RUL Prediction with Encoding and Spatial-Temporal Deep Neural Networks

The objective of this paper is to design and develop an approach to estimate the Remaining Useful Life (RUL) of an industrial equipment evolving in a Cyber-Physical System (CPS). To do so, this work aims to predict failures and malfunctions of an industrial equipment, as well as evaluating all the main underlying causes. The system will also identify the actions to be taken in order to maintain the system at a certain level of performance and guarantee that it will operate in the most efficient way, which implies a study of all aspects: reliability, availability and maintenance. Existing works are often limited to the estimation phase of the RUL, with only a few case studies (such as ball bearings and Proof Of Concept(PoC)) available and a significant lack of data. The innovation of this research work will be to design and develop diagnostic and prognostic approaches based on Deep Learning (DL)(e.g. Boltzmann Machine, Recurrent Networks and Echo State Network (ESN)) for the maintenance and health management of industrial equipment evolving in a CPS.