Efficient Bayesian metamodeling for fine-grained and robust fragility analysis of buildings at a regional scale. (May 2023)

• Record Type: Journal Article
• Authors: Su, Peiyang; Xiong, Feng; Lu, Yang; Hu, Qidan; Zhang, Bowen
• Is Part Of: Structural safety. Volume 102(2023)
• Publication Date: 2023-05
• Subjects: Regional seismic fragility assessment -- Site-city interaction effect -- Bayesian neural network -- Data augmentation algorithm; Structural stability -- Periodicals; Safety factor in engineering -- Periodicals; Reliability (Engineering) -- Periodicals; Constructions -- Stabilité -- Périodiques; Coefficient de sécurité en ingénierie -- Périodiques; Fiabilité -- Périodiques; 620.86
• DOI: 10.1016/j.strusafe.2023.102324
• Languages: English
• ISSNs: 0167-4730

Abstract

Highlights: Bayesian Neural Network is used to derive single building-level seismic fragility. Improved data augmentation algorithm is proposed to sample simulated responses. Epistemic uncertainties in Fragility estimations are quantified. Abstract: A fine-grained seismic fragility analysis of regional buildings considering 'Soil-Structure-Cluster Interaction' (SSCI) effect faces a dilemma between accuracy and efficiency. In the present study, a Bayesian Neural Network (BNN) model is adopted to address this problem. Specifically, the conventional neural network (NN) algorithm and a Bayesian inference are integrated into one approach, where the NN predicts the structural responses of buildings and the Bayesian inference quantifies the epistemic uncertainty of fragility estimations arising from limited response data due to the high computational cost of the structural analysis. Moreover, the Gaussian kernel function-based data augmentation (KDA) algorithm is proposed to sample the simulated structural response data for BNN model training. The proposed framework is implemented on the regional buildings of Sichuan University as a case study. The results show that the BNN can make accurate and robust fragility estimations with high modeling efficiency.