Efficient simulation-based approaches for community-level probabilistic seismic risk assessment

Abstract

A research is required to evaluate seismic risk of community on the probabilistic framework. That is because (a) spatially distributed buildings/infrastructures are critical assets in urban community and (b) there are uncertainties in both natural hazard and structural behavior. While Monte Carlo Simulation (MCS) has been a solution to assess risk at community-level, MCS is not always efficient solution especially when severe hazardous scenarios should be investigated. Although Monte-Carlo simulation (MCS) provides straightforward environment to evaluate seismic risk on the urban assets, it requires a large computational cost to forecast rare event. Such catastrophic situations should be identified to keep our community being sustainable even after urban disasters. To overcome this issue, this thesis proposes alternative simulation-based approaches for probabilistic seismic risk assessment at community-level, (a) cross-entropy-based concurrent adaptive importance sampling (CE-CAIS) and (b) clustering-based approach. These new techniques are designed to establish computationally efficient frameworks for probabilistic seismic risk assessment on urban community. In Chapter 2, CE-CAIS is introduced to identify seismic risk on multi-state, large-scale systems with two dimensionality reduction techniques to expand the applicability of CE-CAIS. In Chapter 3, the clustering-based approach is demonstrated to forecast seismic risk on complex urban road networks with decreased computing resources. Several numerical examples are attached on each chapter to validate our proposals. Through this thesis, further researches are expected to produce other valuable achievements while increasing the communication with urban disaster and resilience.