Vehicle fault predicting and positioning system
Through edge-cloud collaborative architecture and incremental learning technology, combined with fault knowledge graph and multi-model fusion, the real-time, stability and cross-model applicability of the existing vehicle fault prediction system are solved, efficient and reliable fault prediction and positioning are achieved, and maintenance costs are reduced.
Patent Information
- Application Number
- CN202510410718.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-02
- Publication Date
- 2025-07-04
AI Technical Summary
The existing vehicle failure prediction system has problems such as difficulty in taking into account real-time and stability, imperfect model update mechanism, difficult to balance data privacy and utilization, insufficient cross-vehicle knowledge sharing, lack of interpretability, difficulty in system expansion and maintenance, and high labor maintenance costs.
Adopt edge-cloud collaborative architecture, combining incremental learning, edge intelligent preprocessing and data encryption technologies, build a fault knowledge graph to realize cross-vehicle knowledge migration, adopt multi-model convergence architecture and microservice design, and deploy lightweight edge computing nodes to provide emergency fault detection and fault prediction.
It improves the accuracy, real-time and adaptability of vehicle failure prediction, reduces maintenance costs, ensures system reliability and data privacy protection, and achieves efficient fault location and maintenance guidance.
Smart Images
Figure CN120255480A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of vehicle fault detection, and particularly relates to a vehicle fault prediction and positioning system. Background Art
[0002] The modern automotive industry is accelerating its transformation towards electrification and intelligentization. The cost proportion of automotive electronic components is continuously rising, and the massive data generated by electronic control units (ECUs) provides a basis for vehicle fault prediction. Currently, there are various vehicle fault prediction systems, such as those based on static models, combinations of physical models and data-driven approaches, online learning based on cloud platforms, and upgraded versions of OBD-II diagnostic tools. Although these systems have achieved certain results in enhancing user experience, reducing maintenance costs, and ensuring safety, they have limitations such as difficulty in balancing real-time performance and stability, underutilized potential of edge computing, imperfect model update mechanisms, low data fusion efficiency, high susceptibility of system reliability to network influence, insufficient cross-vehicle model knowledge sharing, lack of interpretability, difficulty in balancing data privacy and utilization, difficulties in system expansion and maintenance, and high human maintenance costs, and thus require further technological innovation and optimization. Summary of the Invention
[0003] The purpose of the present invention is to provide a vehicle fault prediction and positioning system, which adopts an edge-cloud collaboration and incremental learning architecture to solve the technical problems of long update cycles, high resource consumption, limited real-time performance, low intelligence level, and poor cross-vehicle applicability of existing vehicle fault prediction systems.
[0004] To solve the above technical problems, the present invention is realized through the following technical solutions:
[0005] The present invention provides a vehicle fault prediction and positioning system, which includes: a vehicle edge node module, a cloud module, and a human-computer interaction module;
[0006] The vehicle edge node module is used to collect and process in-vehicle sensor data, and upload the processed feature data to the cloud module;
[0007] The cloud module is used to store the feature data, and analyze and process the feature data using a preset fault prediction model to locate the fault source and output a fault prediction result;
[0008] The human-computer interaction module is used to display the fault prediction result output by the cloud module, and upload user feedback information to the cloud module.
[0009] In an embodiment of the present invention, the vehicle edge node module includes an in-vehicle sensor network unit, an edge computing node unit, and an edge-cloud communication unit;
[0010] The in-vehicle sensor network unit is used to collect the working parameters of key vehicle systems in real time to obtain the in-vehicle sensor data;
[0011] The edge computing node unit is used to preprocess and extract features from the in-vehicle sensor data to obtain the feature data;
[0012] The edge cloud communication unit is used to upload the feature data to the cloud module for storage.
[0013] In an embodiment of the present invention, the edge node module further includes an edge fault detection unit, which is used to detect emergency faults and send the detection results to the human-machine interaction module to achieve emergency warning.
[0014] In an embodiment of the present invention, the cloud module further includes a fault knowledge graph unit, which is used to explain the fault causes and development mechanisms of the fault prediction results.
