Generation of artificial nonstationary accelerograms with natural variability

The present work, describes a modified version of the method for generating non-stationary accelerograms originally proposed by Sabetta and Pugliese (1996). This method adopts four ground motion parameters – Arias intensity, duration, central frequency and frequency bandwidth – to define an approximated spectrogram which is then used to generate artificial accelerograms. The values of the four parameters defining the spectrogram are estimated using ground-motion prediction equations once a scenario is defined in terms of magnitude, distance and soil stiffness. In the present work, the Sabetta and Pugliese’s method has been revisited and enhanced by i) proposing new forms for the ground-motion prediction equations involved in the model, ii) considering a different regression technique to calibrate the ground-motion prediction equations; ii) using a larger and more comprehensive database of accelerograms; iii) using a more consistent definition of magnitude; iv) proposing a new criterion for taking into account the directions of recording sensors and giving a better description of ground motion variability