On selection and scaling of ground motions for analysis of seismically isolated structures

A broader consensus on the number of ground motions to be used and the method of scaling to be adopted for nonlinear response history analysis (RHA) of structures is yet to be reached. Therefore, in this study, the effects of selection and scaling of ground motions on the response of seismically isolated structures, which are routinely designed using nonlinear RHA, are investigated. For this purpose, isolation systems with a range of properties subjected to bidirectional excitation are considered. Benchmark response of the isolation systems is established using large sets of unscaled ground motions systematically categorized into pulse-like, non-pulse-like, and mixed set of motions. Different subsets of seven to 14 ground motions are selected from these large sets using (a) random selection and (b) selection based on the best match of the shape of the response spectrum of ground motions to the target spectrum. Consequences of weighted scaling (also commonly referred to as amplitude scaling or linear scaling) as well as spectral matching are investigated. The ground motion selection and scaling procedures are evaluated from the viewpoint of their accuracy, efficiency, and consistency in predicting the benchmark response. It is confirmed that seven time histories are sufficient for a reliable prediction of isolation system displacement demands, for all ground motion subsets, selection and scaling procedures, and isolation systems considered. If ground motions are selected based on their best match to the shape of the target response spectrum (which should be preferred over randomly selected motions), weighted scaling should be used if pulse-like motions are considered, either of weighted scaling or spectral matching can be used if non-pulse-like motions are considered, and an average of responses from weighted-scaled and spectrum-matched ground motions should be used for a mixed set of motions. On the other hand, the importance of randomly selected motions in representing inherent variability of response is recognized and it is found that weighted scaling is more appropriate for such motions.