Vehicle fault automatic determination method and device and computer readable storage medium

Abstract

The embodiment of the invention relates to the technical field of artificial intelligence, and discloses a vehicle fault automatic determination method and device and a computer readable storage medium, and the method comprises the steps: obtaining working parameters of a vehicle, and generating working parameter images of parts of the vehicle; based on the working parameter image, determining abnormal image areas in the working parameter image and an abnormal category corresponding to each abnormal image area; determining a corresponding standard image according to the abnormal category; and according to the abnormal image area and a standard image. By applying the technical scheme of the invention, the fault judgment accuracy can be improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of artificial intelligence technology, and specifically to a method and device for automatically determining vehicle faults, and a computer-readable storage medium. Background Art

[0002] With the development of vehicle intelligence, vehicle fault diagnosis has gradually evolved from relying on manual experience and judgment to automatic fault determination by the vehicle.

[0003] At present, electronic devices can be trained to obtain fault diagnosis models based on operating data under fault conditions. After obtaining the operating data of the vehicle, the electronic devices can input the operating data into the fault diagnosis model to determine the fault type of the vehicle.

[0004] However, the above method has the problem of low accuracy in determining the fault category. Summary of the invention

[0005] In view of the above problems, an embodiment of the present invention provides a method, device and computer-readable storage medium for automatically determining vehicle faults, which are used to solve the problem of low accuracy in determining fault categories in the prior art.

[0006] According to one aspect of an embodiment of the present invention, a method for automatically determining a vehicle fault is provided, the method comprising:

[0007] Acquire an operating parameter image of a component in a vehicle, wherein the operating parameter image includes a curve graph of the operating parameter corresponding to the component;

[0008] Determining an abnormal image region in the working parameter image, wherein the abnormal image region is an image region in the working parameter image where the working parameters are abnormal;

[0009] Determining a standard image corresponding to the abnormal image area; wherein the standard image includes a curve graph of the operating parameters corresponding to the abnormal image area of ​​the component of the vehicle during normal operation;

[0010] The damage degree of the failure of the component is determined based on the abnormal image area and the standard image.

[0011] According to another aspect of an embodiment of the present invention, there is provided an automatic determination device for a vehicle fault, comprising:

[0012] An acquisition module, used for acquiring an operating parameter image of a component in a vehicle, wherein the operating parameter image includes a curve graph of an operating parameter corresponding to the component;

[0013] A processing module is used to determine an abnormal image area in the working parameter image, wherein the abnormal image area is an image area in the working parameter image where the working parameters are abnormal; determine a standard image corresponding to the abnormal image area; wherein the standard image includes a curve chart of the working parameters corresponding to the abnormal image area of ​​the component of the vehicle during normal operation; and determine the damage degree of the fault of the component based on the abnormal image area and the standard image.

[0014] According to another aspect of an embodiment of the present invention, there is provided an electronic device, including:

[0015] A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0016] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations such as the above-mentioned method for automatically determining vehicle faults.

[0017] According to another aspect of the embodiments of the present invention, a computer-readable storage medium is provided.

[0018] The storage medium stores at least one executable instruction, and when the executable instruction is executed on the electronic device, the electronic device executes the operation of the automatic determination method of the vehicle fault as described above.

[0019] The embodiment of the present invention obtains the working parameters of the vehicle and generates a working parameter image of the vehicle components; based on the working parameter image, determines the abnormal image area in the working parameter image; obtains the standard image corresponding to the abnormal image area; and determines the damage degree of the vehicle component according to the abnormal image area and the standard image, thereby achieving the effect of improving the accuracy of fault judgment.

[0020] The above description is only an overview of the technical solution of the embodiment of the present invention. In order to more clearly understand the technical means of the embodiment of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the embodiment of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings are only used to illustrate the embodiments and are not to be considered as limiting the present invention. In addition, the same reference symbols are used to represent the same components throughout the accompanying drawings. In the accompanying drawings:

[0022] Figure 1 A schematic flow chart showing a first embodiment of a method for automatically determining a vehicle fault provided by the present invention;

[0023] Figure 2A schematic flow chart showing a second embodiment of the method for automatically determining a vehicle fault provided by the present invention;

[0024] Figure 3 A schematic structural diagram showing a first embodiment of a device for automatically determining a vehicle fault provided by the present invention;

[0025] Figure 4 A schematic structural diagram of an embodiment of an electronic device provided by the present invention is shown. DETAILED DESCRIPTION

[0026] Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0027] Figure 1 FIG. 1 is a flow chart showing a first embodiment of a method for automatically determining a vehicle fault according to the present invention, the method being executed by an electronic device. Figure 1 As shown, the method comprises the following steps:

[0028] Step 110: Acquire an operating parameter image of a component in the vehicle, wherein the operating parameter image includes a curve graph of the operating parameters corresponding to the component.

[0029] The electronic device may obtain the operating parameters of the components in the vehicle through sensors in the vehicle. The electronic device may generate a curve graph according to the operating parameters. The curve graph constitutes an operating parameter image of the component. Optionally, each component may correspond to an operating parameter graph. Optionally, the component may be a basic part of the vehicle such as an engine, a chassis, a body, and an electrical device.

[0030] In one implementation, the electronic device can continuously obtain the operating parameters through sensors on the vehicle. Optionally, the electronic device can realize real-time monitoring of the operating parameters based on large-scale real-time collection of vehicle-mounted CAN bus and other sensor data. Optionally, the electronic device can collect the operating parameters of various vehicle components in real time through vehicle-mounted communication interfaces such as advanced sensor networks and OBD-II interfaces installed inside the vehicle. Optionally, the operating parameters include but are not limited to important indicators such as engine operating status, fuel consumption, brake system performance, battery voltage, coolant temperature, etc.

[0031] Optionally, since the internal communication of the vehicle is mainly realized through the Controller Area Network (CAN) bus, the electronic equipment can make full use of the CAN bus technology to connect and monitor the communication information between the electronic control units (ECU) of the vehicle in real time. Optionally, the working parameters obtained through the CAN bus include but are not limited to engine operating parameters, brake system status, battery management system data, chassis control system signals, vehicle body electronic equipment information, and original readings of various sensors, etc., which are all-round dynamic and static information of the vehicle. Optionally, the engine operating parameters may include speed, load, fault codes, etc.

