Answers are grounded in this paper and cite their sources.
Analysis reveals efficient modeling of seismically isolated buildings to optimize repair costs and downtime.
As seismic design paradigms shift toward functional recovery, there is a need for efficient methods to achieve desired decision variables (DVs), such as probability of replacement, repair cost, and downtime of the building, for seismically isolated buildings. Designing to achieve specific DVs is typically iterative and computationally costly, especially given the wide variety and combinations of isolation and superstructure systems. To directly translate structural design of isolated buildings to performance targets, a surrogate model that maps design variable combinations to their resulting performance is proposed. A database is constructed of steel moment resisting frames and concentrically braced frames, isolated with either friction pendulum bearings or lead rubber bearings, with varying structural configurations and design parameters. Each design is modeled in two dimensions and subjected to a random horizontal time series via nonlinear time history analysis, and the resulting superstructure drifts and accelerations are collected. Probabilistic damage and loss estimation is performed using the FEMA P‐58 framework. Gaussian process‐based models are then constructed for each system to predict DVs using the structures’ input parameters across the design space. A simplified upfront cost optimization is then carried out among designs that achieve the given DV limits. The inverse design framework enables efficient selection of optimal design parameters, such as bearing stiffness, minimum isolation moat gap, and superstructure strength, required to achieve the desired repair cost and/or downtime for structures, allowing engineers to directly design for functionality postearthquakes.
No takes yet. Share an insight, caveat, or question.
Pham et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: