentation of system signals and data interaction, with higher weights for the "associate vehicle modules" and "software and hardware environment configuration" fields and an emphasis on the interaction logic between modules. Through these differentiated adjustments to field weights and expression logic, the final generated defect work order can fully highlight the key defect information under different business scenarios, improving the relevance and readability of the defect content.

[0054] S52, fill the structured and hierarchically reorganized defect content into the matched standard template, and output a standardized defect work order that conforms to the general specifications for vehicle project management.

[0055] Specifically, the defect content, which has undergone structured hierarchical reorganization in step S4, is filled into the corresponding field positions of the matching template, ultimately outputting a standardized defect work order that conforms to the general specifications for vehicle project management. The generated standardized defect work order can be submitted directly to the defect management platform by testers, or used directly by developers for troubleshooting.

[0056] This embodiment utilizes a scenario-adaptive template matching and differentiated weight adjustment mechanism to highlight core defect information for different testing tasks, improving the relevance and readability of defect content. Simultaneously, it ensures a consistent defect work order format across all scenarios, conforming to project management standards, eliminating the need for manual format adjustments, effectively reducing cross-functional understanding costs, and further improving the overall efficiency of defect entry and processing.

[0057] Based on the above embodiments, as a preferred implementation, after generating a standardized defect work order in step S5, the AI ​​intelligent analysis and structured generation method for vehicle-mounted test defects provided by the present invention further includes: Step S6: Archive the generated standardized defect work orders into the database and collect usage feedback from the testing and development ends. Based on the usage feedback, iteratively optimize the element extraction rules of the AI ​​Agent, the Prompt generation rules, and the configuration of the vehicle-mounted test comprehensive knowledge base.

[0058] Specifically, the generated standardized defect work orders are automatically connected to the defect management platform and uniformly archived, forming a traceable and reusable defect asset library, providing historical data support for the subsequent location and handling of similar defects. Feedback is collected simultaneously from both ends, covering improvement suggestions, accuracy evaluations, and usability feedback. For example, testers can provide feedback on whether the business scenario identified by the AIAgent is accurate and whether the extracted core defect elements are complete; developers can provide feedback on whether the generated defect analysis results meet investigation expectations, whether the suspected root causes of the fault are accurate, and whether the readability and understandability of the work orders meet requirements.

[0059] Furthermore, based on collected user feedback, the configuration of the AI ​​Agent's element extraction rules, Prompt generation rules, and the comprehensive vehicle testing knowledge base was iteratively optimized. The element extraction rules guide the AI ​​Agent's behavior logic in extracting core defect elements from raw unstructured data, including determining the dimensions of the extracted elements and the semantic recognition patterns for each dimension. The Prompt is the input instruction that drives the large language model to perform semantic reasoning and defect content generation. The Prompt generation rules standardize the task description, output format requirements, and knowledge base call logic for this instruction. The configuration of the comprehensive vehicle testing knowledge base includes updating the terminology of the general business knowledge base, adjusting the rules of the field specification library, optimizing the templates in the scenario template library, and supplementing and improving the fault association rules.

[0060] This embodiment achieves standardized accumulation and traceable management of defect assets by uniformly archiving standardized defect work orders into a database, providing historical data support for the rapid location and handling of similar defects in the future. By collecting usage feedback from the testing and development ends and iteratively optimizing the Agent element extraction rules, Prompt generation rules, and knowledge base configuration accordingly, the system's defect analysis accuracy and work order generation quality can be continuously improved, realizing a closed-loop value chain from defect entry, development troubleshooting to standardized accumulation of defect assets.

[0061] In a preferred embodiment of the present invention, a complete application scenario is used to illustrate the technical solution provided by the present invention. This embodiment selects an in-vehicle central control interaction test scenario and specifically implements the AI ​​intelligent analysis and structured generation method for in-vehicle test defects of the present invention. The complete implementation steps are as follows: S1. Collect raw, unstructured data during the vehicle-mounted testing process. During the testing of the vehicle's in-vehicle central control interaction function, the system synchronously collects multi-source raw unstructured data generated throughout the entire testing process. This includes testers' touch operation logs, abnormal screens such as lag and blackouts on the in-vehicle central control interface, vehicle system error messages, and central control function click interaction records—all raw test data that has not been standardized. All collected unstructured data is then aggregated and used as the primary input data source for subsequent AI-powered intelligent defect analysis.

