Vehicle fault detection method, vehicle fault detection system, and computer storage medium

Abstract

The present application relates to the technical field of automobile fault diagnosis, and relates to a vehicle fault detection method, a vehicle fault detection system and a computer storage medium.The vehicle fault detection method comprises the following steps: obtaining a target measurement point interface corresponding to a to-be-detected vehicle according to a target vehicle model of the to-be-detected vehicle, wherein the target measurement point interface comprises a connection relationship diagram of a plurality of measurement points; receiving a detection mode selection instruction for the target measurement point interface; executing a corresponding target detection mode according to the detection mode selection instruction; and performing fault detection on the to-be-detected vehicle in the target detection mode to obtain target fault information.The present method can intuitively present the correlation relationship between the vehicle model measurement points through a visual interface, and supports flexible selection of the detection mode, thereby effectively improving the adaptability and flexibility of vehicle fault detection.

Description

Technical Field

[0001] This invention relates to the field of automotive fault diagnosis technology, specifically to a vehicle fault detection method, a vehicle fault detection system, and a computer storage medium. Background Technology

[0002] With the increasing number of new energy vehicles on the road, the demand for fault repair of high-voltage systems (involving complex components such as battery packs and motor controllers) is growing. Existing technologies for handling high-voltage insulation leakage or open-circuit faults mainly suffer from the following problems: First, the information is fragmented and lacks a holistic view, requiring technicians to consult numerous manuals to mentally visualize the vehicle's high-voltage topology, making it difficult to intuitively grasp the connection relationships. Second, the circuit diagrams are disconnected from the actual physical locations on the vehicle, lacking on-site photographic guidance, making it difficult for technicians to quickly locate high-voltage connectors and pins. Third, the fragmented operational procedures create efficiency and safety hazards, requiring technicians to frequently switch between consulting documents and measuring on the actual vehicle, and lacking intelligent tools to plan "binary" test paths, often leading to blind testing and repeated disassembly and assembly, which can easily cause misoperation and the risk of high-voltage electric shock. Finally, troubleshooting is highly dependent on personal experience; junior technicians, unfamiliar with common vehicle problems and easily measurable points, struggle to develop reasonable testing procedures, resulting in high repair costs and long processing times. Summary of the Invention

[0003] One objective of this invention is to provide a vehicle fault detection method, a vehicle fault detection system, and a computer storage medium to address the technical problems of low troubleshooting efficiency, high operational risks, and high technical barriers in the maintenance of high-voltage systems in new energy vehicles, caused by fragmented information, separation of virtual and real data, process interruptions, and reliance on experience.

[0004] In a first aspect, embodiments of the present invention provide a vehicle fault detection method, the method comprising: Based on the target vehicle model to be tested, a target measurement point interface corresponding to the vehicle to be tested is obtained. The target measurement point interface includes a connection relationship diagram of multiple measurement points. Received a detection mode selection instruction for the target measurement point interface; According to the detection mode selection instruction, the corresponding target detection mode is executed, and under the target detection mode, the vehicle to be detected is subjected to fault detection to obtain target fault information.

[0005] In a second aspect, a vehicle fault detection system is provided, the vehicle fault detection system including an electronic device, the electronic device including a memory and a processor, the memory being connected to the processor, the processor being configured to execute one or more computer programs stored in the memory, the processor, when executing the one or more computer programs, causing the vehicle fault detection system to implement the vehicle fault detection method as described in the first aspect.

[0006] In a third aspect, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the vehicle fault detection method as described in the first aspect.

[0007] In a fourth aspect, a vehicle fault detection device is provided, the vehicle fault detection device comprising: The determining unit is used to obtain the target measurement point interface corresponding to the vehicle to be detected based on the target vehicle model obtained. The target measurement point interface includes a connection relationship diagram of multiple measurement points. The receiving unit is configured to receive a detection mode selection instruction for the interface of the target measurement point; The detection unit is used to execute the corresponding target detection mode according to the detection mode selection instruction, and to perform fault detection on the vehicle to be detected under the target detection mode to obtain target fault information.

[0008] In the embodiments implemented by the above-mentioned vehicle fault detection method, vehicle fault detection system, vehicle fault detection device, and computer storage medium, this embodiment achieves visualization of fault detection for different vehicle models by acquiring the target model of the vehicle to be detected and displaying a target measurement point interface containing a connection diagram of multiple measurement points, intuitively presenting the correlation between measurement points; furthermore, by responding to the detection mode selection command for this interface, the corresponding target detection mode is executed to perform fault detection, realizing flexible adaptation of the fault detection method, which can meet the needs of different usage scenarios and operators with different experience; the fault detection is completed and the target fault information is output under the selected detection mode, which effectively simplifies the fault troubleshooting process, improves the adaptability and flexibility of vehicle fault detection, and reduces the dependence on human experience and operational complexity. Attached Figure Description

[0009] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram of the structure of a vehicle fault detection system according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a vehicle fault detection method according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the target measurement point interface in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a vehicle fault detection device according to an embodiment of the present invention. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.

[0012] It should be noted that, unless otherwise specified, the various features in the embodiments of this invention can be combined with each other, all of which are within the protection scope of this invention. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. Moreover, the terms "first," "second," and "third" used in this invention do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0013] The existing technology has the following problems: First, at the underlying information acquisition level, existing maintenance data is fragmented and lacks a global view. Maintenance technicians usually need to consult a large number of maintenance manuals and mentally or even hand-draw the entire vehicle's high-voltage topology from scattered circuit diagrams. They cannot intuitively and systematically obtain the connection relationships of the vehicle's high-voltage components, which constitutes a fundamental obstacle to troubleshooting.

[0014] Secondly, in terms of spatial positioning during the transformation from "blueprints" to "actual vehicles," there is a severe disconnect between circuit logic and physical location. Traditional circuit diagrams lack the correspondence between the actual physical location on the vehicle and the guidance of real-world photographs. This makes it difficult for technicians, even after understanding the theoretical connection relationships, to quickly locate the specific physical location of the high-voltage connector on the vehicle and the definitions of the positive and negative pins, causing troubleshooting to get stuck at the point-finding stage.

[0015] Secondly, at the level of specific measurement operation execution, the fragmented process leads to inefficiency and safety hazards. Technicians must wear insulated protective equipment during the measurement process and frequently switch between consulting documents (paper or electronic) and actual vehicle operation. This constant shift in attention not only significantly reduces work efficiency but also greatly increases the risk of misoperation due to inconvenient information access (such as misplacing connector pins), thereby increasing the safety risk of high-voltage electric shock. At the same time, when faced with complex series and parallel circuits, technicians lack effective auxiliary tools to scientifically plan the "binary method" test path, often only able to blindly test dozens of connection points, resulting in a lengthy troubleshooting process that usually requires 3-5 or more repeated disassembly, assembly, and measurement.

[0016] Finally, at the highest level of experience-based decision-making, existing repair methods rely heavily on individual experience and have high technical barriers. Due to the lack of intelligent path planning tools, junior technicians, when faced with complex faults, do not know which components are prone to failure or where is easy to measure, making it difficult to formulate reasonable testing procedures on their own. This results in a large amount of unnecessary disassembly and assembly time during the repair process, cumbersome diagnostic steps, and seriously limits the improvement of overall repair efficiency in the industry.

[0017] Please see Figure 1 , Figure 1 This is a structural diagram of a vehicle fault detection system. Figure 1 In the vehicle fault detection system 10, there is an electronic device 20, which includes at least one processor 201 and a memory 202.

