An AI-powered intelligent OBD fault diagnosis and early warning system compatible with both gasoline and new energy vehicles

By designing an AI-powered intelligent OBD fault diagnosis and early warning system compatible with both gasoline and new energy vehicles, the system solves the problem that traditional OBD devices cannot detect hidden faults in new energy vehicles. It achieves plug-and-play functionality and local fault warning, improving the reliability and real-time performance of the diagnosis. It is suitable for scenarios such as auto repair shops, fleet management, owner self-inspection, and used car evaluation.

CN123113804APending Publication Date: 2026-11-17ZHUHAI YANGYANG NETWORK TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611262808.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-19
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Traditional OBD diagnostic equipment cannot effectively detect hidden faults in new energy vehicles, has poor compatibility, relies on cloud networks resulting in insufficient real-time performance and data privacy risks, and cannot complete the quantitative rating of hidden faults locally.

Method used

The design incorporates an AI-powered intelligent OBD fault diagnosis and early warning system compatible with both gasoline and new energy vehicles. It includes a hardware adaptation layer, a data acquisition layer, an embedded AI inference layer, and a diagnostic output layer. It supports multiple bus protocols, integrates vibration sensors and AI models, and enables local fault warning and quantitative rating.

Benefits of technology

It enables plug-and-play fault diagnosis for both fuel-powered and new energy vehicles, provides early warning of hidden faults, ensures data privacy and diagnostic reliability, has trend tracking capabilities, and is suitable for various application scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113804A_ABST
    Figure CN123113804A_ABST
Patent Text Reader

Abstract

This invention discloses an AI-powered intelligent OBD fault diagnosis and early warning system compatible with both gasoline and new energy vehicles. It includes a hardware adaptation layer housed within the OBD diagnostic box, containing multi-protocol physical interfaces supporting CAN2.0, CANFD, and FlexRay buses. This invention quantifies and rates latent faults using a multi-dimensional AI model, overcoming the limitations of traditional solutions that rely solely on standard fault codes. The hardware adaptation layer supports three mainstream vehicle bus protocols: CAN2.0, CANFD, and FlexRay, automatically identifying and locking them without requiring manual vehicle model selection, achieving true plug-and-play functionality. A local historical database enables the system to track trends, promptly identifying components experiencing accelerated deterioration and providing a basis for preventative maintenance. The system can be applied to various scenarios, including routine inspections at auto repair shops, fleet batch inspections, owner-managed monitoring, and used car technical condition assessments.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis system technology, and specifically to an AI intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles. Background Technology

[0002] Since the widespread adoption of the OBD-II standard in the 1990s, on-board diagnostic systems (OBD) have become a core tool for vehicle maintenance and repair. Traditional OBD diagnostic equipment determines the location and nature of faults by reading standardized fault codes (DTCs) recorded by the vehicle's electronic control unit (ECU). This mechanism was initially designed primarily for emission-related sensors and actuators, and is quite effective for typical problems where faults have already occurred. However, with the rapid development of automotive electronic and electrical architectures, especially the widespread adoption of new energy vehicles, the limitations of traditional OBD diagnostic systems have become increasingly apparent.

[0003] First, fault code generation relies on threshold judgment logic within the electronic control unit (ECU). The ECU will only record a corresponding fault code and illuminate the malfunction indicator lamp (MIL) when a sensor's signal consistently deviates from a preset range for a certain period. This means that many gradual degradations, such as progressively thickening carbon deposits on the throttle body, slow decline in battery capacity, abnormal noises from chassis components, and a year-on-year decrease in air conditioning compressor efficiency, will not generate any fault codes until the ECU's set hard threshold is reached. The owner and repair personnel will be completely unaware of these issues. By the time the MIL illuminates, the problem has often progressed to the point where parts need to be replaced, missing the optimal maintenance window.

[0004] Secondly, traditional diagnostic equipment suffers from poor vehicle compatibility. Different brands and years of vehicles use different communication protocols and data definitions. While gasoline-powered vehicles commonly use the CAN 2.0 bus, new energy vehicles increasingly employ CAN FD or even FlexRay buses to meet their high bandwidth requirements. To cover common vehicle models, repair shops often need to purchase multiple dedicated diagnostic tools, each costing anywhere from several thousand to tens of thousands of yuan, and they also need to regularly update their vehicle databases. This not only increases operating costs but also leaves many small repair shops unable to afford a complete set of equipment, forcing them to rely solely on experience for diagnosis.

[0005] Third, new energy vehicles present entirely new diagnostic challenges. Battery packs consist of dozens to hundreds of cells connected in series. While performance degradation in a single cell may not trigger vehicle-wide fault codes, long-term accumulation can lead to a decrease in the overall battery pack capacity and even the risk of thermal runaway. The fault modes of high-voltage components such as drive motors and inverters are completely different from those of traditional engines, and mature diagnostic models are lacking. Air conditioning systems in electric vehicles are driven by electric compressors, and their efficiency degradation also lacks effective quantitative monitoring methods. Existing diagnostic equipment has virtually no capability to test these components.

