Interaction models for remaining useful life estimation
The paper deals with the problem of controlling the state of industrial devicesaccording to the readings of their sensors. The current methods rely on oneapproach to feature extraction in which the prediction occurs. We proposed atechnique to build a scalable model that combines multiple different feature ex-tractor blocks. A new model based on sequential sensor space analysis achievesstate-of-the-art results on the C-MAPSS benchmark for equipment remaininguseful life estimation. The resulting model performance was validated includingthe prediction changes with scaling.