[0032] Optionally, the electronic device can also integrate the vehicle telematics processor T-Box as a data bridge, and use the high-speed wireless communication capability of the T-Box to encrypt the working parameters collected through the CAN bus and upload them to the cloud server in real time through 4G / 5G, Wi-Fi or cellular network. Optionally, the electronic device can be a cloud server.

[0033] The method of obtaining the working parameters collected in real time through the CAN bus in the vehicle by the electronic device through the T-Box not only ensures the safety of the working parameters, but also ensures the real-time performance, so that the vehicle can efficiently transmit a large amount of real-time data to the remote data center at any time and anywhere. Optionally, the remote data center is the area in the electronic device that stores the working parameters.

[0034] In one implementation, after obtaining the working parameters uploaded by the vehicle in real time, the electronic device can also electronically implement big data-driven analysis, and conduct in-depth mining and correlation analysis on various parameters during the operation of the vehicle. Optionally, after the electronic device receives the original CAN data transmitted by the T-Box, it performs preliminary cleaning, screening and format conversion on these massive data to eliminate invalid, repeated or erroneous data. Based on the professional knowledge in the field of automotive engineering, the electronic device can extract key performance indicators (KPIs) from the original data, such as engine operating parameters, electrical system status, driving behavior characteristics, etc. These indicators will provide a basis for subsequent data analysis. Optionally, the electronic device can also further process the data to obtain data such as calculating average fuel efficiency, analyzing the response time of the braking system, and monitoring the trend of battery health changes, providing an accurate basis for subsequent fault prediction and diagnosis.

[0035] In one implementation, the electronic device may be provided with a data platform for efficient storage, cleaning, integration and analysis of massive real-time data. Optionally, the data platform may also use machine learning algorithms and artificial intelligence models to predict potential faults, such as identifying abnormal patterns through learning from historical data, and discovering possible problems at an early stage. Optionally, the data platform may be configured to: continuously track the health status of the vehicle and provide early warning of potential faults by real-time collection of CAN bus data combined with historical model analysis, provide forward-looking predictive maintenance recommendations to vehicle owners, and reduce the risks and maintenance costs caused by sudden failures.

[0036] In one implementation, electronic devices can also use digital twin technology to dynamically create a virtual twin network of the automotive system based on the real-time acquired CAN data waveform to simulate the system behavior under actual working conditions. Electronic devices can input working parameters into the twin network to compare the data in the virtual model with the real environment, and achieve early warning and accurate identification of potential faults or performance degradation. The twin network can reflect the state changes of the real vehicle in real time. Electronic devices can simulate actual working conditions and predict vehicle behavior under different conditions by mapping real vehicle data into the twin model. When the model deviates from the measured data, the potential source of the fault can be quickly located.

[0037] Optionally, the electronic device can map the real-time CAN data stream uploaded from the vehicle to the twin network, so that the virtual vehicle in the twin network is synchronized with the real vehicle in real time. When new sensor data is received, the model parameters will be dynamically adjusted to reflect the actual working conditions and performance of the current vehicle.

[0038] Step 120: Determine an abnormal image region in the working parameter image, wherein the abnormal image region is an image region in the working parameter image where the working parameter is abnormal, and the abnormal image region has at least one abnormal category.

[0039] Among them, after acquiring the working parameter image, the electronic device can input the working parameter image into the deep learning model. The electronic device can identify the normal image area and the abnormal image area in the working parameter image through the deep learning model. And, the deep learning module can also identify at least one abnormal category corresponding to the abnormal image area. Optionally, each abnormal category includes at least a damaged part. Optionally, the damaged part can be a more specific part of the component. Optionally, the abnormal category can also include a degree of wear or a damaged area. Optionally, an abnormal category can correspond to a degree of wear of a damaged part, or a damaged area of ​​a damaged part.

[0040] In one implementation, the electronic device can input the working parameter image into a pre-trained recognition network to obtain at least one abnormal image area in the working parameter image; the recognition network is used to segment the normal image area and the abnormal image area in the working parameter image; the abnormal image area corresponds to the abnormal category.

[0041] Wherein, the deep learning network included in the electronic device may be a pre-trained recognition network. The recognition network is used to segment the normal image area and the abnormal image area in the working parameter image. Optionally, the recognition network can also be used to segment the abnormal image areas corresponding to different abnormal categories in the working parameter image. Optionally, when the working parameter image includes multiple abnormalities, the electronic device can obtain multiple abnormal image areas through the recognition network segmentation. When there is no abnormality in the working parameter image, the electronic device will not be able to obtain the abnormal image area from the working parameter image through the recognition network.

[0042] Optionally, the recognition network can be Mask R-CNN. Optionally, the Mask R-CNN is applied to the segmentation of target instances in the prior art. For example, the Mask R-CNN can segment vehicles in an image. However, in the present application, Mask R-CNN can segment abnormal image areas in a working parameter image. Moreover, in the present application, the Mask R-CNN can segment abnormal image areas corresponding to different abnormal categories. Optionally, the Mask R-CNN can accurately locate each component and identify its visual features such as the degree of wear and damage, label them by category, and form a historical model library with rich semantic information.

[0043] Optionally, the training of the recognition network can utilize an instance segmentation model in deep learning technology, and can be based on image data in a historical data set after refined recognition and classification records. For example, the image data can include abnormal images of specific parts of vehicle components.

[0044] Optionally, this embodiment uses a working parameter image composed of a graph of working parameters for processing, which increases the effect of time series analysis and trend prediction compared to the traditional method. The working parameter image can reflect the trend of the performance evolution of each component over time. Through this recognition network, the electronic device can make accurate predictions of possible failure modes and performance degradation trends in the future. That is, the abnormal category corresponding to the abnormal image area extracted in this step can be an abnormality that may occur in the future.

[0045] In one implementation, the electronic device can also use Mask R-CNN to identify and classify historical data. And combine the historical data domain with the working parameters uploaded by the real-time CAN bus to conduct cross-domain and multi-dimensional correlation analysis. Electronic devices can discover the causal relationship between potential faults and hidden performance bottlenecks by comparing and analyzing data changes in different periods and situations, thereby providing more comprehensive and targeted diagnostic suggestions and preventive maintenance plans. Optionally, the use of Mask R-CNN realizes three-dimensional monitoring and intelligent diagnosis of vehicle health status, which effectively promotes the efficient operation of the intelligent vehicle fault diagnosis service system and the improvement of decision-making support capabilities.

