A fault positioning method, cloud server and system of an in-vehicle infotainment system

Abstract

This invention provides a fault location method, cloud server, and system for an in-vehicle infotainment system. It determines feature vectors for various fault characteristics based on fault data, representing the importance of each fault characteristic to each fault type with quantifiable numerical values ​​(correlation values), thus eliminating biases from human judgment. Because the fault diagnosis model can learn the complex patterns between feature combinations and fault rationality, model-based diagnosis can more accurately identify fault types. Since the three-dimensional correlation matrix associates fault types with possible root causes and corresponding original fault characteristics, it can accurately locate the most likely root cause of the fault. Compared to existing fault location methods that require manual inspection of vehicle logs, this invention utilizes a fault diagnosis model and a three-dimensional correlation matrix of fault root causes to complete most of the fault location work, significantly shortening fault diagnosis time, improving fault diagnosis efficiency, and thus ensuring driving safety.

Description

Technical Field

[0001] This invention relates to the field of vehicle infotainment system technology, and in particular to a fault location method, cloud server and system for an in-vehicle infotainment system. Background Technology

[0002] As the automotive industry accelerates its transformation towards intelligence and connectivity, in-vehicle infotainment (IVI) systems have evolved into the core interactive hub of automobiles. The breadth and depth of IVI system functions continue to expand, and the workload and system complexity they undertake are also growing exponentially. The operational stability of IVI systems faces severe challenges, and frequent failures will further exacerbate the imbalance of system stability. Therefore, unprecedentedly stringent requirements are placed on the reliable operation of IVI systems.

[0003] The main types of IVI system malfunctions include "blackout, distorted, and frozen" malfunctions such as blackout dashboard, distorted screen, and freezing or crashing of the central control screen. These malfunctions may interfere with the normal operation of key vehicle functions such as vehicle control, imaging, and intelligent driving, posing potential hazards to driving safety. Therefore, it is essential to analyze and locate IVI system malfunctions.

[0004] In related technologies, the analysis methods for the "black card" fault in the IVI system still rely on the mode of users manually uploading vehicle logs and maintenance personnel manually downloading and parsing them. This mode has many unavoidable shortcomings: First, log acquisition is difficult, relying on users' active operation, which easily leads to problems such as missing logs and untimely uploads; second, the diagnostic efficiency is low, as manual analysis cannot cope with massive amounts of fault data and it is difficult to quickly locate the fault node; third, the fault tracing capability is insufficient. Faced with the complex fault scenarios of multi-module coupling and multi-factor superposition in the IVI system, it is difficult to accurately mine the fault correlation characteristics and find the root cause of the fault.

[0005] It can be seen that existing fault analysis methods are no longer suitable for the complex operating scenarios of intelligent IVI systems. There is an urgent need for an efficient and accurate fault analysis method to enable rapid location and root cause analysis of IVI system faults. Summary of the Invention

[0006] To address the problems existing in the prior art, embodiments of the present invention provide a fault location method, cloud server, and system for in-vehicle infotainment systems, in order to solve or partially solve the technical problem that the prior art cannot efficiently and accurately locate faults in IVI systems and perform root cause analysis, thereby affecting vehicle driving safety.

[0007] A first aspect of the present invention provides a fault location method for an in-vehicle infotainment system, the method comprising: Fault data of the in-vehicle infotainment system is acquired, and feature vectors of various fault characteristics are determined based on the fault data; the feature vectors contain the correlation value of each fault characteristic to each type of baseline fault. The pre-trained fault diagnosis model is used to diagnose the fault type by analyzing the feature vectors of the aforementioned fault features, thereby obtaining the target fault type. Based on the target fault type and the pre-constructed three-dimensional correlation matrix of fault root causes, the root cause of the target fault is located to obtain the target fault root cause corresponding to the target fault type.

[0008] In the above scheme, determining the feature vectors of each fault characteristic based on the fault data includes: Extract fault features associated with each baseline fault type from the fault data; the baseline fault types include black screen, distorted screen, lag, and system crash. For each fault feature, the correlation value of the fault feature with each benchmark fault type is determined based on the historical fault sample library, and the feature vector of the fault feature is determined based on the correlation value of the fault feature with each benchmark fault type.

[0009] In the above scheme, determining the association value of the fault features with each type of baseline fault based on the historical fault sample library includes: According to the formula Determine the association value of the fault characteristics with each baseline fault type; The For the first i Fault characteristics for the first t The associated value of the class baseline fault type, the The first in the historical fault sample library i The fault characteristics are accompanied by the first t The number of samples that occurred for the benchmark fault type, the The first in the historical fault sample library t The total number of samples that occurred for the baseline fault type. t =1 indicates a black screen. t A value of 2 indicates a screen flickering error. t =3 indicates lag, when t =4 indicates a system crash. For the first i The anomaly coefficient of the fault characteristics.

[0010] In the above scheme, before using the pre-trained fault diagnosis model to diagnose the fault type of the feature vectors of the various fault features, the method further includes: Construct a first training set, which includes feature vectors of each historical fault feature of the in-vehicle infotainment system; each historical fault feature has a corresponding fault label. The pre-built fault diagnosis model is iteratively trained using the first training set until the model converges.

[0011] In the above scheme, the step of using a pre-trained fault diagnosis model to diagnose the fault type from the feature vectors of the various fault features includes: Obtain the fault correlation value sub-model in the fault diagnosis model, wherein the fault correlation value sub-model is: ; The fault association value sub-model is used to determine the first... t Fault association values ​​of the baseline fault type ; The fault association values ​​of all baseline types are sorted from largest to smallest to obtain a fault association value sequence; If the maximum fault association value is greater than a preset first threshold and the difference between the maximum fault association value and the second largest fault association value is greater than a preset difference threshold, then the baseline fault type corresponding to the maximum fault association value is determined as the target fault type. If the maximum fault correlation value is less than or equal to a preset first threshold, or the difference between the maximum fault correlation value and the second largest fault correlation value is less than or equal to a preset difference threshold, then the baseline fault type corresponding to the maximum fault correlation value and the baseline fault type corresponding to the second largest fault correlation value are determined as the target fault type; wherein, The For the first i Fault characteristics for the first t The associated value of the class baseline fault type, the For the first i Fault characteristics for the first t The weights of the class-based fault types, the For the first t Reporting correction factor for baseline fault types, t =1 indicates a black screen. t A value of 2 indicates a screen flickering error. t =3 indicates lag, when t =4 indicates a system crash.

[0012] In the above scheme, after obtaining the target fault type, the method further includes: Obtain the first confidence sub-model in the fault diagnosis model, wherein the first confidence sub-model is... ; Determine the first confidence sub-model using the first confidence sub-model First confidence level corresponding to the target fault type The first confidence level is used to assess the credibility of the target fault type; wherein, For the first The fault association value corresponding to the target fault type, the For the first t Fault association values ​​of the baseline fault type The .