[0015] In an embodiment of the present invention, the cloud module further includes a user feedback unit, which is used to receive the user feedback information and use natural language processing technology to extract and label the fault descriptions in the user feedback information.
[0016] In an embodiment of the present invention, the user feedback information includes the user's feedback on the fault prediction results and the actual maintenance results.
[0017] In an embodiment of the present invention, the cloud module further includes an incremental learning unit, which is used to optimize and update the preset fault prediction model according to the user feedback information to achieve version iteration of the preset fault prediction model.
[0018] In an embodiment of the present invention, the performance of the updated preset fault prediction model is evaluated. When it is detected that the performance of the iterated new model decreases, the model version with the optimal performance is restored through an automatic rollback operation.
[0019] In an embodiment of the present invention, the preset fault prediction model is composed of multiple sub-models to capture fault features at different levels respectively, and an adaptive weight adjustment mechanism is used to enhance the generalization ability of the preset fault prediction model.
[0020] In an embodiment of the present invention, the human-machine interaction module can provide a customized information display interface according to different user roles.
[0021] As described above, a vehicle fault prediction and location system provided by the present invention includes a vehicle edge node module, a cloud module, and a human-machine interaction module. The vehicle edge node module is used to collect and process in-vehicle sensor data, and upload the processed feature data to the cloud module. The cloud module is used to store the feature data, and analyze and process the feature data using a preset fault prediction model to locate the fault source and output a fault prediction result. The human-machine interaction module is used to display the fault prediction result output by the cloud module, and upload user feedback information to the cloud module. The vehicle fault prediction and location system deploys lightweight edge computing nodes at the vehicle end, which are responsible for processing emergency fault detection, while the cloud is responsible for the training and in-depth analysis of complex models. Through an intelligent task allocation mechanism, the system combines millisecond-level response with high-precision prediction. At the same time, a complete data processing flow is designed, including data screening, priority sorting, incremental training, model evaluation, and an automatic rollback mechanism, ensuring that the model can be continuously optimized safely and efficiently. The system uses multiple models such as XGBoost, LSTM, and autoencoders to capture fault characteristics at different levels, and improves the generalization ability of the system through an adaptive weight adjustment mechanism. In addition, the edge node has an independent basic fault diagnosis function, ensuring that core services can still be provided during network interruption, and automatically synchronizing data and models after the network is restored. Through transfer learning technology, knowledge sharing between different vehicle models is realized, enabling new vehicle models to draw on the experience of existing vehicle models and accelerating model convergence. The system also combines domain expert knowledge to construct a fault knowledge graph, provides cause analysis for fault prediction, and indicates the reliability of the prediction through a confidence evaluation mechanism. In terms of data processing, the edge node is responsible for data preprocessing and desensitization, and only transmits necessary features and statistical information, effectively balancing the utilization of data value and privacy protection. The system adopts a microservice design and a Docker+Kubernetes deployment method, realizing modular management, flexible expansion, and version control. Based on the fault prediction result, the system provides detailed repair suggestions and illustrated operation guides, reducing the dependence on senior technicians and improving the repair efficiency. Finally, a complete closed-loop of user feedback-system optimization-effect verification is established, and the user feedback is automatically analyzed in combination with a large language model to continuously improve the accuracy of the system.
[0022] The vehicle fault prediction and location system not only completely solves the key technical problems faced by existing automotive fault prediction systems, but also achieves significant improvements in terms of accuracy, real-time performance, adaptability, and reliability. At the same time, it greatly reduces the maintenance cost, providing a comprehensive health management solution for modern intelligent vehicles. Of course, it is not necessary for any product implementing the present invention to simultaneously achieve all the above-mentioned advantages. Brief Description of the Drawings
[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.
[0024] Figure 1 The system block diagram of the vehicle fault prediction and location system provided by an exemplary embodiment of the present application.
[0025] Figure 2 The flowchart of data transmission of the vehicle fault prediction and location system provided by an exemplary embodiment of the present application.