[0046] In one implementation, the electronic device can use Mask R-CNN to identify abnormal image areas in the working parameter image, and for each abnormal image area, generate an accurate pixel-level segmentation Mask. The electronic device can implement segmentation of the abnormal image area based on the Mask. The use of the Mask is essential for accurately identifying vehicle component signals and potential damage areas.

[0047] Optionally, in the Mask R-CNN, classifying each pixel in the working parameter image is a core step. The output of the Mask R-CNN has a dimension of K*m*m, where K represents the total number of preset categories and m*m represents the resolution of the Region of Interest (ROI). Each pixel will be classified into one of the K categories. Optionally, the K categories may include vehicle components such as compressors, fans, batteries, etc.

[0048] Optionally, during the training process of the Mask R-CNN model, in order to ensure that the model can accurately learn and predict the category of each pixel, this embodiment uses the average binary cross entropy loss function as the loss function. This loss function is suitable for binary classification problems and can effectively measure the difference between the model prediction and the actual label, thereby guiding the model training process.

[0049] Optionally, in order to more effectively extract useful information from the working parameter image, this embodiment uses ResNeXt as a feature extractor. ResNeXt is a deep residual network characterized by having multiple independent transformation paths, called "cardinality". In the application, ResNeXt consists of 32 groups of such independent paths, which greatly enhances the diversity of features and the generalization ability of the model.

[0050] Optionally, before inputting the image into Mask R-CNN, the electronic device may also preprocess the working parameter image. The preprocessing may include scaling the original high-resolution image to a 256-bit pixel map to meet the input requirements of the model. In addition, the preprocessing may also include applying a threshold of 0.5 for binarization, converting the image to black and white to highlight key areas in the image, facilitating subsequent segmentation operations by the model.

[0051] Optionally, the Mask R-CNN can realize panoramic segmentation and semantic segmentation of the working parameter image. Specifically, the panoramic segmentation function of the Mask R-CNN can completely segment the normal part and the abnormal part of the image. The Mask R-CNN can also perform semantic segmentation on these segmented abnormal image areas, that is, identify and mark the specific category of each abnormal image area. Optionally, the category is the abnormal category. Optionally, the abnormal category can be distinguished by labeling. For example, the label can be a loss part.

[0052] Optionally, the electronic device can further perform instance segmentation on the abnormal image area to distinguish different individuals in the same category (for example, to distinguish multiple fans). The instance segmentation can accurately select all the damaged parts involved and mark the abnormal instance photos and their specific locations in the damaged image. This is crucial for quickly locating and repairing faults.

[0053] Optionally, the use of Mask R-CNN can not only accurately distinguish normal image areas from abnormal image areas, but also accurately locate and mark the damaged time period. This provides strong support for subsequent fault diagnosis, repair, and prevention of similar faults from happening again. The entire process combines deep learning, image processing, and multiple segmentation technologies to achieve efficient and accurate vehicle fault detection and location.

[0054] In another implementation, when the recognition network is a twin network, the electronic device can input the working parameter image into the pre-trained twin network to simulate and obtain the ideal parameter image of the vehicle in the operating state. The electronic device can compare the working parameter image with the ideal parameter image, and determine the inconsistent area between the working parameter image and the ideal parameter image as the abnormal image area. Among them, the twin network is used to simulate the operating state of the vehicle according to the working parameters of the vehicle. Optionally, the Mask R-CNN network can also be used to construct a twin network.

[0055] Optionally, the electronic device can input the acquired working parameters into a pre-trained twin network. Optionally, the twin network is a model generated according to the vehicle for simulating the operating state of the vehicle. The electronic device can simulate the operation of the vehicle in the twin network according to the working parameters. The twin network can output an ideal parameter image of the vehicle in the operating state. The electronic device can compare the ideal parameter image with the working parameter image. If the ideal parameter image is consistent with the working parameter image, it means that there is no abnormality in the working parameter. If the working parameter image is inconsistent with the ideal parameter image, it means that there is an abnormality in the working parameter. The electronic device can determine the abnormal image area based on the inconsistent working parameter. Optionally, the electronic device can input the control instruction corresponding to the working parameter into the twin network to simulate the vehicle's operating state.

[0056] Optionally, electronic devices can also use digital twin models to perform high-precision system-level simulations to simulate vehicle behavior and reactions under various driving conditions. By simulating potential failure modes, problems that may occur under specific conditions can be predicted and their impact on the overall system can be evaluated.

[0057] Optionally, the twin network can simulate all key subsystems of the car. Optionally, the key subsystems include but are not limited to the power system, electrical system, suspension system, braking system and environmental interaction module. Optionally, the environmental interaction module may include an ADAS advanced driver assistance system.

[0058] In one implementation, the electronic device can also simulate different maintenance strategies or upgrade options in the twin network and obtain the operating status of the twin model to select the best solution. Optionally, the twin model can also be used to optimize component design, improve manufacturing processes, and improve overall vehicle performance.

[0059] Step 130: Determine a standard image corresponding to the abnormal image region; wherein the standard image includes a curve diagram of the working parameters corresponding to the abnormal image region of the vehicle component during normal operation.

[0060] The electronic device may determine a standard image for each abnormal category according to the determined abnormal category. Optionally, the standard image may be a graph of operating parameters generated by collecting operating parameters of components of the vehicle when the abnormal category is determined to exist.

[0061] In one implementation, the electronic device can determine the corresponding standard image based on the preset knowledge graph and the working parameters of the component during normal operation. Optionally, the preset knowledge graph indicates the component, and the association between the abnormality of the component and the working parameters. When the working parameters are related to the absence and the abnormality of the component, the electronic device needs to consider the working parameters when analyzing the abnormality.

[0062] Optionally, the electronic device may store a professional knowledge graph in the automotive field. Optionally, the professional knowledge graph is a preset knowledge graph. Optionally, the electronic device may quickly locate the source of the fault, provide accurate maintenance guidance, and effectively improve the speed and quality of troubleshooting based on the preset knowledge graph through structured knowledge representation and reasoning mechanism.

[0063] Optionally, the preset knowledge graph can be combined with the professional knowledge system in the field of automotive engineering. The preset knowledge graph can include component attributes, relationships, and failure modes. The electronic device can map the data differences obtained during the diagnosis process to the knowledge graph, infer the cause of the failure and the scope of impact in a structured manner, and then provide more accurate maintenance suggestions and preventive maintenance strategies. When the electronic device detects signs of failure, it can use knowledge graph reasoning to find the most likely cause of the failure, and provide accurate maintenance step guidance and preventive maintenance suggestions.

