Numerical modelling of seismic site response at large strains: a parametric study.

The numerical analysis of seismic site response at large strains should adopt constitutive models able to guarantee not only a correct modelling of stiffness and damping properties but also a compatibility with the shear strength of the materials. The traditional hyperbolic models used in nonlinear analyses are generally calibrated on stiffness and damping curves and therefore does not necessarily match the soil shear strength. An inaccurate modelling of shear strength can lead to unrealistic predictions of the seismic site response with results that are not necessarily conservative: underestimation or overestimation of the computed surface response depends on the difference between the maximum shear stress implied by the adopted hyperbolic nonlinear model and the real soil shear strength. In this paper, over 1900 one-dimensional parametric analyses on ideal sand and clay deposits were executed with DEEPSOIL software. A first comparison was undertaken between equivalent linear and nonlinear analyses; then the nonlinear analyses were addressed to study the influence of shear strength as an input parameter on the results of numerical site response analyses. In particular two strategies to take into account the soil shear strength were considered: an adjustment procedure associated to the standard MKZ hyperbolic model and the GQ/H model which allows the shear strength to be explicitly defined as input parameter of the analyses. This parametric study made it possible to define preliminary threshold shear strain values, beyond which it is necessary to execute numerical analyses with more advanced models or procedures, able to capture the real behavior of the soil at large strains. Indicatively above shear strains of 0.1%, traditional nonlinear models neglecting soil strength can provide unrealistic results, with important overestimation of the seismic motion (up to 30% in terms of PGA at the surface).