EUVIMEDEUVIMED entdecken
Crossref

An Interpretable Multiplication Kolmogorov--Arnold Network for Small-Sample Intelligent Equipment State Evaluation

Qihang Wu, Zijia Xu, Miao Zheng, Xiaoxi Ding

IEEE Sensors Journal · 2026-08 · 26 (16) · 24449-24460

DOI: 10.1109/jsen.2026.3707745 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Sensors Journal, 2026-08. Verfasst von Qihang Wu und 3 weiteren Autor:innen. Die Arbeit nennt 24 Literaturangaben. Gefördert von: National Key Research and Development Program of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Focusing on the algorithm failure due to the scarcity of fault data and the ambiguous mapping between model parameters and fault features, this study proposes an interpretable multiplication Kolmogorov–Arnold network (MKAN). It is composed of a simple two-layer structure, including the multiplication Kolmogorov–Arnold layer (MKAL) and the efficient Kolmogorov–Arnold layer (EKAL). In the MKAL, inspired by the multiple modulation mechanisms of fault signals, a feature selection matrix is used to achieve the adaptive initial screening of fault features and establish the correlation between multiscale distribution fault features. Subsequently, a series of learnable global Wiener activation functions (WAFs)—which are prototypes of Wiener filtering kernels—are used to replace traditional activation functions to target enhancing the initially screened features. In the EKAL, inspired by the Kolmogorov–Arnold representation theorem, it is implemented using a two-layer Kolmogorov–Arnold network (KAN) architecture, aiming to achieve an efficient and mathematically clear expression of the fault features extracted by MKAL, thereby guiding the state identification of MKAN. The state identification performance, interpretability, robustness, and generalization of MKAN under small-sample conditions were verified through comparisons with seven advanced models. Based on these, MKAN achieves interpretable fault feature extraction and efficient fault feature representation, which is beneficial for small-sample equipment state evaluation.

Förderung

  • National Key Research and Development Program of China

Zitieren

APA

Wu, Q., Xu, Z., Zheng, M., & Ding, X. (2026). An Interpretable Multiplication Kolmogorov--Arnold Network for Small-Sample Intelligent Equipment State Evaluation. IEEE Sensors Journal, 26(16), 24449-24460. https://doi.org/10.1109/jsen.2026.3707745

Vancouver

Wu Q, Xu Z, Zheng M, Ding X. An Interpretable Multiplication Kolmogorov--Arnold Network for Small-Sample Intelligent Equipment State Evaluation. IEEE Sensors Journal. 2026;26(16):24449-24460. doi:10.1109/jsen.2026.3707745

BibTeX
@article{Wu2026,
  author = {Wu, Qihang and Xu, Zijia and Zheng, Miao and Ding, Xiaoxi},
  title = {An Interpretable Multiplication Kolmogorov--Arnold Network for Small-Sample Intelligent Equipment State Evaluation},
  journal = {IEEE Sensors Journal},
  year = {2026},
  volume = {26},
  number = {16},
  pages = {24449--24460},
  doi = {10.1109/jsen.2026.3707745}
}
Kostenlos starten →