[0062] S2. Identify business scenarios and extract core defect elements through AI Agent semantic parsing. The raw, unstructured data collected by S1 is input into the AI ​​Agent of the core control engine. The AI ​​Agent performs comprehensive semantic analysis on the raw data through its built-in perception and understanding module. By identifying key semantic information in the data, it automatically determines that the current test scenario is an in-vehicle central control interaction test scenario. Based on this scenario, the AI ​​Agent intelligently extracts complete core defect elements according to the business logic of the corresponding scenario. These elements include: abnormal behavior such as central control interface lag and black screen, the complete touch operation path that triggers the fault, the current in-vehicle hardware and software version information, system operating environment parameters, and the in-vehicle functional modules associated with the central control interaction. This achieves the automatic conversion of unstructured test data into structured core defect elements.

[0063] S3. Using a large language model combined with a comprehensive knowledge base for vehicle testing, semantic reasoning is performed to obtain defect analysis results. The pre-built comprehensive vehicle testing knowledge base includes a general vehicle testing business knowledge base, a defect standard field specification library, and a multi-business scenario template library. The defect standard field specification library has uniformly defined fixed fields for standard defects. The system inputs the vehicle central control interaction testing business scenarios identified by S2 and the extracted core defect elements into a large language model. The large language model then collaborates with the general vehicle testing business knowledge base to perform semantic reasoning. First, it automatically completes missing fields in the defect description, standardizes vehicle-related professional terminology, and removes colloquial expressions from the original data to generate a standardized defect description. Then, based on the standardized defect description and core defect elements, it deeply infers the suspected root causes of the fault and the scope of the fault's impact on the functionality of modules such as the central control interaction and vehicle display. Finally, it generates a complete defect analysis result that conforms to the vehicle defect flow specifications.

[0064] S4. Reorganize the defect analysis results in a structured, layered manner according to the developers' troubleshooting logic. Based on the complete defect analysis results obtained from S3, and following the troubleshooting thought process and reading habits of developers—from phenomenon to cause, from operation to environment—the defect information is structurally and hierarchically reorganized. Following a fixed logical order of abnormal phenomena, reproducible operation steps, hardware and software environment configuration, associated vehicle modules, and suspected root causes, various types of information regarding central control interface defects are reordered and hierarchically organized to eliminate information clutter and logical gaps, forming a logically closed-loop, unambiguous structured defect content that aligns with the troubleshooting logic of developers for fault location and root cause analysis.

[0065] S5: Adaptively matches standard templates to generate standardized defect work orders. Based on the in-vehicle central control interaction testing business scenario identified by S2, the system adaptively matches the corresponding exclusive standard defect template from a multi-business scenario template library. Considering the business characteristics of this scenario, which emphasizes operation paths and interface responses, the system differentiates and increases the field weight of the reproduced operation steps, and optimizes the description logic of operation sequence-related content. Subsequently, the structured defect content, reorganized by S4 in layers, is filled into the corresponding field positions of the standard template one by one, ultimately generating a standardized defect work order that conforms to the general specifications for in-vehicle project management, which can be directly used for test submission and development fault diagnosis.

[0066] S6. Defect work orders are archived and stored in the database, and system configuration is iteratively optimized based on two-way feedback. Standardized central control interaction defect work orders generated by S5 are automatically integrated into the defect management platform and uniformly archived into the database, building a traceable and reusable vehicle test defect asset library, providing data support for the rapid handling of similar central control interaction defects in the future. Feedback from both the testing and development ends is collected simultaneously. The testing end provides feedback on the accuracy of scenario identification and the completeness of defect element extraction, while the development end provides feedback on the rationality of root cause inference, work order readability, and troubleshooting practicality. Based on the collected feedback from both ends, three core configurations are iteratively optimized: optimizing the defect element extraction dimensions and semantic recognition mode of the AI ​​Agent, optimizing the Prompt generation rules driving the large language model inference, and updating and improving the terminology, field specifications, scenario templates, and fault association rules of the comprehensive vehicle test knowledge base. This continuously improves the accuracy and clarifiability of system defect generation, forming a closed loop of full-chain business value.