[0013] In the above scheme, the three-dimensional correlation matrix of fault root causes stores the correspondence between fault type, fault root cause, and fault feature. The step of root cause localization based on the target fault type and the pre-constructed three-dimensional correlation matrix of fault root causes to obtain the target fault root cause corresponding to the target fault type includes: Based on the target fault type, a set of corresponding candidate fault root causes is selected from the three-dimensional correlation matrix of fault root causes. Determine a corresponding second confidence level for each candidate fault root cause in the candidate fault root cause set; If the second confidence level is greater than the preset second threshold, then the candidate fault root cause is determined as the target fault root cause corresponding to the target fault type.

[0014] In the above scheme, determining a corresponding second confidence level for each candidate fault root cause in the candidate fault root cause set includes: The second confidence level of each candidate root cause of failure is determined using the second confidence sub-model. ;in, The second confidence sub-model is The In order to meet the number of fault characteristics of the root cause of the fault, the The total number of original fault features corresponding to the root cause of the fault, the The deviation correction coefficient is used to conform to the original fault characteristics of the root cause of the fault. For the first t Fault association values ​​for the baseline fault type, t =1 indicates a black screen. t A value of 2 indicates a screen flickering error. t =3 indicates lag, when t =4 indicates a system crash.

[0015] A second aspect of the present invention provides a fault location cloud server for an in-vehicle infotainment system, the cloud server comprising: The data preprocessing unit is used to acquire fault data of the in-vehicle infotainment system and determine the feature vector of each fault feature based on the fault data; the feature vector contains the correlation value of each fault feature to each type of baseline fault. The fault diagnosis unit is used to perform fault type diagnosis on the feature vectors of the various fault features using a pre-trained fault diagnosis model to obtain the target fault type. The root cause localization unit is used to perform root cause localization based on the target fault type and a pre-constructed three-dimensional correlation matrix of fault root causes, so as to obtain the target fault root cause corresponding to the target fault type.

[0016] A second aspect of the present invention provides a fault location system for an in-vehicle infotainment system, the system comprising: a vehicle, a cloud server, and a data platform; The vehicle is used to collect fault data of the in-vehicle infotainment system using data acquisition sensors and transmit the fault data to the cloud server. The cloud server is used to determine feature vectors for each fault feature based on the fault data; the feature vectors contain the correlation value of each fault feature to each baseline fault type; and the fault diagnosis model is used to diagnose the fault type by analyzing the feature vectors of each fault feature to obtain the target fault type. Based on the target fault type and the pre-constructed three-dimensional correlation matrix of fault root causes, the root cause is located to obtain the target fault root cause corresponding to the target fault type. The data platform is used to display the fault type and target root cause of the fault based on a preset display format.

[0017] This invention provides a fault location method, cloud server, and system for an in-vehicle infotainment system. The method includes: acquiring fault data of the in-vehicle infotainment system; determining feature vectors for various fault features based on the fault data; the feature vectors containing the correlation value of each fault feature to each baseline fault type; using a pre-trained fault diagnosis model to diagnose the fault type of the feature vectors of the various fault features to obtain a target fault type; and performing root cause location based on the target fault type and a pre-constructed three-dimensional correlation matrix of fault root causes to obtain the target fault root cause corresponding to the target fault type. Thus, determining the feature vectors of various fault features based on fault data means that it no longer relies on the experience or vague descriptions of maintenance personnel, but instead uses quantifiable numerical values ​​to represent the importance of each fault feature to each fault type. The correlation value is used to eliminate the bias of human judgment and make the input parameters for fault diagnosis more objective. When using the fault diagnosis model for fault type diagnosis, the model can learn the complex rules between feature combinations and fault rationality, so the model-based diagnosis can more accurately identify the fault type. Finally, since the three-dimensional correlation matrix strongly correlates the fault type with the possible root cause and the corresponding original fault features, the most likely fault root cause can be accurately located when locating the fault root cause through the three-dimensional correlation matrix. In addition, compared with the fault location method in the prior art that requires manual checking of the vehicle log one by one, the present invention can complete most of the fault location work by using the fault diagnosis model and the three-dimensional correlation matrix of fault root cause, which greatly shortens the fault diagnosis time, improves the fault diagnosis efficiency, and thus ensures driving safety. Attached Figure Description

[0018] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A schematic diagram of the overall structure of a fault location system for an in-vehicle infotainment system according to an embodiment of the present invention is shown. Figure 2 A schematic flowchart of a fault location method for an in-vehicle infotainment system according to an embodiment of the present invention is shown. Figure 3 A schematic diagram of the overall structure of a fault location cloud server for an in-vehicle infotainment system according to an embodiment of the present invention is shown. Detailed Implementation

[0019] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0020] To better understand the technical solution of this invention, a fault location system for an in-vehicle infotainment system will be introduced first, such as... Figure 1 As shown, the fault location system has a three-level architecture, including: vehicle 1, cloud server 2, and data platform 3; Vehicle 1 is used to collect fault data of the in-vehicle infotainment system using data acquisition sensors and transmit the fault data to cloud server 2. Cloud Server 2 is used to determine the feature vectors of various fault characteristics based on fault data. The feature vectors contain the correlation values ​​of each fault characteristic to each baseline fault type. The fault diagnosis model is used to diagnose the fault type of each fault characteristic's feature vector to obtain the target fault type. Based on the target fault type and the pre-constructed three-dimensional correlation matrix of fault root causes, the root cause is located to obtain the target fault root cause corresponding to the target fault type. Data Platform 3 is used to display fault types and target root causes based on preset display formats.

[0021] Specifically, this invention is mainly used to identify faults such as black screen, distorted screen, lag and crash in IVI system. When vehicle 1 collects fault data, it can collect data from three aspects: fault identification, collection dimension and collection mechanism.

[0022] For fault identifiers, before data collection, it is necessary to initially identify the type of fault based on preset fault identification rules, and combine this with information actively reported by the user (such as the navigation screen getting stuck) to determine the fault identifier (the initially determined fault type). After the fault identifier is initially identified, during data collection, not all data will be collected indiscriminately. Instead, the priority of fault features that are strongly correlated with the currently occurring fault identifier will be increased, and the fault data corresponding to the fault identifier will be collected first.

[0023] For example, the preset fault identification rules can be: Lag: The frame rate of the central control screen is less than 10fps; Screen flickering: The abnormal proportion of screen pixels exceeds the threshold; Black screen: The screen backlight is off and there is no response; System crash: The system has been unresponsive for longer than the threshold. If the frame rate of the central control screen is determined to be less than 10fps and the user reports that the screen freezes when clicking on the navigation, then the fault can be identified as a lag. Subsequently, fault data related to the lag will be collected first, such as CPU load, frame rate, and application response time.