[0026] Figure 3 The flowchart of incremental learning and model update provided by an exemplary embodiment of the present application.
[0027] The reference numerals are as follows:
[0028] 100 Human-machine interaction module
[0029] 200 Vehicle edge node module
[0030] 300 Cloud module Specific embodiments
[0031] The following illustrates the embodiments of the present invention through specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other.
[0032] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. Therefore, only the components related to the present invention are shown in the diagrams, rather than being drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and ratio of each component in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.
[0033] In the following description, a large number of details are explored to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.
[0034] To solve the technical problems of the existing vehicle fault prediction system, such as long update cycle, large resource consumption, limited real-time performance, low intelligence level, and poor cross-model applicability, the present invention provides a vehicle fault prediction and positioning system. In view of the challenges existing in the existing vehicle fault diagnosis system, such as the difficulty in balancing real-time performance and stability, imperfect model update mechanism, difficulty in balancing data privacy and efficient utilization, insufficient cross-model knowledge sharing mechanism, insufficient interpretability, effective fusion of multi-source heterogeneous data, system scalability and maintainability, high human maintenance cost, and system reliability in an unstable network environment, etc., innovative solutions such as an edge-cloud collaborative architecture, a complete incremental learning framework and model rollback mechanism, edge intelligent preprocessing and data encryption technology, cross-model knowledge transfer technology and hot fault rapid learning strategy, an interpretability improvement method combining knowledge graph and expert system, a multi-model fusion architecture and a dedicated data preprocessing process, a microservices architecture and containerized deployment, and a network-disconnection disaster tolerance mechanism are proposed, effectively solving the above problems, greatly improving the performance, efficiency, and reliability of the vehicle fault diagnosis system, while reducing the maintenance cost and enhancing the user's trust in the system.
[0035] Please refer to Figure 1 As shown in the figure, in an exemplary embodiment of the present application, the vehicle fault prediction and positioning system includes a vehicle edge node module 200, a cloud module 300, and a human-computer interaction module 100. The vehicle edge node module 200 is used to collect and process in-vehicle sensor data, and upload the processed feature data to the cloud module. The cloud module 300 is used to store the feature data and analyze and process the feature data using a preset fault prediction model to locate the fault source and output a fault prediction result. The human-computer interaction module 100 is used to display the fault prediction result output by the cloud module 300 and upload the user feedback information to the cloud module 300.
[0036] It should be noted that the implementation of the vehicle fault prediction and location system covers system initialization and deployment, specifically including edge node deployment, cloud platform construction, communication gateway configuration, and human-computer interaction terminal development. As the data collection layer, the edge node needs to install edge computing hardware integrated with computing, storage, and communication functions on the vehicle, configure in-vehicle sensor network interfaces to connect sensors and collect data, and set up communication modules to ensure that the collected data can be stably and efficiently sent to the cloud platform. As the data processing and analysis layer, the cloud platform first builds a distributed data storage system to store the massive data from the edge nodes. Then, it deploys a data analysis engine to perform real-time processing and analysis on the collected data. In addition, it also sets up a model training platform to train and optimize the fault prediction model to improve the accuracy of prediction. As the bridge connecting the edge nodes and the cloud platform, the communication gateway needs to establish a stable and reliable edge-cloud communication gateway to ensure the smooth transmission of data. As the user interaction interface, the human-computer interaction terminal is provided with an in-vehicle terminal interface, a maintenance system, and a mobile terminal application. Among them, the in-vehicle terminal interface is used to display information such as vehicle status and fault warnings, the maintenance system is used to help maintenance personnel quickly locate faults and formulate maintenance plans, and the mobile terminal application is used to provide convenient functions such as remote monitoring and fault diagnosis for vehicle owners. The vehicle fault prediction and location system realizes the accurate prediction and location of vehicle faults through the close cooperation of each part.