[0018] Among them, electronic device 20 can be a car diagnostic tool, tablet computer, laptop computer, desktop computer, server, etc., and is not limited to one specific device.

[0019] The processor 201 is configured to support the vehicle fault detection system in performing the corresponding functions of the vehicle fault detection method in the above-described method embodiments. The processor 201 can be a central processing unit (CPU), a graphics processing unit (GPU), a network processor (NP), a hardware chip, or any combination thereof. The aforementioned hardware chip can be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The aforementioned PLD can be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0020] The memory 202 is used to store program code and common storage components, including graph databases, vector databases, MySQL, Redis, and MQ, to ensure data storage and management. The memory 202 may include volatile memory (VM), such as random access memory (RAM); it may also include non-volatile memory (NVM), such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); or it may include a combination of the above types of memory.

[0021] Specifically, the processor 201 may include a transmitting card, a receiving card, and a driver chip.

[0022] See Figure 2 , Figure 2 This is a schematic flowchart of a vehicle fault detection method provided in an embodiment of the present invention. The method includes the following steps: S10. Based on the target vehicle model of the vehicle to be tested, obtain the target measurement point interface corresponding to the vehicle to be tested. The target measurement point interface includes a connection relationship diagram of multiple measurement points.

[0023] The vehicle to be tested refers to the physical vehicle connected to the vehicle fault detection system that requires fault diagnosis. It encompasses various types, including pure electric vehicles, hybrid vehicles, and traditional gasoline vehicles. In this embodiment, it is the source of data acquisition, integrating complex electrical systems (such as powertrain systems, thermal management systems, chassis control systems, etc.) and containing numerous electronic control units (ECUs), sensors, actuators, and wiring harnesses. It is the physical object on which the vehicle fault detection system performs analysis and diagnosis.

[0024] The target vehicle model refers to the identification information of the vehicle to be tested, which is the key index for the vehicle fault detection system to index data and match resources. It is also a precise set of configuration parameters, typically including brand, model series, specific model, year, engine / motor type, and configuration level (e.g., "2023 Smart Edition of a certain brand"). Different target vehicle models correspond to completely different electrical topologies, component models, and measurement point definitions.

[0025] The target measurement point interface refers to the human-machine interface that the vehicle fault detection system loads and generates in real time based on the target vehicle model. The target measurement point interface is the core operating platform for users to troubleshoot faults, presented in a visual graphical format. It is not a static image but a dynamic data mapping carrier, uniquely bound to the target vehicle model. It is responsible for displaying the vehicle's internal electrical logic structure and serves as a view container for subsequent detection mode selection and measurement point interaction.

[0026] In this context, a measurement point refers to the smallest detection unit defined in a vehicle fault detection system, corresponding to a specific physical detection location or logical detection interface on the vehicle. A measurement point can be a specific hardware connection point (such as a sensor connector pin or a fuse in a fuse box) or a data read point (such as a frame of a message signal on a CAN bus). Each measurement point is bound to a unique identifier in the system and associated with multi-dimensional attribute information (such as standard voltage value, signal type, and system circuit to which it belongs).

[0027] The connection diagram refers to the topology diagram displayed in the target measurement point interface, which describes the electrical logic connection between various vehicle components and measurement points. It visually presents the transmission path of current or signal in the form of nodes (representing components or measurement points) and lines (representing wiring harnesses or signal flow), helping users understand the fault propagation path.

[0028] Specifically, the acquisition process may include, but is not limited to, automatic identification or manual interaction.

[0029] Optionally, the user connects the diagnostic terminal to the vehicle's OBD diagnostic port via VCI (Vehicle Communication Interface). The vehicle fault detection system sends a wake-up command to establish a communication link with the vehicle gateway. The system then sends a command to read the VIN (Vehicle Identification Number) stored in the vehicle's ECU. The VIN parsing algorithm built into the system breaks down the 17 characters, extracting key information such as the manufacturer, model code, and production year. If the vehicle communication fails and the VIN cannot be read, the user can manually enter the model information on the terminal interface or select it from a drop-down menu (path: brand -> series -> year -> model). The parsed model code is then compared with the model configuration table in the background database to confirm the unique target model ID.

[0030] Furthermore, based on the target vehicle model ID, the vehicle fault detection system loads the corresponding electrical topology data and interface configuration file from the cloud or local database, and renders and generates the target measurement point interface.

[0031] Specifically, using the target vehicle model ID as an index, the system retrieves the pre-stored topology connection data file (such as XML or JSON format) for that vehicle model from the vehicle model database. This file defines the logical coordinates and connection attributes of all key components of the vehicle model (such as water pumps, fans, relays, and controllers). The front-end graphics engine reads the topology data file and draws the nodes and connections on the screen, thus constructing a complete interactive interface. For example, for vehicle model A, the interface might display its unique dual-motor four-wheel drive topology; for vehicle model B, it might display a "range-extended hybrid topology." Once the interface loads, the target measurement point interface is generated.

[0032] Furthermore, while rendering the connection graph, the vehicle fault detection system performs data association operations in parallel, binding the graphical nodes with the multi-dimensional attribute tag library in the background to achieve deep integration of graphics and data.

[0033] Optionally, the vehicle fault detection system traverses each component node in the topology data. For each component, the system further extracts its subordinate measurement point information. For example, for the "electronic water pump" node, the system extracts its associated measurement points: "water pump power positive terminal," "water pump ground negative terminal," and "water pump control signal line." The system then calls upon a multi-dimensional attribute tag library for the measurement points, attaching the attribute data (such as failure rate, disassembly / reassembly difficulty, and standard parameters) of each measurement point to the graphical node. Based on the circuit schematic, the system connects the measurement points with lines of different colors. For example, red lines represent the high-voltage power supply path, green lines represent the low-voltage control signal path, and blue lines represent the CAN communication path. The connections not only show the physical connections but also the logical flow (indicated by arrows). The system registers a click event listener for each measurement point node. The interface displays not just a static diagram but an operable logical map. When a user clicks on a connection line in the diagram (such as "air conditioner compressor power supply line"), the system will highlight all associated measurement points on that line (power supply end, grounding end, intermediate connection plugs) and pop up a floating window to display the overall health status of that path.

[0034] For example, the target measurement point interface can be referenced. Figure 3 , Figure 3 This is a schematic diagram of the interface for the target measurement point.

[0035] As can be seen, this embodiment achieves accurate matching of the target measurement point interface corresponding to the target vehicle model, allowing users to quickly grasp the structure and measurement point distribution of the high-voltage system of the vehicle under test through the connection diagram of the target measurement point interface without having to manually search for fragmented repair manuals. This avoids measurement errors caused by mismatch between data and vehicle model and unintuitive topology in traditional repair, while also reducing the threshold for users to read data and the complexity of operation.

[0036] S20. Received a detection mode selection instruction for the target measurement point interface.

[0037] Among them, the detection mode selection instruction refers to the control signal triggered by the user on the terminal device according to the actual fault diagnosis needs, which is used to specify the subsequent diagnostic logic type of the vehicle fault detection system.

[0038] Specifically, in this embodiment, the instruction corresponds to three different diagnostic logics: First detection mode (large model recommendation mode): Instructs the vehicle fault detection system to call the preset fault detection large model for prediction and automatically generate detection steps.

[0039] The second detection mode (fault tree guided mode) instructs the vehicle fault detection system to call the preset diagnostic knowledge base and troubleshoot according to the standardized process.