[0006] Fourth, some high-end diagnostic instruments have begun to upload collected data to the cloud for analysis, leveraging big data and machine learning models in the background to identify anomalies. While this approach does improve diagnostic accuracy, it also introduces new problems: network dependence, as signal coverage in underground parking spaces in repair shops and remote areas is often unstable, and the equipment cannot function properly when the network is down; data privacy risks, as vehicle operation data involves sensitive information such as the owner's driving trajectory and driving habits, and uploading it to third-party servers poses a risk of leakage; and insufficient real-time performance, as there are delays in the data upload, cloud queuing, and result return processes, making it impossible to perform diagnosis while data is being collected.

[0007] In conclusion, there is an urgent need in the market for an intelligent diagnostic system that can be compatible with both fuel vehicles and new energy vehicles, does not rely on cloud networks, and can complete the quantitative rating of hidden faults locally, in order to meet the actual needs of various scenarios such as auto repair shops, fleet management, vehicle owner self-inspection, and used car evaluation. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention provides an AI-powered intelligent OBD fault diagnosis and early warning system compatible with both gasoline and new energy vehicles. It includes a hardware adaptation layer housed within the OBD diagnostic box, containing multi-protocol physical interfaces supporting CAN2.0, CANFD, and FlexRay buses. These interfaces automatically identify the current bus type and establish a communication link after insertion into the vehicle's OBD interface. A three-axis MEMS vibration sensing unit is included. For new energy and hybrid vehicles, it interfaces with BMS message data and combines it with micro-vibration characteristics collected during battery charging and discharging to provide early warnings of single-cell attenuation and early cell bulging risks. For gasoline vehicles, it captures engine speed fluctuation data under cold start conditions and simultaneously collects timing data signals from engine vibration, battery, starter, alternator, and ignition / intake systems to achieve cold start failure prediction, quantitative and graded early warning of engine carbon buildup, and warnings of hidden fuel consumption increases. This system is designed for all types of vehicles. The system analyzes the chassis vibration spectrum characteristics under vehicle driving conditions to identify early wear and failure of chassis components such as suspension, ball joints, bushings, shock absorbers, and brake pads, as well as potential abnormal noises. It includes an onboard bus transceiver module for data format conversion and signal adaptation between onboard networks with different communication protocols, ensuring stable and real-time information exchange between ECUs. A high-voltage signal isolation protection module establishes an electrical isolation barrier between the high-voltage electrical system and the low-voltage control circuit, blocking dangerous voltage conduction, suppressing common-mode transient interference, and ensuring signal integrity, while meeting automotive-grade functional safety and safety certification requirements. A wireless communication module enables data transmission between devices without physical connections via radio waves. A low-power power management module intelligently regulates the power supply status and voltage level of each sub-module in the electronic system to minimize energy consumption, extend battery life, and improve system efficiency.

[0009] The data acquisition layer, connected to the hardware adaptation layer, is used to collect operating condition data of the vehicle electronic control unit, battery management system, motor controller, engine electronic control unit and air conditioning controller in real time at a configurable sampling frequency. It also collects vibration sensing data of fuel vehicles, fuel vehicle start-up and carbon deposit related data, and adds early warnings for fuel vehicle start-up attenuation related signals and carbon deposit and fuel consumption related signals. At the same time, it collects audio signals from the vibration sensor of the chassis suspension system damper, control arm ball joint and bushing, steering system tie rod ball joint, braking system brake pads and half shaft CV joint, and outputs the multi-source data after aligning them with a unified timestamp.

[0010] An embedded AI inference layer is located inside the OBD diagnostic box and connected to the data acquisition layer. It has multiple lightweight offline AI models built in, which do not rely on cloud computing. After receiving a signal, it first performs unsupervised learning, uploads abnormal data to the cloud for supervised learning, and finally gives an early warning. It is used to run the corresponding quantitative model independently for each core component and output the health score, degradation index or failure probability item by item.

[0011] The diagnostic output layer, connected to the embedded AI inference layer, is used to generate a quantitative rating report based on the output results of each model according to preset grading rules, and push it to the user terminal through a display terminal or wireless communication module.

[0012] The output of the hardware adaptation layer serves as the input of the data acquisition layer, the output of the data acquisition layer serves as the input of the embedded AI inference layer, the output of the embedded AI inference layer serves as the decision basis for the diagnostic output layer, and the diagnostic report and user feedback data generated by the diagnostic output layer serve as a reference source for system parameter updates.

[0013] As a preferred embodiment of the AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles described in this invention, the hardware adaptation layer includes: a protocol identification submodule, which is used to try the CAN2.0, CANFD and FlexRay wake-up sequence in sequence after the system is powered on, and automatically lock the current protocol according to the bus response. The whole process does not require manual configuration.

[0014] The physical interface submodule includes OBD-II pin definitions conforming to the ISO15765-4 standard, and additionally reserves FlexRay differential signal pins, supporting 12V / 24V dual voltage power supply adaptation.

