Remaining Useful Life Prediction of Equipment With Multi-Failure-Mode Degradation Based on Multi-Scale Similarity Matching
Journal of Computational Design and Engineering · 11.11.2026
DOI: 10.1093/jcde/qwag096 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Computational Design and Engineering, online erschienen am 11.11.2026. Verfasst von Qi Zhang und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).
Abstract (Originalsprache)
Accurate remaining useful life (RUL) prediction is essential for equipment reliability management and maintenance decision-making. However, in the presence of multiple fault modes and varying operating conditions, existing methods often struggle to achieve both reliable fault-mode identification and accurate, robust RUL prediction. Therefore, this paper proposes a fault-mode-aware RUL prediction framework based on differentiated multi-scale similarity matching with adaptive focal-scale activation. First, an autoencoder is employed to construct a degradation-sensitive and robust health index from healthy operating data. Subsequently, Temporal and Contextual Contrasting is used for unsupervised fault mode identification, ensuring that only same-mode degradation trajectories are retained as references. Finally, a differentiated multi-scale matching mechanism is developed using mean-residual comparison, dynamic time warping, and spherically weighted dynamic time warping at the short-term, full-life-cycle, and focal scales, respectively. Additionally, a sliding-threshold strategy is employed to facilitate focal-scale matching and accurately capture critical degradation stages. The experimental results indicate that the proposed method is superior to six advanced benchmark methods on the C-MAPSS dataset. Specifically, compared with Bidirectional Long Short-Term Memory Encoder-Decoder, the proposed method reduces the root-mean-square error by 30.1%, 26.4%, 41.9%, and 20.3% across the four C-MAPSS subsets, with corresponding Score reductions of 46.7%, 50.5%, 54.6%, and 30.3%. Furthermore, the proposed method exhibits high prediction accuracy, strong robustness, and excellent generalizability, demonstrating its potential for predictive maintenance of complex industrial equipment.
Zitieren
APA
Zhang, Q., Cheng, X., Tang, S., Pei, H., & Zheng, J. (2026). Remaining Useful Life Prediction of Equipment With Multi-Failure-Mode Degradation Based on Multi-Scale Similarity Matching. Journal of Computational Design and Engineering. https://doi.org/10.1093/jcde/qwag096
Vancouver
Zhang Q, Cheng X, Tang S, Pei H, Zheng J. Remaining Useful Life Prediction of Equipment With Multi-Failure-Mode Degradation Based on Multi-Scale Similarity Matching. Journal of Computational Design and Engineering. 2026. doi:10.1093/jcde/qwag096
BibTeX
@article{Zhang2026,
author = {Zhang, Qi and Cheng, Xinyi and Tang, Shengqi and Pei, Hong and Zheng, Jianfei},
title = {Remaining Useful Life Prediction of Equipment With Multi-Failure-Mode Degradation Based on Multi-Scale Similarity Matching},
journal = {Journal of Computational Design and Engineering},
year = {2026},
doi = {10.1093/jcde/qwag096}
}
EUVIMED