Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
Transfer Learning Performance for Structural Health Monitoring through Boundary Condition Investigation
Submission Author:
Estênio Fuzaro , SP , Brazil
Co-Authors:
Estênio Fuzaro, Samuel da Silva
Presenter: Estênio Fuzaro
doi://10.26678/ABCM.COBEM2023.COB2023-0165
Abstract
This study explores the impact of boundary conditions on the efficacy of transfer learning in Structural Health Monitoring (SHM). By experimenting with beam structures subjected to a range of boundary conditions, we analyze how these variations modulate the success of transfer learning. A key focus of our investigation is the role of similarity analysis, especially when viewed through the prism of Cosine Similarity. Our hypothesis asserts that a nuanced understanding of similarity analysis can substantially bolster the optimization of transfer learning outcomes. The primary aim is to elucidate the intrinsic relationship between heightened similarity indices of source and target features and the ensuing improvement in transfer learning. Such insights hold promise in significantly enhancing the robustness and precision of structural damage detection systems.
Keywords
Similarity Analysis, Domain Adaptation, transfer learning, Twin Structures, Damage detection and structural health monitoring