[0015] As a preferred embodiment of the AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles described in this invention, the embedded AI inference layer includes a cell health model for identifying the degradation of individual battery cells in new energy vehicles, providing a health model for each battery cell. Calculate the health status value using the following formula :

[0016] ;

[0017] in, This is the baseline health value for the battery cells at the factory. This is the current capacity normalized value. For rated capacity, The current temperature. For reference temperature, , The attenuation coefficient is pre-calibrated based on the electrochemical mechanism.

[0018] As a preferred embodiment of the AI ​​intelligent OBD fault diagnosis and early warning system compatible with both gasoline and new energy vehicles described in this invention, the embedded AI inference layer includes a carbon deposit index model for quantitative rating of carbon deposits on the throttle valve of gasoline vehicles, and the carbon deposit index is calculated using the following formula. :

[0019] ;

[0020] in, To measure the intake air mass flow rate, This is the theoretical intake mass flow rate. Intake air temperature, As the reference intake temperature, The rate of change of throttle opening. , , The weighting coefficients are used; the diagnostic output layer is based on... The value classifies the degree of carbon buildup into levels 1 to 5.

[0021] As a preferred embodiment of the AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles described in this invention, the embedded AI inference layer includes a motor health scoring model for drive motor fault analysis, and the motor health score is calculated using the following formula. :

[0022] ;

[0023] in, For the Sigmoid function, for shaft current, Rated peak current, For winding temperature, The rated winding temperature, For shell Shaft vibration acceleration, , , Preset weighting coefficients; The closer it is to 1, the higher the probability of failure.

[0024] As a preferred embodiment of the AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel-powered and new energy vehicles described in this invention, the embedded AI inference layer includes a fuel consumption deviation model for determining excessive fuel consumption, and the fuel consumption deviation rate is calculated using the following formula. :

[0025] ;

[0026] in, This refers to the actual fuel consumption per 100 kilometers. This is the theoretical fuel consumption per 100 kilometers based on the engine MAP chart; when When the preset threshold is exceeded, the diagnostic output layer determines that the fuel consumption is abnormal.

[0027] As a preferred embodiment of the AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles described in this invention, the embedded AI inference layer includes a compressor health degradation model for progressive fault warning of the air conditioning system, and the health degradation rate is calculated using the following formula. :

[0028] ;

[0029] in, This represents the current power of the compressor. The rated power of the compressor. This is the current cooling capacity. This is the rated cooling capacity; when If two consecutive diagnoses show an upward trend, the system determines it to be a progressive failure.

[0030] As a preferred embodiment of the AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles described in this invention, the diagnostic output layer has a built-in four-level quantitative rating rule, which is set into four levels from low to high severity: minor, attention, serious, and critical.

[0031] Mild level: All model outputs are within safe limits; only data is recorded.

[0032] Attention Level: If the output of any model exceeds the normal range but does not reach the alarm threshold, the system will mark it with a yellow warning in the report;

[0033] Severity level: If any model output reaches the alarm threshold, the system will display a red warning on the screen and recommend that it be repaired as soon as possible;

[0034] Critical level: Motor health score ≥0.7 or cell If the system falls below the preset safety threshold (80%), it will immediately issue an audible and visual alarm and recommend stopping operation.

[0035] As a preferred embodiment of the AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles described in this invention, the embedded AI inference layer also includes a vehicle model routing table, which automatically loads the corresponding model combination and parameter configuration based on the bus type and vehicle VIN code identified by the hardware adaptation layer.

[0036] For gasoline vehicles, load the throttle body carbon buildup model, fuel consumption deviation model, start-up anomaly diagnosis model, chassis component abnormal noise failure model and air conditioning attenuation model;

[0037] For new energy vehicles, a battery cell health model, a motor health scoring model, a chassis component noise and failure model, and an air conditioning attenuation model are loaded.

[0038] For hybrid vehicles, all the above models are loaded simultaneously, and the sampling priority is switched according to the current driving mode.

[0039] As a preferred embodiment of the AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles described in this invention, it also includes a local historical diagnostic database. This database stores the original time-series data snapshots of each diagnosis, the intermediate layer output features of each AI model, and the final quantitative rating results. When the system reconnects to the same vehicle, it automatically retrieves the vehicle's historical diagnostic records, compares the current quantitative rating with the historical baseline, generates a degradation trend curve, and highlights components whose deterioration rate exceeds a preset threshold in the diagnostic report.

[0040] The beneficial effects of this invention are reflected in:

[0041] This invention does not rely on the standard fault code system. Instead, it uses a multi-dimensional AI model to quantify and rate latent faults, overcoming the limitations of traditional solutions that rely solely on standard fault codes. The hardware adaptation layer supports three mainstream automotive bus protocols: CAN2.0, CANFD, and FlexRay, automatically identifying and locking them without requiring manual vehicle model selection, achieving true plug-and-play functionality. All AI models run locally on the OBD diagnostic box, independent of cloud networks, ensuring data privacy and diagnostic reliability. A local historical database enables the system to track trends, promptly identifying components experiencing accelerated deterioration and providing a basis for preventative maintenance. The system can be applied to various scenarios, including routine inspections at auto repair shops, batch inspections of fleets, owner-managed monitoring, and used car technical condition assessments.

