A multi-source data-driven approach for evaluating the seismic response of non-ductile reinforced concrete moment frames. (1st March 2023)

• Record Type: Journal Article
• Authors: Chen, Peng-Yu; Guan, Xingquan
• Is Part Of: Engineering structures. Volume 278(2023)
• Publication Date: 2023-03-01
• Subjects: Multi-source data-driven approach -- Image processing -- Convolutional neural network -- Seismic response prediction -- Non-ductile reinforced concrete moment frames; Structural engineering -- Periodicals; Structural analysis (Engineering) -- Periodicals; Construction, Technique de la -- Périodiques; Génie parasismique -- Périodiques; Pression du vent -- Périodiques; Earthquake engineering; Structural engineering; Wind-pressure; Periodicals; 624.105
• DOI: 10.1016/j.engstruct.2022.115452
• Languages: English
• ISSNs: 0141-0296

Abstract

Abstract: As the smart city concept becomes increasingly popular and important in both industries and academia, a broad range of data collected from various sources is employed to assist the policy maker in making more informed decisions. Among these data, some are well-structured and stored in spreadsheets, such as the building and site information stored in archive documents, whereas the rest are unstructured, such as the images or videos taken by unmanned aerial vehicles and written texts extracted from most recent news reports. This paper proposes a multi-source data-driven framework that can rapidly estimate the seismic response of non-ductile reinforced concrete frame buildings by leveraging the images and well-tabulated data. This framework meticulously incorporates computer vision and well-structured data processing techniques. To demonstrate its efficacy, the proposed framework is applied to a comprehensive dataset, which includes 1400 non-ductile reinforced concrete frame designs, their nonlinear structural models, associated seismic responses, and the building exterior images. A thorough review of the application result reveals that the proposed framework is able to efficiently and reliably estimate the seismic drift demands in non-ductile reinforced concrete frames subjected to earthquake scenarios. Such a multi-source data-driven framework would become an essential component in constructing a smart city. Highlights: A multi-source data-driven approach is developed for evaluating seismic response. The proposed approach leverages both building images and ground motion parameters. A comprehensive response database for non-ductile RC moment frames is constructed. Deep CNNs, transfer learning, and XGBoost are integrated into the proposed approach.