[0024] If it is determined that the proportion of abnormal screen pixels exceeds the threshold, the fault can be identified as a screen distortion problem. This will increase the collection frequency and priority of fault data related to screen distortion, such as prioritizing the collection of abnormal screen pixel proportions.

[0025] After initially identifying the fault identifier, corresponding fault data will be collected according to three dimensions: hardware, software, and interaction. At the hardware level, it is necessary to collect fault data that reflects abnormal hardware operation, such as CPU load, memory usage, system power supply status, device temperature, etc. Optional auxiliary data may include screen frame rate, backlight current, etc.

[0026] At the software level, it is necessary to collect fault data that reflects abnormal operation of the running system, such as graphics rendering process logs and display driver error information; optional auxiliary fault data may include diagnostic reports from third-party diagnostic tools (such as third-party diagnostic tools installed in the vehicle).

[0027] In terms of interaction, it can collect user click events and the names of the applications that triggered them when a fault occurs.

[0028] The data acquisition mechanism may include real-time acquisition, command acquisition, and timed acquisition. That is, when collecting fault data, the acquisition mechanism can be selected according to actual needs, and the above-mentioned multi-dimensional fault data can be collected under the acquisition mechanism.

[0029] When the data acquisition mechanism is set to real-time acquisition, fault data acquisition will be triggered when a fault occurs, such as when the dashboard screen goes black or the central control screen does not respond for ≥1 second.

[0030] When the data collection mechanism is command-based, fault data can be collected upon receiving a collection command sent by cloud server 2.

[0031] When the data acquisition mechanism is timed acquisition, fault data can be collected based on a pre-set acquisition cycle.

[0032] It should be noted that Cloud Server 2 can be configured with corresponding data collection modes, data collection periods, data collection frequencies, and data collection dimensions for different vehicle models.

[0033] For example, Cloud Server 2 can remotely issue different data collection triggering methods based on vehicle model and high-incidence fault scenarios: for certain vehicle models, real-time data collection is prioritized, and fault data is collected immediately as soon as a fault occurs; For other vehicle models, timed data collection is configured; alternatively, when it is necessary to investigate specific problems, instructions can be issued to collect data, allowing the vehicle to immediately report the specified data.

[0034] Cloud Server 2 allows you to specify the start and end times for data collection. For example, you can collect only the critical data from 30 seconds before the failure to 1 minute after it, thus avoiding full data collection.

[0035] Cloud Server 2 can also be configured with a sampling frequency. For example, when a screen flickering fault is detected, the cloud server issues an instruction to increase the sampling frequency of abnormal pixel ratios from 1 time / second to 10 times / second; when there is no fault, the sampling frequency is reduced to save bandwidth and storage.

[0036] The cloud server 2 can flexibly configure the data collection dimensions. For example, when troubleshooting hardware problems, the focus can be on collecting hardware-level fault data (CPU, memory, temperature); when troubleshooting software problems, the focus can be on collecting software-level fault data (application logs, kernel logs); it can also be configured to collect only sensor data specific to a particular vehicle model.

[0037] When vehicle 1 collects fault data, it needs to de-identify the fault data to remove private information (such as mobile phone number, ID card number, etc.). After collection, the fault data needs to be transmitted to cloud server 2.

[0038] Vehicle 1 and cloud server 2 use a dual-protocol adaptive switching method for vehicle-to-cloud communication, employing both Message Queuing Telemetry Transport (MQTT) and Hypertext Transfer Protocol (HTTP). For scenarios involving network fluctuations on the vehicle side, the communication protocol can be automatically switched. For example, HTTP is used to transmit large data packets when the network is stable, while MQTT is used to ensure low latency when the network fluctuates (e.g., packet loss rate > 1%). In the event of a network interruption, faulty data is temporarily stored locally, awaiting network recovery for resumed transmission.

[0039] It should be noted that during data transmission, the TLS 1.3 protocol is used to encrypt faulty data to ensure end-to-end data security. After receiving faulty data, Cloud Server 2 will use a symmetric encryption algorithm (such as AES-256 encryption algorithm) to encrypt and store the faulty data.

[0040] Continue to refer to Figure 1 The cloud server 2 includes a data preprocessing unit 21, a fault diagnosis unit 22, and a root cause localization unit 23; The data preprocessing unit 21 determines the feature vectors of each fault feature based on the fault data; the feature vectors contain the correlation value of each fault feature to each type of baseline fault. The fault diagnosis unit 22 uses a pre-trained fault diagnosis model to diagnose the fault type of the feature vectors of various fault features and obtain the target fault type. The root cause localization unit 23 is used to perform root cause localization based on the target fault type and the pre-constructed three-dimensional correlation matrix of fault root causes, so as to obtain the target fault root cause corresponding to the target fault type.

[0041] The fault diagnosis model includes a fault correlation value sub-model and a first confidence sub-model; the root cause localization unit 23 includes a second confidence sub-model. Specifically, the root cause localization unit 23 can use the second confidence sub-model to perform root cause localization based on the target fault type and the pre-constructed three-dimensional correlation matrix of fault root causes, and obtain the target fault root cause corresponding to the target fault type.

[0042] The specific implementation logic of cloud server 2 can be found in the detailed description of the subsequent cloud server-side implementation, so it will not be repeated here.

[0043] After determining the target fault type and root cause, cloud server 2 sends the target fault type and root cause to data platform 3. Data platform 3 displays the target fault type and root cause based on a preset display format; it also extracts the corresponding solution from the historical solution library based on the fault type and pushes the solution to the target user's terminal device.

[0044] Specifically, data platform 3 also needs to acquire fault data collected by vehicle 1, and then integrate the target fault type, target fault root cause, and fault data to achieve fault visualization. This includes displaying the distribution of fault type and fault root cause on the "black card" in multi-dimensional chart form, retrieving fault data based on fault type, time range, and vehicle information, and displaying the analysis chain of a single fault to achieve fault tracing; the analysis chain is: original fault data - fault characteristics - fault type diagnosed by the model - root cause analysis.

[0045] The following is an example of a scenario where the distribution of fault types and root causes of "blackened cards" is displayed in a multi-dimensional chart format: Suppose that the maintenance personnel of an automaker need to analyze the overall distribution of "black card" faults in Model B of Brand A in the first quarter of 2025. The visualization logic of Data Platform 3 is as follows: Automotive maintenance personnel can select the brand and model dimensions (Brand A, Model B) and the time dimension (Q1 2025) on the human-machine interface of Data Platform 3, and set the filter condition to "screen display error" as the fault type. Data Platform 2 can automatically generate the screen display error situation for Brand A, Model B in Q1 2025 based on the received filter conditions. The frequency of screen flickering failures each week in the first quarter is displayed using a bar chart, clearly showing the frequency of screen flickering failures each week in the first quarter; A pie chart can be used to show the distribution of root causes of screen flickering failures during a given period. For example, a pie chart can show that GPU rendering process crashes account for 65%, abnormal display driver versions account for 25%, and damaged display hardware pixels account for 10%.