[0064] Optionally, the electronic device can construct a knowledge graph that includes vehicle components, sensor data, failure modes, maintenance methods, and other related knowledge, so as to represent the vehicle parts, component status and functions, dependencies between components, and fault propagation paths in a structured manner. For example, the knowledge graph can include connections generated by relationships such as "component A may cause failure of component B" or "when sensor C reads abnormally, it may mean that there is a problem with component D." The knowledge graph can also link discrete data points to form a networked knowledge system by integrating multi-source information such as vehicle maintenance manuals, historical maintenance records, and expert experience. This enables isolated data to be interconnected, enhancing the system's comprehensive understanding and reasoning capabilities. Such a model helps to derive potential root causes from complex failure phenomena.

[0065] Optionally, based on the determined abnormality category, the electronic device can determine the possible fault of the vehicle according to the knowledge graph. The electronic device can obtain the standard image corresponding to each fault. Optionally, in the diagnostic reasoning process, the electronic device can also obtain other abnormal signals of the vehicle. The electronic device can use the knowledge graph to perform traversal search and logical reasoning on the graph according to the received abnormal signals to locate possible faults. Optionally, the electronic device can also provide possible cause analysis and maintenance suggestions based on the knowledge graph.

[0066] In one implementation, the knowledge graph can have the function of dynamic updating and self-improvement. With the learning of new fault cases and the accumulation of maintenance practices, electronic devices can continuously add new entities, attributes and relationships in the knowledge graph, so that the system has the ability to self-update and evolve, thereby improving the accuracy and efficiency of fault diagnosis. For example, an electronic device can obtain the fault of a component reported by a user, as well as the working parameters corresponding to the fault. If the working parameter does not appear in the knowledge graph and there is no association between the fault and the component, the association between the working parameter and the fault and the component is added to the knowledge graph.

[0067] Optionally, the introduction of this knowledge graph provides powerful background knowledge support and intelligent reasoning methods for the intelligent vehicle fault diagnosis system, greatly enhancing the system's diagnostic accuracy and decision-making effectiveness, and plays a vital role in achieving efficient and accurate vehicle fault diagnosis.

[0068] Step 140: Determine the damage degree of the component failure based on the abnormal image area and the standard image.

[0069] After acquiring the standard image corresponding to each abnormal image area, the electronic device can determine the probability of the vehicle having the fault corresponding to the standard image. The electronic device can determine the most likely fault category of the vehicle component based on the probability corresponding to each abnormal image area.

[0070] In one implementation, the electronic device inputs the abnormal image area and the standard image into a preset network, and calculates the log-likelihood value between the abnormal image area and the standard image. The electronic device can determine the degree of damage of the component failure based on the log-likelihood value. Optionally, the preset network can be a deep learning network that evaluates the similarity between the abnormal image area and the standard image. Alternatively, the preset network can also simulate the probability of the abnormal image area evolving to the standard image through an adversarial neural network. Optionally, the electronic device can more accurately identify minor deviations and potential problems in the normal data stream by using an adversarial neural network for anomaly detection, thereby improving the sensitivity and accuracy of fault judgment and reducing the possibility of false alarms and missed alarms. Specifically, when the probability of the abnormal image area evolving to the standard image is small, it means that the anomaly is a problem or potential problem caused by a minor deviation.

[0071] In summary, the present application obtains the working parameters of the vehicle and generates the working parameter image of the vehicle components. Based on the working parameter image, the electronic device can determine the abnormal image area in the working parameter image. The electronic device can determine the corresponding standard image according to the component and the working parameter. The electronic device can determine the degree of damage of the fault of the vehicle component according to the abnormal image area and the standard image. In the above process, the electronic device effectively integrates advanced technical means such as big data analysis, digital twins, adversarial neural networks, and knowledge graphs to form a comprehensive and efficient vehicle health status assessment and fault diagnosis process, which effectively promotes the development of automotive aftermarket services in a more intelligent and refined direction. This integrated design breaks the limitations of traditional diagnostic methods and realizes real-time monitoring, accurate prediction and efficient diagnosis of vehicle status. In addition, the use of this set of solutions not only innovates the technical means of vehicle fault diagnosis, but also greatly improves the service efficiency and service quality of the entire automotive aftermarket, and has significant practical value and technological leadership in promoting the development of smart transportation and intelligent networked vehicles.

[0072] On the basis of the above examples, the electronic device can determine the component that needs to be determined for fault according to the user's needs. Alternatively, the electronic device can also determine the component that needs to be determined for fault and the fault corresponding to the abnormal reminder according to the abnormal reminder when the abnormal reminder of the vehicle is obtained. The electronic device can use the knowledge graph to find at least one working parameter related to the component and the fault. Optionally, the working parameter is positively correlated with the component and the fault. For example, the working parameter can be a parameter such as power, current, voltage, temperature, etc. The electronic device can obtain the parameter value of the related working parameter. Optionally, the electronic device can trace the working parameter from the current moment to the first moment when the abnormal reminder is obtained. Optionally, the first moment can be the moment when the working parameter has not yet been abnormal. For example, for the abnormality of the battery temperature being too high, the first moment can be the temperature of the battery temperature within the threshold range before the current moment. Or, the first moment can be the moment when the vehicle is started. Optionally, the electronic device can continue to obtain the parameter value of the working parameter from the current moment until the second moment when the abnormal reminder is obtained. For example, the second moment can be the moment corresponding to 2 minutes after the current moment. For example, the second moment can be the moment when the vehicle is turned off. The electronic device can generate an operating parameter image according to the parameter value.

