Research on Vehicle State Segmentation and Failure Prediction Based on Big Data
2022-01-0254
03/29/2022
- Event
- Content
- Vehicle failure prediction technology is an important part of PHM technology, which is of great significance to the safety maintenance of vehicles and to improve driving safety. This article relies on the research group's production-university-research project, based on the vehicle operating data collected by the telematics system on-board terminal (T-box), and conducts research on the state of vehicle failure. First of all, this paper conducts statistical analysis on vehicle historical fault data from T-box, specifically performs preprocessing such as cleaning, integration, and protocol, divides the data set, and establishes an association matrix between fault categories and specific signal parameters that cause such faults. Three angles of time, frequency, and proximity are selected to construct a vehicle state subdivision system. In addition, the K-means algorithm is used to classify different vehicle categories from the perspective of vehicle value, and to establish labeling information for vehicles in different states. Furthermore, this article uses text mining technology to extract features of historical fault texts, establishes high-frequency fault types through cluster analysis, and combines tag cloud technology to visualize short texts of key fault types. Moreover, taking engine faults as an example, using gray correlation analysis method to establish key fault characteristic parameters, this article builds fault prediction model based on self-organizing mapping network theory (SOM), learns the fault data of the vehicle, and obtains the characteristic differences between different states , completes the prediction or evaluation based on such differences. Finally, some test data is selected to verify the prediction accuracy of the model.
- Citation
- Zheng, L., "Research on Vehicle State Segmentation and Failure Prediction Based on Big Data," SAE Technical Paper 2022-01-0254, 2022, .