[0046] The following is an example of a scenario that demonstrates the analysis of a single fault: Suppose a 4S dealership repair technician receives a repair order for a C-brand D-model vehicle. The user reports that the navigation app displays a black screen after clicking on it. The logic for displaying a single fault path is as follows: The repair personnel enter the vehicle's VIN code and the time the fault occurred to retrieve the complete fault data, specifically: Original fault data: Displaying fault data collected from the vehicle: When the navigation APP starts, the CPU load reaches 98%, the memory usage rate is 92%, and the screen backlight voltage is 0V; Fault characteristics: The correlation value between CPU overload and black screen failure is 0.91, and the correlation value between memory overflow and black screen failure is 0.88. Model fault diagnosis: determined to be a single fault: black screen, with a black screen fault correlation value of 0.95; Root cause analysis: The results of the 3D correlation matrix matching show that the root cause of the failure is a system process crash caused by a memory leak in the navigation app.

[0047] In addition, Data Platform 3 can also extract corresponding solutions from the historical solution library based on the fault type and push the solutions to the target user's terminal device. Specifically, Data Platform 3 can extract solutions containing the fault root cause, solution, and operation steps from the historical solution library based on the target fault root cause and target fault type output by Cloud Server 2, and can push them to different objects (such as car owners, 4S store personnel, etc.). After the fault is resolved, the repair conclusion and success rate of the solution are recorded and pushed to the car manufacturer's operation and maintenance platform to trigger the update of the solution.

[0048] For example, when a car owner is driving a B model of brand A, the car's infotainment system displays a screen malfunction message. After diagnosis, the root cause of the problem is a temporary abnormality in the display driver (confidence level 0.89), which is a minor fault that can be repaired remotely.

[0049] Data Platform 3 matches the corresponding root cause of the problem to the owner's self-service repair solution from the historical solution database. The solution includes: root cause of the problem, one-click repair steps and precautions.

[0050] The solution is pushed out via a pop-up window on the in-vehicle infotainment system and / or a message in the owner's app. The content is simplified to steps that the owner can take, and may include: Root cause of the fault: Temporary malfunction of the vehicle's display driver; Solution: Restart the display service with one click; The steps to implement the solution are as follows: 1. Click on vehicle infotainment settings - system - display service; 2. Click "one-click restart"; 3. Wait 10 seconds for the screen to return to normal.

[0051] After the car owner performed the operation, the vehicle's infotainment system automatically detected that the screen had returned to normal. The car owner then checked the app and confirmed that the problem had been resolved. Data Platform 3 recorded the solution as successful, updating the success rate to 96.5%, and pushed this result to the vehicle manufacturer's operations and maintenance platform, confirming the solution's effectiveness and indicating that no update was needed.

[0052] Taking the solution for 4S store repair personnel as an example, when a C brand D model comes to the store for repair, after the repair personnel scan the vehicle's VIN code, the data platform 3 obtains a list of root causes of the fault: poor contact of the display cable (confidence level 0.93), which is a hardware fault that requires professional operation.

[0053] The data platform matches the corresponding professional repair solutions from the historical solution database, which include: the root cause of the fault, professional repair steps, required tools, precautions, and after-sales reporting requirements.

[0054] The solution is delivered via maintenance terminal work orders and WeChat push notifications, containing professional and standardized procedures, mainly including: Root cause of the fault: Poor contact between the vehicle's central control screen cable and the motherboard interface; Solution: Re-plug and reinforce the display cable; Operating steps: 1. Disconnect the power and remove the outer frame of the central control screen; 2. Disconnect the display cable and clean the oxide layer on the interface; 3. Reinsert the cable and tighten the fixing clips; 4. Power on and test that the screen display is normal; 5. Reset the outer frame and clean up the site.

[0055] After the maintenance personnel complete the operation, they enter the repair conclusion as successful in the terminal device and upload test photos. Data platform 3 records that the solution was successfully repaired this time, maintaining a cumulative success rate of 98.2%, and pushes the result to the vehicle manufacturer's operation and maintenance platform; if three consecutive repairs are recorded as failures, the operation and maintenance platform triggers the solution update process, adding a backup step of replacing the wiring harness.

[0056] This approach uses multi-dimensional charts to visually present the distribution of fault types and root causes, helping maintenance personnel quickly identify high-incidence issues, high-risk vehicle models, and time periods, enabling a shift from reactive response to proactive prevention. Based on fault type, intelligent matching of solutions is provided, with precise recommendations to different stakeholders such as vehicle owners and repair personnel, saving time spent consulting manuals and relying on experience-based judgment, thus shortening the fault repair cycle. Furthermore, the system records solution repair conclusions and success rates, driving iterative updates to the historical solution library and fault diagnosis model, forming a closed loop of diagnosis-repair-reverse-optimization, continuously improving the overall system capabilities.

[0057] Based on the same inventive concept as the foregoing embodiments, the present invention also provides a fault location method for an in-vehicle infotainment system, which is applied in a cloud server, such as... Figure 2 As shown, the method includes the following steps: S210, acquire fault data of the in-vehicle infotainment system, and determine feature vectors for each fault feature based on the fault data; the feature vectors contain the correlation value of each fault feature to each type of baseline fault.

[0058] As mentioned above, after the vehicle collects fault data of the in-vehicle infotainment system from multiple dimensions, it will transmit the fault data to the cloud server through the MQTT / HTTP dual protocol, so that the cloud server can obtain the fault data of the in-vehicle infotainment system.

[0059] Since the fault data is encrypted, it needs to be decrypted after acquisition to obtain multi-dimensional fault data collected from the vehicle and corresponding basic fault identifiers. These basic fault identifiers include: vehicle information, fault occurrence timestamp, vehicle infotainment system version, and hardware information. Then, the faulty data is cleaned. Data cleaning mainly includes processing invalid data, data standardization, and filtering low-value data.

[0060] When dealing with invalid data, missing values ​​caused by anomalies during the collection process can be supplemented by using the mean of historical normal data in the same dimension or by interpolation. When standardizing data, keywords can be extracted and encoded for text data. When filtering low-value data, normal data or other irrelevant abnormal data that are not in the fault occurrence range can be removed, and only data that is strongly correlated with the fault can be retained, thereby making the subsequent fault diagnosis model more accurate.