[0042] This invention integrates bus data with physical vibration sensing software and hardware, without relying on additional sensors added by the vehicle manufacturer. It outputs graded warnings in the early stages of fault occurrence, applicable to aftermarket fault diagnosis and warning scenarios for passenger cars and commercial vehicles. For new energy and hybrid vehicles, it interfaces with BMS message data and combines micro-vibration characteristics collected during battery charging and discharging to complete early risk warnings for single-cell attenuation and cell bulging. For gasoline vehicles, it captures engine speed fluctuation data under cold start conditions and simultaneously collects timing data signals from engine vibration, battery, starter, generator, and ignition and intake systems to achieve cold start failure prediction and quantitative graded warnings for engine carbon buildup. For all vehicle models, it analyzes the chassis vibration spectrum characteristics under vehicle driving conditions to identify early wear failures and abnormal noise hazards in chassis components such as suspension, ball joints, bushings, shock absorbers, and brake pads. Attached Figure Description

[0043] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0044] Figure 1 This is a schematic diagram of the overall system architecture of the present invention;

[0045] Figure 2 This is a flowchart of the intelligent diagnosis and quantitative rating process of the present invention;

[0046] Figure 3 This is a flowchart illustrating the multi-scenario closed-loop application of the present invention. Detailed Implementation

[0047] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0048] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0049] Example 1

[0050] like Figure 1 , Figure 2 , Figure 3As shown, an AI intelligent OBD fault diagnosis and early warning system compatible with both gasoline and new energy vehicles includes a hardware adaptation layer located within the OBD diagnostic box. This layer includes multi-protocol physical interfaces supporting CAN2.0, CANFD, and FlexRay buses, used to automatically identify the current bus type and establish a communication link after insertion into the vehicle's OBD interface. It also includes a three-axis MEMS vibration sensing unit. For new energy and hybrid vehicles, it interfaces with BMS message data and combines it with micro-vibration characteristics collected during battery charging and discharging to provide early warnings of single-cell attenuation and early cell bulging risks. For gasoline vehicles, it captures engine speed fluctuation data under cold start conditions and simultaneously collects engine vibration signals, battery, starter, alternator, and ignition / intake system timing data signals to achieve cold start failure prediction, quantitative and graded early warning of engine carbon buildup, and warnings of hidden fuel consumption increases. It is applicable to all vehicle models and analyzes vehicle data. The system identifies chassis vibration spectrum characteristics under vehicle driving conditions, enabling early wear and failure of chassis components such as suspension, ball joints, bushings, shock absorbers, and brake pads, as well as potential abnormal noises. It includes an onboard bus transceiver module for data format conversion and signal adaptation between onboard networks using different communication protocols, ensuring stable and real-time information exchange between ECUs. A high-voltage signal isolation protection module establishes an electrical isolation barrier between the high-voltage electrical system and the low-voltage control circuit, blocking dangerous voltage conduction, suppressing common-mode transient interference, and ensuring signal integrity, while meeting automotive-grade functional safety and safety certification requirements. A wireless communication module enables data transmission between devices without physical connections via radio waves. A low-power power management module intelligently regulates the power supply status and voltage level of each sub-module in the electronic system to minimize energy consumption, extend battery life, and improve system efficiency.

[0051] The data acquisition layer, connected to the hardware adaptation layer, is used to collect operating condition data of the vehicle electronic control unit, battery management system, motor controller, engine electronic control unit and air conditioning controller in real time at a configurable sampling frequency. It also collects vibration sensing data of fuel vehicles, fuel vehicle start-up and carbon deposit related data, and adds early warnings for fuel vehicle start-up attenuation related signals and carbon deposit and fuel consumption related signals. At the same time, it collects audio signals from the vibration sensor of the chassis suspension system damper, control arm ball joint and bushing, steering system tie rod ball joint, braking system brake pads and half shaft CV joint, and outputs the multi-source data after aligning them with a unified timestamp.

[0052] An embedded AI inference layer is located inside the OBD diagnostic box and connected to the data acquisition layer. It has multiple lightweight offline AI models built in, which do not rely on cloud computing. After receiving a signal, it first performs unsupervised learning, uploads abnormal data to the cloud for supervised learning, and finally gives an early warning. It is used to run the corresponding quantitative model independently for each core component and output the health score, degradation index or failure probability item by item.

[0053] The diagnostic output layer, connected to the embedded AI inference layer, is used to generate a quantitative rating report based on the output results of each model according to preset grading rules, and push it to the user terminal through a display terminal or wireless communication module.

[0054] The output of the hardware adaptation layer serves as the input of the data acquisition layer, the output of the data acquisition layer serves as the input of the embedded AI inference layer, the output of the embedded AI inference layer serves as the decision basis for the diagnostic output layer, and the diagnostic report and user feedback data generated by the diagnostic output layer serve as a reference source for system parameter updates.

[0055] Preferably, the hardware adaptation layer includes a protocol identification submodule, which attempts the CAN2.0, CANFD and FlexRay wake-up sequence sequentially after the system is powered on, and automatically locks the current protocol based on the bus response. The entire process requires no manual configuration.

