Scholar: Rola El-Nimri
When: Friday, October 31st, 1:00pm
Where: KH A549
Zoom: https://unl.zoom.us/j/92030314867
Title: Truss Bridge Damage Detection and Prediction Using Enhanced Transfer Learning and Proper Orthogonal Decomposition
Abstract: Bridges are critical transportation infrastructure components. Ensuring their safety and longevity requires gathering reliable information. Structural Health Monitoring (SHM) is a technology that, in theory, allows for continuous collection of smart, reliable, and continuous health information. This research presents a transfer learning framework that enhances SHM performance under varying structural and environmental conditions. Proper Orthogonal Modes (POMs) were employed as damage-sensitive features and enabled knowledge transfer between data types and sets. A developed Singular Value Decomposition–based Subspace Alignment Domain Adaptation (SVD-SADA) method aligned modal features between source and target domains, substantially improving damage prediction accuracy. A new, physics-based Modal Orthogonality Loss Index (MOLI) was introduced to efficiently quantify damage severity. The integration of POM-based feature alignment and MOLI demonstrated accurate, generalizable, and computationally efficient damage detection for different bridge, enabling development of transferable and data-efficient SHM frameworks for real-world applications.