[0040] The third detection mode (manual detection mode) indicates that the vehicle fault detection system has free measurement permissions, and the detection is led by a human.

[0041] Optionally, a detection mode selection control can be drawn in a preset area of ​​the target measurement point interface (such as the bottom dock, side floating window, or blank space in the topology map). For example, an AI intelligent diagnosis (recommended) button (corresponding to the first mode), an expert fault tree button (corresponding to the second mode), and a free manual detection button (corresponding to the third mode) can be displayed. The system background registers an event listener for each control to continuously scan for touch signals, mouse click signals, or external keyboard input signals on the screen.

[0042] Furthermore, the system converts the underlying input events into structured instruction objects that can be recognized by the business logic layer. These encapsulated instructions are then distributed to the scheduling center, which activates the corresponding functional modules based on the instruction type, preparing for the specific detection process.

[0043] As can be seen, in this embodiment, by receiving the mode selection command, manual / AI / expert mode can be flexibly selected; furthermore, the corresponding detection logic is entered according to the user's intention, avoiding blind detection and improving the efficiency and safety of vehicle fault detection.

[0044] S30. According to the detection mode selection instruction, execute the corresponding target detection mode, and perform fault detection on the vehicle to be detected under the target detection mode to obtain target fault information.

[0045] The target detection modes include a first detection mode, a second detection mode, and a third detection mode. Specifically, the first detection mode corresponds to S310, the second detection mode corresponds to S360, and the third detection mode corresponds to S370, respectively.

[0046] The specific process of S30 can be combined with S310-S370, and will not be described again here. Furthermore, S310, S360, and S370 are parallel schemes.

[0047] S310. In one embodiment, the target detection mode includes a first detection mode. The step of executing the corresponding target detection mode according to the detection mode selection instruction, and performing fault detection on the vehicle to be detected under the target detection mode to obtain target fault information, includes: acquiring multi-dimensional attribute labels for each measurement point in the connection graph to obtain measurement point data for each measurement point; inputting the measurement point data of each measurement point into a preset fault detection large model for prediction to obtain a first fault detection step list output by the preset fault detection large model, the first fault detection step list including multiple first detection paths, the first fault detection step list being arranged according to a preset sorting rule; executing the first target detection path in the first fault detection step list to obtain a first detection result corresponding to the first target detection path, the first target detection path being the first first detection path in the first fault detection step list; determining whether the first detection result meets a preset detection stop condition; if the first detection result meets the preset detection stop condition, then obtaining target fault information based on the first detection result.

[0048] S320. In one embodiment, after determining whether the first detection result meets the preset detection stop condition, the method further includes: if the first detection result does not meet the preset detection stop condition, inputting the first detection result into the preset fault detection large model for prediction to obtain a second fault detection step list output by the preset fault detection large model, the second fault detection step list including multiple second detection paths, the second fault detection step list being arranged according to the preset sorting rule; executing the second target detection path in the second fault detection step list to obtain the second detection result corresponding to the second target detection path, the second target detection path being the first second detection path in the second fault detection step list; determining whether the second detection result meets the preset detection stop condition; repeating this process until a target detection result meets the preset detection stop condition, then obtaining target fault information based on the target detection result.

[0049] In conjunction with S320, the implementation process of S310 can be one of the following descriptions A1-A9, for example only: A1. Obtain the multi-dimensional attribute labels of each measurement point in the connection relationship graph to obtain the measurement point data of each measurement point.

[0050] A2. Input the measurement data of each measurement point into a preset fault detection model for prediction to obtain a first fault detection step list output by the preset fault detection model. The first fault detection step list includes multiple first detection paths and is arranged according to a preset sorting rule.

[0051] A3. Execute the first target detection path in the first fault detection step list to obtain the first detection result corresponding to the first target detection path. The first target detection path is the first detection path in the first fault detection step list.

[0052] A4. Determine whether the first detection result meets the preset detection stop condition.

[0053] A5. If the first detection result meets the preset detection stop condition, then the target fault information is obtained based on the first detection result.

[0054] or, A6. If the first detection result does not meet the preset detection stop condition, the first detection result is input into the preset fault detection large model for prediction to obtain the second fault detection step list output by the preset fault detection large model. The second fault detection step list includes multiple second detection paths and is arranged according to the preset sorting rules.

[0055] A7. Execute the second target detection path in the second fault detection step list to obtain the second detection result corresponding to the second target detection path. The second target detection path is the first second detection path in the second fault detection step list.

[0056] A8. Determine whether the second detection result meets the preset detection stop condition.

[0057] A9. Repeat this process until a target detection result meets the preset detection stop condition. Then, based on the target detection result, obtain the target fault information.

[0058] Among them, multi-dimensional attribute labels refer to the structured feature data attached to each measurement point. Multi-dimensional attribute labels may include, but are not limited to: topological attributes, maintenance attributes, physical attributes, and the fault phenomena that will occur when each measurement point fails.

[0059] Furthermore, the multi-dimensional attribute labels for each measurement point are obtained by consulting various sources, such as online repair forums, official TSB repair announcements, actual vehicle parts disassembly and assembly, official repair manuals, and the experience judgment of professional repair technicians.

[0060] Specifically, topology attributes refer to the node's hierarchy in the circuit diagram (e.g., the upper level is a fuse, the lower level is a load); maintenance attributes refer to historical failure probability (based on big data statistics), disassembly and assembly difficulty (e.g., "requires removal of interior trim panels"), and testing time. Physical attributes refer to standard electrical parameters (standard voltage, resistance range) and signal type (PWM / CAN / LIN).

[0061] The measurement point data is a data packet that has been vectorized or structured from the above labels, and is used as the feature vector input to the preset fault detection large model.

[0062] In the specific implementation of A1, the vehicle fault detection system traverses all high-voltage system nodes on the connection diagram in the current target measurement point interface (e.g., node A - high-voltage distribution box PDU output, node B - air conditioning compressor high-voltage input, node C - PTC heater high-voltage input). For each node, the vehicle fault detection system performs an index query in the background database based on its unique ID: Extracting node A's tags: {Type: High-voltage distribution box PDU, Rated voltage: 400V DC, Failure rate: Low, Disassembly difficulty: Medium (requires disconnection of 12V low-voltage battery and high-voltage maintenance switch), Insulation requirements: 500V megohmmeter measurement, Safety level: High}; Extracting node B's tags: {Type: Electric compressor, Rated voltage: 400V DC, Failure rate: High, Disassembly difficulty: High (involves refrigerant piping disassembly and vacuuming), Interlock type: High-voltage interlock signal, Standard insulation resistance: >1MΩ}; Extracting node C's tags: {Type: PTC heater, Rated voltage: 400V DC, Failure Rate: Medium, Disassembly / Assembly Difficulty: Medium (located in thermal management circuit), Interlock Type: High-voltage interlock signal, Standard Voltage: 400V DC. The retrieved multi-dimensional attribute tags containing high-voltage safety and process characteristics are assembled into a standardized measurement point dataset, ready to be input into a pre-defined fault detection large-scale model.

[0063] For example, when analyzing a pure electric SUV, the system reads the PDU output to the compressor's high-voltage wiring harness measurement point and finds that its label indicates "Failure Rate: High (the high-voltage connector of this model is prone to water ingress, leading to insulation degradation)" and "Disassembly / Removal Difficulty: Medium (accessible only by disconnecting the high-voltage maintenance switch MSD)." These characteristics will directly guide the AI ​​model to prioritize testing the insulation status of this point, rather than directly disassembling the compressor body, which is extremely difficult to disassemble.

