Deconvolution method between kinematic interaction and dynamic interaction of soil-foundation systems based on observed data

Abstract

To prove the effect of kinematic interaction experimentally, earthquake records both at the ground and at the foundation are needed. Since the foundation motion is effected not only by the kinematic interaction but also by the dynamic interaction of a soil-foundation system, a general deconvolution method between the kinematic interaction and the dynamic interaction of a soil-foundation system based on the observed data is derived in this paper, where a structure is modelled by a multiple-degree-of-freedom (MDOF) system which represents the dynamic interaction, and the unknown foundation input motion is expressed by a moving average model (MA-model) of the ground motion, which represents the filtering effect of kinematic interaction. Generally, as the parameters of a MDOF system can not be determined exactly, these parameters are treated as random variables on trial using the Monte Carlo simulation method. Next, the coefficients of the MA-model are identified applying the Kalman filter. On the many trials, the result which minimizes the root-mean-square error between the observed response and calculated response is determined as the best estimation. Examples of (1) two single-degree-of-freedom systems arranged in a series and (2) a four-storey reinforced concrete school building are demonstrated to verify the proposed method.