Anomaly detection in sensor data provided by combine harvesters

Modern combine harvesters often stay beyond their theoretic optimal performance during harvesting operations. Explanations and remedies for this reduced efficiency are difficult to find, as the actual performance is influenced by a variety of different parameters. This paper presents a continuous analysis of machine-provided data streams in order to assess potential reasons (i.e. anomalies) and to identify suggestions for optimization. To this end, new sensor data-based machine learning algorithms are being developed, applied and evaluated.