Seismic risk of aging bridge networks by Cross-entropy-based Stationary Proposal Importance Sampling

The feasibility of traditional Monte Carlo simulation techniques in structural reliability analysis and lifeline risk assessment can be limited in practice when intensive numerical analyses are required to estimate small failure probabilities over complex limit state domains. Life-cycle assessment of deteriorating systems may also involve the modeling of complex stochastic processes, further exacerbating the computational effort. This paper presents a novel mathematical framework based on Importance Sampling (IS) to estimate timevariant seismic risk metrics of aging infrastructure systems. The simulation procedure allows adaptively selecting a near-optimal Stationary Proposal (SP) sampling density from a chosen family of parametric distributions by minimizing the Kullback–Leibler (KL) Cross-Entropy (CE). Potentialities and limitations of the proposed numerical approach is investigated in comparison with traditional simulation approaches to address the multi-hazard risk of a deteriorating bridge network exposed to regional seismic hazard.