Vehicle fault analysis method and system based on artificial intelligence agency, medium and equipment

Abstract

The invention provides a vehicle fault analysis method and system based on artificial intelligence agency, a medium and equipment. The method comprises the following steps: acquiring multi-source heterogeneous data of a vehicle; constructing a vehicle fault knowledge base and a vehicle maintenance knowledge base; obtaining fault object-fault mode association information corresponding to the multi-source heterogeneous data based on a vehicle fault knowledge base; performing duplicate removal and aggregation on the fault object-fault mode associated information to obtain effective fault object-fault mode associated information; based on a vehicle fault knowledge base, obtaining a fault cause corresponding to the effective fault object-fault mode association information; based on a vehicle maintenance knowledge base, obtaining a maintenance scheme corresponding to the effective fault object-fault mode association information; and generating a vehicle fault analysis report by adopting a large language model based on the fault cause and the maintenance scheme. According to the vehicle fault analysis method and system based on the artificial intelligence agent, the medium and the equipment, automatic analysis of the fault is realized based on the multi-source heterogeneous data of the vehicle.

Description

Technical Field

[0001] This invention belongs to the technical field of vehicle diagnostics, and in particular relates to a vehicle fault analysis method, system, medium, and device based on artificial intelligence agents. Background Technology

[0002] With the continuous growth of motor vehicle ownership and the increasing stringent requirements for vehicle emission control, the efficiency of vehicle fault diagnosis and repair is of great significance for ensuring road traffic safety and improving emission compliance. Currently, remote vehicle diagnostics mainly relies on fault code information (such as SAE-compliant J2012 and J1939 protocols) provided by on-board diagnostic systems (OBD) for preliminary judgment.

[0003] However, the existing diagnostic system has several significant problems that urgently require technological innovation to improve.

[0004] (1) Traditional diagnostic methods do not make sufficient use of dynamic monitoring data, thus limiting their diagnostic capabilities. Existing systems generally focus on reading Diagnostic Trouble Codes (DTCs) as the basis for judgment, while failing to effectively integrate and analyze the large amount of dynamic monitoring data generated during vehicle operation (such as engine speed, load, fuel consumption, after-treatment efficiency, and emission concentration). This makes it impossible to accurately identify potential faults based on actual operating conditions, especially when complex problems such as efficiency decline, regeneration abnormalities, and urea injection failure occur in emission after-treatment systems (such as selective catalytic reduction systems SCR and diesel particulate filters DPF). It is difficult to make a comprehensive judgment based solely on static fault codes.

[0005] (2) Enterprise-defined fault code systems are complex and diverse, with low standardization, making unified identification difficult. In addition to the common J2012 and J1939 fault code standards, different vehicle manufacturers and engine manufacturers (such as Weichai, Yuchai, Cummins, etc.) have generally established their own proprietary fault code systems, which may use different coding rules and meanings for the same fault. This multi-source and heterogeneous fault code system increases the difficulty of diagnosis, and traditional static fault code tables cannot meet the needs of unified identification and interpretation.

[0006] (3) The causes of failures and maintenance recommendations are constantly changing, and static knowledge bases are difficult to adapt to actual needs. With the increasing complexity of vehicle operating conditions and the continuous application of new technologies, the causes of some failures may change due to factors such as environment, oil quality, and vehicle aging. In addition, maintenance experience and methods are also constantly being accumulated and updated. Traditional fault handling methods that mainly rely on fixed rule matching or manual judgment are difficult to reflect new fault modes and maintenance paths in actual operation in a timely manner.

[0007] (4) Repair recommendations still rely on personal experience and lack systematic knowledge support. In current repair practice, most repair personnel rely on personal experience or subjective judgment to handle problems, lacking systematic integration of knowledge sources such as OEM repair manuals, official operating guidelines, and online platforms (such as repair communities and forums). The uneven level of experience leads to unstable repair quality, and may even result in misjudgment or repeated repairs, increasing repair costs.

[0008] In summary, existing vehicle fault diagnosis and repair suggestion systems still suffer from problems such as insufficient information integration, untimely knowledge updates, and inadequate analytical capabilities. Summary of the Invention

[0009] In view of the shortcomings of the prior art described above, the purpose of this invention is to provide a vehicle fault analysis method, system, medium, and device based on artificial intelligence agents, which realizes automated analysis of vehicle faults based on multi-source heterogeneous vehicle data, thereby providing accurate and comprehensive diagnostic reports and maintenance suggestions.

[0010] In a first aspect, the present invention provides a vehicle fault analysis method based on an artificial intelligence agent. The method includes the following steps: acquiring multi-source heterogeneous data of a vehicle; constructing a vehicle fault knowledge base and a vehicle maintenance knowledge base; the vehicle fault knowledge base is used to record fault object-fault mode-fault cause association information; the vehicle maintenance knowledge base is used to record fault object-fault mode-repair solution association information; based on the vehicle fault knowledge base, acquiring fault object-fault mode association information corresponding to the multi-source heterogeneous data; deduplicating and aggregating the fault object-fault mode association information to obtain valid fault object-fault mode association information; based on the vehicle fault knowledge base, acquiring the fault cause corresponding to the valid fault object-fault mode association information; based on the vehicle maintenance knowledge base, acquiring the repair solution corresponding to the valid fault object-fault mode association information; and generating a vehicle fault analysis report using a large language model based on the fault cause and the repair solution.

[0011] In one implementation of the first aspect, the multi-source heterogeneous data includes remote online monitoring data, a fault code list, engine brand, fault protocol type, and textual fault description.

[0012] In one implementation of the first aspect, obtaining the fault object-fault mode association information corresponding to the multi-source heterogeneous data based on the vehicle fault knowledge base includes the following steps: Based on the calibration code ID and calibration verification number in the vehicle's electronic control unit, it is determined whether the electronic control unit has been tampered with, so as to obtain the electronic control unit's fault object-fault mode association information for the tampered fault object; Based on the online monitoring data, extract the corresponding fault object-fault mode association information; Based on the textual fault description, extract the association information between the fault object and the fault symptom; Based on the fault codes in the fault code list, the engine brand, and the fault protocol type, the corresponding fault object-fault mode association information is searched in the vehicle fault knowledge base.

[0013] In one implementation of the first aspect, the electronic control unit is determined to have been tampered with based on whether the edit distance offset between the calibration code ID and the calibration verification number exceeds a preset value.

[0014] In one implementation of the first aspect, constructing a vehicle fault knowledge base includes the following steps: Obtain the vehicle's fault code database; the fault code database records the fault code, fault description, and fault cause for each fault; Based on the fault code database, a large language model is used to generate fault object-fault mode-fault cause association information for each fault.

[0015] In one implementation of the first aspect, constructing a vehicle maintenance knowledge base includes the following steps: Obtain vehicle maintenance manuals, vehicle technical bulletins, and repair case studies; Convert the vehicle maintenance manual, the vehicle technical bulletin, and the maintenance case into plain text information; Construct a prompt word template, wherein the prompt words are extracted from plain text information and related to faults and repairs, and output according to a preset format; Based on the plain text information and the prompt word template, a large language model is used to generate information related to fault objects, fault modes, and repair solutions.

[0016] In one implementation of the first aspect, the process of deduplicating and aggregating the fault object-fault mode association information to obtain valid fault object-fault mode association information includes the following steps: The fault object-fault mode association information is vectorized based on the SBERT model or the Transformer model; Cluster the vectorized fault object-fault mode association information to obtain multiple fault clusters; For each fault cluster, select the fault object-fault mode association information corresponding to the fault as the effective fault object-fault mode association information.