[0067] Figure 2 The structural block diagram of the AI ​​intelligent analysis and structured generation system for vehicle-mounted test defects provided by the present invention is shown below. Figure 2 The system includes: The test data acquisition module 201 is used to collect raw unstructured data during the vehicle-mounted testing process; The scene element extraction module 202 is used to perform semantic parsing on the original unstructured data through the AI ​​Agent, identify the business scene to which the current test belongs, and extract the core elements of the defect. The defect semantic reasoning module 203 is used to perform semantic reasoning by calling a large language model and associating it with a pre-built vehicle test comprehensive knowledge base based on the business scenario and the core elements of the defect, so as to obtain the normalized defect analysis results. The defect layering and reorganization module 204 is used to perform structured layering and reorganization according to the defect analysis results and the developer's troubleshooting logic. The work order template adaptation module 205 adaptively matches the corresponding standard template according to the business scenario, fills the structured and layered defect content into the standard template, and generates a standardized defect work order. The defect closed-loop iteration module 206 is used to uniformly archive the generated standardized defect work orders into the database, collect usage feedback from the testing and development ends, and iteratively optimize the element extraction rules, prompt generation rules, and configuration of the vehicle-mounted test comprehensive knowledge base based on the usage feedback.

[0068] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the above system and modules can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0069] Figure 3 A structural block diagram of the electronic device provided by the present invention, such as Figure 3 As shown, the present invention also provides an electronic device, which can be a mobile terminal, desktop computer, laptop, handheld computer, server, or other computing device. The electronic device 300 includes a processor 301 and a memory 302, wherein the memory 302 stores an AI intelligent analysis and structured generation program 303 for vehicle-mounted test defects.

[0070] In some embodiments, memory 302 may be an internal storage unit of a computer device, such as a hard drive or memory. In other embodiments, memory 302 may be an external storage device of a computer device, such as a plug-in hard drive, smart media card (SMC), secure digital card (SD), flash card, etc. Further, memory 302 may include both internal and external storage units of the computer device. Memory 302 is used to store application software and various types of data installed on the computer device, such as program code installed on the computer device. Memory 302 can also be used to temporarily store data that has been output or will be output. In one embodiment, when the AI ​​intelligent analysis and structured generation program 303 for vehicle-mounted test defects is executed by processor 301, the following steps are implemented: Collect raw, unstructured data during vehicle-mounted testing; The AI ​​Agent performs semantic parsing on the raw unstructured data to identify the business scenario to which the current test belongs and extract the core elements of the defect. Based on the business scenario and the core elements of the defect, the large language model is invoked to associate with the pre-built vehicle test comprehensive knowledge base for semantic reasoning, and the normalized defect analysis results are obtained. Based on the defect analysis results, the system is restructured and layered according to the developer's troubleshooting logic. Based on the business scenario, the corresponding standard template is adaptively matched, and the structured and layered defect content is filled into the standard template to generate a standardized defect work order.

[0071] In some embodiments, processor 301 may be a central processing unit (CPU), microprocessor or other data processing chip, used to run program code stored in memory 302 or process data, such as executing an AI intelligent analysis and structured generation program for vehicle test defects.

[0072] This embodiment also provides a computer-readable storage medium storing an AI intelligent analysis and structure generation program for vehicle test defects. When the AI ​​intelligent analysis and structure generation program for vehicle test defects is executed by a processor, it implements the steps of the AI ​​intelligent analysis and structure generation method for vehicle test defects.

[0073] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An AI-powered intelligent analysis and structured generation method for vehicle-mounted test defects, characterized in that, include: Collect raw, unstructured data during vehicle-mounted testing; The AI ​​Agent performs semantic parsing on the raw unstructured data to identify the business scenario to which the current test belongs and extract the core elements of the defect. Based on the business scenario and the core elements of the defect, the large language model is invoked to associate with the pre-built vehicle test comprehensive knowledge base for semantic reasoning, and the normalized defect analysis results are obtained. Based on the defect analysis results, the system is restructured and layered according to the developer's troubleshooting logic. Based on the business scenario, the corresponding standard template is adaptively matched, and the structured and layered defect content is filled into the standard template to generate a standardized defect work order.

2. The AI ​​intelligent analysis and structured generation method for vehicle-mounted test defects according to claim 1, characterized in that, Before collecting raw, unstructured data during the vehicle-mounted testing process, the method further includes: Construct a comprehensive knowledge base for vehicle-mounted testing; the comprehensive knowledge base for vehicle-mounted testing includes a general vehicle-mounted testing business knowledge base, a defect standard field specification library, and a multi-business scenario template library; Standard defect fixed fields are uniformly defined in the defect standard field specification library. These standard defect fixed fields include problem phenomenon, preconditions, operation reproduction steps, probability of anomaly recurrence, hardware and software environment configuration, associated vehicle module, and suspected root cause of the fault.

