BROWSE
Detail
Conference Paper
Application of Transfer Learning to Enhance Scalability of Virtual Sensors for Bridge Health Monitoring
Cited - time in 
Share 
Share
- Authors
- SeungSae Joo, WonKyu Choi, Ji-Hoon Bae
- Issue Date
- 2025-08
- Citation
- ADINTECH 2025, pp.1-6
- Publisher
- 한국정보기술학회
- Language
- English
- Type
- Conference Paper
- Abstract
- Recent advancements in IoT and sensor technologies have significantly improved the capabilities of structural health monitoring (SHM) systems. Among critical components of transportation infrastructure, bridges are increasingly being equipped with various physical sensors to assess structural safety and performance in real time. However, physical sensors often suffer from practical limitations, including power consumption, durability, susceptibility to noise, and signal drift. To overcome these challenges, virtual sensors have emerged as a promising alternative, enabling the estimation of physical responses without the need for extensive hardware deployment. Despite their potential, most existing virtual sensor models are tailored to single structures or specific environments, making it difficult to generalize them across different conditions or locations. To address this limitation, this study proposes a transfer learning-based virtual sensor model aimed at enhancing generalizability and predictive accuracy. The proposed approach was validated using data collected from two bridges located in Gunsan and Busan, South Korea, with the objective of estimating acceleration responses. The model achieved a mean absolute error of approximately 0.0984, showing a 41.32% improvement over models without transfer learning. These results highlight the effectiveness of transfer learning in improving the robustness and applicability of virtual sensor models for diverse infrastructures.
- Keyword
- Deep Learning, Virtual Sensor, Transfer Learning, Cycle-Consistent Adversarial Networks.
- KSP Keywords
- Bridge Health Monitoring, Critical components, Data collected, Different conditions, Learning-based, Mean Absolute Error, Physical sensors, Power Consumption, Sensor Technology, Sensor model, South Korea