Ricerc@Sapienza

Predicting the seismic site effects in complex real case histories is a non-trivial scientific and technical task. In fact, it requires the overall geometry of the problem (i.e. surface and underground) to be properly described and combined to adequate soil models, accounting for the peculiar aspects of the dynamic soil behaviour. These latter should include the dependency of the initial stiffness on the stress state and its non-linear evolution with strain amplitude, together with the corresponding one of the damping ratio. A further ingredient to be considered when approaching this class of problems is that of selecting a robust and accurate numerical scheme, thus a suitable numerical code, to solve it. All the above issues are addressed in this paper, which describes the different numerical approaches that were adopted to analyse the seismic site-effects expected at a real site. This latter is characterized by a soft soil valley and rock outcrop hills of relatively complex geometry. It was analysed by performing time domain 1D and 2D Finite Element (FE) analyses through two different codes: QUAKE/W, based on the equivalent-linear approach, and PLAXIS 3D, adopting the Hardening Soil model with small strain stiffness available in the material model library. 3D simulations were also carried out by means of the code PLAXIS 3D. A real accelerogram was selected as the outcrop motion, characterized by a duration of 22 seconds and a frequency content ranging from 0 to 10 Hz. In order to validate the two FE codes, the results of 1D analyses relative to specific soil columns extracted by the site were preliminary compared to those obtained by the code EERA, based on the equivalent-linear approach in the frequency domain. Analogous results were obtained, indicating a low impact of the constitutive assumptions in the investigated case study, provided non-linearity was similarly accounted for in the different adopted approaches. Next, the influence of the geometrical schematization was analysed, revealing that the seismic site response at the same location can strongly depend on the employed dimensional approach (i.e. 1 to 3D). In general, assuming the outcrop signal as the reference motion, greater amplifications were observed at the centre of the valley and, with a minor extent, at the upper sides of the hills: those trends were significantly influenced by the multidimensional schematization (2 and 3D), while they were less evident or barely detectable in the 1D approach.