Sea state uncertainty-aware monitoring of underwater mooring systems using domain-adapted deep learning techniques
Document Type
Conference Proceeding
Publication Date
5-13-2025
Department
Department of Mechanical and Aerospace Engineering
Abstract
Underwater mooring systems are essential for marine infrastructure safety but face stiffness reduction and potential failure due to long-term environmental loads like waves and currents, requiring timely and accurate health monitoring. Data-driven deep learning techniques, which identify mooring system health from dynamic responses, offer more efficient and cost-effective solutions compared to traditional methods. However, the high complexity and uncertainty of the sea state pose challenges for effective monitoring tasks utilizing general deep learning models. While ocean wave spectra are often considered invariant over relatively short time scales, the exact time series of free sea surface elevation remains inherently unpredictable, which interferes with the health-related features to degrade the monitoring reliability and accuracy. To address this challenge, this study integrates domain adaptation techniques into deep learning models to mitigate distribution discrepancies and enhance the generalization of these models. Case studies demonstrate that the model exhibits significantly improved generalization ability, showing great potential for managing mooring system monitoring in practical applications.
Publication Title
Proceedings Health Monitoring of Structural and Biological Systems XIX
Recommended Citation
Liu, Y.,
Zou, S.,
Ye, X.,
&
Zhou, K.
(2025).
Sea state uncertainty-aware monitoring of underwater mooring systems using domain-adapted deep learning techniques.
Proceedings Health Monitoring of Structural and Biological Systems XIX,
13437.
http://doi.org/10.1117/12.3045956
Retrieved from: https://digitalcommons.mtu.edu/michigantech-p2/1918