A multi?fidelity Bayesian framework for robust seismic fragility analysis

Fragility analysis of structures via numerical methods involves a complex trade-off between the desired accuracy, explicit consideration uncertainties (both epistemic and aleatory) related to structural model available computational performance. This paper introduces framework for deriving fragility relationships based on multi-fidelity non-linear models structure under investigation response-analysis types. The proposed aims reduce burden while achieving accuracy estimates without neglecting aleatory uncertainties. approach is an extension well-known robust (RF) framework. Different classes, each characterised by increasing refinement, are used define polynomial expansions parameters. Each result then considered as ‘new observation’ in Bayesian update coefficients expansions. An adaptive sampling algorithm also futher improve performance Specifically, such relies partitioning sample space Kullback–Leibler divergence find optimal path. allows analyst specify different criteria parameters regions, thus further improving procedure. illustrated archetype reinforced concrete (RC) frame which two classes developed/analysed: simple lateral mechanism (SLaMA), coupled with capacity spectrum method, dynamic analysis. Both involve cloud-based employing unscaled real (i.e. recorded) ground motions. derived procedure finally compared those calculated using only most advanced/high-fidelity (HF) class, quantifying highlighting research needs.