Earthquake Accelerogram Selection and Scaling Procedures for Estimating the Distribution of Drift Response

The problem of selecting a suite of earthquake accelerograms for time-domain analyses is of particular practical and academic interest. Research in this field has led to numerous approaches for compiling suites of accelerograms that may be used to robustly estimate the median structural response. However, many applications in earthquake engineering require the estimation of the full distribution of a structural response parameter for a particular predefined scenario. This article presents an efficient procedure whereby the distributions of interstory or roof drifts may be well approximated. The procedure makes use of three-point approximations to continuous distributions and the strong correlation that exists between the spectral acceleration at the initial fundamental period of the structure and the drift response. The distributions obtained under the proposed approach are compared with a reference distribution assumed to represent the true underlying distribution of drift response. The reference distribution is defined through a regression analysis conducted on the results of time-domain analyses of a six-story reinforced-concrete frame building subjected to 1,666 unsealed natural accelerograms. The results indicate that robust estimates of the first and second moments of the distribution of logarithmic drift may be obtained by subjecting the structure to several accelerograms scaled to match three target spectra over a range of periods. The target spectra are defined by the numbers of standard deviations above or below the median 5%-damped spectral acceleration and correspond to the roots of a third-order Hermite polynomial. The results demonstrate that consideration of fifth-order Hermite polynomials does not lead to a significantly improved performance of the approac