The purpose of this study is to investigate the effects of the significant duration of ground motions on responses of base-isolated nuclear power plants (NPPs). Two sets of ground motion records with short duration (SD) and long duration (LD) motions, scaled to match the target response spectrum, are used to perform time-history analyses. The reactor containment building in the Advanced Power Reactor 1400 (APR1400) NPP is numerically modeled using lumped-mass stick elements in SAP2000. Seismic responses of the base-isolated NPP are monitored in forms of lateral displacements, shear forces, floor response spectra of the containment building, and hysteretic energy of the lead rubber bearing (LRB). Fragility curves for different limit states, which are defined based on the shear deformation of the base isolator, are developed. The numerical results reveal that the average seismic responses of base-isolated NPP under SD and LD motion sets were shown to be mostly identical. For PGA larger than 0.4g, the mean deformation of LRB for LD motions was bigger than that for SD ones due to a higher hysteretic energy of LRB produced in LD shakings. Under LD motions, median parameters of fragility functions for three limit states were reduced by 12% to 15% compared to that due to SD motions. This clearly indicates that it is important to select ground motions with both SD and LD proportionally in the seismic evaluation of NPP structures.

The phase properties of ground acceleration records from Mw 5.5~6.5 earthquakes are analyzed. The interrelationships between phase properties and significant durations, as well as PGA, are clarified through both of theoretical and empirical approaches. The probabilistic characteristics of phase information is also discussed based on previous studies and it is shown that circular normal distribution is the most appropriate probability distribution for the phase angle and phase difference. Whereas those variates can be modeled by Gaussian random variables. From the survey results on the frequency dependency of the phase statistics, a simple model is introduced, which is possible to express the frequency dependency of phase information. It is also shown that the significant duration can be controlled by appropriately chosen standard deviation of phase difference for 4~8Hz frequency band and additional consideration of phase scattering in higher frequency band through a series of Monte Carlo simulations. The source of phase scattering effect is also pointed out and discussed.