Incorporating site response analysis and associated uncertainties into the seismic hazard assessment of nuclear facilities
Abstract
The development of a site-specific seismic hazard curve for a soil site requires the incorporation of site effects into the hazard calculation through the use of a site-specific amplification function. This study investigates the effect on the resulting soil hazard curves of different approaches to compute the site-specific amplification function. Amplification functions and their standard deviations can be developed using equivalent linear site response analyses. This study investigates the amplification function predictions of one-dimensional (1D) and two-dimensional (2D) site response analyses. For 1D analysis, one set of analyses are performed using time series (TS) input motions while a second set is performed using random vibration theory (RVT). One-dimensional site response analyses are performed for a shallow and a deep soil site and the results are compared for seismic hazard predictions. The influence of spatial variability introduced through randomization of site shear wave velocity (V[subscript S]) is also investigated. Shear wave velocity profile randomization does not significantly change the predicted amplification functions, except for the RVT analysis near the site period. At these periods, (V[subscript S]) randomization reduces the amplification function predicted by RVT making it more similar to the TS analysis prediction. The surface hazard at a site is dependent on the median amplification factor and its associated standard deviation. Spatial variability and uncertainties in soil properties across a site are often taken into account by modeling multiple 1D profiles in 1D site response analyses. However, this approach assumes that analyzing multiple 1D profiles captures accurately the effects of the true multi-dimensional spatial variability of the soil properties. This study investigates the results of two-dimensional (2D) site response analyses that incorporate spatial variability in the (V[subscript S]) profile through Monte Carlo simulation. Two-dimensional site response analyses are performed for 2D random fields generated with various statistical parameters (i.e. vertical and horizontal correlation distances) to investigate the effect of different levels of spatial variability on surface response across a region of interest (ROI). Two-dimensional site response analyses are performed for a shallow site. Results indicate that horizontal correlation distance has more influence on the analyses results than the vertical correlation distance. As the horizontal correlation distance increases, the median surface response spectrum across the ROI decreases. This reduction in median surface response is more pronounced around the site period. The influence of the vertical correlation distance is more pronounced when the horizontal correlation distance is large. As the vertical correlation distance increases, the median surface response spectrum across the ROI increases, which is more pronounced around the period of the motion. The predictions of 1D and 2D site response analyses modeling the (V[subscript S]) variability are compared. 1D analyses are performed on separately generated 1D (V[subscript S]) profiles (infinite horizontal correlation) and on the (V[subscript S]) profiles across the ROI of each 2D (V[subscript S]) field realization generated for 2D analysis (finite horizontal correlation). The results indicate that both sets of 1D analyses predict lower median response than 2D analyses. The 1D analyses with finite horizontal correlation display comparable levels of variability in the site response, however 1D analyses with infinite horizontal correlation display higher variability.