From Measurements to Anomalies: MLOps-Supported Anomaly Detection in Connected Vehicle Components

Due to the complexity of connected vehicle components, the automotive industry increasingly employs intelligent systems such as Machine Learning (ML) and Deep Learning (DL) algorithms. These algorithms help analyze large data sets more efficiently and quickly by identifying unknown anomalies. However, the complexity of these algorithms necessitates making them practical for users in engine development. The Machine Learning Operations (MLOps) paradigm provides a solution by aiming to reliably deploy and maintain machine learning models in production. Furthermore, MLOps facilitates the use of these algorithms by dividing their application into multiple components and abstracting them, enabling vehicle experts to perform complex analyses. Therefore, a semi-automated MLOps pipeline will be developed to support experienced vehicle experts in detecting anomalous events from time series measurement data. Based on this, the data will be compressed to a manageable level. Compression will extract essential information for anomaly detection, saving valuable analysis time for vehicle experts.