GNGTS 2022 - Atti del 40° Convegno Nazionale

GNGTS 2022 Sessione 2.2 323 The stages from 4 to 7 were carried out through a computer programme purposely written in FORTRAN, which incorporates a routine for the simulation of non-stationary stochastic ground motions, according to the method proposed by Sabetta et al. (2021), derived from the code provided by the authors. This programme, first examines the real accelerograms extracted from the available databases during the stage 3), finding the four whose SA values best fit those of the design spectrum at selected values of periods between 0.05 and 2 s. Then the target spectrum for the simulated accelerograms is calculated according to the criterion defined for the stage 6 and, for each M/D combination defined at the stage 2), 20 randomly varying accelerograms are generated. For each scenario, the procedure selects the accelerogram whose SA values best fit the target spectrum at the same periods used for the selection of the real accelerograms and, finally, the 3 scenarios providing the best fitting supply the chosen accelerograms. Discussion. In this study we focused our attention on simulated and natural accelerograms. As a matter of fact, natural records are not always available (or not in a sufficient number) for predefined scenarios. Sometimes the range of the parameters (earthquake magnitude, distance and style of fault) of the scenarios to be represented by the selected seismic inputs need to be enlarged to find enough waveforms among which to select those best fitting the spectrum defined at stage 1), so simulated accelerograms may be the only way to obtain more appropriate records for the representation of relevant scenarios. According to the Italian National Guidelines for the SM Studies (Gruppo di Lavoro MS, 2008) the minimum size of the accelerogram set used as seismic input should be 7 records, of which at least 4 should be real accelerograms; these records should be representative of the ground motion recorded on a stiff ground. The most recent update of the National Guidelines for the execution of SM studies (CTMS, 2020), specifies that, as final product, these studies should draw three maps reporting a subdivision of the investigated territory into “Microzones” characterised by different amplifications of spectral accelerations for 3 period ranges (0.1-0.5 s, 0.4-0.8 s, 0.7-1.1 s). Thus, the seismic input from which numerical simulations derive the estimate of such amplification should show a compatibility with the elastic design spectrum at the site of the simulation, at least in a range of periods from 0.1 to 1.1 s. However, in the stage 4) and 6) of the proposed procedure we have slightly enlarged this range to avoid the selection of accelerograms whose acceleration response spectrum deviates significantly from the design spectrum immediately outside the period range of interest. The selection of real and simulated accelerograms best fitting the target spectra is based on finding the minimum misfit between the acceleration response spectra of the examined records and of the target spectrum, according to the formula proposed by Iervolino et al. (2010). This formulation was directly used for the selection of the 4 real accelerograms best fitting the design spectrum. For the simulated seismograms, considering that they are generated according to a stochastic approach, the adopted procedure admits the possibility of scaling the generated accelerogram by a constant factor K calculated as that minimizing the misfit. However, during the examination of all the accelerograms generated, only the 60 ones for which the factor K has the values closest to 1 are considered as candidates for the selection of those providing the minimum misfit. This expedient avoids taking into consideration accelerograms derived from scenarios that provide SA values too much far from the target spectrum. The use, as target spectrum, of one modified with respect to the local design spectrum, according to the criterion defined for the stages 5) and 6), was motivated by the consideration that the median of the SA values for the real accelerograms selected could present a significant discordance at some period intervals. In view of obtaining a final set of 7 accelerograms for