[0061] Then, based on the cleaned fault data, feature vectors for each fault feature are determined; the feature vectors contain the correlation value of each fault feature to each baseline fault type.

[0062] Since this invention primarily focuses on extracting fault features corresponding to fault types such as black screen, screen distortion, lag, and system crash, in one embodiment, feature vectors for each fault feature are determined based on fault data, including: Extract fault features associated with each baseline fault type from the fault data; baseline fault types include black screen, distorted screen, lag, and system crash. For each fault feature, the correlation value between the fault feature and each benchmark fault type is determined based on the historical fault sample library, and the feature vector of the fault feature is determined based on the correlation value between the fault feature and each benchmark fault type.

[0063] In one implementation, determining the association value of the fault features with each baseline fault type based on a historical fault sample library includes: The correlation value of the fault characteristics with each baseline fault type is determined according to formula (1): (1) In formula (1), For the first i Fault characteristics for the first t The associated value of the baseline fault type, The first in the historical fault sample library i The fault characteristics are accompanied by the first t The number of samples that occurred for the baseline fault type. The first in the historical fault sample library t The total number of samples that occurred for the baseline fault type. t =1, 2, 3, 4, For the first i The anomaly coefficient of a fault characteristic. For example, when t =1 indicates a black screen; when t =2 indicates a screen flickering fault; when t =2 indicates a stutter; when t =4 indicates a system crash.

[0064] Specifically, quantifiable fault features strongly correlated with the "blackened card" can be extracted from the cleaned fault data, and then the feature vector of each quantifiable fault feature can be determined.

[0065] For example, when the baseline fault type is black screen, quantifiable fault characteristics strongly associated with black screen may include: screen backlight voltage: backlight voltage = 0V, lasting ≥1s. Display controller register status: Register value = 0x00 (no video signal output). Display screen power supply status: The display screen power supply circuit is disconnected or the voltage is abnormal. Graphics rendering process status: The rendering process has terminated abnormally or is a zombie process. System log keywords: display driver crash, HDMI display exception, etc.

[0066] When the baseline fault type is screen flickering, the quantifiable fault characteristics strongly correlated with screen flickering can include: Display pixel anomaly ratio: Abnormal pixel ratio ≥5%, lasting ≥3 frames; GPU load rate: GPU load ≥95%, duration ≥500ms; Frame buffer data checksum: The checksum deviates from the baseline value by ≥20%; Render queue blocking: The length of the render queue exceeds the threshold, causing screen tearing.

[0067] When the baseline fault type is stuttering, quantifiable fault characteristics strongly correlated with stuttering can include: Central control screen frame rate: Frame rate <10fps, lasting ≥2s; CPU load rate: Single core load ≥90%, lasting ≥1 second; Memory usage: ≥95%; Application response time: Application thread response timeout ≥ 500ms.

[0068] When the baseline fault type is a system crash, quantifiable fault characteristics strongly associated with a system crash may include: System power supply status: power supply voltage fluctuation ≥ ±10%, lasting ≥ 1 second; Equipment core temperature: Overheat protection is triggered when the core temperature is ≥105℃; Kernel panic log: The kernel log contains keywords such as "Kernel Panic".

[0069] For each fault feature (quantifiable fault feature), the correlation value of the fault feature to each benchmark fault type is determined using formula (1) based on the historical sample library. The feature vector of the fault feature is obtained by combining the correlation values ​​of the fault feature to each benchmark fault type.

[0070] For example, suppose the IVI system malfunctions and collects multiple fault data points. The first fault characteristic is a memory occupancy rate of 96% (…). i =1), then use formula (1) to calculate the correlation value between 96% memory usage and black screen. The correlation between 96% memory usage and screen flickering. The correlation between memory usage of 96% and lag. The correlation between 96% memory usage and system crashes Therefore, the feature vector of the first fault feature is: .

[0071] by For example, suppose there were a total of 1000 black screen failures in the historical fault sample database ( Of these 1000 black screen failures, 600 also exhibited an anomaly of memory usage exceeding 96% (i.e., When memory usage exceeds 96%, the anomaly coefficient is 1.2. The same method can be used to calculate... , and The specific values ​​of will not be described in detail here.

[0072] Each correlation value ranges from [0,1]. The closer the correlation value is to 1, the higher the correlation between the feature and the fault. The closer it is to 0, the lower the correlation.

[0073] For example, the feature vector of the first fault feature is: This indicates that a memory usage rate of 96% is most strongly correlated with the type of lag or stuttering.

[0074] To eliminate redundant and low-value feature vectors, making subsequent fault diagnosis models more efficient and accurate, the method further includes, after determining the feature vectors of various fault features based on the fault data: For any feature vector, determine whether all associated values ​​in the feature vector are less than a preset associated threshold. If so, delete the feature vector. If not, then retain the feature vector; where the association threshold can be 0.3.

[0075] After determining the feature vectors of each fault characteristic based on the fault data, the method also includes: For any two feature vectors, if the difference in their correlation values ​​for the same baseline type is less than a preset threshold, then the two feature vectors are considered very similar, and one of them is deleted. The threshold can be 0.05.

[0076] For example, the feature vector of fault feature A is [0.72, 0.05, 0.85, 0.60], and the feature vector of fault feature B is [0.71, 0.06, 0.84, 0.59]. The difference in the correlation between the two for the four fault types is less than 0.05, so one of the feature vectors is deleted.

[0077] This process eliminates low-value features irrelevant to the "black card" problem, avoiding interference with diagnosis; and reduces the input dimensions of subsequent fault diagnosis models, improving computational efficiency and generalization ability.

[0078] S211, using a pre-trained fault diagnosis model to diagnose the fault type of the feature vectors of the aforementioned fault features, thereby obtaining the target fault type.

[0079] Then, the pre-trained fault diagnosis model is used to diagnose the fault type by analyzing the feature vectors of each fault feature, resulting in the target fault type, including: Obtain the fault correlation value sub-model from the fault diagnosis model. The fault correlation value sub-model is... ; The fault association value sub-model determines the first t Fault association values ​​of the baseline fault type , t =1, 2, 3, 4; The fault association values ​​of all baseline types are sorted from largest to smallest to obtain a fault association value sequence; If the maximum fault correlation value is greater than the preset first threshold and the difference between the maximum fault correlation value and the second largest fault correlation value is greater than the preset difference threshold, then the baseline fault type corresponding to the maximum fault correlation value is determined as the target fault type. If the maximum fault correlation value is less than or equal to a preset first threshold, or the difference between the maximum fault correlation value and the second largest fault correlation value is less than or equal to a preset difference threshold, then the baseline fault type corresponding to the maximum fault correlation value and the baseline fault type corresponding to the second largest fault correlation value are determined as the target fault types; where, For the first i Fault characteristics for the first t The associated value of the baseline fault type, For the first i Fault characteristics for the first t Weights of the baseline fault types, For the first t Reporting correction coefficients for baseline fault types.

