A hybrid IF–LOF machine learning framework for detecting GNSS-RTK outliers in cable-stayed bridges displacement monitoring
Transport and Communications Science Journal · 07.03.2026 · 77 (5) · 672-686
DOI: 10.47869/tcsj.77.5.5 ↗Kurz gefasst
Fachartikel in der Zeitschrift Transport and Communications Science Journal, online erschienen am 07.03.2026. Verfasst von Giang Lê Khánh und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Global Navigation Satellite System Real-Time Kinematic (GNSS-RTK) technology is widely used in structural health monitoring (SHM) of long-span bridges because it enables continuous three-dimensional displacement measurements with millimetre-level accuracy. However, GNSS-RTK time-series data often contain abnormal observations caused by multipath effects, signal interruptions, and environmental disturbances. This study addresses the problem of detecting and removing outliers in GNSS-RTK displacement data used for cable-stayed bridge monitoring. A hybrid machine-learning framework combining Isolation Forest (IF) and Local Outlier Factor (LOF) is proposed. In the first stage, IF performs global screening to identify potentially anomalous observations. In the second stage, LOF evaluates the local density of these candidate samples to confirm true outliers through a dual-confirmation mechanism. The method was validated using a GNSS-RTK dataset collected from the Can Tho cable-stayed bridge in Vietnam, consisting of 12,615 displacement samples. The hybrid framework detected 177 outliers (1.40%), compared with 253 detected by IF and 379 detected by LOF. Furthermore, by restricting the LOF analysis to a subset of suspicious observations identified by IF, the proposed framework substantially reduced the computational burden of the local density evaluation stage while preserving most valid structural responses. The results demonstrate that the proposed IF–LOF framework provides an efficient preprocessing approach for improving the reliability of GNSS-based bridge monitoring systems
Zitieren
APA
Khánh, G. L., Lan, H. H. T., & Thuỳ, L. N. (2026). A hybrid IF–LOF machine learning framework for detecting GNSS-RTK outliers in cable-stayed bridges displacement monitoring. Transport and Communications Science Journal, 77(5), 672-686. https://doi.org/10.47869/tcsj.77.5.5
Vancouver
Khánh GL, Lan HHT, Thuỳ LN. A hybrid IF–LOF machine learning framework for detecting GNSS-RTK outliers in cable-stayed bridges displacement monitoring. Transport and Communications Science Journal. 2026;77(5):672-686. doi:10.47869/tcsj.77.5.5
BibTeX
@article{Khnh2026,
author = {Khánh, Giang Lê and Lan, Hương Hồ Thị and Thuỳ, Linh Nguyễn},
title = {A hybrid IF–LOF machine learning framework for detecting GNSS-RTK outliers in cable-stayed bridges displacement monitoring},
journal = {Transport and Communications Science Journal},
year = {2026},
volume = {77},
number = {5},
pages = {672--686},
doi = {10.47869/tcsj.77.5.5}
}
EUVIMED