Transfer Learning for SHM Using a Representative Bridge

Continuous monitoring of civil engineering structures has become increasingly important due to the growing impact of environmental actions and aging effects on structural perfor-mance. In this context, sensor-based Structural Health Monitoring (SHM) systems provide essential real-time information to support preventive maintenance and informed decision-making. This study focuses on the development of predictive models capable of estimating certain structural parameters using limited information from similar structures. In fact, the moni-tored structure is a prestressed concrete overpass representative of many similar infrastruc-tures built in Portugal during the same period, making it suitable as a reference structure for Transfer Learning applications. The experimental campaign included ambient vibration tests for modal identification, com-plemented by continuous monitoring using customized temperature and displacement sen-sors installed at critical sections. The collected data were decomposed into daily and annual trend components. The phase delay effects between temperature and displacement were iden-tified and corrected, and regression models were developed to describe the relationship be-tween thermal actions and structural response. The results demonstrate a strong correlation between temperature variations and structural displacements, enabling accurate estimation of displacement components using temperature data. The proposed methodology supports the idea of transfer of behavioral relationships between similar structures and contributes to the development of scalable and cost-effective monitoring strategies for large networks of roadway infrastructures.