[0080] For example, suppose the fault correlation value for a black screen is... =0.65, the fault correlation value for screen flickering is... =0.48, the fault correlation value for stuttering is =0.05, the fault correlation value for the system crash is... =0.02.

[0081] The four fault correlation values ​​are sorted in descending order, with the largest fault correlation value being [value missing]. =0.65, the second largest fault correlation value =0.48. If the first threshold is 0.60 and the preset difference threshold is 0.15, since... =0.65>0.6, - If 0.17 > 0.15, then the black screen is identified as the target fault type.

[0082] like =0.58, =0.55, =0.05, If the value is 0.02, then black screen and distorted screen are identified as the target fault types.

[0083] In one implementation, after obtaining the target fault type, the method further includes: Obtain the first confidence sub-model in the fault diagnosis model. The first confidence sub-model is... ; Determine the first confidence sub-model First confidence level corresponding to the target fault type The first confidence level is used to assess the credibility of the target fault type; where, For the first The fault association value corresponding to the target fault type. For the first t Fault association values ​​of the baseline fault type The .

[0084] Continuing with the example above, if the black screen is confirmed to be the target fault type, =1, then .

[0085] After diagnosis, the system outputs the target fault type, the confidence level of the target fault type, and a list of key fault characteristics corresponding to the target fault type.

[0086] The list of key failure features is obtained by filtering quantifiable failure features that are strongly correlated with the target failure type. For each quantifiable failure feature, the following is determined: Value, will Sort the values ​​from largest to smallest and select the top M. The quantifiable fault characteristics corresponding to the values ​​are used as a list of key fault characteristics. Here, M can be 10.

[0087] S212, based on the target fault type and the pre-constructed three-dimensional correlation matrix of fault root causes, root cause localization is performed to obtain the target fault root cause corresponding to the target fault type.

[0088] Once the target fault type is determined, root cause localization is performed based on the target fault type and a pre-constructed three-dimensional correlation matrix of fault root causes to obtain the target fault root causes corresponding to the target fault type, including: Based on the target fault type, the corresponding set of candidate fault root causes is selected from the three-dimensional correlation matrix of fault root causes. Determine the corresponding second confidence level for each candidate root cause in the candidate root cause set; If the second confidence level is greater than the preset second threshold, then the candidate fault root cause is determined as the target fault root cause corresponding to the target fault type.

[0089] In one implementation, determining a corresponding second confidence level for each candidate root cause in the candidate root cause set includes: The second confidence level of each candidate root cause of failure is determined using the second confidence sub-model. ;in, The second confidence sub-model is The second confidence level of each candidate root cause of failure is determined using the second confidence sub-model. ;in, To meet the number of fault characteristics that constitute the root cause of the fault, This represents the total number of original fault features corresponding to the root cause of the fault. The deviation correction coefficient is used to match the original fault characteristics of the root cause of the fault. For the first t Fault association values ​​for the baseline fault type.

[0090] Specifically, the three-dimensional correlation matrix of fault root causes stores the correspondence between fault type, fault root cause, and fault feature. For example, a freeze (fault type) _ application freeze (fault root cause) _ [single application CPU usage > 80% for 5 seconds; application process unresponsive code issued more than 2 times] (fault feature). Therefore, a set of candidate fault root causes can be matched from the three-dimensional correlation matrix based on the fault type.

[0091] The second confidence level sub-model is then used to calculate the corresponding second confidence level for each candidate root cause of the failure. Candidate root causes with a second confidence level greater than a second threshold (which can be 0.8) are retained. If all second confidence levels are less than the second threshold, the second confidence levels are sorted from highest to lowest, and the top three candidate root causes with the highest second confidence levels are selected as the target root causes. Finally, the target root causes and their corresponding second confidence levels are output to facilitate the display of failure distribution on the data platform.

[0092] Taking application freezing as an example of a candidate root cause of a lag issue, the key fault list corresponding to a lag issue includes: 1. Navigation application CPU usage of 92% for 6 seconds; 2. ANR detected in the system log: incom.autonavi.amapauto; 3. Memory usage of 95%; 4. Central control frame rate of 8fps. Of these four key fault characteristics, two match application freezing (the first and second). =2.

[0093] The original fault characteristics of application freezing / lag are defined as follows: 1. A single application's CPU usage exceeds 80% for 5 seconds; 2. The application process displays unresponsive code more than twice; 3. The system log contains the ANR keyword. Therefore... =3.

[0094] Assuming the correlation value for a stuttering fault is 0.85 and the bias correction factor is 1, then the second confidence level for applying the stuttering as a candidate root cause is: .

[0095] If the second confidence level of other candidate root causes under the lag fault is greater than 0.8, then the other candidate root causes are determined as the target root cause.

[0096] It should be noted that, to ensure the accuracy of the fault diagnosis model, before using the pre-trained fault diagnosis model to diagnose the fault type based on the feature vectors of various fault features, the method also includes: Construct the first training set, which includes the feature vectors of each historical fault feature of the in-vehicle infotainment system; each historical fault feature has a corresponding fault label. The pre-built fault diagnosis model is iteratively trained using the first training set until the model converges; The fault diagnosis model includes: a fault association value sub-model and a first confidence sub-model; the root cause localization unit contains a second confidence sub-model. When training the fault diagnosis model, the training primarily focuses on the second confidence sub-model within the fault association value sub-model. i Fault characteristics for the first t The weights of the baseline fault types, the first t The reporting correction coefficients for the baseline fault type are optimized until the model converges.

[0097] The training process for a fault diagnosis model can be as follows: Construct a system that includes fault types, true fault root causes, and multi-dimensional fault features; the first training set contains the labeled samples mentioned above. initialization , : In determining Initial values ​​can be determined based on the fault types reported during data acquisition and historical accuracy.

[0098] Taking a black screen fault as an example, during data collection, a black screen fault is reported when the detected backlight voltage exceeds a threshold. By analyzing historical fault samples, the reporting accuracy of black screen faults is statistically calculated. Assuming the accuracy rate of reporting black screen faults due to backlight abnormalities is 90%, then the accuracy rate for reporting black screen faults is... The initial value is 0.9.

[0099] For each fault sample in the labeled sample library, perform forward calculations: use the fault association value sub-model and the first confidence sub-model in sequence to calculate, and output the fault type with the highest first confidence as the prediction result.

[0100] The predicted fault types are compared with the actual fault types to calculate the root cause localization error (such as cross-entropy loss), and the fault type detection rate and false alarm rate are also calculated.