[0017] Secondly, the present invention provides a vehicle fault analysis system based on artificial intelligence agents, the system comprising a first acquisition module, a construction module, a second acquisition module, a third acquisition module, a fourth acquisition module, a fifth acquisition module, and an analysis module; The first acquisition module is used to acquire multi-source heterogeneous data of the vehicle; The construction module is used to build a vehicle fault knowledge base and a vehicle repair knowledge base; the vehicle fault knowledge base is used to record the association information of fault object-fault mode-fault cause; the vehicle repair knowledge base is used to record the association information of fault object-fault mode-repair solution. The second acquisition module is used to acquire fault object-fault mode association information corresponding to the multi-source heterogeneous data based on the vehicle fault knowledge base; The third acquisition module is used to deduplicate and aggregate the fault object-fault mode association information to obtain valid fault object-fault mode association information. The fourth acquisition module is used to acquire the fault cause corresponding to the effective fault object-fault mode association information based on the vehicle fault knowledge base; The fifth acquisition module is used to acquire the repair plan corresponding to the effective fault object-fault mode association information based on the vehicle maintenance knowledge base; The analysis module is used to generate a vehicle fault analysis report based on the cause of the fault and the repair plan, using a large language model.

[0018] Thirdly, the present invention provides an electronic device, the electronic device comprising: a processor and a memory; The memory is used to store computer programs; The processor is used to execute the computer program stored in the memory, so that the electronic device performs the above-described vehicle fault analysis method based on artificial intelligence agents.

[0019] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by an electronic device, implements the above-described vehicle fault analysis method based on an artificial intelligence agent.

[0020] As described above, the vehicle fault analysis method, system, medium, and equipment based on artificial intelligence agents of the present invention have the following beneficial effects.

[0021] (1) Based on a large language model and artificial intelligence agent, it can achieve automated and in-depth vehicle fault analysis based on multi-source heterogeneous data from the vehicle (including static fault codes, dynamic monitoring data and user descriptions), and can obtain accurate and comprehensive diagnostic reports and maintenance suggestions.

[0022] (2) It can integrate dynamic monitoring data, standard and custom fault codes, real-time semantic analysis and knowledge base reasoning capabilities to achieve multi-level and dynamic fault diagnosis, thereby promoting the realization of intelligent vehicle diagnosis and remote refined operation and maintenance. Attached Figure Description

[0023] Figure 1 The flowchart shown is an embodiment of the vehicle fault analysis method based on artificial intelligence agents of the present invention.

[0024] Figure 2 The diagram shown is a structural schematic of an embodiment of the vehicle fault analysis system based on artificial intelligence agents of the present invention.

[0025] Figure 3 The diagram shown is a structural schematic of an embodiment of the electronic device of the present invention. Detailed Implementation

[0026] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0027] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0028] Remote online monitoring technology has the potential to intelligently diagnose and push maintenance suggestions to a massive number of vehicles within a region. In recent years, with the construction and widespread application of remote emission monitoring platforms for heavy-duty diesel vehicles in many places, the operating status, emission levels, and fault information of vehicles have been continuously, in real time, and online. Based on this data, combined with artificial intelligence technology and the semantic understanding and reasoning capabilities of large-scale models, it is possible to achieve intelligent diagnosis and push maintenance suggestions to a large number of vehicles within a region, breaking through the limitations of traditional manual vehicle-by-vehicle analysis and greatly improving diagnostic efficiency and accuracy.

[0029] The technical solutions of the present invention will now be described in detail with reference to the accompanying drawings.

[0030] like Figure 1 As shown, in one embodiment, the vehicle fault analysis method based on artificial intelligence agent of the present invention includes steps S1-S7.

[0031] Step S1: Obtain multi-source heterogeneous data of the vehicle.

[0032] Specifically, this invention collects vehicle fault data from various sources to construct the multi-source heterogeneous data, ensuring the comprehensiveness of diagnostic information. In one embodiment, the multi-source heterogeneous data includes remote online monitoring data, fault code lists, engine brands, fault protocol types, and textual fault descriptions.

[0033] The fault protocol type and engine brand are the initial contextual information for the vehicle fault diagnosis process, such as the vehicle's communication protocol standard (e.g., CAN bus protocol) and the specific engine model. This information is crucial for limiting the subsequent search scope of the vehicle fault knowledge base or serving as context for generating prompts in a large language model. Table 1 shows the vehicle's static basic information, based on which information such as vehicle model, engine model, and bus protocol can be obtained.

[0034] Table 1. Basic Static Information of the Vehicle Serial Number project 1 license plate number 2 VIN code 3 Vehicle model 4 Total mass 5 Vehicle type 6 Engine Model 7 Engine power 8 Engine displacement 9 Vehicle CAN bus protocol .

[0035] The system generates standardized (J1939 / J2012) and enterprise-defined fault code lists. These fault codes adhere to industry standards such as SAE J1939 and SAE J2012, as well as enterprise-specific fault codes. This invention utilizes an artificial intelligence agent to directly search these fault codes one by one in a fault knowledge base, quickly obtaining the association information between fault code, fault object, and fault symptom. Table 2 shows the information provided by the on-board diagnostic system. In this invention, fault-related information for all vehicles in the region is read according to the requirements of "Limits and Measurement Methods for Pollutant Emissions from Heavy-Duty Diesel Vehicles (China VI)" (GB 17691—2018).

[0036] Table 2. Information provided by the on-board diagnostic system project Data types 1 OBD Diagnostic Protocol BYTE 2 MIL status BYTE 3 Diagnostic readiness WORD 4 Vehicle Identification Number (VIN) STRING 5 Software calibration identification number (CAL ID) STRING 6 CAPTCHA (Verification Code No.) STRING 7 IUPR value DSTRING 8 Total number of fault codes BYTE 9 Fault Code Information List N*BYTE(4) .

[0037] The remote online monitoring data refers to various performance parameters and sensor readings recorded in real time or historically during vehicle operation. This data reflects the actual operating status of various vehicle components. This invention, by invoking an artificial intelligence agent, transforms the problems reflected in this online data into vectorized fault object-fault mode association information, thereby converting dynamic, numerical data into structured information that can be processed by the diagnostic system. For example, remote online monitoring data for all vehicles in the region is read according to the requirements of "Limits and Measurement Methods for Pollutant Emissions from Heavy-Duty Diesel Vehicles (China VI)" (GB 17691—2018). The remote online monitoring data is shown in Table 3.

[0038] Table 3. Remote Online Monitoring Data Serial Number Data Items unit Data types 1 Speed km / h WORD 2 Atmospheric pressure kPa BYTE 3 Engine net output torque % BYTE 4 Friction torque % BYTE 5 Engine speed r / min WORD 6 Engine fuel flow L / h WORD 7 Output value of NOx sensor upstream of SCR <![CDATA[10 -6 ]]> WORD 8 SCR downstream NOx sensor output value <![CDATA[10 -6 ]]> WORD 9 Reagent balance % BYTE 10 intake volume kg / h WORD 11 SCR inlet temperature ℃ WORD 12 SCR outlet temperature ℃ WORD 13 DPF pressure difference kPa WORD 14 Engine coolant temperature ℃ BYTE 15 fuel tank level % BYTE 16 Location status BYTE 17 longitude ° DWORD 18 latitude ° DWORD 19 Cumulative Mileage km DWORD .

[0039] The textual fault descriptions come from unstructured text descriptions by users, such as sensory information like "abnormal noise" or "smoke" entered through a guided interface.

[0040] Step S2: Construct a vehicle fault knowledge base and a vehicle repair knowledge base; the vehicle fault knowledge base is used to record the association information of fault object-fault mode-fault cause; the vehicle repair knowledge base is used to record the association information of fault object-fault mode-repair solution.

[0041] Specifically, by leveraging the semantic understanding capabilities of the Large Language Model (LLM), traditional, unstructured vehicle fault and maintenance data are processed into machine-readable, highly structured vehicle fault knowledge bases and vehicle maintenance knowledge bases.

[0042] In one embodiment, building a vehicle fault knowledge base includes the following steps.

