Comparison of approaches for generation of fully non-stationary artificial accelerograms

The modelling of the seismic input is a critical aspect when non-linear time-history analyses (NLTHAs) are carried out. As a matter of fact, seismic response of structures is very sensitive to the input excitation time history. The present work aims to highlight the differences in the input modelling and the assessment of seismic response of three r.c. structures employing four generation methods of fully non-stationary artificial accelerogram sets at a given construction site. For each method, seven accelerograms are generated and employed to perform NLTHAs on three r.c. structures having irregular mass and stiffness distributions. The original contribution of the paper relies in the criterion for generation method effectiveness evaluation. Efficiency is evaluated by carrying out a comparison of the structure response obtained by input modelling according to the analysed generation methods, against that evaluated by using seven recorded ground motions registered in the neighbourhood of a selected site during a same event, assumed as target accelerograms. The results point out that spectrum-compatibility based generation methods provide accelerograms able to match maximum structural displacements registered with target ground motions, but in most of them the assessed response exhibits total energy and potentially induced damage higher than that contained by the target ones. Conversely, non-spectrum-compatible method generates accelerograms having a total energy history comparable to that held by target accelerograms, but leading to wrong interstorey drift ratio assessments. The comparison claims to formulate a method of generation that combines the merits of the examined methods.