CALIBRATION OF 2D NUMERICAL MODELS FOR THE CASE OF SOIL-STRUCTURE SYSTEMS WITH SURFACE FOUNDATIONS

Finite element simulations are widely used in soil-structure interaction (SSI) problems. It is usual that the semi-infinite soil medium becomes truncated to a manageable size through artificial boundaries with finite degrees of freedom. However, ambiguous points still exist in different aspects of reasonable modeling of such boundaries. Fixed boundary condition in static analysis can be used with appropriate distance from structure without sacrificing much in terms of accuracy. Using fixed boundary condition in dynamic analysis neglects the propagation of waves into infinity. From this point, the mentioned ambiguity starts with free field motion analysis which normally precedes each SSI problem. Although considering wide soil domain with material damping can help vanish the reflected waves, it is computationally expensive. The current practice suggests viscous or viscoelastic boundaries including spring and dashpot elements. Introducing input motion to the system through such boundary elements should guarantee the correct free field motion in the medium before structural positioning. However, seismic loading from such borders faces frequency content distortions which need to be treated appropriately. In the second stage, the boundaries should be able to properly simulate the outward propagation of waves, emanated by SSI, i.e., radiation damping. Although viscous boundary condition is capable of damping out most of the reflecting waves due to the assumption of 1D wave propagation in setting damper coefficients, it is not able to absorb whole of the body waves with different angles of incident. This weakness becomes augmented in the case of surface waves. Here, the existing limitations on seismic loading in boundaries are discussed. Then, practical solutions are suggested that can improve the arrival and transmitting out of seismic waves from viscoelastic boundaries, properly. The major contribution of this research belongs to new approaches to, first, remove frequency distortion of input motion caused by viscoelastic boundaries; second, highlight special cautions about minimum element size to cover proper frequency range of interest, and third, to introduce the idea of extending the boundaries to the main model in order to substantially increase their absorbing