[0043] 21) Obtain the vehicle's fault code database; the fault code database records the fault code, fault description, and fault cause for each fault.

[0044] The raw data will be sourced from SAE standards and the company’s own fault code database (which includes fault codes, fault descriptions, and fault causes).

[0045] For example, the SAE J1939 database (SAE J1939 fault database, SAE J1939 system enterprise-specific fault database) specifically stores fault codes conforming to the SAE J1939 standard and enterprise-specific fault codes conforming to this system. The fault database contains the following fields: 1. Serial number; 2. Fault code; 3. Fault description; 4. Fault cause.

[0046] The SAE J2012 database (SAE J2012 Fault Database, SAE J2012 System Enterprise-Specific Fault Database) is similar to the J1939 database. This fault database stores fault codes conforming to the SAE J2012 standard and enterprise-specific fault codes conforming to this system. The fault database contains the following fields: 1. Serial Number; 2. Fault Code; 3. Fault Description; 4. Fault Cause.

[0047] It should be noted that the fault database needs to be formatted before being used as input to the large language model. For example, for the source data: 45, P0135, O2 sensor heater circuit fault (row 1, sensor 1), heater element damaged or circuit open, it can be concatenated into the following input text: Fault code: P0135. Fault description: O2 sensor heater circuit fault (row 1, sensor 1). Fault cause: heater element damaged or circuit open.

[0048] 22) Based on the fault code database, a large language model is used to generate fault object-fault mode-fault cause association information for each fault.

[0049] The faulty object refers to the physical entity (such as an oxygen sensor or EGR valve) or logical entity (such as a CAN communication bus or ignition control module software) that malfunctions. The fault mode refers to the specific state or behavior exhibited by the faulty object (such as excessively high signal voltage, stuck in the open position, communication loss, or slow response). The fault cause refers to the reason why the faulty object exhibits the fault mode (such as a broken CAN cable or sensor aging).

[0050] Design specific prompt templates to guide the large language model to automatically decompose and extract standard "fault object-fault mode-fault cause" triple association information for each fault record. These prompt templates need to teach the large language model how to accurately identify "who is broken (fault object)," "how it broke (fault mode)," and "the specific cause" (fault cause). For example, the prompt might be: "You are a professional automotive fault diagnosis engineer, proficient in decomposing complex fault descriptions into standardized 'fault objects' and 'fault modes.' Your task is to analyze the fault data I provide and, strictly following the specified JSON format, extract the 'fault code,' 'fault object,' and 'fault mode.' The fault object should be a specific physical or logical component. The fault mode should be the specific failure state exhibited by that component. Both must be directly extracted from the provided information, not guessed."

[0051] For example, the process of generating the association information between fault object, fault mode, and fault cause is as follows: **Case 1:** **Input Text:** Fault code: P0108. Fault description: High input to manifold absolute pressure / atmospheric pressure sensor circuit. Cause of failure: Internal short circuit to power supply in the sensor. **Output JSON:** { "fault_code": "P0108", "fault_object": "Manifold absolute pressure / atmospheric pressure sensor circuit", "fault_mode": "High input (short circuit to power supply)" "fault_reason": "Power short circuit" }

[0052] **Case 2:** **Input Text:** Fault Code: P0404. Fault Description: Exhaust Gas Recirculation (EGR) control circuit range / performance. Cause: The EGR valve is stuck due to carbon buildup and cannot reach the target position. **Output JSON:** { "fault_code": "P0404", "fault_object": "Exhaust Gas Recirculation (EGR) Valve", "fault_mode": "Control scope / performance issues" "fault_reason": "Carbon buildup and jamming" }

[0053] **Case 3:** **Input Text:** Fault code: U0100. Fault description: Loss of communication with ECM / PCM 'A'. Cause of the fault: Open circuit in the CAN bus line or faulty ECM module. **Output JSON:** { "fault_code": "U0100", "fault_object": "Communication link with ECM / PCM 'A'", "fault_mode": "Lost communication" "fault_reason": "CAN bus line open circuit"}.

[0054] In one embodiment, building a vehicle maintenance knowledge base includes the following steps.

[0055] 21a) Obtain vehicle maintenance manuals, vehicle technical bulletins, and maintenance case studies.

[0056] This includes obtaining official repair manuals from the target engine and vehicle manufacturers. These manuals may be in PDF, XML, HTML, or proprietary formats. Relevant Technical Bulletins (TSBs), repair case studies, and other supplementary data should also be collected.

[0057] 22b) Convert the vehicle maintenance manual, the vehicle technical bulletin, and the maintenance case into plain text information.

[0058] The above data is preprocessed as follows: (1) Format conversion: All documents are converted to plain text format. For PDF files, a high-precision OCR (Optical Character Recognition) tool is required, and special attention should be paid to preserving the structure of tables and lists. (2) Text cleaning: Noise information such as headers, footers, watermarks, and irrelevant image descriptions is removed. (3) Text chunking: The lengthy manual is divided into smaller, more LLM-friendly text chunks according to chapters, fault codes (DTCs), or logical units (such as "Cooling System Fault Diagnosis"). A text chunk should contain a relatively complete fault description and repair process.

[0059] 23c) Construct a prompt word template, wherein the prompt words are extracted from plain text information and related to faults and maintenance, and output according to a preset format.

[0060] Specifically, a highly structured prompt template, including role definitions, format constraints, a small number of sample demonstrations, and a dynamic input area, guides the large language model to accurately extract structured data containing "fault object," "fault mode," and repair plan (including required tools, spare parts, steps, and safety precautions). The structured data extracted by the large language model, including fault object, fault mode, and required tools, spare parts, repair steps, and safety precautions from the repair plan, are stored in the corresponding fields of the vehicle repair knowledge base, thus forming a structured vehicle repair knowledge base.

[0061] The prompt word template includes the following content.

[0062] I. Role and Task Definition Zone.

[0063] Objective: To define precise identities, task objectives, and behavioral guidelines for large language models. This is the highest-level instruction, setting the tone for all subsequent behaviors.

[0064] Components: (a) Persona: Clearly state your expert status. For example: "You are an experienced automotive repair technician and data analyst with expertise in all types of vehicles." (b) Core Task: Clearly describe the work that needs to be done. For example: "Your sole task is to extract key information about a specific fault from the text snippets of the repair manual I provide." (c) Behavioral Constraints: Set strict rules to exclude unnecessary creative, inferential, or conversational behaviors. For example: "You must strictly follow the JSON format I define for output. You are absolutely not allowed to add any explanations, comments, or text unrelated to the JSON format. If information for a field does not exist in the original text, use the specified value (such as 'not mentioned'). You are prohibited from inferring or creating information on your own."

[0065] II. Output Format Constraint Zone.

[0066] Objective: To define the "skeleton" of data extraction, ensuring that the output is machine-readable, directly parseable structured data. This is the foundation for automation.

[0067] constitute: Format declaration: JSON format is explicitly required.

[0068] Schema Definition: Provides a complete JSON structure template with annotations. The annotations explain the meaning of each field to the LLM, ensuring its understanding is correct.

[0069] JSON { / / (a) The macroscopic component or system where the failure occurred, such as: SCR aftertreatment system, engine cooling system "fault_object": "string", / / (b) A detailed description of the fault, which may include fault codes (DTCs) and symptoms, for example: DTC P20EE - Low SCR catalyst efficiency. "fault_mode": "string", / / (c) Object containing all maintenance plan information "repair_solution": { / / (c1) List of all tools required during the repair process "required_tools": [ "string" ], / / (c2) A list of spare parts that need to be replaced should include part numbers if possible. "required_spare_parts": [ "string (part number)" ], / / (c3) A detailed, step-by-step list of repair procedures, each step being a separate string. "repair_steps": [ "string" ], / / List of safety regulations that must be followed during maintenance "safety_precautions": [ "string" ]}}.

