Long Short Term Memory-based anomaly detection applied to an industrial dosing pump

Performing predictive maintenance in the case of equipment failures is a relatively recent and difficult topic, in part due to the heterogeneity of industrial processes but also to the increasing amounts of information that can be gathered. We investigate the topic of anomaly detection using Long Short Term Memory for failures of an industrial dosing pump. We show that we are able to obtain accurate results corresponding to real failures leading the way to set up actions in order to avoid such failures in the future.