[0037] Please refer to Figure 1 and Figure 2 As shown, in an exemplary embodiment of the present application, the vehicle edge node module 200 includes an in-vehicle sensor network unit, an edge computing node unit, and an edge-cloud communication unit. The in-vehicle sensor network unit is used to collect the working parameters of the vehicle's key systems in real time to obtain the in-vehicle sensor data. The edge computing node unit is used to preprocess and extract features from the in-vehicle sensor data to obtain the feature data. The edge-cloud communication unit is used to upload the feature data to the cloud module for storage.
[0038] Specifically, the vehicle edge node module 200 collects the working parameters of key systems such as the engine, transmission, and chassis through the in-vehicle sensor network in real time. The working parameters are preprocessed in the edge computing node. The preprocessing process includes data cleaning, denoising, and normalization processing, aiming to convert the raw data into effective feature vectors and reduce the data dimension, so as to obtain concise feature data. Subsequently, the feature data is transmitted to the cloud data storage center through the edge-cloud communication gateway.
[0039] Please continue to refer to Figure 1 and Figure 2As shown, in an exemplary embodiment of the present application, the edge node module 200 further includes an edge fault detection unit, which is used to detect emergency faults and send the detection results to the human-machine interaction module to achieve emergency warning.
[0040] Specifically, the edge fault detection deployment includes lightweight model deployment, real-time detection configuration, local cache mechanism, and vehicle-mounted actuator connection. The lightweight model deployment is to deploy the trained lightweight model to the vehicle edge computing node. The real-time detection configuration involves setting the detection threshold and execution strategy of the edge fault monitoring engine. The local cache mechanism ensures the normal operation of the system even in the case of network disconnection by configuring the data cache policy of the edge node. It should be noted that in this embodiment, through the network-disconnection disaster tolerance mechanism, the edge node is capable of independently running the basic fault diagnosis function, ensuring the reliable operation of the system in various network environments. The vehicle-mounted actuator connection mainly connects the fault detection results with the vehicle-mounted execution controller to achieve the emergency response function.
[0041] Please continue to refer to Figure 1 and Figure 2 As shown, in an exemplary embodiment of the present application, the cloud module 300 further includes a fault knowledge graph unit and a user feedback unit. The fault knowledge graph unit is used to explain the fault cause and development mechanism of the fault prediction result, and the user feedback unit is used to receive the user feedback information and extract and annotate the fault description from the user feedback information using natural language processing technology. It should be noted that in this embodiment, the user feedback information includes the user's feedback on the fault prediction result and the actual maintenance result.
[0042] Specifically, in this embodiment, the cloud module 300 includes two major functions: model construction and initial training, and cloud prediction and decision generation. Model construction and initial training specifically cover basic model creation, knowledge graph establishment, model fusion layer development, and initial model training. Among them, basic model creation includes building XGBoost, LSTM, and autoencoders, etc. in the cloud. Knowledge graph establishment involves building an automotive fault knowledge graph, integrating expert knowledge and fault correlation information. Model fusion layer development implements mechanisms such as stacking integration and weighted voting to integrate the outputs of multiple models. Initial model training uses historical data to train the model and establish the baseline performance. Cloud prediction and decision generation specifically cover in-depth data analysis, multi-model comprehensive prediction, fault source location, and decision generation. Among them, in-depth data analysis discovers potential fault patterns by deeply mining historical and real-time data. Multi-model comprehensive prediction uses the model fusion layer to generate comprehensive fault prediction results. Fault source location distinguishes primary faults and secondary faults by combining the knowledge graph. Decision generation proposes maintenance suggestions and preventive measures based on the fault prediction results and historical maintenance data.
[0043] Please continue to refer to Figure 1 and Figure 2 As shown, in an exemplary embodiment of the present application, the cloud module 300 further includes an incremental learning unit, which is used to optimize and update the preset fault prediction model according to the user feedback information to achieve version iteration of the preset fault prediction model. It should be noted that after the preset fault prediction model is updated, it is necessary to evaluate the performance of the updated preset fault prediction model. When it is detected that the performance of the iterated new model decreases, an automatic rollback operation is performed to restore to the model version with the optimal performance. It is worth noting that in this embodiment, the preset fault prediction model is composed of multiple sub-models to capture fault characteristics at different levels respectively, and an adaptive weight adjustment mechanism is used to enhance the generalization ability of the preset model.
