CAG-BRB: An open-set belief rule base method for health state assessment via joint unknownness discrimination and reliable clustering discovery
Journal of King Saud University Computer and Information Sciences · 2026-11 · 38 (9)
DOI: 10.1007/s44443-026-01382-2 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of King Saud University Computer and Information Sciences, online erschienen am 06.11.2026. Verfasst von Shuyin Nan und 5 weiteren Autor:innen. Die Arbeit nennt 45 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Complex systems are susceptible to operating-condition variations, component degradation, and external disturbances during long-term operation, which may lead to new faults or abnormal states that have not been predefined. Traditional health state assessment methods are usually developed based on a fixed set of known states. When unknown states occur, they may be misclassified as known states, thereby reducing the reliability of assessment results. The Belief Rule Base (BRB) can integrate expert knowledge and sample data, and has strong capabilities in uncertainty representation and interpretable reasoning. However, traditional BRB methods still suffer from insufficient unknown-state rejection capability and limited ability to discover unknown structures. To address these issues, this paper proposes an open-set BRB method integrating CAG-based joint unknownness discrimination and reliable clustering discovery, denoted as CAG-BRB. First, a BRB health state assessment model is constructed using known-state samples, and an Elite-guided Adaptive Crested Porcupine Optimizer (EA-CPO) is proposed to optimize model parameters, thereby improving parameter search accuracy and reasoning performance. Second, a CAG-based joint unknownness discrimination mechanism is designed using output confidence, rule activation concentration, and known-class geometric range, enabling unified discrimination between known-state identification and unknown-state rejection. Furthermore, for samples rejected as unknown states, input-space features and BRB rule activation vectors are fused to construct a joint representation, and reliable clustering discovery is performed to discover potential unknown fault structures. Finally, experiments are conducted on a new energy vehicle fault diagnosis dataset and a milling-machine operating-state dataset. The results show that, on the new energy vehicle dataset, the overall accuracy, known-class recognition accuracy, unknown-class rejection rate, and Macro-F1 of CAG-BRB reach 0.9415, 0.9732, 0.8854, and 0.9396, respectively; on the milling-machine dataset, the corresponding values reach 0.9636, 0.9640, 0.9626, and 0.9702, respectively. These results indicate that the proposed method can effectively reject unknown states while maintaining known-state recognition performance, and can reveal latent structures within rejected samples.
Zitieren
APA
Nan, S., Yu, Y., Ma, N., Yan, X., Duan, X., & Chen, M. (2026). CAG-BRB: An open-set belief rule base method for health state assessment via joint unknownness discrimination and reliable clustering discovery. Journal of King Saud University Computer and Information Sciences, 38(9). https://doi.org/10.1007/s44443-026-01382-2
Vancouver
Nan S, Yu Y, Ma N, Yan X, Duan X, Chen M. CAG-BRB: An open-set belief rule base method for health state assessment via joint unknownness discrimination and reliable clustering discovery. Journal of King Saud University Computer and Information Sciences. 2026;38(9). doi:10.1007/s44443-026-01382-2
BibTeX
@article{Nan2026,
author = {Nan, Shuyin and Yu, Yan and Ma, Ning and Yan, Xingchi and Duan, Xiping and Chen, Manlin},
title = {CAG-BRB: An open-set belief rule base method for health state assessment via joint unknownness discrimination and reliable clustering discovery},
journal = {Journal of King Saud University Computer and Information Sciences},
year = {2026},
volume = {38},
number = {9},
doi = {10.1007/s44443-026-01382-2}
}
EUVIMED