Robust Prediction of Remaining Useful Life Using Physics-Guided Multi-Fidelity Approaches
IEEE Access · 2026 · 14 · 95137-95151
DOI: 10.1109/access.2026.3704966 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Yoojeong Noh und 1 weiteren Autor:innen. Die Arbeit nennt 38 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Long-Term Overseas Dispatch of PNU’s Tenure-Track Faculty, in 2023.
Abstract (Originalsprache)
Data-driven approaches are known to have difficulty in predicting system’s response in the extrapolation region where no data are available. In this paper, we present physics-guided multi-fidelity approaches that can integrate low-fidelity physical information with data-driven approaches to enhance prediction capability in the unsampled region. The proposed multi-fidelity model is composed of a low-fidelity model with tunable parameters and the discrepancy function that represents an additive correlation surrogate. The discrepancy function is modeled using either polynomial response surfaces or neural networks. The proposed multi-fidelity models are applied to predicting the remaining useful life of cooling fan bearing degradation. Numerical examples show that low-fidelity models and neural networks alone cannot accurately predict the remaining useful life, while the proposed multi-fidelity model performs better. It turned out that the accuracy of low-fidelity model does not affect the accuracy of multi-fidelity model much because the bias error can easily be compensated by discrepancy. Instead, the model parameters of low-fidelity model were calibrated to simplify the discrepancy function, thereby improving extrapolation robustness.
Förderung
- Long-Term Overseas Dispatch of PNU’s Tenure-Track Faculty, in 2023
Zitieren
APA
Noh, Y., & Kim, N. H. (2026). Robust Prediction of Remaining Useful Life Using Physics-Guided Multi-Fidelity Approaches. IEEE Access, 14, 95137-95151. https://doi.org/10.1109/access.2026.3704966
Vancouver
Noh Y, Kim NH. Robust Prediction of Remaining Useful Life Using Physics-Guided Multi-Fidelity Approaches. IEEE Access. 2026;14:95137-95151. doi:10.1109/access.2026.3704966
BibTeX
@article{Noh2026,
author = {Noh, Yoojeong and Kim, Nam Ho},
title = {Robust Prediction of Remaining Useful Life Using Physics-Guided Multi-Fidelity Approaches},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {95137--95151},
doi = {10.1109/access.2026.3704966}
}
EUVIMED