[0070] 24d) Based on the plain text information and the prompt word template, a large language model is used to generate the association information of fault object-fault mode-maintenance plan.

[0071] In addition, this invention also provides a Few-shot Demonstration Zone for "pre-job training" of large language models. By providing 1 to 3 high-quality "input->output" examples, the large language model learns in practice how to apply the rules defined in the output role and task definition area and the format constraint area.

[0072] The examples include the following.

[0073] Input text: A real, pre-processed maintenance manual text.

[0074] Output JSON: A "perfect answer" manually written by human experts based on the input text, strictly adhering to the JSON structure defined in the format constraints.

[0075] Examples are key to improving extraction accuracy. The language in repair manuals can be interpreted in multiple ways; examples teach large language models how to choose the most appropriate interpretation in a specific context. Through examples, large language models can learn the names of specific parts and tools, as well as typical expressions of repair procedures. Compared to simply looking at the format definition, actually seeing a completed example allows large language models to better internalize and adhere to the output format.

[0076] In one embodiment, when the "repair plan" based on the enterprise's repair manual cannot provide a matching result for a given "fault object-fault mode", the present invention will activate the secondary reasoning generation mechanism of the vehicle repair knowledge base. This mechanism relies on a diesel vehicle fault knowledge base with a structure of "fault object-fault mode-fault cause" and uses a large language model to intelligently and creatively generate a complete and structured set of repair suggestions based on the known "fault causes".

[0077] In this embodiment, a prompt word template is first generated. The core purpose of the prompt word template is no longer extraction, but to guide the large language model in logical reasoning and knowledge creation. The prompt word template section includes the following content.

[0078] I. Role and Task Definition Zone.

[0079] Objective: To elevate the role of the large language model from a "data analyst" to a "diagnostic master and technical trainer," whose task is no longer "extraction" but "reasoning and creation" based on limited information, and to set a professional, rigorous and logical tone for subsequent actions.

[0080] constitute: (a) Persona: Assign a higher level of expert status. For example: "You are a top diesel vehicle diagnostic master and technical trainer with over 20 years of experience." (b) Core Task: Clearly describe the creative work required to complete the task. For example: "Your task is to reason and generate a complete, structured, logical, and safe preliminary repair recommendation based on the known 'failure objects,' 'failure modes,' and 'failure causes' that I provide." (c) Behavioral Constraints & Reasoning Guide: In addition to existing format constraints, a clear logical reasoning framework is added to guide the thought process. For example: "You must strictly follow the JSON format I define for output. Furthermore, your generation process should follow these logical guidelines: 1. Confirmation and Diagnosis: First, consider how to confirm the cause of the failure. 2. Repair Preparation: Infer the required tools and spare parts based on the cause of the failure. 3. Repair Steps: Generate the steps in the order of 'Preparation-Disassembly-Core Repair-Installation-Testing'. 4. Safety First: Consider all potential risks and generate safety precautions."

[0081] II. Output Format Constraint Zone.

[0082] Objective: To ensure that the maintenance recommendations generated by the large language model are fully compatible in data structure with the standard recommendations extracted from the manual, so that the front-end application can process and display them uniformly. Additionally, metadata fields are added to distinguish data sources and credibility.

[0083] constitute: Format declaration: JSON format is explicitly required.

[0084] Schema Definition: Provides a complete JSON structure template with newly added metadata fields.

[0085] JSON { / / (a) Clearly state the source of this recommendation to distinguish it from official data in the repair manual. "generation_source": "AI-Generated Suggestion", / / (b) The reliability assessment given by LLM based on the explicitness of the input information and its own knowledge base "confidence_score": "High / Medium / Low", / / (c) An object containing all maintenance plan information, with a structure consistent with the extract template. "repair_suggestion": { / / (c1) List of tools deduced based on the cause of the fault and repair procedures "required_tools": [ "string" ], / / (c2) List of potentially needed spare parts inferred from the cause of the failure "required_spare_parts": [ "string (part number)" ], / / (c3) Step-by-step maintenance process generated based on reasoning logic "repair_steps": [ "string" ], / / (c4) Safety specifications derived from maintenance procedures and objects "safety_precautions": [ "string" ] } }

[0086] Similarly, this embodiment also needs to provide a few examples for "advanced training" of the large language model, not only teaching the format, but also teaching a reasoning and creative process from "cause" to "solution". These examples demonstrate to the large language model how to deduce a complex set of "solutions" (e.g., need for cleaning agent, disassembly steps, cleaning methods, installation torque, etc.) from a simple "cause" (e.g., excessive carbon buildup), helping the large language model establish the implicit relationship between "fault causes" and "required tools / spare parts". High-quality examples ensure that the repair steps generated by the large language model conform to actual operational logic and safety specifications. Accordingly, the structured data extracted by the large language model, including fault objects, fault modes, and the required tools, spare parts, repair steps, and safety precautions in the derived repair solutions, are stored in the corresponding fields of the vehicle repair knowledge base, thus forming a structured vehicle repair knowledge base.

[0087] Step S3: Based on the vehicle fault knowledge base, obtain the fault object-fault mode association information corresponding to the multi-source heterogeneous data.

[0088] Specifically, based on the vehicle fault knowledge base, obtaining the fault object-fault mode association information corresponding to the multi-source heterogeneous data includes the following steps.

[0089] 31) Determine whether the electronic control unit (ECU) has been tampered with based on the calibration code ID and calibration verification number in the ECU, so as to obtain the fault object-fault mode association information of the ECU for the tampered fault.

[0090] The process involves comparing the strings corresponding to the vehicle control unit's calibration code (CAL) ID and calibration verification number (CVN) with the initially recorded string value. If a change occurs and the edit distance deviation from the initially recorded string exceeds a set value (adjusted as needed), the vehicle ECU is considered tampered with, forming a "ECU-tampered" fault object-fault mode association information. Preferably, the edit distance deviation K(%) = max(Loriginal, Lnew) / D(Soriginal, Snew) × 100%, where Soriginal represents the original string recorded initially, Snew represents the newly acquired string, D(Soriginal, Snew) represents the edit distance (Levenshtein Distance), referring to the minimum number of single-character editing operations (insertion, deletion, or replacement) required to convert Soriginal to Snew, and max(Loriginal, Lnew) represents the length of the longer of the two strings. Using the length of the longer string as the numerator, the result can be normalized to between 0% (completely identical) and 100% (completely different).

[0091] 32) Extract the corresponding fault object-fault mode association information based on the online monitoring data.

[0092] In this invention, the online monitoring data and issues related to key components are extracted, and then the Agent transforms these online data issues into vectorized fault object-fault mode association information, thereby ensuring that real-time or historical monitoring data can be effectively integrated into the diagnostic process.

[0093] Online monitoring data can be extracted into corresponding fault object-fault mode association information through single-index statistical rules, multi-parameter ratio rules, one-dimensional feature similarity rules, and two-dimensional feature similarity rules. Single-index statistical rules determine whether individual sensor data is abnormal based on statistical parameters (such as quantiles), for example, judging the degree of SCR carrier failure based on the quantile value of the NOx sensor upstream of the SCR. Multi-parameter ratio rules calculate the ratios between cumulative parameters (such as cumulative mileage and cumulative fuel consumption) to form key diagnostic indicators (such as NOx emission factor per unit power, NOx conversion rate, and urea-fuel ratio) to evaluate system performance. Feature similarity rules use algorithms such as Ring's Consistency Correlation Coefficient (CCC) to compare the consistency of key vehicle parameters with data from standard vehicles under different operating conditions to evaluate performance such as thermal management.

[0094] Specifically, the single-indicator statistical rule categorizes a segment of data into a single indicator, and determines the severity of the problem based on the range of values ​​for that indicator. Possible statistical values ​​include the mean, median, and x-quantile. Statistical indicators and values ​​are shown in Table 4.