[0064] Among them, the pre-set fault detection big model refers to a deep learning model (such as the Transformer architecture) trained on massive amounts of repair cases and circuit topology data. The pre-set fault detection big model has logical reasoning ability similar to human experts and can comprehensively weigh fault probability, repair difficulty and topology path.

[0065] The first fault detection step list refers to the recommended detection schemes output by the preset fault detection big model, which is an ordered set of instructions.

[0066] The first detection path refers to each item in the list, which represents a specific detection action (such as "measuring the insulation resistance of the PDU output terminal to ground" or "measuring the continuity of the compressor interlock circuit").

[0067] Among them, the preset sorting rule refers to the decision logic inside the preset fault detection model, which usually follows a balance strategy of "lowest cost first" (testing the easy ones first) and "maximum likelihood first" (testing the much worse ones first).

[0068] In the specific implementation of A2, all high-voltage measurement point data (including high-voltage topology connection information under normal conditions) obtained in step A1 are encoded into tensors readable by the preset fault detection model. The preset fault detection model, combined with the vehicle's current fault symptoms (e.g., "the vehicle cannot engage the high-voltage OK position and reports an insulation fault"), performs path deduction in a neural network: "Although directly measuring the battery pack output can pinpoint the overall insulation problem, the PDU output has a higher failure rate and is relatively easier to disassemble and reassemble; moreover, measuring the insulation resistance from the PDU end can simultaneously verify the insulation status of the PDU's internal circuits and downstream compressors, PTCs, etc., resulting in higher troubleshooting efficiency." The preset fault detection model scores and ranks possible detection paths based on an efficiency reward function (comprehensively considering high-voltage safety risks, disassembly and reassembly time, and fault probability), generating a JSON-formatted list. Step 1 (First Detection Path): Disconnect the PDU output connector and measure the insulation resistance to ground on the PDU side (sorting criteria: high failure rate, no need to disassemble core components, one-step isolation and verification of the high voltage distribution circuit insulation).

[0069] Step 2 (First Detection Path): Measure the continuity and resistance of the high-voltage interlock circuit of the electric compressor (sorting criteria: high voltage failure to power on is often caused by interlock disconnection; measuring the interlock is a low-voltage operation, with extremely low safety risk and fast troubleshooting speed).

[0070] Step 3 (First Detection Path): Measure the insulation resistance to ground at the high voltage input terminal of the PTC heater (priority: leakage of PTC heater coolant leads to a high rate of insulation failure, but requires disassembly of surrounding pipelines, and the disassembly and assembly difficulty is moderate, so it is the second best inspection point).

[0071] The first target detection path refers to the first step in the list, which is the detection action that the preset fault detection model considers to have the lowest current security risk, the highest troubleshooting efficiency, and should be executed first. The first detection result refers to the actual measurement data returned by the user after performing the measurement.

[0072] In the specific implementation of A3, the vehicle fault detection system highlights the "High Voltage Distribution Unit (PDU) Output Connector" location on the interface and displays the following guidance: "Safety Warning: Please ensure the high voltage maintenance switch (MSD) is disconnected and wear insulated gloves! Please use a megohmmeter (insulation resistance tester) on the 500V range to measure the insulation resistance between the A-phase pin of the PDU output terminal and the vehicle ground wire." The maintenance personnel perform the measurement according to high voltage safety regulations and system guidance, inputting the measured value "0.5MΩ" via voice, manual input, or automatic feedback from the Bluetooth smart megohmmeter. The system records "0.5MΩ (below the standard threshold of 1MΩ, abnormal)" as the first detection result. The preset detection stop condition refers to the logical threshold for determining whether the fault point has been located. The preset detection stop condition may include, but is not limited to: a measured value significantly deviating from the standard range, the detection of a clear open / short circuit characteristic, or the location of a specific faulty component.

[0073] In the specific implementation of A4, the vehicle fault detection system compares the measured value "0.5MΩ" with the standard parameter ">1MΩ (high-voltage system insulation safety threshold)". Judgment logic: The insulation resistance is only 0.5MΩ, which is far below the safety threshold, indicating that there is serious leakage from the high-voltage circuit to the vehicle body, meeting the stop condition of "finding a clear abnormality". The fault is locked in the insulation failure at the high-voltage power supply end (such as internal creepage in the PDU or damage to the downstream wiring harness to ground).

[0074] In the specific implementation of A5, the vehicle fault detection system confirmed the cause of the fault as "high voltage insulation failure", generated target fault information, and displayed: "Faulty component: High voltage distribution box (PDU) to downstream output circuit; Fault cause: Insulation resistance is lower than the safety threshold, and there is leakage to ground; Recommended troubleshooting: Check whether there is condensation / creep traces inside the PDU, or check whether the downstream high voltage harness is damaged and grounded", and the detection closed loop is completed.

[0075] In the specific implementation of A6, for example, suppose that in step A3, the maintenance personnel measured the insulation resistance at the PDU output terminal to be "500MΩ (normal)". The vehicle fault detection system judges: high-voltage insulation is normal, the stopping condition is not met (leakage fault not found). The vehicle fault detection system takes the first detection result: insulation is normal as a new feature vector and inputs it into the preset fault detection model. The preset fault detection model updates its understanding: "the high-voltage main circuit insulation is good, ruling out leakage problems between the power battery and PDU to ground". Therefore, the preset fault detection model recalculates: "Since the insulation is not a problem, it is highly likely that the high-voltage interlock signal is missing, causing the BMS to cut off the high voltage, or the inverter has an internal open circuit", and outputs a new list: the preset fault detection model outputs the second fault detection step list: Step 1 (Second Detection Path): Measure the circuit resistance of the PDU high-voltage interlock terminal.

[0076] Step 2 (Second detection path): Measure the high voltage interlock signal voltage at the DC input terminal of the inverter.

[0077] Therefore, the vehicle fault detection system can adjust subsequent strategies in real time based on the successful troubleshooting results of the previous step, avoiding ineffective operations such as blindly disassembling high-voltage components for insulation testing.

[0078] In the specific implementation of A7, the vehicle fault detection system selects the first path in the second fault detection step list as the second target detection path; the interface highlights the location of the PDU interlock terminal and pops up a prompt: "Please use a multimeter in resistance mode to measure the continuity of the PDU interlock pins"; the user completes the measurement and provides feedback on the result (such as "infinity - open circuit"); the system records it as the second detection result.

[0079] Furthermore, A8 and A4 are completely identical in logic. The system compares the second detection result with the stopping condition to determine whether the detection can be terminated.

[0080] In the specific implementation of A9, if the current result still does not meet the stopping condition, the result is continued to be input into the vehicle fault detection system to update the next round of detection steps list; the following steps are repeated: model update path, execution of the first path, obtaining results, and determining whether to stop, until a certain round of detection results meets the preset stopping condition; the system infers and outputs the target fault information based on the target detection result that meets the condition in that round; the detection process ends.

[0081] For example, regarding the fault "vehicle cannot engage the high-voltage OK position": Round 1: Measure the insulation resistance at the PDU output terminal → Result is normal (no leakage) → Continue; Round 2: Measure the resistance of the PDU high-voltage interlock circuit → Result is normal (interlock is active) → Continue; Round 3: Measure the insulation resistance to ground at the PTC heater high-voltage terminal → Result is abnormal (insulation resistance 0.1MΩ, severe leakage) → Stop condition met; System output: The target fault information is an insulation fault at the high-voltage terminal of the PTC heater (it is recommended to check whether the PTC coolant is leaking and causing a short circuit).

