Proceedings of the

The 33rd European Safety and Reliability Conference (ESREL 2023)
3 – 8 September 2023, Southampton, UK

Learning Linearized Degradation of Health Indicators using Deep Koopman Operator Approach

Sergei Garmaeva and Olga Finkb

Laboratory of intelligent maintenance and operations systems, EPFL, Switzerland.


In this study, we showcase the successful application of the Deep Koopman operator approach to model the dynamics of industrial systems at multiple time scales. Specifically, we demonstrate its effectiveness in modeling the rapidly changing operating conditions as well as the slowly evolving degradation of the systems. Furthermore, we propose a novel approach inspired by Koopman theory to model the degradation of controlled dynamical systems. The proposed algorithm allows to predict the degradation trend with a limited number of full run-to-failure trajectories.

Keywords: Deep Koopman operator, Remaining useful lifetime, Predictive maintenance.

Download PDF