[0095] Table 4. Statistical Indicators and Values ​​for Single-Indicator Statistical Rules

[0096] Accordingly, as shown in Tables 5 and 6, in the "Fault Object - Fault Mode" output rules of the single-index statistical rules, when the x-quantile is less than 1200 ppm, a corresponding symptom judgment is given, such as "SCR - Moderate carrier failure"; when the x-quantile is less than 2 kPa, a corresponding symptom judgment is given, such as "DPF - Moderate carrier blockage". It should be noted that Tables 5 and 6 only provide some parameter rules as examples; other parameters can be set according to the actual application scenario.

[0097] Table 5. Output Rules for Fault Object-Fault Mode in Single-Indicator Statistical Rules of SCR Diagnosis Statistical parameters 99% quantile <1200 ppm 85th percentile <1200 ppm 70th percentile <1200 ppm 70% percentile ≥1200ppm Output value of NOx sensor upstream of SCR SCR normal SCR slight failure SCR moderate failure SCR severe failure .

[0098] Table 6. Output Rules for Fault Object-Fault Mode in Single-Indicator Statistical Rules of DPF Diagnosis 60th quantile <2kPa 50% quantile <2kPa 40% quantile <2kPa 40% percentile ≥2kPa DPF pressure difference DPF is normal. DPF slightly blocked DPF moderate blockage DPF severe blockage .

[0099] The parameters for which the multi-parameter ratio rule applies are shown in Table 7.

[0100] Table 7. Parameters corresponding to the multi-parameter ratio rule

[0101] The first row and first column of Table 7 list the seven core basic data points that the system accumulates in real time, as follows.

[0102] (1) Accumulated mileage (Mil_AC): The total distance traveled by the vehicle.

[0103] Mil_AC=SUM(V / 3600).

[0104] (2) Cumulative fuel consumption (Mfuel_AC): The total amount of fuel consumed by the engine.

[0105] Mil_AC=SUM(Lfuel*0.85 / 3600).

[0106] (3) Cumulative power (Power_AC): The total power output of the engine during operation.

[0107] Mil_AC=SUM(NE*Tq_max*TQ / 9550 / 3600).

[0108] (4) Cumulative NOx emissions downstream of SCR (MNOxD_AC): The cumulative amount of nitrogen oxides (NOx) emitted into the atmosphere after treatment by the SCR (Selective Catalytic Reduction) system. The lower this value, the better.

[0109] MNOxD_AC=SUM((0.38185 • 𝐿𝑓𝑢𝑒l+ 0.44445 • Mair ) • 𝛷𝑁𝑂• 10−6).

[0110] (5) Cumulative NOx consumption upstream of SCR (MNOxU_AC): The cumulative nitrogen oxide (NOx) emissions before the engine enters the SCR system.

[0111] (6) Cumulative fuel level drop percentile (SOT_AC): The cumulative percentage of fuel consumed in the fuel tank. It is only accumulated when the fuel level drops.

[0112] SOT_AC=-SUM(ΔSOT) when ΔSOT<0.

[0113] (7) Cumulative Reactive Agent Level Drop Percentage (SOR_AC): This usually refers to the cumulative percentage of reactive agent (urea solution) used in the SCR system in the urea (AdBlue®) tank. It is only accumulated when the level drops.

[0114] SOR_AC=-SUM(ΔSOR) when ΔSOR<0.

[0115] The middle section of Table 7 contains the core "ratio parameters," i.e., the diagnostic criteria. Details are as follows.

[0116] (1) Fuel consumption per unit mile (Mfuel_AC / Mil_AC): This refers to fuel consumption per 100 kilometers. An abnormal increase in this value may indicate a decrease in engine efficiency, insufficient tire pressure, or abnormal vehicle load.

[0117] (2) Power per unit distance (Power_AC / Mil_AC): This reflects the work done by the vehicle per unit distance and can indirectly reflect the road conditions (such as higher power when climbing hills or driving at high speeds).

[0118] (3) Downstream / upstream NOx emission factor per unit mileage (MNOxD_AC / Mil_AC or MNOxUAC / Mil_AC): This represents the amount of NOx emitted per kilometer driven. This is a key environmental indicator for assessing whether vehicle emissions meet standards.

[0119] (4) Percentile drop in urea tank level per unit mileage (RT_AC / Mil_AC): This reflects the percentage of urea consumed per kilometer traveled.

[0120] (5) Specific fuel consumption / thermal efficiency (Power_AC / Mfuel_AC): This is a very critical engine performance indicator. It represents the work produced per unit mass of fuel consumed. The higher this value, the better the engine's fuel economy (thermal efficiency). If this value decreases, it means that the engine's efficiency in converting fuel into power has decreased.

[0121] (6) Downstream / Upstream NOx Emission Factor per Unit Fuel Consumption (MNOxD_AC / Mfuel_AC or MNOxU_AC / Mfuel_AC): This represents the amount of NOx generated per unit of fuel consumed. This indicator eliminates the influence of driving mileage and more directly reflects the NOx generation characteristics of the engine under specific operating conditions and the processing efficiency of the SCR system.

[0122] (7) Percentile of urea tank level drop per unit fuel consumption (RT_AC / Mfuel_AC): measures the ratio of fuel consumption to urea consumption.

[0123] (8) Downstream / Upstream NOx Emission Factor per Unit Power (MNOxD_AC / Power_AC or MNOxU_AC / Power_AC): This represents the amount of NOx emitted per unit of power output by the engine (g / kWh). This is the core standard unit for measuring NOx emissions in heavy-duty diesel vehicle regulations. Monitoring this indicator is crucial for regulatory compliance.

[0124] (9) NOx conversion rate (1-MNOxD_AC / MNOxU_AC): This is the most crucial indicator for evaluating the performance of an SCR system. It directly assesses the efficiency of the SCR system by calculating the proportion of NOx entering the SCR that is successfully converted. If this value is significantly lower than the design value (usually above 90%), it clearly indicates a serious malfunction in the SCR system, such as catalyst aging, urea injection system failure, or sensor inaccuracy.

[0125] (10) NOx reduction per unit urea level drop ((MNOxU_AC-MNOxD_AC) / SOT_AC): This ratio measures how much NOx reduction is achieved by consuming a unit percentage of urea. It can be used to diagnose the accuracy of urea injection or the quality of the urea solution. If urea is consumed but the NOx reduction is low, it may mean insufficient injection volume, nozzle blockage, or the use of inferior / diluted urea.

[0126] (11) Urea-to-Fuel Ratio (SOR_AC / SOF_AC): This ratio monitors the relative proportion of urea consumption to fuel consumption. Under normal engine operating conditions and SCR system operation, this ratio should be within a relatively fixed range (e.g., 2%-8%, depending on the specific technology). If the ratio is abnormal, such as too high, it may indicate a urea system leak or over-injection. Too low or zero may indicate that the urea injection system is not working at all, the pipeline is blocked, a cheating device is being used, or the urea tank is empty. This is a parameter that regulatory agencies closely monitor to prevent emissions cheating.

[0127] In SCR diagnostics, the range of values ​​for the "downstream NOx emission factor per unit power" provides corresponding symptom judgments. For example, if the downstream NOx emission factor per unit power = 5 g / kWh, then "SCR - Slightly High Emissions". The fault object-fault mode output rules for multi-parameter ratio rules are shown in Table 8. It should be noted that Table 8 only provides some parameter rules as examples; other parameters can be set according to the actual application scenario.

[0128] Table 8. Fault Object-Fault Mode Output Rules for Multi-Parameter Ratio Rules in SCR Diagnosis Statistical parameters <2g / kWh <3.5g / kWh <7g / kWh ≥7g / kWh Downstream NOx emission factor per unit power SCR normal SCR has slightly high emissions. SCR moderate to high emissions SCR has high emissions. .

