The science of smooth driving: a taxonomy for anomaly detection in endurance run time series data

The complexity of divers’ system components in vehicles causes errors that are difficult to identify. In order to prevent these errors the automotive manufacturers observe the system components with the help of fast analyses based on huge data volumes. In this regard, big data approaches support the development processes of experienced engineers within the driveability development for balanced engines. This paper classifies anomalies that occur in endurance run time series data in the field of anomaly detection. As both science and practice continue to evolve, anomalies and any terms associated with them are often mixed and misunderstood. As a result, anomaly detection might be inaccurate due to the choice of inappropriate big data methods. Therefore, it is crucial to establish a taxonomy and apply it to the driveability-relevant acceleration anomaly called hesitation from the Mercedes-Benz cars engine development environment to ensure a clear differentiation of such events and prevent misunderstandings.