[0082] Optionally, in A9, if the number of cycles exceeds a preset threshold (e.g., 10 times) and the problem is still not located (e.g., all downstream components such as the compressor, PTC, and DC-DC have been checked and their insulation and interlocks are normal, but the system still cannot be powered on), the system will trigger "High-voltage difficult fault transfer to manual prompt". It is recommended that the maintenance personnel contact the OEM technical support or conduct a high-voltage wiring harness flying wire test to prevent the system from falling into a dead loop and to avoid the risk of blind high-voltage operation.

[0083] As can be seen, in this embodiment, by first acquiring multi-dimensional attribute data of measurement points, generating a priority-ordered list of detection steps by reasoning from a preset fault detection model, executing the first path and dynamically iterating and updating the detection path according to the actual measurement results, looping until the fault location conditions are met, and finally outputting the target fault information, unnecessary measurements are greatly reduced while ensuring detection accuracy, thereby improving the fault diagnosis efficiency of the vehicle fault detection system.

[0084] S330. In one embodiment, before obtaining the multi-dimensional attribute labels of each measurement point in the connection relationship graph and obtaining the measurement point data of each measurement point, the method further includes: obtaining historical maintenance records and component structure information of multiple vehicle models; constructing a multi-dimensional attribute label library for each vehicle model based on the historical maintenance records and component structure information of each vehicle model, wherein the multi-dimensional attribute label library includes each vehicle model and the multi-dimensional attribute labels corresponding to each measurement point in each vehicle model.

[0085] Historical maintenance records refer to the complete data archive generated during the vehicle's past maintenance and repair processes. This includes not only the final fault diagnosis (such as "PTC heater insulation fault"), but also intermediate data during the diagnostic process (such as "insulation resistance to ground 0.5MΩ", "high-voltage interlock circuit open"), descriptions of fault symptoms, names of replaced parts, and repair man-hours.

[0086] Component structural information refers to the technical information describing the physical layout and logical connections of the vehicle's hardware. This information includes circuit diagrams, wiring harness layout diagrams, component assembly location diagrams (e.g., the high-voltage distribution box is located above the rear subframe), component interface definitions (high-voltage pin and interlock pin definitions), and disassembly procedure instructions (including high-voltage power-off and voltage testing procedures). The component structural information determines the physical attributes of measurement points (e.g., location, difficulty of disassembly / reassembly).

[0087] Among them, the measurement point multidimensional attribute label library refers to the structured database constructed by the vehicle fault detection system, which is indexed by vehicle model-measurement point-label value.

[0088] Among them, multidimensional attributes mean that the label is no longer a single dimension, and may include, but is not limited to: fault dimension (historical fault probability, common failure modes such as insulation degradation or interlock disconnection), structural dimension (installation location, disassembly and assembly complexity (such as needing to disconnect the high-voltage maintenance switch MSD, drain the coolant), detection accessibility), and electrical dimension (rated high voltage range, standard insulation resistance threshold, high voltage interlock (HVIL) type, dangerous voltage label).

[0089] In practice, the vehicle fault detection system connects to the OEM's production database, the after-sales service station (4S store)'s DMS system (dealer management system), and the electronic repair manual system through interfaces; it extracts repair work orders, fault codes (DTC), fault phenomenon descriptions, final failed components, high-voltage insulation test data, and interlock status data streams for specific models over the past few years; and further imports the high-voltage system circuit diagram files and component disassembly and assembly manuals for the vehicle model.

[0090] Extract topology connections: such as "High-voltage distribution unit (PDU) -> High-voltage DC bus -> Electric compressor" and "BMS interlock output -> PDU interlock pin -> Compressor interlock pin". Extract physical location and attribute information: such as "The electric compressor is located at the bottom of the front nacelle. The high-voltage maintenance switch (MSD) must be disconnected and the voltage tested first. The interlock status must be confirmed before disconnecting the high-voltage connector."

[0091] Furthermore, using big data analytics algorithms and natural language processing (NLP) technology, structured tags are extracted from the messy raw data and written into the database.

[0092] Optionally, based on the component structure information of each vehicle model, the component installation position and associated obstructing components corresponding to each measurement point in each vehicle model are determined; based on the number of associated obstructing components and the disassembly process time, a disassembly and assembly complexity label for each measurement point is generated; based on the historical maintenance records, the failure frequency of associated components at each measurement point is counted, and a failure probability value is calculated based on the failure frequency to generate a component failure rate label for each measurement point.

[0093] Specifically, categorized by vehicle model, the "measurement point ID, measurement point location, associated components, disassembly / assembly complexity label (level + code), and component failure rate label" for each model are integrated into structured data, forming a subset of multi-dimensional attribute labels for measurement points specific to that model. The label subsets for all models are then aggregated to construct a complete multi-dimensional attribute label library for measurement points. This library uses a quickly searchable format (such as database tables), establishing an index for model-measurement point-label to ensure the system can quickly retrieve the multi-dimensional attribute labels for all measurement points of a target model. The completed label library is pre-stored in the system database of the fault detection equipment, and linked and bound to data such as the connection relationship diagram and measurement point location diagram for each model, providing data support for subsequent steps (obtaining measurement point data) and ensuring the system can quickly retrieve the corresponding labels.

[0094] As can be seen, this embodiment further improves the inference accuracy of the preset fault detection large model by constructing a multi-dimensional attribute label library for measurement points; avoids manual annotation bias, reduces invalid measurements, and improves fault detection efficiency; and adapts to the differences in high-voltage systems of multiple vehicle models, enhancing the practicality and scalability of the solution.

[0095] S340. In one embodiment, after obtaining the multi-dimensional attribute labels of each measurement point in the connection relationship graph and obtaining the measurement point data of each measurement point, the method further includes: obtaining the fault diagnosis code and / or fault description text corresponding to the target vehicle; inputting the fault diagnosis code and / or fault description text corresponding to the target vehicle and the measurement point data of each measurement point into a preset fault detection large model for prediction, and obtaining the fault detection step list output by the preset fault detection large model.

[0096] S340 is a parallel scheme of A2 mentioned above.

[0097] Specifically, the following steps are performed: First, obtain the fault diagnosis code and / or fault description text corresponding to the target vehicle; second, input the fault diagnosis code and / or fault description text corresponding to the target vehicle, along with the measurement point data of each measurement point, into a preset fault detection model for prediction, resulting in a first fault detection step list output by the preset fault detection model. This first fault detection step list includes multiple first detection paths, arranged according to a preset sorting rule; third, execute the first target detection path in the first fault detection step list to obtain a first detection result corresponding to the first target detection path, where the first target detection path is the first first detection path in the first fault detection step list; fourth, determine whether the first detection result meets a preset detection stop condition; fifth, if the first detection result meets the preset detection stop condition, obtain the target fault information based on the first detection result. This process is repeated sequentially.

[0098] Diagnostic fault codes (DFCs) are standardized error codes (such as P0A0F and U0100) generated by a vehicle's ECU (Electronic Control Unit) when it detects an anomaly during self-testing. DFCs typically point to a failure in a subsystem or functional logic (such as "high-voltage interlock fault" or "CAN communication timeout"), and are highly structured and specific.

[0099] Among them, fault description text refers to the subjective description of the fault phenomenon entered by the user in natural language (such as "power interruption during driving" or "charging gun cannot be unplugged"). Fault description text is an important supplement to fault codes and is particularly suitable for scenarios where there are symptoms but no fault codes. It belongs to unstructured or semi-structured data.