[0129] The one-dimensional and two-dimensional feature similarity rules employ Lin's Concordance Correlation Coefficient (CCC) to assess the consistency between two sets of data. It simultaneously measures the correlation (slope) and the degree of deviation from the perfect concordance line (y=x) (absolute value). The CCC value ranges from -1 to 1; the closer to 1, the better the consistency between the two sets of data, indicating high similarity in both trend and absolute value. The formula for the Lin's Concordance Correlation Coefficient is as follows: ,in x It is a set of certain parameters of a standard vehicle. y It is a set of parameters of a vehicle to be evaluated. σ x 2 and σ y 2 It is the variance of the two sets of data. μ x and μ y It is the mean of the two sets of data.

[0130] The correlation between variables can be determined using either Rule Type 1 or Rule Type 2, as shown in Tables 9 and 10, respectively. Rule Type 1 compares the correlation between vehicle parameters and a standard vehicle under different operating conditions. A better correlation indicates better vehicle health. This can be an "input 1-output" pattern or an "input 1-input 2-output" pattern. Rule Type 2 compares the correlation between vehicle parameters. Better correlation between some parameters indicates poorer performance in others, while other parameters are correlated. For example, a better correlation between instantaneous fuel consumption and NOx indicates that the SCR is not activated; when the correlation reaches 0.8 or higher, the SCR is completely ineffective.

[0131] Table 9, Rule Type 1

[0132] Table 10, Rule Type 2

[0133] Therefore, the fault object-fault mode output rules are derived based on the relationships between variables. In the thermal management evaluation of exhaust gas purification devices, the value ranges of "speed - engine net output torque - SCR outlet temperature" for the vehicle under evaluation and "speed - engine net output torque - SCR outlet temperature" under all operating conditions for the standard vehicle provide corresponding symptom judgments, as shown in Table 11. For example, if the CCC of "speed - engine net output torque - SCR outlet temperature" under all operating conditions for both the vehicle under evaluation and the standard vehicle is <0, the output is "SCR - thermal management failure". It should be noted that Table 11 only provides some parameter rules as examples; other parameters can be set according to the actual application scenario.

[0134] Table 11. Fault Object-Fault Mode Output Rules Based on Variable Relationships Statistical parameters CCC<0 CCC<0.5 CCC<0.9 CCC≥0.9 CCC evaluation of the vehicle under evaluation and the standard vehicle under all operating conditions: "RPM - net engine output torque - SCR outlet temperature". SCR thermal management failure SCR has poor thermal management. SCR thermal management is normal. SCR has excellent thermal management. .

[0135] 33) Extract the association information between the fault object and the fault symptom based on the text fault description.

[0136] In this process, users or maintenance personnel input the "sensory-object-symptom" characteristics of the vehicle. For example, the user is guided to the first key decision point. Four options (which can be added as needed) are provided to the user: "abnormal noise," "smoke," "lights," and "odor." Based on the user's selection, different targeted guidance paths are initiated. The core task of each path is to identify the "object" of the fault and refine the description of the "symptom." Specific guidance paths can be expanded as needed. The specific guidance process for the fault object-fault symptom association information is shown in Table 12.

[0137] Table 12. Guidance Information on the Association Between Fault Object and Fault Symptoms

[0138] 34) Based on the fault codes in the fault code list, the engine brand and the fault protocol type, search for the corresponding fault object-fault mode association information in the vehicle fault knowledge base.

[0139] Specifically, based on the fault codes in the fault code list, the engine brand, and the fault protocol type, the corresponding fault object-fault mode association information is searched in the vehicle fault knowledge base, thereby ensuring immediate response and matching of standardized fault information.

[0140] For example, when the fault code begins with P, C, B, or U, the system searches the vehicle fault knowledge base conforming to the SAE J2012 standard. If it is not found in SAE J2012, the system searches the manufacturer-specific vehicle fault knowledge base based on the input vehicle or engine manufacturer information. As another example, when the fault code begins with SPN, the system searches the strategy fault knowledge base conforming to the SAE J1939 standard. If it is not found in SAE J2012, the system searches the manufacturer-specific vehicle fault knowledge base based on the input vehicle or engine manufacturer information.

[0141] Step S4: Deduplicate and aggregate the fault object-fault mode association information to obtain valid fault object-fault mode association information.

[0142] Specifically, the deduplication and aggregation of the fault object-fault mode association information aims to solve the problem of repeated or inconsistent fault descriptions caused by multi-source data. Natural language processing technology is used to mine the inherent correlation and aggregate them into core faults.

[0143] The steps to deduplicate and aggregate the fault object-fault mode association information to obtain valid fault object-fault mode association information include the following steps.

[0144] 41) Vectorize the fault object-fault mode association information based on the SBERT model or the Transformer model.

[0145] This involves calling a pre-trained language model f. embed Each fault description phrase ti ("fault object-fault mode") is processed and mapped to a high-dimensional numerical vector vi. This process can be represented as: vi = f embed (ti). Where vi∈Rd, meaning the vector vi is a point in d-dimensional real space, where d is a preset vector dimension (e.g., d=768). This high-dimensional vector vi is the feature vector or text embedding of the phrase ti. The high-dimensional vector can represent the corresponding phrase in the model f. embed In the defined semantic space, phrases with similar semantics have vectors that are also geometrically close in that space.

[0146] In one embodiment, the pre-trained language model f embed This could be a sentence embedding model (Sentence-BERT, SBERT) optimized for sentence similarity tasks. In another embodiment, a general Transformer-based language model (such as BERT) could also be used, and the output of a specific layer (e.g., the [CLS] flag in the last layer) could be extracted as the feature vector vi representing the text.

[0147] 42) Cluster the vectorized fault object-fault mode association information to obtain multiple fault clusters.

[0148] Specifically, cluster analysis is performed on the feature vector set V = {v1, v2, ..., vn}. This clustering algorithm groups the vectors based on their spatial distance, with each group forming a fault cluster. Fault descriptions within the same fault cluster are semantically considered highly correlated. First, the semantic similarity between any two vectors vi and vj is defined. In this embodiment, cosine similarity is used for calculation, with the formula: .

[0149] Here, vik is the k-th component of vector vi. Sim(vi,vj) ranges from [-1, 1], with values ​​closer to 1 indicating greater semantic similarity. Subsequently, the distance D(vi,vj) between vectors is defined based on similarity, using the following formula: D(vi,vj) = 1 − Sim(vi,vj). The distance D ranges from [0, 2], with smaller values ​​indicating closer proximity and greater semantic relevance.

[0150] In a preferred embodiment, the clustering algorithm may employ the density-based clustering algorithm (DBSCAN). The DBSCAN algorithm operates using two core parameters: neighborhood radius. And the minimum number of samples, MinPts, that constitute the core object. The algorithm will satisfy D(vi,vj)≤ Vector pairs are considered density-reachable, and clusters are formed based on this. The advantage of this algorithm is that it does not require pre-specifying the number of clusters, can automatically discover fault clusters Gk based on the distribution characteristics of the data itself, and can effectively identify outliers (marked as noise) as independent faults. For example, calculations will show that the pairwise distances D between the vectors va, vb, and vc corresponding to the phrases "Cylinder 1 - Misfire," "Engine - Vibration," and "Engine - Weak Acceleration" are all less than a preset threshold. Therefore, they are grouped into the same fault cluster.

[0151] 43) For each fault cluster, select the fault object-fault mode association information corresponding to the fault as the effective fault object-fault mode association information.

[0152] For each generated fault cluster G kBy applying a preset representative selection strategy, a fault is selected from the "fault object-fault mode" {t1,t2,...,tm} contained in the fault cluster as the representative fault of the fault cluster, thereby generating a final fault list after aggregation and deduplication.