3. The AI-powered intelligent analysis and structured generation method for vehicle-mounted test defects according to claim 1, characterized in that, The raw, unstructured data collected during the vehicle-mounted testing process includes: Collect test operation logs, abnormal screens of the vehicle interface, system error messages and function interaction records during the vehicle testing process as raw unstructured data.

4. The AI ​​intelligent analysis and structured generation method for vehicle-mounted test defects according to claim 1, characterized in that, The AI ​​Agent performs semantic parsing on the raw unstructured data to identify the business scenario to which the current test belongs and extracts the core elements of the defects, including: The AI ​​Agent's perception and understanding module performs semantic parsing on the raw unstructured data to automatically identify the business scenario to which the current test belongs; In the aforementioned business scenario, the core elements of the defect are intelligently extracted. These core elements include abnormal behavior, complete operation path, software and hardware version, operating environment, and associated vehicle function modules.

5. The AI ​​intelligent analysis and structured generation method for vehicle-mounted test defects according to claim 2, characterized in that, Based on the aforementioned business scenario and the core elements of the defect, a large language model is invoked to perform semantic reasoning in conjunction with a pre-built comprehensive knowledge base for vehicle testing, resulting in a standardized defect analysis outcome, specifically including: By calling the large language model and associating it with the general vehicle testing business knowledge base, semantic reasoning is performed on the core elements of the defect, automatically filling in missing fields, unifying industry professional terms, and eliminating colloquial expressions to obtain a standardized defect description. Based on the core elements of the defect and the regularized defect description, a large language model is invoked to deduce the suspected root causes of the fault and the scope of functional impact, thereby obtaining the defect analysis results.

6. The AI-powered intelligent analysis and structured generation method for vehicle-mounted test defects according to claim 1, characterized in that, Based on the defect analysis results, a structured, layered reorganization was performed according to the developer's troubleshooting logic, including: Based on the conventional thinking sequence of developers troubleshooting vehicle malfunctions, the defect content in the defect analysis results is reordered and hierarchically organized so that the defect content is presented in the order of abnormal phenomenon, reproduction operation steps, software and hardware environment, related modules, and suspected root cause.

7. The AI ​​intelligent analysis and structured generation method for vehicle-mounted test defects according to claim 1, characterized in that, After generating standardized defect work orders, the method further includes: The generated standardized defect work orders are uniformly archived and stored in the database; Collect usage feedback from both the testing and development teams; Based on the user feedback, the feature extraction rules of the AI ​​Agent, the Prompt generation rules, and the configuration of the in-vehicle test comprehensive knowledge base are iteratively optimized.

8. An AI-powered intelligent analysis and structured generation system for vehicle-mounted test defects, the system being used to execute the AI-powered intelligent analysis and structured generation method for vehicle-mounted test defects as described in any one of claims 1-7, characterized in that, The system includes: The test data acquisition module is used to collect raw, unstructured data during the vehicle-mounted testing process; The scenario element extraction module is used to perform semantic parsing on the original unstructured data through an AI Agent, identify the business scenario to which the current test belongs, and extract the core elements of the defect. The defect semantic reasoning module is used to perform semantic reasoning by calling a large language model and associating it with a pre-built vehicle test comprehensive knowledge base based on the business scenario and the core elements of the defect, so as to obtain the normalized defect analysis results. The defect layering and reorganization module is used to perform structured layering and reorganization based on the defect analysis results and the developer's troubleshooting logic. The work order template adaptation module adaptively matches the corresponding standard template according to the business scenario, fills the structured and layered defect content into the standard template, and generates a standardized defect work order. The defect closed-loop iteration module is used to uniformly archive the generated standardized defect work orders into the database, collect usage feedback from the testing and development ends, and iteratively optimize the element extraction rules, prompt generation rules, and configuration of the vehicle-mounted test comprehensive knowledge base based on the usage feedback.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the AI ​​intelligent analysis and structured generation method for vehicle-mounted test defects as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the AI ​​intelligent analysis and structured generation method for vehicle test defects as described in any one of claims 1 to 7.