A deep learning-aided seismic fragility analysis method for bridges. (June 2022)

• Record Type: Journal Article
• Authors: Wang, Mengge; Zhang, Hao; Dai, Hongzhe; Shen, Luming
• Is Part Of: Structures. Volume 40(2022)
• Page Start: 1056
• Page End: 1064
• Publication Date: 2022-06
• Subjects: Seismic -- Fragility analysis -- Structural Reliability -- Probabilistic analysis -- Deep Learning -- Neural Networks; Structural engineering -- Periodicals; 624.1
• DOI: 10.1016/j.istruc.2022.04.058
• Languages: English
• ISSNs: 2352-0124

Abstract

Abstract: Bridges are critical but vulnerable components in a transportation network as they are exposed to the threats induced by long-term aging effects as well as natural hazards such as earthquakes. The traditional seismic fragility analysis is associated with high computational cost, making it infeasible for the cases requiring multiple fragility analyses, such as evaluating time-dependent seismic fragility for deteriorating facilities, or a transportation network involving many bridges. In this study, a deep learning-aided seismic fragility analysis method is proposed to improve the computational efficiency. Fragility analysis is transformed into a binary classification problem. An improved deep neural network classification algorithm with a new activation function is proposed and benchmarked with traditional deep neural networks and other machine learning counterparts. The accuracy and the robustness of the new method are demonstrated by examples.