Time-domain deconvolution procedure for elastoplastic materials: Application to the Treasure Island site during the 1989 Loma Prieta earthquake

When dynamic soil-structure interaction (DSSI) analyses are performed, e.g. using the finite element (FE) method, the input signal is required at the base of the model. Nevertheless, acceleration records are usually available at the surface and, therefore, the desired motion must be deconvolved to the base. The latter is usually performed through the solution of one-dimensional propagation of shear waves in an elastic medium, in the frequency domain. Herein, nonlinear behavior is generally incorporated through the equivalent-linear method, by iteratively reducing the stiffness and increasing the critical damping ratio as a function of the maximum strains attained in each iteration. However, if complex material models are adopted to characterize the soil, the input motion derived with the equivalent linear method will not be compatible due to the simplified approach used to represent the nonlinear behavior. In this article, the use of a procedure to perform a time-domain deconvolution in non-linear elastoplastic materials is demonstrated. The goal is to generate input accelerograms at the base of a FE model to perform DSSI analyses. The procedure is based on the iterative modification of the motion at the base according to the relative differences between the propagated and target surface spectra. To illustrate the use of the methodology, it was applied to a FE model of the Treasure Island site (San Francisco, US), to derive the required motion at the base from a record of the Loma Prieta earthquake. • This article provides a useful guideline to optimize the use of the deconvolution procedure to derive input motions for dynamic FE analyses considering nonlinear elastoplastic materials.