[0044] Specifically, please refer to Figure 3 As shown, incremental learning and continuous optimization specifically cover feedback data annotation, model incremental update, model performance evaluation, and version management and rollback. Among them, the feedback data annotation includes classifying and annotating the collected feedback data to form an incremental training data set. The model incremental update includes using an incremental learning framework to update the model online without full retraining. The model performance evaluation includes multi-dimensional performance evaluation of the updated new model. The version management and rollback include recording model version information and performing an automatic rollback operation when necessary (such as when a performance decrease, prediction error, or other abnormal conditions are detected) to restore to the model version with the optimal performance.
[0045] Please continue to refer to Figure 1 and Figure 2 As shown, in an exemplary embodiment of the present application, the human-computer interaction module 100 can provide a customized information display interface according to different user roles.
[0046] Specifically, the human-computer interaction module 100 covers prediction result display, user operation interface, feedback collection, and preliminary feedback processing. Among them, the prediction result display part is responsible for pushing the fault prediction results and maintenance suggestions output by the cloud module 300 to the corresponding human-computer interaction terminals. The user operation interface provides a customized interface according to different user roles (such as car owners, maintenance technicians, or administrators, etc.). The feedback collection link collects the feedback of users on the fault prediction results and the actual maintenance results through various terminals. The preliminary feedback processing uses natural language processing technology to structurally process the text feedback of users for subsequent data analysis and processing.
[0047] It should be noted that the vehicle fault prediction and location system further includes cross - model knowledge mapping, transfer learning application, system function extension, and continuous integration and deployment. Among them, the cross - model knowledge mapping refers to establishing the mapping relationship between sensors and parameters among different vehicle models, enabling the system to understand and process data from different vehicle models. The transfer learning application refers to migrating the existing model knowledge to new vehicle models to reduce the time and data volume required for training new vehicle models. Specifically, when a new vehicle model is introduced, due to the lack of sufficient historical data, traditional machine learning methods may face the cold start problem (i.e., the model performs poorly in the initial stage). Transfer learning can utilize the data and knowledge of existing vehicle models to accelerate the training and optimization of new vehicle models. The system function extension refers to gradually expanding the system function modules based on the microservice architecture to meet the changing business requirements. The continuous integration and deployment refers to achieving continuous iteration and smooth upgrade of system components to ensure the stability and availability of the system.
[0048] In summary, a vehicle fault prediction and location system provided by the present invention includes a vehicle edge node module 200, a cloud module 300, and a human-machine interaction module 100. The vehicle edge node module 200 is used to collect and process in-vehicle sensor data, and upload the processed feature data to the cloud module. The cloud module 300 is used to store the feature data, and analyze and process the feature data using a preset fault prediction model to locate the fault source and output a fault prediction result. The human-machine interaction module 100 is used to display the fault prediction result output by the cloud module, and upload user feedback information to the cloud module. The vehicle fault prediction and location system deploys lightweight edge computing nodes at the vehicle end, which are responsible for processing emergency fault detection, while the cloud is responsible for the training and in-depth analysis of complex models. Through an intelligent task allocation mechanism, the system realizes the combination of millisecond-level response and high-precision prediction. At the same time, a complete data processing flow is designed, including data screening, priority sorting, incremental training, model evaluation, and automatic rollback mechanism, to ensure that the model can be continuously optimized safely and efficiently. The system uses multiple models such as XGBoost, LSTM, and autoencoders to capture fault features at different levels, and improves the generalization ability of the system through an adaptive weight adjustment mechanism. In addition, the edge node has an independent basic fault diagnosis function to ensure that core services can still be provided when the network is interrupted, and data and models are automatically synchronized after the network is restored. Through transfer learning technology, knowledge sharing between different vehicle models is realized, enabling new vehicle models to draw on the experience of existing vehicle models and accelerating model convergence. The system also combines domain expert knowledge to construct a fault knowledge graph, provides cause analysis for fault prediction, and indicates the reliability of the prediction through a confidence evaluation mechanism. In terms of data processing, the edge node is responsible for data preprocessing and desensitization, and only transmits necessary features and statistical information, effectively balancing the utilization of data value and privacy protection. The system adopts a microservice design and a Docker+Kubernetes deployment method to achieve modular management, flexible expansion, and version control. Based on the fault prediction result, the system provides detailed repair suggestions and illustrated operation guides, reducing the dependence on senior technicians and improving the repair efficiency. Finally, a complete closed-loop of user feedback-system optimization-effect verification is established, and the user feedback is automatically analyzed in combination with a large language model to continuously improve the accuracy of the system.