[0153] This selection process can be achieved by scoring each fault description within the cluster using a scoring function Score(ti) and selecting the item with the highest score. The scoring function can be composed of a weighted average of one or more sub-policies, and its general form can be defined as follows: .

[0154] in, This represents a fault description within a fault cluster. This represents the weight coefficient of each strategy. f prio ( source(t i ) ) is the scoring function for the data source priority strategy, which is based on the fault... Source (Such as on-board diagnostic systems (OBD), on-board monitoring modules (OBM), and owner-reported data) return a preset value. For example, it can be set to... fprio (OBD)=10, f prio (OBM)=5, f prio (According to the car owner) = 1. This is a scoring function for semantic specificity strategies, used to evaluate the level of detail and clarity in fault descriptions, returning a quantified score. For example, descriptions containing specific part numbers (such as "cylinder number 1") or precise numerical values ​​will receive a higher score than macroscopic descriptions (such as "engine").

[0155] Therefore, the above method can effectively aggregate fault information from multiple sources that appears different but is essentially related. For example, aggregating multiple characteristics (misfire, vibration, weak acceleration) to the root cause of "misfire in cylinder 1" greatly improves the accuracy and automation of fault diagnosis, provides clear and non-redundant input for subsequent maintenance decisions, and overcomes the limitations of existing technologies that rely on human experience for correlation analysis.

[0156] Step S5: Based on the vehicle fault knowledge base, obtain the fault cause corresponding to the effective fault object-fault mode association information.

[0157] Specifically, by inputting the fault object-fault mode association information into the vehicle fault knowledge base, the corresponding fault cause can be retrieved.

[0158] Step S6: Based on the vehicle maintenance knowledge base, obtain the maintenance plan corresponding to the effective fault object-fault mode association information.

[0159] Specifically, by inputting the fault object-fault mode association information into the vehicle maintenance knowledge base, the corresponding maintenance solution can be retrieved.

[0160] Step S7: Based on the cause of the fault and the repair plan, generate a vehicle fault analysis report using a large language model.

[0161] Specifically, based on the causes of the fault and the repair plan, the large language model automatically integrates the vehicle's basic information, the main problems (the core fault after aggregation), the cause analysis of each problem, and detailed solutions (including repair steps, required tools and spare parts) according to a preset professional template, to generate a professional and easy-to-understand comprehensive diagnostic and repair suggestion report for end users (such as repair technicians or car owners).

[0162] The first step is to generate a prompt word template for the large language model, guiding it to act as a professional technical communicator and organically organize scattered data points into a logically clear and complete technical document. The prompt word template includes the following:

[0163] I. Role and Task Definition Zone.

[0164] Objective: To define the final output role for the large language model—a professional automotive technical writer and senior service consultant whose task is to transform technical data into fluent, professional written reports.

[0165] constitute: Persona: Clearly define your expert identity, for example: "You are a professional automotive technical writer and senior service advisor, skilled at transforming complex technical diagnostic data into clear, standardized, and user-friendly reports." Core Task: Clearly describe the work to be done, for example: "Your task is to generate a complete 'Vehicle Diagnosis and Repair Recommendation Report' based strictly on the diagnostic data in JSON format that I provide. The report must be clearly structured, professionally written, and include information from all input data." Behavioral Constraints: Set strict output rules, such as: "You must strictly follow the Markdown format I define for output. No comments, explanations, or text unrelated to the report format are allowed. All information must come from the input data I provide and may not be deleted or created by yourself."

[0166] II. Output Format Constraint Zone.

[0167] Objective: To define the precise "skeleton" of the final report, ensuring that the structure, headings, and format of all reports remain highly consistent for easy reading and archiving.

[0168] Structure: Provides a complete report structure template defined using the Markdown language.

[0169] Similarly, this invention provides a few examples to help the large language model generate qualified vehicle fault analysis reports. For example, it provides the large language model with a complete "input JSON -> output Markdown report" example, allowing it to intuitively learn how to populate structured data into a formatted report template. This example includes the following.

[0170] • Composition: Input JSON: A complete JSON object containing all diagnostic outputs.

[0171] Output Markdown Report: A "perfect report example" hand-written by human experts based on the input JSON, strictly following the Markdown template defined in the format constraints.

[0172] It should be noted that each step of the vehicle fault analysis method based on artificial intelligence agents of the present invention can be implemented using a corresponding agent, thereby realizing vehicle fault analysis based on artificial intelligence agents. The scope of protection of the vehicle fault analysis method based on artificial intelligence agents described in the embodiments of the present invention is not limited to the execution order of the steps listed in this embodiment. Any solutions implemented by adding, deleting, or replacing steps in the prior art based on the principles of the present invention are included within the scope of protection of the present invention.

[0173] This invention also provides a vehicle fault analysis system based on artificial intelligence agents. The vehicle fault analysis system based on artificial intelligence agents can implement the vehicle fault analysis method based on artificial intelligence agents described in this invention. However, the implementation device of the vehicle fault analysis system based on artificial intelligence agents described in this invention includes, but is not limited to, the structure of the vehicle fault analysis system based on artificial intelligence agents listed in this embodiment. All structural modifications and substitutions of the prior art made in accordance with the principles of this invention are included within the protection scope of this invention.

[0174] like Figure 2 As shown, in one embodiment, the vehicle fault analysis system based on artificial intelligence agent of the present invention includes a first acquisition module 21, a construction module 22, a second acquisition module 23, a third acquisition module 24, a fourth acquisition module 25, a fifth acquisition module 26 and an analysis module 27.

[0175] The first acquisition module 21 is used to acquire multi-source heterogeneous data of the vehicle.

[0176] The construction module 22 is used to construct a vehicle fault knowledge base and a vehicle repair knowledge base; the vehicle fault knowledge base is used to record the association information of fault object-fault mode-fault cause; the vehicle repair knowledge base is used to record the association information of fault object-fault mode-repair plan.

[0177] The second acquisition module 23 is connected to the first acquisition module 21 and the construction module 22, and is used to acquire fault object-fault mode association information corresponding to the multi-source heterogeneous data based on the vehicle fault knowledge base.

[0178] The third acquisition module 24 is connected to the second acquisition module 23 and is used to deduplicate and aggregate the fault object-fault mode association information to obtain valid fault object-fault mode association information.

[0179] The fourth acquisition module 25 is connected to the construction module 22 and the third acquisition module 24, and is used to acquire the fault cause corresponding to the effective fault object-fault mode association information based on the vehicle fault knowledge base.

[0180] The fifth acquisition module 26 is connected to the construction module 22 and the third acquisition module 24, and is used to acquire the repair plan corresponding to the effective fault object-fault mode association information based on the vehicle maintenance knowledge base.

[0181] The analysis module 27 is connected to the fourth acquisition module 25 and the fifth acquisition module 26, and is used to generate a vehicle fault analysis report based on the fault cause and the repair plan using a large language model.

[0182] The structure and principle of the first acquisition module 21, the construction module 22, the second acquisition module 23, the third acquisition module 24, the fourth acquisition module 25, the fifth acquisition module 26 and the analysis module 27 correspond one-to-one with the steps in the above-mentioned vehicle fault analysis method based on artificial intelligence agents, so they will not be described again here.

[0183] In the embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, or methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of modules / units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or units may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of apparatuses or modules or units may be electrical, mechanical, or other forms.

[0184] The modules / units described as separate components may or may not be physically separate. The components shown as modules / units may or may not be physical modules; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules / units can be selected to achieve the objectives of the embodiments of the present invention, depending on actual needs. For example, the functional modules / units in the various embodiments of the present invention may be integrated into one processing module, or each module / unit may exist physically separately, or two or more modules / units may be integrated into one module / unit.

