Comparison of fully non-stationary artificial accelerogram generation methods in reproducing seismicity at a given site

Seismic input modelling is a crucial step when Non-Linear Time-History Analyses (NLTHAs) are performed, the seismic response of structures being highly responsive to the input employed. When natural accelerograms able to represent local seismicity are not available, the use of generated accelerograms is an efficient solution for input modelling. The aim of the present paper is to compare four methods for generating fully non-stationary artificial accelerograms on the basis of a target spectrum, identified using seven recorded accelerograms registered in the neighbourhood of the construction site during a single event, assumed as target accelerograms. For each method, seven accelerograms are generated and used to carry out NLTHAs on three RC structures, having irregular mass and stiffness distributions. The generation methods are evaluated in terms of both input modelling and seismic response of three RC structures. The results point out that maximum interstorey drift ratios are well reproduced only by generation methods based on spectrum-compatible criterion, while energy contained by the target set is only simulated by methods based on records of natural signals. Moreover, none of the method here investigated is able to reproduce the hysteretic energy dissipated by the structures subjected to the target accelerograms, in terms of both total amount and cumulative trend over the duration.