Ensemble of optimized echo state networks for remaining useful life prediction

Autor: Piero Baraldi, Enrico Zio, Marco Rigamonti, Scott Poll, Indranil Roychoudhury, Kai Goebel
Přispěvatelé: Chaire Sciences des Systèmes et Défis Energétiques EDF/ECP/Supélec (SSEC), EDF R&D (EDF R&D), EDF (EDF)-EDF (EDF)-CentraleSupélec-SUPELEC-Ecole Centrale Paris, Laboratoire Génie Industriel - EA 2606 (LGI), CentraleSupélec, Politecnico di Milano [Milan] (POLIMI), Ecole Centrale Paris-Ecole Supérieure d'Electricité - SUPELEC (FRANCE)-CentraleSupélec-EDF R&D (EDF R&D), EDF (EDF)-EDF (EDF), NASA Ames Research Center (ARC)
Jazyk: angličtina
Rok vydání: 2018
Předmět:
0209 industrial biotechnology
Computer science
Cognitive Neuroscience
02 engineering and technology
Machine learning
computer.software_genre
Prediction Intervals
[SPI]Engineering Sciences [physics]
020901 industrial engineering & automation
Artificial Intelligence
Prediction uncertainty
0202 electrical engineering
electronic engineering
information engineering

Mean variance
Ensembles
ComputingMilieux_MISCELLANEOUS
Echo state networks
business.industry
Echo (computing)
Prediction interval
Differential Evolution
Computer Science Applications1707 Computer Vision and Pattern Recognition
Computer Science Applications
Recurrent neural network
Recurrent neural networks
Differential evolution
020201 artificial intelligence & image processing
Noise (video)
State (computer science)
Artificial intelligence
business
computer
Zdroj: Neurocomputing
Neurocomputing, Elsevier, 2018, 281, pp.121-138. ⟨10.1016/j.neucom.2017.11.062⟩
ISSN: 0925-2312
DOI: 10.1016/j.neucom.2017.11.062⟩
Popis: International audience; The use of Echo State Networks (ESNs) for the prediction of the Remaining Useful Life (RUL) of industrial components, i.e. the time left before the equipment will stop fulfilling its functions, is attractive because of their capability of handling the system dynamic behavior, the measurement noise, and the stochasticity of the degradation process. In particular, in this paper we originally resort to an ensemble of ESNs, for enhancing the performances of individual ESNs and providing also an estimation of the uncertainty affecting the RUL prediction. The main methodological novelties in our use of ESNs for RUL prediction are: i) the use of the individual ESN memory capacity within the dynamic procedure for aggregating of the ESNs outcomes; ii) the use of an additional ESN for estimating the RUL uncertainty, within the Mean Variance Estimation (MVE) approach. With these novelties, the developed approach outperforms a static ensemble and a standard MVE approach for uncertainty estimation in tests performed on a synthetic and two industrial datasets.
Databáze: OpenAIRE