EFFICIENT PROBABILISTIC SEISMIC SOIL-STRUCTURE INTERACTION (SSI) ANALYSIS FOR NUCLEAR STRUCTURES USING A REDUCED-ORDER MODELING IN PROBABILISTIC SPACE

摘要

The paper investigates different reduced-order modeling approaches for performing fast probabilistic seismic SSI analyses. Three types of ROM approaches are considered: ⅰ) Latin Hypercube Sampling (LHS) simulation, ⅱ) Random Vibration Theory-based (RVT) simulation, and ⅲ) Stochastic Reduced Order Modeling (SROM) simulation. While LHS and RVT are classical approaches, currently recommended by the new ASCE 04-2013 standard, SROM is a recently developed, new approach based on an advanced probabilistic modeling. Comparative results are shown for the EPRI AP1000 RB stick model for both soil and rock sites. The paper shows that all the three ROM approaches are useful tools for probabilistic seismic SSI analysis. However, it is noted that the RVT results are quite sensitive to the particular analytical approach employed for estimating maximum probabilistic response. The newer SROM approach appears to be highly efficient, apparently with a faster statistical convergence than LHS.