The development processes for electric vehicles are becoming increasingly demanding, partly due to the high complexity of the battery systems. Furthermore, the developers have to handle various usage requirements of the systems and diverse user behavior. Here, inductive development approaches can help developers deal with high complexities by identifying interrelationships. This work introduces a Data Mining Process Model (DMPM) that has been adapted for the battery system's highly technical domain to support the developers in their work. The model represents an interdisciplinary process requiring data science and battery system expertise. It systematically supports the quick usage of fleet data, the building of algorithms, and the investigation of them and their outputs. The first application studies are discussed, where fleet data was used to investigate the system behavior under actual operating conditions and thus discovered rare anomalies in the battery systems. Machine learning clustering algorithms can find these anomalies, which are investigated afterward. In this way, the developers are supported in their work, learning complex interrelationships and previously unknown knowledge about the battery system through the application of the DMPM. Consequently, the DMPM is a helpful tool that supports the systematic use of inductive development approaches in battery systems.
Session:
Poster
|