[0073] Figure 2 A flow chart showing another embodiment of the method for automatically determining a vehicle fault of the present invention is shown, and the method is executed by an electronic device. Figure 1 Based on the embodiment shown, when the electronic device can obtain the similarity score between the abnormal image area and the standard image through the generation network calculation in step 140, such as Figure 2 The following steps may be included:

[0074] Among them, the generative network is part of the adversarial neural network. The adversarial neural network may include a generative network and a discriminative network. During the training process, the generative network generates realistic normal data samples, and the discriminative network is responsible for distinguishing between real data and generated data. Optionally, the judgment network in the adversarial neural network can be used to distinguish between data in normal state and abnormal state. That is, when the actual collected data cannot be judged as normal, it is considered that there is a fault signal, and further analysis is performed to determine the type and degree of the fault. The electronic device can use the Glow generative model in the adversarial neural network to compare the segmented vehicle pixel image with the original normal image. As a reversible, flow-based generative model, the Glow generative model not only supports accurate latent variable inference and log-likelihood evaluation, but also can efficiently process reasoning and synthesis, provide available potential space for downstream tasks, and significantly save memory space. The use of the Glow generative model in step 140 realizes the creative application of adversarial neural networks in the intelligent vehicle fault diagnosis process, and enhances the detection capability of vehicle faults and the accuracy of fault identification. Optionally, the application of the adversarial neural network in fault detection is actually to generate a network through training, imitate the data distribution under normal conditions, and compare the actual detected data with it to find possible deviations and abnormal signals, thereby improving the anti-interference ability of the diagnostic system and the accuracy of fault identification.

[0075] Step 210: Use the generative network to extract and calculate features of the abnormal image region to obtain a first feature, which is used to characterize the image features of the abnormal image region. Use the generative network to extract and calculate features of the standard image to obtain a second feature, which is used to characterize the image features of the standard image.

[0076] The electronic device may input the abnormal image region and the standard image into the generation network respectively. The generation network may extract and calculate features of the abnormal image region and the standard image to obtain a first feature of the abnormal image region and a second feature of the standard image. Optionally, the first feature is a feature of the abnormal image region. The second feature may be a feature of the standard image. Optionally, the feature may be a feature obtained by calculating the maximum likelihood of the latent variable.

[0077] Optionally, this step mainly realizes the extraction of features. The extraction of the features can be performed for each image. That is, the electronic device can obtain the first feature of the abnormal image area and the second feature of the standard image by inputting the abnormal image area and the standard image one by one. Optionally, one or more abnormal image areas can be included. And one or more standard images can be included.

[0078] Optionally, the adversarial neural network constructs a framework based on generative adversarial networks (GANs). The adversarial neural network consists of two parts: a generator and a discriminator. The generator is used to simulate the distribution of CAN bus data under normal conditions, and the discriminator is responsible for distinguishing between real data and generated data. The generator generates pseudo samples that are close to the real ones by learning a large number of working parameters during normal operation. After training, the discriminator can accurately distinguish whether the data actually uploaded is real data from the vehicle or simulated data generated by the generator.

[0079] In one implementation, when the real-time uploaded working parameters are analyzed by the discriminator, if they cannot be accurately judged as real data, they are judged as potential abnormal signals. This indicates that the measured data may contain fault characteristics or abnormal operating conditions, triggering further in-depth diagnostic processes. Optionally, the electronic device can also combine the digital twin model and the knowledge graph to conduct a detailed analysis of the abnormal data identified by the discriminator, find the most likely source of the fault, and determine the specific location and type of the fault by comparing historical models, the correlation between components, and the fault propagation path. The adversarial neural network effectively improves the intelligence level of the intelligent vehicle fault diagnosis system, realizes accurate monitoring of the vehicle's health status and efficient fault diagnosis, and provides strong technical support for automobile maintenance.

[0080] In one implementation, the adversarial neural network has the ability to self-learn. With the input of more real data and the iterative update of the model, its ability to identify abnormal situations will be continuously optimized and improved. At the same time, the adversarially generated samples can be used to enrich the training set and improve the generalization performance of the fault diagnosis algorithm.

[0081] In one implementation, the electronic device may also use a generative network to extract and calculate features of the abnormal image region to obtain a first feature. The process may include the following steps:

[0082] Step 211: Activate neurons in the generation network and normalize the abnormal image area.

[0083] Wherein, the electronic device can obtain the initial value of each neuron in the generated network when loading the generated network for the first time. The electronic device can normalize the neurons according to the initial value of each neuron. Optionally, the process can be an actnorm step. Through this step, the electronic device can adjust the scale of the data to make it more suitable for subsequent processing.

[0084] Specifically, after obtaining the initial value of each neuron, the electronic device can calculate the mean μ and standard deviation σ of the neurons in the generated network. The electronic device can normalize the initial value of the neuron according to the mean and standard deviation. The formula can be:

[0085] y=(x-μ) / σ

[0086] Where x represents the abnormal image area of ​​the input generation network. μ represents the mean of all neurons. σ represents the standard deviation of all neurons. y represents the normalized abnormal image area.

[0087] Step 212: construct a convolution function, where the convolution function includes a convolution region and a weight for convolving the abnormal image region.

[0088] The convolution function may include a reversible 1×1 convolution layer. The setting of the convolution layer helps the model learn the intrinsic structure of the data while maintaining the reversibility of the model.

[0089] Step 213: Use the convolution function to perform affine transformation on the abnormal image area to obtain convolution features.

[0090] In the process of using the convolution function to convolve the abnormal image area, a layer of convolution can be regarded as a layer of affine transformation. Through affine transformation, the model can further extract the features of the input data and prepare for subsequent comparison and generation tasks.

[0091] In this process, the electronic device may first determine a weight matrix according to a convolution function. The weight matrix represents the weight of the pixel points in the abnormal image area in the convolution region when the convolution is performed in the abnormal image area. Optionally, the weight is the convolution kernel. The electronic device may perform an affine transformation on the abnormal image area according to the weight matrix and preset parameters to obtain a convolution feature. Among them, the preset parameters represent the bias value. The process of the affine transformation can be shown in the following formula:

[0092] z=Wx+b

[0093] Wherein, W represents a weight matrix. b is a preset parameter. x is input data. Optionally, the input data may be the normalized abnormal image region y calculated in step 211. z represents the transformed output. Optionally, the output is a convolution feature.

[0094] Since the generation network can include multiple layers of convolution, the electronic device can complete the feature extraction of the abnormal image region through the above steps and form a new data stream. This data stream can efficiently process the input abnormal image region and gradually restore it to a changed form.

[0095] Step 214: transform and extract the convolution feature to obtain the first feature.

[0096] Among them, electronic devices can use the multi-scale architecture of the Glow generative network to combine information at different scales to model latent variables more accurately. In this way, the exact log-likelihood value of the latent variable can be obtained. The calculation formula for the log-likelihood value can be:

[0097] logp(x)=logpz(f(x))+log|detJ f (x)|

[0098] Where p(x) is the probability density of data point x. f(x) is the mapping function from data space to latent space. f (x) is the Jacobian matrix of f(x). pz(f(x)) is the probability density of the latent variable.

