Numbers of scaled and matched accelerograms required for inelastic dynamic analyses

Selecting, scaling and matching accelerograms are critically important to engineering design and assessment, enabling structural response to be determined with greater confidence and through fewer analyses than if unscaled accelerograms are employed. This paper considers the response of an 8-storey multiple-degree-of-freedom reinforced concrete structure to accelerograms selected, linearly scaled or spectrally matched using five different techniques. The first method consists of selecting real records on the basis of seismological characteristics, while the remaining methods make an initial selection on the basis of magnitude and spectral shape before (1) scaling to the target spectral acceleration at the initial period; (2) scaling to the target spectrum over a range of periods; (3) using wavelet adjustments to match the target spectrum and (4) using wavelet adjustments to match multiple target spectra for multiple damping ratios. The analyses indicate that the number of records required to obtain a stable estimate of the response decreases drastically as one moves through these methods. The exact number varies among damage measures and is related to the predictability of the damage measure. For measures such as peak roof and inter-storey drift, member end rotation and the Park and Ang damage index, as few as one or two records are required to estimate the response to within ±5% (for a 64% confidence level) if matching to multiple damping ratios is conducted. Bias checks are made using predictive equations of the expected response derived from the results of 1656 nonlinear time-domain analyses of the structure under the action of unscaled accelerograms. Copyright © 2008 John Wiley & Sons, Ltd.