[0100] Among them, prediction refers to the process by which a pre-set fault detection model deduces the best troubleshooting path based on the input fault codes / text and measurement point data.

[0101] In practice, the user connects the fault detection device to the OBD interface of the target vehicle, ensuring normal communication between the device and the vehicle's ECU. The system automatically sends a fault code reading command to the ECU and receives the fault diagnostic code returned by the ECU. It further analyzes the fault diagnostic code to identify the corresponding fault type (such as insulation abnormality) and associates it with the target vehicle and measurement point data for storage. The system then displays an input box, guiding the user to input the actual fault symptoms of the vehicle. The user inputs a natural language description (such as "cannot power on at startup, high voltage alarm on the dashboard"), and the system segments and encodes the text, extracting core fault information. The processed fault description text is then associated with the fault diagnostic code (if any) and measurement point data to supplement the fault information.

[0102] For example, if a user inputs "cannot power on, high voltage alarm," the system extracts the core information and associates it with the diagnostic fault code P0A80 to form complete fault information. The system integrates the target vehicle's diagnostic fault code and / or fault description text (already encoded) with all the measurement point data (including attribute labels) generated above to form model input data in a unified format.

[0103] In practice, the integrated input data is fed into the preset fault detection model, triggering the preset fault detection model prediction process. The preset fault detection model combines fault information to analyze the disassembly and assembly complexity and component failure rate of each measurement point. Combined with the connection relationship of the vehicle fault detection system, it generates multiple detection paths. The preset fault detection model outputs a sorted list of fault detection steps, clarifying the execution order of each detection path, and displays it on the target measurement point interface to guide the user to carry out the detection.

[0104] As can be seen, this embodiment supplements fault information by introducing fault codes and description text, and combines them with measurement point data to enable the preset fault detection model to accurately output the optimal detection steps, reduce blind measurements, and improve the efficiency and accuracy of fault detection.

[0105] S350. In one embodiment, before inputting the measurement point data of each measurement point into a preset fault detection large model for prediction to obtain the first fault detection step list output by the preset fault detection large model, the method further includes: acquiring a training database, the training database including topology connection data samples, preset multidimensional attribute label samples, and fault diagnosis logic samples; pre-training an initial fault detection model based on the topology connection data samples to obtain the pre-trained initial fault detection model; and fine-tuning the pre-trained initial fault detection model based on the preset multidimensional attribute label samples, fault diagnosis logic samples, and a preset efficiency reward function to obtain the preset fault detection large model.

[0106] The training database refers to a massive dataset used for model learning, which includes topology connection data samples, preset multidimensional attribute label samples, and fault diagnosis logic samples.

[0107] Among them, the topology connection data sample refers to the structured data describing the connection relationship between various components of the vehicle (e.g., relay A is connected to fuse B, and fuse B is connected to load C); the preset multi-dimensional attribute label sample refers to the characteristic data that marks each measurement point (e.g., the failure rate of a certain measurement point is "high" and the disassembly and assembly difficulty is "difficult"); the fault diagnosis logic sample refers to the troubleshooting path and final conclusion recorded in real repair cases (e.g., in a certain repair, points A and B were measured in sequence, and the fault was finally found at point C).

[0108] The initial fault detection model refers to an untrained neural network architecture (such as Transformer, Graph Neural Network GNN, etc.), which has strong potential for feature extraction and pattern recognition, but has not yet mastered the knowledge of automotive fault diagnosis.

[0109] Pre-training refers to the self-supervised learning process performed on large-scale unlabeled or weakly labeled data.

[0110] Fine-tuning refers to targeted training using labeled data for a specific task, based on a pre-trained model.

[0111] The pre-defined efficiency reward function is a mathematical formula used to evaluate the quality of the model's output during the fine-tuning phase (especially in reinforcement learning or reward model training). It defines what constitutes a "good detection step" (e.g., fewer steps, shorter time, lower cost) and guides the model to output the optimal policy.

[0112] Furthermore, the efficiency reward function is configured to perform weighted calculations based on the disassembly and assembly complexity parameters and fault probability parameters of each measurement point in the detection path to generate a reward value, thereby guiding the model to generate a detection path with fewer detection steps and lower disassembly and assembly difficulty.

[0113] In practical implementation, the vehicle fault detection system collects raw data from the repair case database, cleans, labels, and structures it to construct a training set conforming to the model input format; it parses circuit diagram files of tens of thousands of vehicle models, extracts the relationship data of nodes (components) and edges (connections), and transforms it into graph structure data samples; it maps the fault frequency in historical repair records and the disassembly and assembly hours in repair manuals to specific measurement point IDs, generating attribute label samples. It extracts the link data of "fault phenomenon -> detection step sequence -> fault result" from historical repair work orders as diagnostic logic samples.

[0114] In the specific implementation of pre-training the initial fault detection model based on the topology connection data samples to obtain the pre-trained initial fault detection model, a self-supervised learning task is designed. The topology connection data samples are input into the initial model, and the model parameters are adjusted through the backpropagation algorithm so that it can accurately restore and predict the circuit topology.

[0115] Furthermore, in the process of fine-tuning the pre-trained initial fault detection model based on the preset multi-dimensional attribute label samples, fault diagnosis logic samples, and preset efficiency reward function to obtain the preset large-scale fault detection model, the preset efficiency reward function is defined. Formula logic: .

[0116] A high bonus score is given if the model recommends fewer steps, has low disassembly and assembly difficulty, and can pinpoint the fault.

[0117] Furthermore, the input consists of a fault diagnosis logic sample (problem) and attribute labels (constraints). The model generates a recommended list of detection steps. The system calculates the score of this list based on a reward function. Finally, it uses a policy gradient algorithm (such as PPO) to encourage the model to generate high-scoring policies and suppress low-scoring policies. After multiple iterations, the detection step list output by the model reaches its optimal balance between efficiency and accuracy. The model parameters are then fixed, generating a pre-defined large-scale fault detection model.

[0118] Optionally, the preset fault detection large model can employ a large model illusion avoidance mechanism. After the large model outputs the measurement path, specific rule checks are required, including but not limited to the following rules: Rule 1, Node Existence Verification: Verify whether all measurement point identifiers included in the initial detection path exist in the node set of the connection graph corresponding to the target vehicle model. If any non-existent measurement point identifier is included, the verification fails. Rule 2, Connection Validity Verification: Verify whether there are valid connection edges between adjacent measurement points in the initial detection path that conform to circuit logic. If there are skip paths without connection relationships, the verification fails. Rule 3, Target Consistency Verification: Verify whether the attribute labels of the measurement points pointed to by the initial detection path match the target of the current detection task. For example, when the detection task is to troubleshoot electric compressor faults, if the initial detection path does not contain measurement points associated with the electric compressor circuit label, the verification fails. If the initial detection path fails any of the above rule checks, the output of the preset fault detection large model is considered invalid, and the preset fault detection large model is triggered to regenerate the detection path.