[0099] The above process can be divided into the following steps:

[0100] Step 2141: Map the convolution feature from the data space to the latent space according to the first mapping function. The first mapping function is used to achieve the mapping from the data space to the latent space. The latent space represents the low-dimensional vector space of the abnormal image area.

[0101] Step 2142: Calculate the probability density of the latent variables in the latent space according to the first probability function. The first probability function represents the possibility of each latent variable in the latent space taking a value. The latent variable is a feature in the latent space. The probability density represents the probability distribution of the latent variable.

[0102] Step 2143: Determine the first feature based on the logarithm of the probability density of the latent variable and the logarithm of the determinant of the Jacobian matrix of the convolution feature.

[0103] Optionally, the calculation process of the second feature of the standard image is the same as the calculation process of the first feature.

[0104] Step 220: Determine a similarity score between the abnormal image region and the standard image based on the difference value between the first feature and the second feature.

[0105] The electronic device may calculate the first feature of each abnormal image region and the second feature of each standard image, and then determine the corresponding relationship between the first feature and the second feature according to the corresponding relationship between the abnormal image region and the standard image. The electronic device may calculate the difference value between the first feature and the second feature that have a corresponding relationship, and further calculate the similarity score between the abnormal image region and the standard image according to the difference value.

[0106] Step 230: The abnormal category corresponding to the standard image with the highest similarity score between the abnormal image region and the multiple standard images is taken as the fault category.

[0107] The electronic device may select the abnormal category corresponding to the standard image with the largest similarity score as the fault category of the vehicle. Alternatively, the electronic device may also be provided with a similarity threshold. When the similarity score is greater than the similarity threshold, the electronic device may select the abnormal category corresponding to the standard image as the fault category of the vehicle.

[0108] In summary, the present application uses a generative network to extract features from the abnormal image area and calculate the first feature, and uses a generative network to extract features from the standard image and calculate the second feature. The electronic device determines the similarity score between the abnormal image area and the standard image based on the difference between the first feature and the second feature. The electronic device can use the abnormal category corresponding to the standard image with the highest similarity score between the abnormal image area and multiple standard images as the fault category. In the above process, the electronic device uses the generative network in the adversarial neural network to achieve matching between the abnormal image area and the standard image, thereby improving the accuracy of fault judgment.

[0109] In one implementation, the electronic device may also obtain a normal image of the vehicle under normal conditions after obtaining the fault category of the vehicle. The normal image may include a curve chart of the working parameters of the vehicle under normal driving conditions. The electronic device may make a more accurate damage estimate of the degree of damage to the vehicle by comparing the difference in log-likelihood values ​​between the abnormal image area and the normal image. Through the above steps, the electronic device may accurately estimate the degree of damage to the vehicle by calculating the log-likelihood value of the latent variable. This method combines the efficiency of deep learning with the accuracy of mathematical models, and provides a new solution for vehicle damage assessment.

[0110] In another implementation, taking a motor operation monitoring application as an example, the specific execution process of the above method may include:

[0111] The electronic device monitors the motor signal in real time. The motor signal is the working parameter. The electronic device can generate a working parameter image based on the motor signal. The working parameter image is a grid voltage waveform. When the grid voltage waveform shows an abnormal waveform, the electronic device can obtain a standard image of the grid voltage abnormality through the knowledge graph.

[0112] Electronic devices can obtain the voltage sag waveform corresponding to abnormal type A, the voltage swell waveform corresponding to abnormal type B, and the voltage fluctuation waveform corresponding to abnormal type C from the knowledge graph. Figure 3 The three abnormal waveforms are obtained by classifying the collected historical motor detection data. The three abnormal waveforms represent three common grid voltage anomalies.

[0113] The electronic device can preprocess the abnormal image area with abnormal waveforms in the working parameter image and the three abnormal waveform images in the gallery. The preprocessing includes normalization, denoising, etc. The electronic device can input the preprocessed abnormal image area and the three abnormal waveform images into the trained Glow model. The glow model is a generative network model. The electronic device can extract and encode the features of the abnormal image area and the three abnormal waveform images respectively to obtain their respective latent variable representations. The electronic device can calculate the similarity of the abnormal image area and the three abnormal waveform images based on the log-likelihood value or other similarity measurement indicators generated by the Glow model. The Glow model will give a similarity score for the abnormal image area and each abnormal waveform image. This score can be based on the difference in log-likelihood values, the Euclidean distance between latent variables, or other similarity measurement methods. Assume that the Glow model gives scores of 85, 70, and 92 for abnormal type A, abnormal type B, and abnormal type C, respectively. Then the electronic device can judge that the abnormal image area is close to abnormal type C based on the highest score.

[0114] The electronic device may output the judgment result, and inform the user that the most likely abnormality currently occurring is abnormality type C. Optionally, the electronic device may also provide corresponding processing suggestions or further operation guidance based on the knowledge graph.

[0115] On the basis of the above embodiments, the electronic device may also be provided with a user interaction and service platform. The user interaction and service platform may develop a supporting mobile application or a web platform. The user interaction and service platform is used to provide users with intuitive vehicle health reports, early warning information, and detailed maintenance guides. Users can obtain instant vehicle condition assessment results through the user interaction and service platform, and communicate and consult with professional technicians to achieve remote diagnosis and service appointments. That is, the user interaction and service platform may include a comprehensive platform that integrates intelligent fault diagnosis result display, user-friendly operation, and convenient service appointment functions.

[0116] Optionally, based on the user interaction and service platform, the electronic device can realize the display of real-time monitoring information and the push of monitoring reports. Optionally, the user interaction and service platform can have a real-time data visualization function, which can intuitively present various key indicators and fault warning information of the vehicle operation status in the form of charts, dashboards, etc. In addition, when the electronic device detects an abnormality or completes the fault diagnosis, the electronic device can also automatically push a detailed diagnostic report to the user through a mobile application or a web terminal.

[0117] Optionally, the user interaction and service platform may include a user-friendly interactive interface. The interactive interface is designed to be simple and easy to use. Users can use the interactive interface to implement graphical operations, thereby easily viewing the vehicle health status, historical maintenance records, and predictive maintenance recommendations. At the same time, the interactive interface may also support personalized settings. For example, the personalized settings may include reminder threshold settings, focus component screening, etc. The personalized settings can meet the usage needs of different users.