[0056] The physical interface submodule includes OBD-II pin definitions conforming to the ISO15765-4 standard, and additionally reserves FlexRay differential signal pins, supporting 12V / 24V dual voltage power supply adaptation.

[0057] Preferably, the embedded AI inference layer includes a cell health model for identifying the degradation of individual battery cells in new energy vehicles, providing data for each cell. Calculate the health status value using the following formula :

[0058] ;

[0059] in, This is the baseline health value for the battery cells at the factory. This is the current capacity normalized value. For rated capacity, The current temperature. For reference temperature, , The attenuation coefficient is pre-calibrated based on the electrochemical mechanism.

[0060] Preferably, the embedded AI inference layer includes a carbon deposit index model for quantitative rating of carbon deposits on the throttle body of fuel vehicles, and the carbon deposit index is calculated using the following formula. :

[0061] ;

[0062] in, To measure the intake air mass flow rate, This is the theoretical intake mass flow rate. Intake air temperature, As the reference intake temperature, The rate of change of throttle opening. , , The weighting coefficients are used; the diagnostic output layer is based on... The value classifies the degree of carbon buildup into levels 1 to 5.

[0063] Preferably, the embedded AI inference layer includes a motor health scoring model for drive motor fault assessment, which calculates the motor health score using the following formula. :

[0064] ;

[0065] in, For the Sigmoid function, for shaft current, Rated peak current, For winding temperature, The rated winding temperature, For shell Shaft vibration acceleration, , , Preset weighting coefficients; The closer it is to 1, the higher the probability of failure.

[0066] Preferably, the embedded AI inference layer includes a fuel consumption deviation model for determining excessive fuel consumption, and the fuel consumption deviation rate is calculated using the following formula. :

[0067] ;

[0068] in, This refers to the actual fuel consumption per 100 kilometers. This is the theoretical fuel consumption per 100 kilometers based on the engine MAP chart; when When the preset threshold is exceeded, the diagnostic output layer determines that the fuel consumption is abnormal.

[0069] Preferably, the embedded AI inference layer includes a compressor health degradation model for progressive fault warning of the air conditioning system, and the health degradation rate is calculated using the following formula. :

[0070] ;

[0071] in, This represents the current power of the compressor. The rated power of the compressor. This is the current cooling capacity. This is the rated cooling capacity; when If two consecutive diagnoses show an upward trend, the system determines it to be a progressive failure.

[0072] Preferably, the diagnostic output layer has a built-in four-level quantitative rating rule, which is set to four levels from low to high severity: mild, concern, severe, and critical.

[0073] Mild level: All model outputs are within safe limits; only data is recorded.

[0074] Attention Level: If the output of any model exceeds the normal range but does not reach the alarm threshold, the system will mark it with a yellow warning in the report;

[0075] Severity level: If any model output reaches the alarm threshold, the system will display a red warning on the screen and recommend that it be repaired as soon as possible;

[0076] Critical level: Motor health score ≥0.7 or cell If the system falls below the preset safety threshold (80%), it will immediately issue an audible and visual alarm and recommend stopping operation.

[0077] Preferably, the embedded AI inference layer also includes a vehicle model routing table, which automatically loads the corresponding model combination and parameter configuration based on the bus type and vehicle VIN code identified by the hardware adaptation layer.

[0078] For gasoline vehicles, load the throttle body carbon buildup model, fuel consumption deviation model, start-up anomaly diagnosis model, chassis component abnormal noise failure model and air conditioning attenuation model;

[0079] For new energy vehicles, a battery cell health model, a motor health scoring model, a chassis component noise and failure model, and an air conditioning attenuation model are loaded.

[0080] For hybrid vehicles, all the above models are loaded simultaneously, and the sampling priority is switched according to the current driving mode.

[0081] Preferably, a local historical diagnostic database is also included, which stores the original time-series data snapshots of each diagnosis, the intermediate layer output features of each AI model, and the final quantitative rating results. When the system reconnects to the same vehicle, it automatically retrieves the vehicle's historical diagnostic records, compares the current quantitative rating with the historical baseline, generates a degradation trend curve, and highlights components whose deterioration rate exceeds a preset threshold in the diagnostic report.

[0082] Example 2

[0083] A chain auto repair shop has adopted this system for routine vehicle inspections. A technician inserted the OBD diagnostic box into the OBD-II port of a hybrid SUV. Upon power-up, the system automatically completed bus protocol identification. The vehicle is equipped with a 2.0T gasoline engine and a permanent magnet synchronous motor. The battery pack is a ternary lithium battery pack with 96 cells connected in series. The system completed the dual-bus handshake between CAN2.0 and CANFD within 2.3 seconds and began collecting real-time data from the engine ECU, motor controller MCU, battery management system BMS, and air conditioning controller at a frequency of 50Hz.

