Real, Scaled, Adjusted and Artificial Records: A Displacement and Cyclic Response Assessment

Different procedures to obtain sets of spectrum compatible records for nonlinear dynamic analysis of structures are compared in terms of post-elastic structural response. Six typologies of records are considered: un-scaled real records, real moderately linearly scaled, real significantly linearly scaled, real adjusted by wavelets, artificial generated by two different procedures. The study is spectral shape-based, that is, all sets of records considered, either generated or selected, match individually (artificial and adjusted) or on average (real records), the same design spectrum of a case-study site in Italy. Bilinear non-degrading single degree of freedom systems were used to evaluate the nonlinear response for the compared sets at different non-linearity levels; i.e., demand spectra in terms of peak and cyclic response were derived for two strength reduction factors. Results show that artificial or adjusted records may underestimate, at high non-linearity levels, the displacement-related non-linear response if compared to real records, which are considered as a benchmark. Conversely, if the cyclic response is considered, artificial record sets show a (more evident) overestimation of the demand, while wavelet-adjusted do not display a significant bias. Finally the two groups of linearly scaled records seem to show no systematic bias for both types of response considered, suggesting that scaling does not impair estimation of seismic response if the spectral shape is controlled.