Vision Transformer Based Damage State Assessment of Bridges using Wavelets of Earthquakes

Thasneem A M, Fayiza (2023) Vision Transformer Based Damage State Assessment of Bridges using Wavelets of Earthquakes. Masters thesis, Indian Institute of Technology Hyderabad.

Text
CE21MTECH11014.pdf - Submitted Version
Restricted to Registered users only until 29 June 2027.
Download (5MB) | Request a copy

Abstract

Bridges play a crucial role in facilitating transportation and economic growth, but their vulnerability to earthquakes necessitates effective damage detection methods. This thesis explores the application of Continuous Wavelet Transforms (CWT) in combination with Vision Transformers (ViTs) for automated detection and assessment of structural damage in bridges. In the conventional approach, analytical fragility curves, which quantify the likelihood of different damage states based on ground shaking intensity, are employed to predict bridge damage. The study addresses the limitations of conventional methods and introduces deep learning techniques to enhance computational efficiency and accuracy. By analyzing a dataset of ground motion records and training the ViT model, the research aims to advance automated bridge inspection techniques and enable early damage detection. The results demonstrate the effectiveness of the CWT-based ViT model in assessing the damage state of bridges. The proposed approach holds promise for proactive maintenance strategies, improving the safety and durability of critical infrastructure. Additionally, the thesis discusses the influence of various factors such as rupture distance on different bridge components and highlights the advantages of employing ViT models in terms of computational cost. Future work involves expanding the dataset to improve prediction accuracy and establishing a more uniform distribution of data for each damage category.

[error in script]

IITH Creators: