TY - GEN
T1 - Vision Transformers-Based Structural Health Monitoring System for Bridge Safety
AU - Choudari, Sudheer
AU - Gosavi, Shrikrishna
AU - Ramu, B.
AU - Hemalatha, J.
AU - Pawar, Saish
AU - Al Nidal Al Said, Nidal
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2025
Y1 - 2025
N2 - In this paper, a Vision Transformers-Based Structural Health Monitoring System to Bridge Safety, a Hybrid Vision Transformer (ViT) + CNN architecture, multi-sensor fusion, and Reinforcement Learning (RL) are used to improve the accuracy, efficiency, and scalability of bridges monitoring. The proposed system combines ViTs, used to extract global features, and CNNs, used to detect local anomalies, up to 95 % accuracy in identifying damage. The multi-sensor fusion adds strength to the model as it is less sensitive to the weather and lighting conditions. Moreover, a Reinforcement Learning agent will self-optimize maintenance schedules, minimizing unnecessary maintenance and increasing bridge lifespan. The system demonstrates the use of PyTorch, Hugging Face Transformers and OpenAI Baselines to provide fast processing and real time performance using edge cloud hybrid architecture. This paper has shown that such a system results in an improvement of the detection accuracy, as well as the efficiency of the operations, which are capable of providing a holistic solution to the long-term management of bridges safety.
AB - In this paper, a Vision Transformers-Based Structural Health Monitoring System to Bridge Safety, a Hybrid Vision Transformer (ViT) + CNN architecture, multi-sensor fusion, and Reinforcement Learning (RL) are used to improve the accuracy, efficiency, and scalability of bridges monitoring. The proposed system combines ViTs, used to extract global features, and CNNs, used to detect local anomalies, up to 95 % accuracy in identifying damage. The multi-sensor fusion adds strength to the model as it is less sensitive to the weather and lighting conditions. Moreover, a Reinforcement Learning agent will self-optimize maintenance schedules, minimizing unnecessary maintenance and increasing bridge lifespan. The system demonstrates the use of PyTorch, Hugging Face Transformers and OpenAI Baselines to provide fast processing and real time performance using edge cloud hybrid architecture. This paper has shown that such a system results in an improvement of the detection accuracy, as well as the efficiency of the operations, which are capable of providing a holistic solution to the long-term management of bridges safety.
KW - Bridge Safety
KW - CNN
KW - Multi-Sensor Fusion
KW - PyTorch
KW - Reinforcement Learning
KW - Structural Health Monitoring
KW - Vision Transformers
UR - https://www.scopus.com/pages/publications/105034729855
U2 - 10.1109/ICECONF65644.2025.11379672
DO - 10.1109/ICECONF65644.2025.11379672
M3 - Conference contribution
AN - SCOPUS:105034729855
T3 - Proceedings ICECONF 2025 - 2025 2nd International Conference on Artificial Intelligence and Knowledge Discovery in Concurrent Engineering
BT - Proceedings ICECONF 2025 - 2025 2nd International Conference on Artificial Intelligence and Knowledge Discovery in Concurrent Engineering
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2nd International Conference on Artificial Intelligence and Knowledge Discovery in Concurrent Engineering, ICECONF 2025
Y2 - 9 October 2025 through 10 October 2025
ER -