A multi-head attention-based deep learning method for condition monitoring in rotating machinery diagnostics using RVMD-FNDNLP feature preprocessing
Measurement Science and Technology · 13.11.2026 · 37 (45) · 455005
DOI: 10.1088/1361-6501/aebab5 ↗Kurz gefasst
Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 11.11.2026. Verfasst von Chen Zhang und 3 weiteren Autor:innen. Die Arbeit nennt 27 Literaturangaben. Gefördert von: Thousands of Young and Middle-Aged Backbone Teachers Training Program for Guangxi Higher Education, National Natural Science Foundation of China.
Abstract (Originalsprache)
Rotating machinery condition monitoring is essential for reliable fault detection and maintenance decision-making in industrial systems. However, nonstationary bearing vibration measurements often produce statistical feature representations with nonlinear coupling, redundancy, and class overlap, thereby limiting diagnostic reliability. In addition, conventional feature extraction and dimensionality reduction methods often suffer from feature redundancy and limited class separability. To address these issues, this paper proposes a bearing condition-monitoring method based on regularized variational mode decomposition (RVMD) and farthest-nearest distance-based neighborhood and locality projections (FNDNLP). By incorporating a regularization strategy into variational mode decomposition, RVMD improves decomposition efficiency and enhances fault-relevant modal information. Multiple statistical features are then extracted from the selected informative modes to construct a high-dimensional monitoring feature set. Subsequently, FNDNLP learns a supervised low-dimensional projection from the training data to reduce redundant and coupled information while improving intraclass compactness and interclass separability. Finally, a temporal convolutional network with multi-head attention (TCN-MHA) models dependencies in the projected representation. The framework is evaluated on the Case Western Reserve University (CWRU) and HUST bearing platforms using complete-pipeline comparisons, controlled classifier evaluations, reduced-dimension sensitivity analysis, and component-wise ablation. It achieves mean test accuracies of 98.89± 0.03% and 98.10± 0.42% on CWRU and HUST, respectively. The results support the effectiveness of the integrated framework under the evaluated datasets and conditions.
Förderung
- Thousands of Young and Middle-Aged Backbone Teachers Training Program for Guangxi Higher Education
- National Natural Science Foundation of China
Zitieren
APA
Zhang, C., Shen, J., Yang, L., & Zhang, Q. (2026). A multi-head attention-based deep learning method for condition monitoring in rotating machinery diagnostics using RVMD-FNDNLP feature preprocessing. Measurement Science and Technology, 37(45), 455005. https://doi.org/10.1088/1361-6501/aebab5
Vancouver
Zhang C, Shen J, Yang L, Zhang Q. A multi-head attention-based deep learning method for condition monitoring in rotating machinery diagnostics using RVMD-FNDNLP feature preprocessing. Measurement Science and Technology. 2026;37(45):455005. doi:10.1088/1361-6501/aebab5
BibTeX
@article{Zhang2026,
author = {Zhang, Chen and Shen, Jiqiang and Yang, Linfeng and Zhang, Qi},
title = {A multi-head attention-based deep learning method for condition monitoring in rotating machinery diagnostics using RVMD-FNDNLP feature preprocessing},
journal = {Measurement Science and Technology},
year = {2026},
volume = {37},
number = {45},
pages = {455005},
doi = {10.1088/1361-6501/aebab5}
}
EUVIMED