Machine Learning at the Edge: a few applicative cases of Novelty Detection on IIoT gateways

Giovanni Burresi, Antonio Rizzo, Martino Lorusso, Sara Ermini, Alessandro Rossi, Francesco Cariaggi

Abstract: In the paper we present the design and development of the Machine Learning (ML) modules for two case studies. In both cases we developed a ML model to learn the system's normal behavior so to identify whichever abnormal condition may arise. Such a framework is usually referred to as Anomaly Detection (also known as Fault Detection or Novelty Detection). Our models succeeded at identifying the injected anomalies. In addition, no anomalies were observed when the model was fed with normal data. The results are discussed considering the trade-off between type of sensors, learning algorithm, training effort, computational demands.