[0119] As can be seen, in this embodiment, by constructing a training database that includes topological connections, multi-dimensional attribute labels, and fault diagnosis logic, the model is pre-trained and then fine-tuned by combining an efficiency reward function. This enables the pre-set fault detection model to accurately learn the system structure and diagnostic rules, thereby outputting more efficient fault detection steps that are more in line with actual maintenance scenarios. S360. In one embodiment, the target detection mode includes a second detection mode. The step of executing the corresponding target detection mode according to the detection mode selection instruction, and performing fault detection on the vehicle to be detected under the target detection mode to obtain target fault information, includes: acquiring a preset diagnostic knowledge base, the fault diagnostic code and / or fault description text corresponding to the target vehicle; matching the fault diagnostic code and / or fault description text corresponding to the target vehicle in the preset diagnostic knowledge base to obtain a target fault tree node, wherein the preset diagnostic knowledge base contains a mapping relationship between fault phenomena and troubleshooting step sequences; acquiring a preset fault detection step list corresponding to the target fault tree node, wherein the preset fault detection step list includes multiple target detection steps; performing fault detection according to the order of the preset fault detection step list to obtain target detection data corresponding to the currently executed target detection step; comparing the target detection data with preset standard parameters to obtain a comparison result; if the comparison result meets preset fault judgment conditions, then outputting target fault information; or, if the comparison result does not meet the preset fault judgment conditions, then executing the next target detection step.

[0120] The second detection mode refers to a standardized detection mode based on a pre-set diagnostic knowledge base. In this mode, the vehicle fault detection system does not perform complex large-scale model reasoning, but instead strictly follows the fault tree logic pre-set by the OEM or expert experience for troubleshooting. The second detection mode is suitable for scenarios where the fault symptoms are clear and a specific fault code (DTC) has been read, and it is characterized by rigorous logic and standardized procedures.

[0121] The pre-built diagnostic knowledge base refers to a pre-constructed structured database that stores a massive mapping relationship between fault phenomena / fault codes and troubleshooting steps. The core data structure of the pre-built diagnostic knowledge base is usually represented as a fault tree, which is a hierarchical decision path from the root node (fault phenomenon) to the leaf node (fault point).

[0122] In this context, the target fault tree node refers to the specific branch entry point of the fault tree located in the knowledge base based on the fault code or description of the current vehicle. For example, if there is a tree in the knowledge base for "P0A0F fault code", this tree becomes the target fault tree node after matching.

[0123] The preset fault detection step list refers to the specific sequence of operation instructions attached to the target fault tree node. The preset fault detection step list is static and predefined (e.g., the first step is to measure voltage, the second step is to measure resistance), unlike the dynamically generated list corresponding to the S310 mode.

[0124] The preset fault judgment conditions refer to the thresholds or logical rules for judging whether the measured data is normal. For example, "voltage value <11V" is judged as abnormal, and "resistance value is infinite" is judged as open circuit.

[0125] Specifically, the vehicle fault detection system reads pre-set diagnostic knowledge base data from the cloud or local storage and constructs an index tree. It reads fault codes stored in the vehicle's ECU (e.g., "P0130: Oxygen sensor circuit fault") through the diagnostic interface; or receives fault description keywords input by repair personnel (e.g., "engine vibration"). The system uses string matching or semantic analysis technology to search for the corresponding fault entry point in the knowledge base. If a completely matching fault code node is found, the fault is successfully located. If no fault code is found, the system matches the "vibration" sub-node under the "engine fault" category in the knowledge base based on the description "engine vibration".

[0126] Furthermore, the vehicle fault detection system retrieves the standard operating procedures (SOPs) list stored under this node. The preset list contents are displayed: Step 1: Check if the oxygen sensor wiring harness connector is loose; Step 2: Measure the oxygen sensor signal line voltage; Step 3: Measure the oxygen sensor heating resistance.

[0127] Furthermore, the vehicle fault detection system guides maintenance personnel to perform operations step by step and collects measured data. The system compares the measured values ​​with preset "standard values / ranges" in the knowledge base. If the measured value exceeds the standard range (meets the fault determination conditions), the system determines that the component indicated by this step is faulty, outputs fault information (such as "oxygen sensor heating resistor open circuit, sensor replacement recommended"), and the process ends. If the measured value is normal (does not meet the fault determination conditions), it indicates that the currently detected component is not faulty. The system moves the current step pointer to obtain the preset fault detection step list corresponding to the target fault tree node. It returns to perform fault detection according to the order of the preset fault detection step list, obtains the target detection data corresponding to the currently executed target detection step, and repeats the process until the fault point is found or the list has been traversed.

[0128] As can be seen, the second detection mode in this embodiment achieves standardization and normalization of the fault detection process through standardized matching and fault tree traversal of the preset diagnostic knowledge base, ensuring the accuracy and traceability of the diagnostic results, effectively reducing the reliance on the personal experience of maintenance personnel, and improving the efficiency of troubleshooting routine faults.

[0129] S370. In one embodiment, the target detection mode includes a third detection mode. The step of executing the corresponding target detection mode according to the detection mode selection instruction, and performing fault detection on the vehicle to be detected under the target detection mode to obtain target fault information, includes: according to at least one received detection instruction for the connection relationship diagram; obtaining the target measurement point corresponding to each detection instruction; performing fault detection on the target measurement point to obtain at least one detection result; and obtaining the target fault information based on the at least one detection result.

[0130] The third detection mode refers to the user-led manual detection mode.

[0131] Among them, the test command refers to the trigger signal generated when maintenance personnel operate on the measurement point on the interactive interface. Test commands may include, but are not limited to, clicking on a node to view parameters, triggering a multimeter reading, requesting oscilloscope waveforms, etc.

[0132] In this context, the target measurement point refers to the specific node that needs to be tested, selected by maintenance personnel in the connection diagram through clicking or selecting a box. The system will provide auxiliary information based on the node's multi-dimensional attribute labels.

[0133] The test results refer to the data obtained after actual measurement of the target measurement point (such as voltage value, resistance value, waveform diagram or on / off state).

[0134] Specifically, the vehicle fault detection system switches the connection diagram from a display state to an interactive state, monitors user actions in real time, captures touchscreen or mouse click events, and generates structured instructions. The system parses these structured instructions, identifies the specific physical object to be tested, and retrieves and displays its detailed information. With the system's assistance, the user performs physical measurements, and the system records and analyzes the measured data. The system summarizes the results of multiple manual tests, performs simple logical integration, or outputs the results directly to assist repair personnel in drawing conclusions.

[0135] As can be seen, the third detection mode in this embodiment grants users flexible autonomous detection rights by opening up direct interaction permissions to the connection relationship graph, breaking the limitations of fixed processes, greatly improving the verification efficiency and review convenience for specific fault points, and realizing efficient human-machine collaborative operation.

[0136] This embodiment achieves visualized fault detection for different vehicle models by acquiring the target vehicle model to be tested and displaying a target measurement point interface containing a connection diagram of multiple measurement points, intuitively presenting the correlation between measurement points. Furthermore, by responding to the detection mode selection command on this interface, the corresponding target detection mode is executed to perform fault detection, realizing flexible adaptation of the fault detection method and meeting the needs of different usage scenarios and operators with different experience. Fault detection is completed and target fault information is output under the selected detection mode, effectively simplifying the fault troubleshooting process, improving the adaptability and flexibility of vehicle fault detection, and reducing the dependence on human experience and operational complexity.

[0137] It should be noted that in the above embodiments, there is no necessarily a certain order between the steps. Those skilled in the art can understand from the description of the embodiments of this application that the above steps may have different execution orders in different embodiments, that is, they may be executed in parallel or in turn, etc.

[0138] As another aspect of the embodiments of this application, this application provides a vehicle fault detection device. The vehicle fault detection device can be a software module, which includes several instructions stored in a memory. A processor can access the memory, call the instructions, and execute them to complete the vehicle fault detection methods described in the various embodiments above.