[0084] The data collection lasted 15 minutes, covering multiple operating conditions including idling, accelerating to 3000 rpm while stationary, air conditioning at maximum cooling, and engine off and stationary. The embedded AI inference unit sequentially invoked five lightweight offline models: the cell health model calculated the health of each of the 96 battery cells. The results showed that cell 37 had the lowest SOH (86.9%), while the remaining cells were all above 92%, indicating a slight consistency deviation in this cell. The throttle body carbon buildup model calculates the carbon buildup index based on the changes in intake air volume, intake air temperature, and throttle opening. =5.66, corresponding to level 3 medium carbon buildup; drive motor model calculation =0.013, judged as healthy; the fuel consumption model compares actual fuel consumption with theoretical fuel consumption, and concludes... =+8.2%, slightly higher than the normal range, suggesting possible influence from driving habits or minor carbon buildup; the air conditioning model calculates the compressor's health degradation rate. =+26.1%, which indicates a gradual decline, and we recommend paying attention to it.

[0085] After the diagnosis, the system automatically generates a structured report, including quantitative values ​​for each indicator, rating levels, maintenance recommendations, and a trend comparison with the vehicle's last diagnosis (three months ago). The technician prints out the report and gives it to the owner, who then schedules a throttle body cleaning and air conditioning system inspection based on the recommendations. The entire diagnostic process takes less than 20 minutes, requires no internet connection, and does not require reading any fault codes, thus completing the screening of hidden faults that traditional equipment would require multiple visits to the shop to detect.

[0086] Example 3

[0087] This embodiment details the plug-and-play adaptation process of the system on multi-bus vehicle models. A certain new energy logistics vehicle uses the FlexRay bus as the power domain backbone and CANFD as the body domain auxiliary bus. After the OBD diagnostic box is inserted, the main MCU (STM32H743) first attempts to send a standard ISO15765-4 request frame at a rate of 500kbps. If there is no response within 2ms, it automatically switches to CANFD to retry; if there is still no response, it calls the FlexRay wake-up sequence, sends a specific wake-up frame (0xFFFF), and waits for the synchronization frame. Actual measurements show that the average time from wake-up to establishing stable communication on the FlexRay bus is 180ms, far lower than the tens of seconds required for manual protocol selection in traditional devices.

[0088] After a successful handshake, the system loads the vehicle-specific routing table, mapping signals such as motor torque, speed, and IGBT temperature from the FlexRay to standardized Diagnostic Data Identifiers (DIDs). Simultaneously, the system automatically identifies the battery type as lithium iron phosphate and switches the preset parameters of the cell health model. =0.12, =0.006, =30°C). The entire adaptation process requires no manual intervention, truly achieving "plug and play".

[0089] Example 4

[0090] This embodiment demonstrates the quantitative calculation process of cell SOH and throttle body carbon deposits using specific numerical values. The 37th cell of the aforementioned hybrid SUV is selected, and its current normalized capacity value is... =0.823, rated capacity =1.0, current temperature =38°C, reference temperature =25°C. Substitute into the corrected formula:

[0091] =1.0−0.15×(1−0.823)−0.008×(38−25)=1.0−0.15×0.177−0.008×13=1.0−0.02655−0.104=0.86945

[0092] Rounding to the nearest whole number =86.9%. This value is higher than the safety lower limit of 80%, but lower than the average value of other cells in the same group (94.2%). It is judged to be a slight L1 abnormality, and it is recommended to perform equalization charging during the next maintenance.

[0093] Throttle body carbon buildup calculation: Actual measurement under idling conditions =18.5g / s, theoretical =22.0g / s, intake air temperature =42°C, reference temperature =25°C, throttle opening change rate =3.2% / s. Substituting into the formula:

[0094] =0.5×(18.5 / 22.0−1)+0.3×(42−25)+0.2×3.2=0.5×(−0.159)+0.3×17+0.2×3.2=−0.0795+5.1+0.64=5.6605

[0095] Rounding to the nearest whole number =5.66. Based on preset grading: <2 indicates Level 1 (Clean), 2≤ <4 indicates Level 2 (mild), 4≤ <7 indicates Level 3 (Medium), 7≤ <10 is Level 4 (Severe). A value of ≥10 indicates Level 5 (blockage). A value of 5.66 falls into Level 3, the medium carbon buildup range, and throttle body cleaning is recommended.

[0096] Example 5

[0097] This embodiment demonstrates a combined diagnosis of drive motor fault assessment and air conditioning compressor degradation warning. When the vehicle accelerates to 3000 rpm and maintains that speed, the motor controller feeds back the following data: shaft current =185A (rated peak value I_q,max=300A), winding temperature =128°C (rated temperature) =105°C), shell vibration Axial acceleration =4.8 m / s². Substituting into the formula:

[0098] =2.5×(185 / 300)+0.03×(128−105)+0.4×4.8=2.5×0.617+0.03×23+1.92=1.5425+0.69+1.92=4.1525

[0099] =sigmoid(4.1525)=1 / (1+e⁻ 4 ·¹ 5 ² 5 )≈0.9847

[0100] This value far exceeds the high-risk threshold of 0.7, indicating a severe L3 fault. The system immediately displays "Drive motor overheating abnormality, immediate shutdown and inspection recommended" on the instrument panel and records the complete waveform data at the time of the fault for subsequent analysis.

