Vehicles are subject to effects that lead to a more or less rapid degradation of functions. On the one hand, aging effects lead to a gradual reduction in functional capability. On the other hand, unexpected defects in components or modules can lead to functional losses. Therefore faults in the powertrain or vehicle dynamics can cause hazards for the driver as well as uninvolved road participants. They must therefore be detected and intercepted as early as possible in order to maintain vehicle function even in critical situations until a safe operating mode can be established without any risk. This contribution presents an approach that can improve both the product development process and the operation of electrified vehicles by dynamically detecting abnormal situations in the powertrain and vehicle dynamics to infer the fault in the system. For this purpose, the vehicle is connected via 5G to an intelligent digital twin located in the Edge backend. The core component is the fault detection system, which is initially trained on the basis of synthetic data. The data sets used for the training are generated in CarMaker for real-world powertrain issues such as demagnetization and open-/short-switch failures based on detailed mathematical models of the powertrain and the vehicle dynamics. The reaction of defined failures is given by a static implemented look-up table. The evaluation of the prototype shows that when a failure occurs, after a detection time, it is identified with high accuracy and a reaction of the vehicle software minimizes the failure impact during operation.
Session:
MODELLING & AI
|
| 08:30 - 09:00