[0049] The vehicle fault prediction and location system not only thoroughly solves the key technical problems faced by existing automotive fault prediction systems, but also achieves significant improvements in accuracy, real-time performance, adaptability, and reliability. At the same time, it greatly reduces the maintenance cost, providing a comprehensive health management solution for modern intelligent vehicles.
[0050] The above embodiments are only illustrative of the principles and effects of the present invention and are not intended to limit the present invention. Any person familiar with this technology can modify or change the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made by those with ordinary knowledge in the technical field without departing from the spirit and technical ideas disclosed by the present invention should still be covered by the claims of the present invention.
Claims
1. A vehicle fault prediction and location system, characterized in that, It includes: a vehicle edge node module, a cloud module, and a human-machine interaction module; The vehicle edge node module is used to collect and process in-vehicle sensor data, and upload the processed feature data to the cloud module; The cloud module is used to store the feature data, and analyze and process the feature data using a preset fault prediction model to locate the fault source and output a fault prediction result; The human-machine interaction module is used to display the fault prediction result output by the cloud module, and upload the user feedback information to the cloud module.
2. The vehicle fault prediction and location system according to claim 1, characterized in that, The vehicle edge node module includes an in-vehicle sensor network unit, an edge computing node unit, and an edge cloud communication unit; The in-vehicle sensor network unit is used to collect the working parameters of the vehicle's key systems in real time to obtain the in-vehicle sensor data; The edge computing node unit is used to preprocess and extract features from the in-vehicle sensor data to obtain the feature data; The edge cloud communication unit is used to upload the feature data to the cloud module for storage.
3. The vehicle fault prediction and location system according to claim 1, wherein The edge node module further includes an edge fault detection unit, which is used to detect emergency faults and send the detection results to the human-machine interaction module to achieve emergency warning.
4. The vehicle fault prediction and positioning system according to claim 1, characterized in that, The cloud module further includes a fault knowledge graph unit, which is used to explain the fault cause and development mechanism of the fault prediction result.
5. The vehicle fault prediction and location system according to claim 1, wherein, The cloud module further includes a user feedback unit, which is used to receive the user feedback information and extract and annotate the fault description from the user feedback information using natural language processing technology.
6. The vehicle fault prediction and location system according to claim 5, wherein, The user feedback information includes the user's feedback on the fault prediction result and the actual repair result.
7. The vehicle fault prediction and positioning system according to claim 1, characterized in that, The cloud module further includes an incremental learning unit, which is used to optimize and update the preset fault prediction model according to the user feedback information to achieve version iteration of the preset fault prediction model.
8. The vehicle fault prediction and location system according to claim 7, wherein Perform performance evaluation on the updated preset fault prediction model. When it is detected that the performance of the new model after iteration decreases, the model version with the optimal performance is restored through an automatic rollback operation.
9. The vehicle fault prediction and positioning system according to claim 1, characterized in that The preset fault prediction model is composed of multiple sub-models to capture fault features at different levels respectively, and enhance the generalization ability of the preset fault prediction model through an adaptive weight adjustment mechanism.
10. The vehicle fault prediction and location system according to claim 1, wherein, The human-machine interaction module can provide a customized information display interface according to different user roles.