[0101] The air conditioning system measured in maximum cooling mode: current compressor power =1850W, rated power =2200W; Current cooling capacity =3200W, rated cooling capacity =4800W. Substitute into the formula:

[0102] =(1850 / 2200) / (3200 / 4800)−1=(0.8409) / (0.6667)−1=1.2614−1=0.2614

[0103] The compressor health degradation rate is +26.1%, indicating a significant decrease in the actual output of the compressor relative to its cooling capacity, possibly due to internal wear or refrigerant leakage. The system has been flagged as an L1 progressive warning, and it is recommended to check the air conditioning system's sealing during the next maintenance.

[0104] Example 6

[0105] This embodiment describes the differentiated output of the system in four typical application scenarios. Scenario 1: Auto repair shop system, generating a complete diagnostic report PDF, including quantitative values, trend charts, and repair suggestion codes for each indicator, which technicians can directly print for car owners. Scenario 2: Fleet management system, pushing diagnostic summaries to the fleet administrator's mobile app via Bluetooth, highlighting all L2 and above warning items for all vehicles, and automatically sorting them by urgency. Scenario 3: Owner self-monitoring, the system displays concise prompts on the OLED screen such as "Battery cell health: Good," "Throttle body carbon buildup: Moderate, cleaning recommended," and "Air conditioning compressor: Slight degradation," which car owners can view details by pressing a button. Scenario 4: Used car appraisal system, generating a "Vehicle Technical Condition Certificate," including quantitative indicators such as battery SOH distribution map, motor health score, and engine combustion efficiency index, which can be used as a transaction reference document.

[0106] The system's built-in local historical database automatically saves snapshots of each diagnostic. When the same vehicle is reconnected, it automatically calculates the degradation rate of various indicators. For example, the throttle body carbon buildup index of a certain vehicle three months ago... =4.2 (Level 2 Mild), this time =5.66 (Level 3, Medium), the degradation rate is approximately 0.487 / month, exceeding the preset threshold of 0.3 / month. The system will highlight "Throttle body carbon buildup is deteriorating too quickly; it is recommended to shorten the cleaning cycle" in the report. This dynamic early warning capability based on historical comparison is not available in traditional OBD diagnostic equipment.

[0107] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. An AI-powered intelligent OBD fault diagnosis and early warning system compatible with both gasoline and new energy vehicles, characterized in that: This includes a hardware adaptation layer located within the OBD diagnostic box, containing multi-protocol physical interfaces supporting CAN2.0, CANFD, and FlexRay buses. This allows for automatic identification of the current bus type and establishment of a communication link after insertion into the vehicle's OBD interface. It also features a three-axis MEMS vibration sensing unit. For new energy and hybrid vehicles, it interfaces with BMS message data and combines it with micro-vibration characteristics collected during battery charging and discharging to provide early warnings of single-cell attenuation and cell bulging risks. For gasoline vehicles, it captures engine speed fluctuation data under cold start conditions and simultaneously collects timing data signals from engine vibration, battery, starter, alternator, and ignition / intake systems to achieve cold start failure prediction, quantitative and graded warnings of engine carbon buildup, and warnings of hidden fuel consumption increases. For all vehicle models, it analyzes the chassis vibration spectrum characteristics under vehicle driving conditions to identify… It prevents early wear and failure of chassis components such as suspension, ball joints, bushings, shock absorbers, and brake pads, as well as potential abnormal noises; it is equipped with an on-board bus transceiver adapter module for data format conversion and signal adaptation between on-board networks with different communication protocols, ensuring stable and real-time information exchange between ECUs; it is equipped with a high-voltage signal isolation protection module to establish an electrical isolation barrier between the high-voltage power system and the low-voltage control circuit, blocking the conduction of dangerous voltages, suppressing common-mode transient interference, and ensuring signal integrity, while meeting automotive-grade functional safety and safety certification requirements; it is equipped with a wireless communication module for data transmission between devices without physical connection via radio waves; and it is equipped with a low-power power management module to intelligently regulate the power supply status and voltage level of each sub-module in the electronic system to minimize energy consumption, extend battery life, and improve system energy efficiency. The data acquisition layer, connected to the hardware adaptation layer, is used to collect operating condition data of the vehicle electronic control unit, battery management system, motor controller, engine electronic control unit and air conditioning controller in real time at a configurable sampling frequency. It also collects vibration sensing data of fuel vehicles, fuel vehicle start-up and carbon deposit related data, and adds early warnings for fuel vehicle start-up attenuation related signals and carbon deposit and fuel consumption related signals. At the same time, it collects audio signals from the vibration sensor of the chassis suspension system damper, control arm ball joint and bushing, steering system tie rod ball joint, braking system brake pads and half shaft CV joint, and outputs the multi-source data after aligning them with a unified timestamp. An embedded AI inference layer is located inside the OBD diagnostic box and connected to the data acquisition layer. It has multiple lightweight offline AI models built in, which do not rely on cloud computing. After receiving a signal, it first performs unsupervised learning, uploads abnormal data to the cloud for supervised learning, and finally gives an early warning. It is used to run the corresponding quantitative model independently for each core component and output the health score, degradation index or failure probability item by item. The diagnostic output layer, connected to the embedded AI inference layer, is used to generate a quantitative rating report based on the output results of each model according to preset grading rules, and push it to the user terminal through a display terminal or wireless communication module. The output of the hardware adaptation layer serves as the input of the data acquisition layer, the output of the data acquisition layer serves as the input of the embedded AI inference layer, the output of the embedded AI inference layer serves as the decision basis for the diagnostic output layer, and the diagnostic report and user feedback data generated by the diagnostic output layer serve as a reference source for system parameter updates.

