Unsupervised learning for structural health monitoring: A multi-station GNSS-RTK approach for operational state detection in cable-stayed bridges
Transport and Communications Science Journal · 04.04.2026 · 77 (9) · 1259-1274
DOI: 10.47869/tcsj.77.9.10 ↗Kurz gefasst
Fachartikel in der Zeitschrift Transport and Communications Science Journal, online erschienen am 04.04.2026. Verfasst von Cong Tran Duc und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Monitoring the operational condition of cable-stayed bridges is essential for ensuring structural safety and long-term serviceability. GNSS-RTK technology has emerged as an effective solution for continuous, high-precision displacement measurement in structural health monitoring (SHM). However, high-frequency GNSS data are often noisy and unlabeled, posing significant challenges for traditional analysis methods. This study addresses the problem of identifying structural operational states from unlabeled GNSS-RTK data using an unsupervised learning framework. Displacement data collected from two GNSS stations installed symmetrically on a cable-stayed bridge were segmented into hourly windows and preprocessed using Hampel filtering and Kalman smoothing. Feature vectors were then standardized and reduced in dimensionality using Principal Component Analysis, while t-SNE was applied for visualization. Clustering algorithms, including K-Means, hierarchical agglomerative clustering, Gaussian mixture models, and DBSCAN, were implemented and evaluated using multiple internal validation indices and stability metrics. The results reveal a consistent two-state structure in the horizontal displacement component, with K-Means achieving the best performance (Silhouette score up to 0.7006) and perfect stability (ARI = 1.0). Importantly, the clustering results are highly consistent across both stations, indicating strong spatial reliability. These findings demonstrate that unsupervised learning can effectively extract meaningful operational states from GNSS-RTK data, providing a robust and scalable approach for structural health monitoring of large-scale bridges.
Zitieren
APA
Duc, C. T., Khanh, G. L., & Lan, H. H. T. (2026). Unsupervised learning for structural health monitoring: A multi-station GNSS-RTK approach for operational state detection in cable-stayed bridges. Transport and Communications Science Journal, 77(9), 1259-1274. https://doi.org/10.47869/tcsj.77.9.10
Vancouver
Duc CT, Khanh GL, Lan HHT. Unsupervised learning for structural health monitoring: A multi-station GNSS-RTK approach for operational state detection in cable-stayed bridges. Transport and Communications Science Journal. 2026;77(9):1259-1274. doi:10.47869/tcsj.77.9.10
BibTeX
@article{Duc2026,
author = {Duc, Cong Tran and Khanh, Giang Le and Lan, Huong Ho Thi},
title = {Unsupervised learning for structural health monitoring: A multi-station GNSS-RTK approach for operational state detection in cable-stayed bridges},
journal = {Transport and Communications Science Journal},
year = {2026},
volume = {77},
number = {9},
pages = {1259--1274},
doi = {10.47869/tcsj.77.9.10}
}
EUVIMED