Automated Generation of Model and Seismic Fragility Inventories of School Buildings in Türkiye

Abstract

This study focuses on creating structure-specific seismic fragility inventories for reinforced concrete (RC) school buildings using a combination of analytical and artificial intelligence (AI) techniques. Building model inventories are critical components of regional seismic risk analyses, particularly for risk prioritization and mitigation efforts, as well as rapid post-event damage assessment. For this pilot study, a set of school buildings in the Istanbul province was selected, which are representative of this type of building across Türkiye. The OpenSees-based SimCenter EE-UQ tool was used to obtain the fragility curves. The building model inventories required for these analyses were generated using novel combinations of several AI and computer vision-based methods, which were applied to facade images and other publicly available data and metadata, which automatically produced model metadata, such as number of floors, building height, and footprint. The accuracies of the fragility curves were evaluated by comparing them with those obtained using analysis models created using structural drawings, enabling model and method validation. Uncertainty quantification analyses were carried out to examine the sensitivities of the fragility curves to the uncertainties/variations in the metadata used for model generation. The findings indicated that the automated framework provides satisfactory results and is feasible for regional-scale studies. While the techniques and tools developed in the present study focus on school buildings in Istanbul, they can be adapted for and applied to other building types and seismic zones.