The recent concerns regarding the seismic safety of the existing building stock have highlighted the need for an improvement of current seismic assessment procedures. Alongside with the development of more advanced commercial software tools and computational capacities, nonlinear dynamic analysis is progressively becoming a common and preferable procedure in the seismic assessment of buildings. Besides the complexity associated with the formulation of the mathematical model, major issues arise related with the definition of the seismic action, which can lead to different levels of uncertainty in terms of local and global building response. Aiming to address this issue, a comparative study of different code-based record selection methods proposed by Eurocode 8, ASCE41-13 and NZS1170.5:2004 is presented herein. The various methods are employed in the seismic assessment of four steel buildings, designed according to different criteria, and the obtained results are compared and discussed. Special attention is devoted to the influence of the number of real ground motion records selected on the estimation of the mean seismic response and, importantly, to the efficiency that is achieved when an additional selection criteria, based on the control of the spectral mismatch of each individual record with respect to the reference response spectrum, is adopted. The sufficiency of the methods with respect to the pairs of M-R of the selected group of records and the robustness of the scaling procedure are also examined. The paper closes with a study which demonstrates the suitability of a simplified probability-based approach recently proposed for estimating mean seismic demands. Copyright (c) 2015 John Wiley & Sons, Ltd.