[0185] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0186] This invention also provides a computer-readable storage medium. Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing a processor. The program can be stored in a computer-readable storage medium, which is a non-transitory medium, such as random access memory, read-only memory, flash memory, hard disk, solid-state drive, magnetic tape, floppy disk, optical disk, and any combination thereof. The storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. This available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., digital video disc (DVD)), or a semiconductor medium (e.g., solid-state drive (SSD)).

[0187] This invention also provides an electronic device. The electronic device includes a processor and a memory.

[0188] The memory is used to store computer programs.

[0189] The memory includes various media capable of storing program code, such as ROM, RAM, magnetic disk, USB flash drive, memory card, or optical disk.

[0190] The processor is connected to the memory and is used to execute the computer program stored in the memory, so that the electronic device performs the above-described vehicle fault analysis method based on artificial intelligence agents.

[0191] Preferably, the processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0192] like Figure 3As shown, the electronic device of the present invention is embodied in the form of a general-purpose computing device. The components of the electronic device may include, but are not limited to: one or more processors or processing units 31, a memory 32, and a bus 33 connecting different system components (including the memory 32 and the processing unit 31).

[0193] Bus 33 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0194] Electronic devices typically include a variety of computer-readable media. These media can be any available media that can be accessed by the electronic device, including volatile and non-volatile media, and removable and non-removable media.

[0195] Memory 32 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 321 and / or cache memory 322. The electronic device may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 323 may be used to read and write non-removable, non-volatile magnetic media (…). Figure 3 Not shown; usually referred to as a "hard drive"). Although Figure 3 Not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disc drive for reading and writing to a removable non-volatile optical disc (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 33 via one or more data media interfaces. Memory 32 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0196] A program / utility 324 having a set (at least one) of program modules 3241 may be stored, for example, in memory 32. Such program modules 3241 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 3241 typically perform the functions and / or methods described in the embodiments of the present invention.

[0197] The electronic device can also communicate with one or more external devices (e.g., keyboard, pointing device, display, etc.), one or more devices that enable a user to interact with the electronic device, and / or any device that enables the electronic device to communicate with one or more other computing devices (e.g., network interface card, modem, etc.). This communication can be performed through input / output (I / O) interface 34. Furthermore, the electronic device can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) through network adapter 35. Figure 3 As shown, network adapter 35 communicates with other modules of the electronic device via bus 33. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the electronic device, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0198] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A vehicle fault analysis method based on artificial intelligence agents, characterized in that, The method includes the following steps: Acquire multi-source heterogeneous data about vehicles; Construct a vehicle fault knowledge base and a vehicle repair knowledge base; the vehicle fault knowledge base is used to record the association information of fault object-fault mode-fault cause; the vehicle repair knowledge base is used to record the association information of fault object-fault mode-repair solution. Based on the vehicle fault knowledge base, obtain the fault object-fault mode association information corresponding to the multi-source heterogeneous data; The fault object-fault mode association information is deduplicated and aggregated to obtain valid fault object-fault mode association information; Based on the vehicle fault knowledge base, obtain the fault causes corresponding to the effective fault object-fault mode association information; Based on the vehicle maintenance knowledge base, obtain the maintenance plan corresponding to the effective fault object-fault mode association information; Based on the causes of the fault and the repair plan, a vehicle fault analysis report is generated using a large language model. The multi-source heterogeneous data includes remote online monitoring data, fault code list, engine brand, fault protocol type and text fault description; Based on the vehicle fault knowledge base, obtaining the fault object-fault mode association information corresponding to the multi-source heterogeneous data includes the following steps: Based on the calibration code ID and calibration verification number in the vehicle's electronic control unit, it is determined whether the electronic control unit has been tampered with, so as to obtain the electronic control unit's fault object-fault mode association information for the tampered fault. Based on the online monitoring data, extract the corresponding fault object-fault mode association information; Based on the textual fault description, extract the association information between the fault object and the fault symptom; Based on the fault codes in the fault code list, the engine brand, and the fault protocol type, the corresponding fault object-fault mode association information is searched in the vehicle fault knowledge base.

2. The vehicle fault analysis method based on artificial intelligence agent according to claim 1, characterized in that: The electronic control unit is determined to have been tampered with based on whether the offset of the edit distance between the calibration code ID and the calibration verification number exceeds a preset value.

3. The vehicle fault analysis method based on artificial intelligence agent according to claim 1, characterized in that: Building a vehicle fault knowledge base includes the following steps: Obtain the vehicle's fault code database; the fault code database records the fault code, fault description, and fault cause for each fault; Based on the fault code database, a large language model is used to generate fault object-fault mode-fault cause association information for each fault.

4. The vehicle fault analysis method based on artificial intelligence agent according to claim 1, characterized in that: Building a vehicle maintenance knowledge base includes the following steps: Obtain vehicle maintenance manuals, vehicle technical bulletins, and repair case studies; Convert the vehicle maintenance manual, the vehicle technical bulletin, and the maintenance case into plain text information; Construct a prompt word template, wherein the prompt words are extracted from plain text information and related to faults and repairs, and output according to a preset format; Based on the plain text information and the prompt word template, a large language model is used to generate information related to fault objects, fault modes, and repair solutions.

5. The vehicle fault analysis method based on artificial intelligence agent according to claim 1, characterized in that: The process of deduplicating and aggregating the fault object-fault mode association information to obtain valid fault object-fault mode association information includes the following steps: The fault object-fault mode association information is vectorized based on the SBERT model or the Transformer model; Cluster the vectorized fault object-fault mode association information to obtain multiple fault clusters; For each fault cluster, select the fault object-fault mode association information corresponding to the fault as the effective fault object-fault mode association information.

6. A vehicle fault analysis system based on artificial intelligence agents, characterized in that, The system includes a first acquisition module, a construction module, a second acquisition module, a third acquisition module, a fourth acquisition module, a fifth acquisition module, and an analysis module; The first acquisition module is used to acquire multi-source heterogeneous data of the vehicle; The construction module is used to build a vehicle fault knowledge base and a vehicle repair knowledge base; the vehicle fault knowledge base is used to record the association information of fault object-fault mode-fault cause; the vehicle repair knowledge base is used to record the association information of fault object-fault mode-repair solution. The second acquisition module is used to acquire fault object-fault mode association information corresponding to the multi-source heterogeneous data based on the vehicle fault knowledge base; The third acquisition module is used to deduplicate and aggregate the fault object-fault mode association information to obtain valid fault object-fault mode association information. The fourth acquisition module is used to acquire the fault cause corresponding to the effective fault object-fault mode association information based on the vehicle fault knowledge base; The fifth acquisition module is used to acquire the repair plan corresponding to the effective fault object-fault mode association information based on the vehicle maintenance knowledge base; The analysis module is used to generate a vehicle fault analysis report based on the fault cause and the repair plan, using a large language model. The multi-source heterogeneous data includes remote online monitoring data, fault code list, engine brand, fault protocol type and text fault description; Based on the vehicle fault knowledge base, obtaining the fault object-fault mode association information corresponding to the multi-source heterogeneous data includes the following steps: Based on the calibration code ID and calibration verification number in the vehicle's electronic control unit, it is determined whether the electronic control unit has been tampered with, so as to obtain the electronic control unit's fault object-fault mode association information for the tampered fault. Based on the online monitoring data, extract the corresponding fault object-fault mode association information; Based on the textual fault description, extract the association information between the fault object and the fault symptom; Based on the fault codes in the fault code list, the engine brand, and the fault protocol type, the corresponding fault object-fault mode association information is searched in the vehicle fault knowledge base.

7. An electronic device, characterized in that, The electronic device includes: a processor and a memory; The memory is used to store computer programs; The processor is used to execute the computer program stored in the memory to cause the electronic device to perform the vehicle fault analysis method based on artificial intelligence agent as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by an electronic device, the program implements the vehicle fault analysis method based on an artificial intelligence agent as described in any one of claims 1 to 5.

Images