[0101] Based on the positioning error, adjustments are made using algorithms such as gradient descent. This allows for higher weighting of strongly correlated features; Adjustment based on fault type detection rate and false alarm rate The bias caused by the imbalance of calibration samples; The above process is repeated iteratively. When the iteration ends under the given condition, the iteration stops and the trained fault diagnosis model is output.

[0102] Similarly, before determining the second confidence level for each candidate root cause of failure using the second confidence sub-model, the second confidence sub-model also needs to be trained. During the training of the second confidence sub-model... This mainly refers to the deviation correction coefficient for the original fault characteristics that conform to the stated root cause of the fault. The optimization and training process is as follows: Construct a second training set, which contains historical fault data samples with true root cause labels; each fault sample is labeled with a fault type label and a true fault root cause label. initialization :Sure When setting the initial value, the following is implemented: if the fault characteristic deviates from the threshold by less than 10%, the coefficient is set to 0.90; if the fault characteristic deviates from the threshold by 10-30%, the coefficient is set to 1.20.

[0103] For example, taking the application freeze characteristic of a lag / stuttering problem (a single application's CPU usage > 80% for 5 seconds) as an example, the threshold is 80%; if the actual CPU usage in the fault sample is 92%, then the deviation is (92% - 80%) / 80% = 15%. The initial value should be 1.2.

[0104] For each fault sample, perform forward computation: use the second confidence sub-model to calculate and output the fault root cause with the highest second confidence as the prediction result.

[0105] The predicted root causes of failures are compared with the actual root causes to calculate the failure repair success rate. Adjustments are then made based on the failure repair success rate. This makes the impact of the second confidence level more consistent with the actual repair results.

[0106] The above process is repeated iteratively. When the iteration ends under the given condition, the iteration stops and the trained second confidence sub-model is output.

[0107] The fault diagnosis model and the second confidence sub-model can be optimized periodically. For example, new fault samples can be added to the training set each month, and the fault diagnosis model can be trained and optimized using the new training set. , and .

[0108] The three-dimensional correlation matrix of fault root causes can also be optimized periodically. Specifically, based on the fault root cause repair records, the repair success rate corresponding to each original fault feature can be statistically analyzed, invalid fault features with low success rates can be eliminated, and the correlation between fault type, fault root cause and fault feature in the three-dimensional correlation matrix can be corrected and updated to improve diagnostic accuracy.

[0109] Similarly, fault characteristic thresholds (e.g., memory usage ≥95% calibrated to memory usage ≥90%) and confidence levels (first threshold, second threshold, and difference threshold) can be recalculated and calibrated periodically based on fault characteristics and maintenance results to improve the accuracy of fault diagnosis reasoning and root cause localization.

[0110] This invention determines the feature vectors of various fault characteristics based on fault data. This means that it no longer relies on the experience or vague descriptions of maintenance personnel, but instead represents the importance of each fault characteristic to each fault type with quantifiable values, eliminating the bias of human judgment and making the input parameters for fault diagnosis more objective. When using the fault diagnosis model for fault type diagnosis, because the fault diagnosis model can learn the complex rules between feature combinations and fault rationality, model-based diagnosis can more accurately identify the fault type. Finally, because the three-dimensional correlation matrix strongly correlates the fault type with the possible root causes and the corresponding original fault characteristics, the most likely fault root cause can be accurately located when locating the fault root cause through the three-dimensional correlation matrix. In addition, compared with the fault location method in the prior art that requires manual checking of the vehicle log one by one, this invention can complete most of the fault location work by using the fault diagnosis model and the three-dimensional correlation matrix of fault root causes, which significantly shortens the fault diagnosis time, improves the fault diagnosis efficiency, and thus ensures driving safety.

[0111] Furthermore, this invention can increase the coverage of strongly correlated fault features to 100% when collecting fault data. Dual-protocol transmission further improves the data transmission success rate, and end-to-end encryption avoids the leakage of user privacy data. The reasoning logic of fault diagnosis avoids the impact of invalid data on the diagnosis of "black card" problems, improving the diagnostic accuracy. The root cause location accuracy can be improved from the module level to the specific component level. The visualized data platform provides a comprehensive display and prediction capability for "black card" problems, and targeted solution push saves time from repeatedly investigating such problems. Cloud configuration and continuous model iteration support improve the adaptability to new vehicle models, meet the long-term operation and maintenance needs of car companies, and have strong scalability.

[0112] Based on the same inventive concept as in the foregoing embodiments, this embodiment also provides a fault location cloud server for an in-vehicle infotainment system, such as... Figure 3 As shown, the cloud server includes: Data preprocessing unit 21 is used to acquire fault data of the in-vehicle infotainment system and determine feature vectors of various fault features based on the fault data; the feature vectors contain the correlation value of each fault feature to each type of baseline fault. The fault diagnosis unit 22 is used to perform fault type diagnosis on the feature vectors of the various fault features using a pre-trained fault diagnosis model to obtain the target fault type. The root cause localization unit 23 is used to perform root cause localization based on the target fault type and the pre-constructed three-dimensional correlation matrix of fault root causes, so as to obtain the target fault root cause corresponding to the target fault type.

[0113] Since the apparatus described in the embodiments of this invention is used for implementing the fault location method of the in-vehicle infotainment system of the embodiments of this invention, those skilled in the art can understand the specific structure and variations of the apparatus based on the method described in the embodiments of this invention, and therefore will not be described in detail here. All apparatuses used in the methods of the embodiments of this invention fall within the scope of protection of this invention.

[0114] Through one or more embodiments of the present invention, the present invention has the following beneficial effects or advantages: This invention provides a fault location method, cloud server, and system for an in-vehicle infotainment system. The method includes: acquiring fault data of the in-vehicle infotainment system; determining feature vectors for various fault features based on the fault data; the feature vectors containing the correlation value of each fault feature to each baseline fault type; using a pre-trained fault diagnosis model to diagnose the fault type of the feature vectors of the various fault features to obtain a target fault type; and performing root cause localization based on the target fault type and a pre-constructed three-dimensional correlation matrix of fault root causes to obtain the target fault root cause corresponding to the target fault type. Thus, determining the feature vectors of various fault features based on fault data means that it no longer relies on the experience or vague descriptions of maintenance personnel, but instead uses quantifiable data to represent the importance of each fault feature to each fault type. The values ​​are displayed to eliminate the bias of human judgment and make the input parameters for fault diagnosis more objective. When using the fault diagnosis model for fault type diagnosis, the model can learn the complex rules between feature combinations and fault rationality, thus the model-based diagnosis can more accurately identify the fault type. Finally, since the three-dimensional correlation matrix strongly correlates the fault type with the possible root causes and the corresponding original fault features, the most likely fault root cause can be accurately located when locating the fault root cause through the three-dimensional correlation matrix. In addition, compared with the fault location method in the prior art that requires manual checking of the vehicle logs one by one, this invention can complete most of the fault location work by using the fault diagnosis model and the three-dimensional correlation matrix of fault root causes, which greatly shortens the fault diagnosis time, improves the fault diagnosis efficiency, and thus ensures driving safety.