[0118] Optionally, the interactive interface can also display interpretation and guidance of the fault diagnosis results. For complex fault diagnosis results, the user interaction and service platform can use knowledge graph technology to conduct in-depth interpretation and convert the fault diagnosis results into easy-to-understand language and graphic descriptions for display. This method can provide users with clear fault cause analysis and corresponding maintenance solution suggestions.

[0119] Optionally, the user interaction and service platform can also be used for appointment maintenance and service management. The user interaction and service platform can integrate online and offline resources, allowing users to directly make online appointments with nearby authorized service centers or professional technicians for maintenance based on the diagnosis results. At the same time, the user interaction and service platform can also provide service progress tracking, cost estimation, work order management and other functions to achieve a one-stop service process from diagnosis to maintenance.

[0120] Optionally, the user interaction and service platform can also have a data security and privacy protection module. The setting of this module fully considers the data security and privacy protection needs of users. The implementation of this module adopts advanced encryption technology and strict authority control mechanism to ensure the safe transmission and storage of vehicle operation data, and at the same time complies with relevant laws and regulations to handle user data in compliance with regulations.

[0121] In summary, the user interaction and service platform not only provides comprehensive and accurate vehicle fault diagnosis information, but also realizes efficient interaction with users, allowing users to easily and quickly obtain and respond to diagnostic results, thereby promoting the intelligent, efficient and humanized development of automobile repair services. In addition, electronic equipment effectively improves the intelligent level of vehicle fault diagnosis through the above method, greatly reducing the time cost and misdiagnosis rate of troubleshooting. The combination of the above method and the user interaction and service platform has promoted the improvement of service quality and efficiency of the entire automotive aftermarket.

[0122] Figure 3 FIG. 2 is a schematic diagram showing the structure of an embodiment of the automatic determination device for vehicle faults of the present invention. Figure 3 As shown, the automatic determination device 300 for vehicle faults includes: an acquisition module 310 and a processing module 320 .

[0123] The acquisition module 310 is used to acquire an operating parameter image of a component in the vehicle, wherein the operating parameter image includes a curve graph of the operating parameter corresponding to the component.

[0124] The processing module 320 is used to determine an abnormal image area in the working parameter image, wherein the abnormal image area is an image area in the working parameter image where the working parameter is abnormal, and the abnormal image area has at least one abnormal category. Determine a standard image corresponding to the abnormal category. The standard image includes a curve chart of the working parameters of the vehicle component when an abnormality corresponding to the abnormal category occurs. Determine the fault category of the component based on the abnormal image area and the standard image.

[0125] In an optional manner, the processing module 320 is used to:

[0126] The abnormal image region and the standard image are input into the preset network, and the similarity score between the abnormal image region and the standard image is calculated.

[0127] The abnormal category corresponding to the standard image with the highest similarity score between the abnormal image area and multiple standard images is taken as the fault category.

[0128] In an optional manner, the preset network is a generated network. The processing module 320 is used to:

[0129] The generative network is used to extract and calculate features of the abnormal image region to obtain a first feature, which is used to characterize the image features of the abnormal image region. The generative network is used to extract and calculate features of the standard image to obtain a second feature, which is used to characterize the image features of the standard image.

[0130] A similarity score between the abnormal image region and the standard image is determined according to the difference value between the first feature and the second feature.

[0131] In an optional manner, the processing module 320 is used to:

[0132] Activate neurons in the generative network and normalize abnormal image regions;

[0133] A convolution function is constructed, wherein the convolution function includes a convolution area and a weight for convolving the abnormal image area.

[0134] The convolution function is used to perform affine transformation on the abnormal image area to obtain the convolution feature.

[0135] The convolution feature is transformed and feature extracted to obtain the first feature.

[0136] In an optional manner, the processing module 320 is used to:

[0137] Activate the neurons in the generated network and get the initial value of each neuron.

[0138] Based on the initial values, the mean and standard deviation of the neurons in the generated network are calculated.

[0139] The initial values ​​of the neurons are normalized according to the mean and standard deviation.

[0140] In an optional manner, the processing module 320 is used to:

[0141] The weight matrix is ​​determined according to the convolution function. The weight matrix represents the weight of the pixel points in the abnormal image area within the convolution area when the convolution is performed in the abnormal image area.

[0142] According to the weight matrix and preset parameters, an affine transformation is performed on the abnormal image area to obtain the convolution feature, and the preset parameters represent the bias value.

[0143] In an optional manner, the processing module 320 is used to:

[0144] According to the first mapping function, the convolution feature is mapped from the data space to the latent space. The first mapping function is used to achieve the mapping from the data space to the latent space. The latent space represents a low-dimensional vector space of the abnormal image area.

[0145] According to the first probability function, the probability density of the latent variables in the latent space is calculated. The first probability function represents the possibility of each latent variable in the latent space taking a value. The latent variable is a feature in the latent space. The probability density represents the probability distribution of the latent variable.

[0146] The first feature is determined according to the logarithm of the probability density of the latent variable and the logarithm of the determinant of the Jacobian matrix of the convolution feature.

[0147] In an optional manner, the processing module 320 is used to:

[0148] The working parameter image is input into a pre-trained recognition network to obtain at least one abnormal image region in the working parameter image; the recognition network is used to segment the normal image region and the abnormal image region in the working parameter image; the abnormal image region corresponds to the abnormal category.

[0149] In an optional manner, the abnormality category includes at least one of the damaged location, the degree of wear, and the damaged area.

[0150] In an optional manner, the processing module 320 is used to:

[0151] According to the abnormal category, a standard image corresponding to the abnormal category is determined based on a preset knowledge graph, wherein the preset knowledge graph indicates the standard image corresponding to the abnormal category.

[0152] In an optional manner, the processing module 320 is used to:

[0153] Get actual failures of components reported by users.

[0154] If the actual fault does not appear in the knowledge graph, the working parameter image and the actual fault are added to the knowledge graph. The actual fault is an abnormal category in the knowledge graph.

[0155] In an optional manner, the acquisition module 310 is used to:

[0156] Collect operating parameters of vehicle components.

[0157] The working parameters are input into the pre-trained twin network to simulate the operating state of the vehicle and obtain the working parameter image of the vehicle under the operating state.

