A novel framework for seismic fragility analysis with the combination of Box-Cox transformation and Bayesian inference. (15th February 2023)

• Record Type: Journal Article
• Authors: Guo, Junjun; Zhang, Penghui; Wang, Jingquan; Li, Shuai; Guan, Zhongguo
• Is Part Of: Engineering structures. Volume 277(2023)
• Publication Date: 2023-02-15
• Subjects: Seismic fragility -- Box-Cox transformation -- Bayesian inference -- Cable-stayed bridge; 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.2022.115436
• Languages: English
• ISSNs: 0141-0296

Abstract

Highlights: A novel framework combing the Box-Cox transformation and Bayesian inference is proposed to conduct the seismic fragility analysis of bridges. The proposed method has the ability to construct nonlinear probabilsitic seismic demand models. A robust fragility model can be derived with the proposed approach, and the assumptions of the cloud-based fragility model are eliminated. Abstract: Fragility curves describe the conditional failure probability that the structural demand reaches or exceeds a limit state under a given intensity measure, which is extensively used in performance-based earthquake engineering. To improve the performance of current methodologies, a novel framework for seismic fragility analysis with the combination of Box-Cox transformation and Bayesian inference is proposed in the present study. A long-span cable-stayed bridge is taken as a case study, and the numerical model of the bridge is established within the OpenSees platform. The probabilistic seismic demand models are established with the Bayesian inference for the Box-Cox transformed data and developing the fragility models with binary Bayesian regression analysis. The numerical results reveal that the proposed framework can establish the nonlinear probabilistic seismic demand models and improve the performance of the classical methods. In addition, the binary Bayesian logistic regression-based fragility model eliminates the assumptions of the classical analytical approaches, and robust results can be obtained. Based on the derived fragility curves, the classical cloud method usually underestimates the failure probability of the components in severe damage states. In contrast, the proposed framework can accurately predict seismic demands at a large intensity level.