Improving the Simulation Accuracy of the Substructure Approach for Soil-Structure-Interaction Analysis Using a Refined Impedance Function

Abstract

The objective of this study is exploring the limitations of the substructure approach for time-domain soil-structure-interaction (SSI) analysis. The key assumption in the substructure approach is that the soil-structure system can be partitioned into two subsystems: the superstructure and the soil-foundation. The responses of these two subsystems are assumed to be uncoupled, which allows applying the principle of superposition. The soil-foundation subsystem is then replaced by a set of springs and dashpots, i.e., the impedance function, that represents the soil-foundation flexibility and damping, respectively. To calculate the impedance function, the behavior of a massless rigid foundation resting on/embedded in the soil medium, i.e., the soil-foundation subsystem, is studied.

In the soil-structure system, however, the structural inertia causes coupling between the foundation horizontal translation and rotation around the flexible-base fundamental frequency. This creates soil displacement field and wavefield different from those of the soil-foundation subsystem in the absence of the superstructure, in which the foundation horizontal translation and rotation are uncoupled. Since the superstructure changes the soil-foundation subsystem behavior, the impedance function shall be developed considering the presence of the superstructure. In this paper, we examine the validity of this proposed solution by means of a numerical study on a two-dimensional linear elastic frame structure resting on the surface of a linear elastic half-space. The refined impedance function for this system is estimated using a Bayesian model inversion technique. The objective of the estimation algorithm is to minimize the discrepancy between the responses of the substructure approach and the more accurate, direct approach. Subsequently, the seismic response of this system is analyzed using the direct approach, the substructure approach developed using the traditional impedance function, and the substructure approach developed using the refined impedance function. The results show that the refined impedance function improves the simulation accuracy of the substructure approach and makes it capable of accurately reproducing the simulated response of the direct approach.