[0158] Among them, the twin network is used to simulate the operating state of the vehicle according to the vehicle's operating parameters.

[0159] In an optional manner, the acquisition module 310 is used to:

[0160] The working parameters are input into the pre-trained twin network to predict the vehicle operating status, and a working parameter image is generated based on the predicted operating status.

[0161] The automatic vehicle fault determination device 300 provided in the embodiment of the present application can execute the above method embodiment. Its specific implementation principle and technical effects can be found in the above method embodiment, and this embodiment will not be repeated here.

[0162] Figure 4 The schematic diagram of the structure of an embodiment of the electronic device of the present invention is shown, and the specific embodiment of the present invention does not limit the specific implementation of the electronic device.

[0163] like Figure 4 As shown, the electronic device may include: a processor (processor) 402 , a communication interface (Communications Interface) 404 , a memory (memory) 406 , and a communication bus 408 .

[0164] The processor 402, the communication interface 404, and the memory 406 communicate with each other via the communication bus 408. The communication interface 404 is used to communicate with other devices such as a client or other server network elements. The processor 402 is used to execute the program 410, which can specifically execute the relevant steps in the above-mentioned XXXX method embodiment.

[0165] Specifically, the program 410 may include program code including computer executable instructions.

[0166] The processor 402 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the electronic device may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0167] The memory 406 is used to store the program 410. The memory 406 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0168] The program 410 can be specifically called by the processor 402 to enable the electronic device to execute the method shown in the above embodiment.

[0169] The electronic device provided in this embodiment can be used to execute the above method, and its implementation method and technical effect are similar, which will not be described in detail in this embodiment.

[0170] An embodiment of the present invention provides a computer-readable storage medium, which stores at least one executable instruction. When the executable instruction is executed on an electronic device, the electronic device executes a method in any of the above method embodiments. The implementation method and technical effect are similar, and this embodiment will not be repeated here.

[0171] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system or other device. In addition, the embodiments of the present invention are not directed to any particular programming language.

[0172] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. Similarly, in order to simplify the present invention and help understand one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, the various features of the embodiments of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. Wherein, the claims that follow the specific embodiment are hereby expressly incorporated into the specific embodiment, wherein each claim itself is a separate embodiment of the present invention.

[0173] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and further may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive.

[0174] It should be noted that the above embodiments illustrate the present invention rather than limit it, and that those skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be construed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention may be implemented by means of hardware comprising a number of different elements and by means of a suitably programmed computer. In a unit claim enumerating a number of devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be understood as limitations on the order of execution.

Claims

1. A method for automatically determining a vehicle fault, characterized in that: The method comprises: Acquiring an operating parameter image of a component in a vehicle, wherein the operating parameter image includes a graph of the operating parameters corresponding to the component; Determining an abnormal image region in the working parameter image, wherein the abnormal image region is an image region in the working parameter image where the working parameters are abnormal; Determining a standard image corresponding to the abnormal image area; wherein the standard image includes a curve diagram of the operating parameters corresponding to the abnormal image area of ​​the component of the vehicle during normal operation; The damage degree of the component failure is determined based on the abnormal image area and the standard image.

2. The method according to claim 1, characterized in that The determining of the abnormal image area in the working parameter image includes: The working parameter image is input into a pre-trained recognition network to obtain at least one abnormal image region in the working parameter image; the recognition network is used to segment the normal image region and the abnormal image region in the working parameter image.

3. The method according to claim 2, characterized in that The recognition network is a twin network; the inputting of the working parameter image into the pre-trained recognition network to obtain at least one abnormal image region in the working parameter image includes: Inputting the working parameter image into a pre-trained twin network to simulate and obtain an ideal parameter image of the vehicle under the operating state; comparing the working parameter image with the ideal parameter image, and determining an inconsistent region between the working parameter image and the ideal parameter image as an abnormal image region; The twin network is used to simulate the operating state of the vehicle according to the operating parameters of the vehicle.

4. The method according to any one of claims 1 to 3, characterized in that The determining of the standard image corresponding to the abnormal image area includes: The operating parameters of the component during normal operation are obtained from the knowledge graph to generate the standard image.

5. The method according to any one of claims 1 to 3, characterized in that Determining the damage extent of the component failure according to the abnormal image area and the standard image includes: Inputting the abnormal image region and the standard image into a preset network, and calculating a log-likelihood value between the abnormal image region and the standard image; The damage degree of the component failure is determined according to the log-likelihood value.

6. The method according to any one of claims 1 to 3, characterized in that The obtaining of the working parameter image of the component in the vehicle includes: Obtaining an abnormality alert of the vehicle, and determining a faulty component based on the abnormality alert; Determining, based on the knowledge graph, at least one operating parameter associated with the abnormality corresponding to the abnormality alert of the component; A parameter value corresponding to the operating parameter is obtained from the vehicle, and the operating parameter image is generated according to the parameter value.

7. The method according to claim 6, characterized in that The obtaining, from the vehicle, a parameter value corresponding to the operating parameter, includes: Obtain the parameter value of the working parameter from a first moment to a second moment; wherein, the first moment is the moment before the vehicle obtains the abnormal reminder, and the second moment is the moment after the vehicle obtains the abnormal reminder; the parameter value of the working parameter does not show any abnormality at the first moment.

8. The method according to any one of claims 1 to 3, characterized in that The method further comprises: Obtaining abnormal working parameters corresponding to the abnormal reminder of the component reported by the user; If the operating parameter and the component and the anomaly corresponding to the anomaly reminder are not associated in the knowledge graph, an association between the operating parameter and the component and the anomaly corresponding to the anomaly reminder is added in the knowledge graph.

9. An automatic determination device for vehicle faults, characterized in that: The device comprises: an acquisition module, configured to acquire an operating parameter image of a component in a vehicle, wherein the operating parameter image includes a graph of the operating parameters corresponding to the component; A processing module is used to determine an abnormal image area in the working parameter image, wherein the abnormal image area is an image area in the working parameter image where the working parameters are abnormal; determine a standard image corresponding to the abnormal image area; wherein the standard image includes a curve diagram of the working parameters corresponding to the abnormal image area of ​​the component of the vehicle during normal operation; and determine the damage extent of the fault of the component based on the abnormal image area and the standard image.

10. An electronic device, characterized in that: include: A processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform the operation of the method according to any one of claims 1 to 8.

Images