[0115] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0116] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A fault location method for an in-vehicle infotainment system, characterized in that, The method includes: Fault data of the in-vehicle infotainment system is acquired, and feature vectors of various fault characteristics are determined based on the fault data; the feature vectors contain the correlation value of each fault characteristic to each type of baseline fault. The pre-trained fault diagnosis model is used to diagnose the fault type by analyzing the feature vectors of the aforementioned fault features, thereby obtaining the target fault type. Based on the target fault type and the pre-constructed three-dimensional correlation matrix of fault root causes, the root cause of the target fault is located to obtain the target fault root cause corresponding to the target fault type.

2. The method as described in claim 1, characterized in that, The step of determining the feature vectors of each fault feature based on the fault data includes: Extract fault features associated with each baseline fault type from the fault data; the baseline fault types include black screen, distorted screen, lag, and system crash. For each fault feature, the correlation value of the fault feature with each benchmark fault type is determined based on the historical fault sample library, and the feature vector of the fault feature is determined based on the correlation value of the fault feature with each benchmark fault type.

3. The method as described in claim 2, characterized in that, The process of determining the association value of the fault features with each baseline fault type based on the historical fault sample library includes: According to the formula Determine the association value of the fault characteristics with each baseline fault type; The For the first i Fault characteristics for the first t The associated value of the class baseline fault type, the The first in the historical fault sample library i The fault characteristics are accompanied by the first t The number of samples that occurred for the benchmark fault type, the The first in the historical fault sample library t The total number of samples that occurred for the baseline fault type. t =1 indicates a black screen. t A value of 2 indicates a screen flickering error. t =3 indicates lag, when t =4 indicates a system crash. For the first i The anomaly coefficient of the fault characteristics.

4. The method as described in claim 1, characterized in that, Before using the pre-trained fault diagnosis model to diagnose the fault type from the feature vectors of the various fault features, the method further includes: Construct a first training set, which includes feature vectors of each historical fault feature of the in-vehicle infotainment system; each historical fault feature has a corresponding fault label. The pre-built fault diagnosis model is iteratively trained using the first training set until the model converges.

5. The method as described in claim 1, characterized in that, The step of using a pre-trained fault diagnosis model to diagnose the fault type from the feature vectors of the various fault features includes: Obtain the fault correlation value sub-model in the fault diagnosis model, wherein the fault correlation value sub-model is: ; The fault association value sub-model is used to determine the first... t Fault association values ​​of the baseline fault type ; The fault association values ​​of all baseline types are sorted from largest to smallest to obtain a fault association value sequence; If the maximum fault association value is greater than a preset first threshold and the difference between the maximum fault association value and the second largest fault association value is greater than a preset difference threshold, then the baseline fault type corresponding to the maximum fault association value is determined as the target fault type. If the maximum fault correlation value is less than or equal to a preset first threshold, or the difference between the maximum fault correlation value and the second largest fault correlation value is less than or equal to a preset difference threshold, then the baseline fault type corresponding to the maximum fault correlation value and the baseline fault type corresponding to the second largest fault correlation value are determined as the target fault type; wherein, The For the first i Fault characteristics for the first t The associated value of the class baseline fault type, the For the first i Fault characteristics for the first t The weights of the class-based fault types, the For the first t Reporting correction factor for baseline fault types, t =1 indicates a black screen. t A value of 2 indicates a screen flickering error. t =3 indicates lag, when t =4 indicates a system crash.

6. The method as described in claim 5, characterized in that, After obtaining the target fault type, the method further includes: Obtain the first confidence sub-model in the fault diagnosis model, wherein the first confidence sub-model is... ; Determine the first confidence sub-model using the first confidence sub-model First confidence level corresponding to the target fault type The first confidence level is used to assess the credibility of the target fault type; wherein, For the first The fault association value corresponding to the target fault type, the For the first t Fault association values ​​of the baseline fault type The .

7. The method as described in claim 1, characterized in that, The three-dimensional correlation matrix of fault root causes stores the correspondence between fault type, fault root cause, and fault feature. The root cause localization based on the target fault type and the pre-constructed three-dimensional correlation matrix of fault root causes, to obtain the target fault root cause corresponding to the target fault type, includes: Based on the target fault type, a set of corresponding candidate fault root causes is selected from the three-dimensional correlation matrix of fault root causes. Determine a corresponding second confidence level for each candidate fault root cause in the candidate fault root cause set; If the second confidence level is greater than the preset second threshold, then the candidate fault root cause is determined as the target fault root cause corresponding to the target fault type.

8. The method as described in claim 7, characterized in that, Determining a corresponding second confidence level for each candidate root cause in the candidate root cause set includes: The second confidence level of each candidate root cause of failure is determined using the second confidence sub-model. ;in, The second confidence sub-model is The In order to meet the number of fault characteristics of the root cause of the fault, the The total number of original fault features corresponding to the root cause of the fault, the The deviation correction coefficient is used to conform to the original fault characteristics of the root cause of the fault. For the first t Fault association values ​​for the baseline fault type, t =1 indicates a black screen. t A value of 2 indicates a screen flickering error. t =3 indicates lag, when t =4 indicates a system crash.

9. A cloud server for fault location in an in-vehicle infotainment system, characterized in that, The cloud server includes: The data preprocessing unit is used to acquire fault data of the in-vehicle infotainment system and determine the feature vector of each fault feature based on the fault data; the feature vector contains the correlation value of each fault feature to each type of baseline fault. The fault diagnosis unit is used to perform fault type diagnosis on the feature vectors of the various fault features using a pre-trained fault diagnosis model to obtain the target fault type. The root cause localization unit is used to perform root cause localization based on the target fault type and a pre-constructed three-dimensional correlation matrix of fault root causes, so as to obtain the target fault root cause corresponding to the target fault type.

10. A fault location system for an in-vehicle infotainment system, characterized in that, The system includes: vehicles, cloud servers, and a data platform; The vehicle is used to collect fault data of the in-vehicle infotainment system using data acquisition sensors and transmit the fault data to the cloud server. The cloud server is used to determine feature vectors for each fault feature based on the fault data; the feature vectors contain the correlation value of each fault feature to each baseline fault type; and the fault diagnosis model is used to diagnose the fault type by analyzing the feature vectors of each fault feature to obtain the target fault type. Based on the target fault type and the pre-constructed three-dimensional correlation matrix of fault root causes, the root cause is located to obtain the target fault root cause corresponding to the target fault type. The data platform is used to display the fault type and target root cause of the fault based on a preset display format.

Images