[0139] See Figure 4 , Figure 4 This is a schematic diagram of the structure of a vehicle fault detection device provided in an embodiment of this application. Figure 4 As shown, the vehicle fault detection device 400 includes: The determining unit 401 is used to obtain a target measurement point interface corresponding to the vehicle to be detected based on the target vehicle model of the vehicle to be detected. The target measurement point interface includes a connection relationship diagram of multiple measurement points. The receiving unit 402 is used to receive a detection mode selection instruction for the target measurement point interface; The detection unit 403 is used to execute the corresponding target detection mode according to the detection mode selection instruction, and to perform fault detection on the vehicle to be detected under the target detection mode to obtain target fault information.

[0140] This embodiment achieves visualized fault detection for different vehicle models by acquiring the target vehicle model to be tested and displaying a target measurement point interface containing a connection diagram of multiple measurement points, intuitively presenting the correlation between measurement points. Furthermore, by responding to the detection mode selection command on this interface, the corresponding target detection mode is executed to perform fault detection, realizing flexible adaptation of the fault detection method and meeting the needs of different usage scenarios and operators with different experience. Fault detection is completed and target fault information is output under the selected detection mode, effectively simplifying the fault troubleshooting process, improving the adaptability and flexibility of vehicle fault detection, and reducing the dependence on human experience and operational complexity.

[0141] It should be noted that the above-described vehicle fault detection device can execute the vehicle fault detection method provided in the embodiments of this application, and has the corresponding functional modules and beneficial effects of the method. Technical details not described in detail in the embodiments of the vehicle fault detection device can be found in the vehicle fault detection method provided in the embodiments of this application.

[0142] This application also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the vehicle fault detection method as described in the foregoing embodiments.

[0143] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0144] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. A vehicle fault detection method, characterized in that, The method includes: Based on the target vehicle model to be tested, a target measurement point interface corresponding to the vehicle to be tested is obtained. The target measurement point interface includes a connection relationship diagram of multiple measurement points. Received a detection mode selection instruction for the target measurement point interface; According to the detection mode selection instruction, the corresponding target detection mode is executed, and under the target detection mode, the vehicle to be detected is subjected to fault detection to obtain target fault information.

2. The method according to claim 1, characterized in that, The target detection mode includes a first detection mode. The step involves selecting a target detection mode according to the selected mode, executing the corresponding target detection mode, and performing fault detection on the vehicle to be detected under the target detection mode to obtain target fault information, including: Obtain the multi-dimensional attribute labels of each measurement point in the connection graph to obtain the measurement point data of each measurement point; The measurement data of each measurement point is input into a preset fault detection model for prediction, and a first fault detection step list output by the preset fault detection model is obtained. The first fault detection step list includes multiple first detection paths and is arranged according to a preset sorting rule. Execute the first target detection path in the first fault detection step list to obtain the first detection result corresponding to the first target detection path, where the first target detection path is the first first detection path in the first fault detection step list. Determine whether the first detection result meets the preset detection stop condition; If the first detection result meets the preset detection stop condition, then the target fault information is obtained based on the first detection result.

3. The method according to claim 2, characterized in that, After determining whether the first detection result meets the preset detection stop condition, the method further includes: If the first detection result does not meet the preset detection stop condition, the first detection result is input into the preset fault detection big model for prediction, and a second fault detection step list output by the preset fault detection big model is obtained. The second fault detection step list includes multiple second detection paths, and the second fault detection step list is arranged according to the preset sorting rules. Execute the second target detection path in the second fault detection step list to obtain the second detection result corresponding to the second target detection path. The second target detection path is the first second detection path in the second fault detection step list. Determine whether the second detection result meets the preset detection stop condition; The process is repeated until a target detection result meets the preset detection stop condition. Then, the target fault information is obtained based on the target detection result.

4. The method according to claim 2, characterized in that, Before obtaining the multi-dimensional attribute labels of each measurement point in the connection graph and obtaining the measurement point data of each measurement point, the method further includes: Obtain historical repair records and component structure information for multiple vehicle models; Based on the historical maintenance records and component structure information of each vehicle model, a multi-dimensional attribute tag library for measurement points is constructed for each vehicle model. The multi-dimensional attribute tag library for measurement points includes each vehicle model and the multi-dimensional attribute tags corresponding to each measurement point in each vehicle model.

5. The method according to claim 2, characterized in that, After obtaining the multi-dimensional attribute labels of each measurement point in the connection graph and obtaining the measurement point data of each measurement point, the method further includes: Obtain the fault diagnostic code and / or fault description text corresponding to the target vehicle; The fault diagnosis code and / or fault description text corresponding to the target vehicle, as well as the measurement point data of each measurement point, are input into a preset fault detection model for prediction, thereby obtaining a list of fault detection steps output by the preset fault detection model.

6. The method according to claim 2, characterized in that, Before inputting the measurement point data of each measurement point into a preset fault detection large model for prediction to obtain the first fault detection step list output by the preset fault detection large model, the method further includes: Obtain a training database, which includes topology connection data samples, preset multidimensional attribute label samples, and fault diagnosis logic samples; The initial fault detection model is pre-trained based on the topology connection data sample to obtain the pre-trained initial fault detection model. The pre-trained initial fault detection model is fine-tuned based on the preset multi-dimensional attribute label samples, fault diagnosis logic samples, and preset efficiency reward function to obtain the preset large fault detection model.

7. The method according to claim 1, characterized in that, The target detection mode includes a second detection mode. The step involves selecting a detection mode and executing the corresponding target detection mode. Under this target detection mode, fault detection is performed on the vehicle to be detected to obtain target fault information, including: Obtain a preset diagnostic knowledge base, the fault diagnostic code and / or fault description text corresponding to the target vehicle; Based on the fault diagnosis code and / or fault description text corresponding to the target vehicle, a match is made in the preset diagnostic knowledge base to obtain the target fault tree node. The preset diagnostic knowledge base contains the mapping relationship between fault phenomena and troubleshooting steps. Obtain a list of preset fault detection steps corresponding to the target fault tree node, wherein the list of preset fault detection steps includes multiple target detection steps; Fault detection is performed in the order of the preset fault detection step list to obtain the target detection data corresponding to the currently executed target detection step. The target detection data is compared with preset standard parameters to obtain the comparison result; If the comparison result meets the preset fault determination conditions, then the target fault information is output; or, If the comparison result does not meet the preset fault determination condition, then the next target detection step is executed.

8. The method according to claim 1, characterized in that, The target detection mode includes a third detection mode. The step involves selecting a target detection mode according to the selected mode, executing the corresponding target detection mode, and performing fault detection on the vehicle to be detected under the target detection mode to obtain target fault information, including: Based on at least one detection instruction received for the connection graph; Obtain the target measurement point corresponding to each detection command; Fault detection is performed on the target measurement point to obtain at least one detection result; The target fault information is obtained based on the at least one detection result.

9. A vehicle fault detection system, characterized in that, The system includes an electronic device, which includes a memory and a processor. The memory is connected to the processor, and the processor is configured to execute one or more computer programs stored in the memory. When the processor executes the one or more computer programs, it causes the vehicle fault detection system to implement the vehicle fault detection method as described in any one of claims 1-8.

10. A computer storage medium, characterized in that, The computer storage medium stores a computer program, which includes program instructions that, when executed by a processor, cause the processor to perform the vehicle fault detection method as described in any one of claims 1-8.

Images