2. The AI intelligent OBD fault diagnosis and early warning system compatible with fuel and new energy vehicles according to claim 1, characterized in that: The hardware adaptation layer includes a protocol identification submodule, which attempts the CAN2.0, CANFD and FlexRay wake-up sequence sequentially after the system is powered on, and automatically locks the current protocol based on the bus response. The entire process requires no manual configuration. The physical interface submodule includes OBD-II pin definitions conforming to the ISO15765-4 standard, and additionally reserves FlexRay differential signal pins, supporting 12V / 24V dual voltage power supply adaptation.

3. The AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles according to claim 1, characterized in that: In the embedded AI inference layer, for new energy vehicle battery cell attenuation identification, a battery cell health model is provided, which is used to calculate the health state value of each battery cell The health state value is calculated according to the following formula : ; in, This is the baseline health value for the battery cells at the factory. This is the current capacity normalized value. For rated capacity, The current temperature. For reference temperature, , The attenuation coefficient is pre-calibrated based on the electrochemical mechanism.

4. The AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles according to claim 1, characterized in that: In the embedded AI inference layer, a carbon deposit index model is set up for the quantitative rating of carbon deposits on the throttle body of fuel vehicles, and the carbon deposit index is calculated according to the following formula. : ; in, To measure the intake air mass flow rate, This is the theoretical intake mass flow rate. Intake air temperature, As the reference intake temperature, The rate of change of throttle opening. , , The weighting coefficients are used; the diagnostic output layer is based on... The value classifies the degree of carbon buildup into levels 1 to 5.

5. The AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles according to claim 1, characterized in that: In the embedded AI inference layer, a motor health scoring model is provided for drive motor fault assessment, and the motor health score is calculated according to the following formula. : ; in, For the Sigmoid function, for shaft current, Rated peak current, For winding temperature, The rated winding temperature, For shell Shaft vibration acceleration, , , Preset weighting coefficients; The closer it is to 1, the higher the probability of failure.

6. The AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles according to claim 1, characterized in that: In the embedded AI inference layer, a fuel consumption deviation model is set up for determining excessive fuel consumption, and the fuel consumption deviation rate is calculated according to the following formula. : ; in, This refers to the actual fuel consumption per 100 kilometers. This is the theoretical fuel consumption per 100 kilometers based on the engine MAP chart; when When the preset threshold is exceeded, the diagnostic output layer determines that the fuel consumption is abnormal.

7. The AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles according to claim 1, characterized in that: In the embedded AI inference layer, a compressor health degradation model is provided for progressive fault warning of the air conditioning system, and the health degradation rate is calculated according to the following formula. : ; in, This represents the current power of the compressor. The rated power of the compressor. This is the current cooling capacity. This is the rated cooling capacity; when If two consecutive diagnoses show an upward trend, the system determines it to be a progressive failure.

8. The AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles according to claim 1, characterized in that: The diagnostic output layer has a built-in four-level quantitative rating rule, which is set into four levels from low to high severity: mild, attention, severe, and critical. Mild level: All model outputs are within safe limits; only data is recorded. Attention Level: If the output of any model exceeds the normal range but does not reach the alarm threshold, the system will mark it with a yellow warning in the report; Severity level: If any model output reaches the alarm threshold, the system will display a red warning on the screen and recommend that it be repaired as soon as possible; Critical level: Motor health score ≥0.7 or cell If the system falls below the preset safety threshold (80%), it will immediately issue an audible and visual alarm and recommend stopping operation.

9. The AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles according to claim 1, characterized in that: The embedded AI inference layer also includes a vehicle model routing table, which automatically loads the corresponding model combination and parameter configuration based on the bus type and vehicle VIN code identified by the hardware adaptation layer. For gasoline vehicles, load the throttle body carbon buildup model, fuel consumption deviation model, start-up anomaly diagnosis model and air conditioning attenuation model; For new energy vehicles, a battery cell health model, a motor health scoring model, and an air conditioning degradation model are implemented. For hybrid vehicles, all the above models are loaded simultaneously, and the sampling priority is switched according to the current driving mode.

10. The AI ​​intelligent OBD fault diagnosis and early warning system compatible with both fuel and new energy vehicles according to claim 1, characterized in that: It also includes a local historical diagnostic database, which stores a snapshot of the raw time-series data for each diagnosis, the intermediate layer output features of each AI model, and the final quantitative rating results. When the system reconnects to the same vehicle, it automatically retrieves the vehicle's historical diagnostic records, compares the current quantitative rating with the historical baseline, generates a degradation trend curve, and highlights components whose deterioration rate exceeds a preset threshold in the diagnostic report.