Short-time frequency-domain method for truly nonlinear dynamic ground response analysis: The equivalent-nonlinear approach

Abstract

Evaluating soil nonlinearity during cyclic loading is one of the most significant challenges in ground response analysis, especially when dealing with the inverse problem of deconvolution. Different schemes have already been developed for dynamic ground response analysis, both in the time and the frequency domain. The most accurate method to account for soil nonlinearity is the nonlinear dynamic analysis in the time domain. This approach is based on nonlinear constitutive models capable of accurately simulating highly nonlinear problems like soil liquefaction. However, the time-domain analysis is suitable only for the convolution analysis to define the ground motion at the free surface of a soil deposit from the bedrock motion. The frequency-domain analysis is the most common solution for the inverse problem called deconvolution, which is used to define the bedrock motion from the free surface ground motion. A well-known approach developed in the frequency domain for ground response analysis is the equivalent-linear method (EQL). This approach adopts an iterative procedure to define elastic shear modulus and damping ratio compatible with the induced strain level. Still, it presents some limitations, especially for highly nonlinear soil response, due to the use of strain-compatible but constant soil properties. This article presents a new scheme to conduct truly nonlinear dynamic analysis in the frequency domain based on the new concept of the short-time transfer function. Unlike the EQL method, which uses a constant transfer function, the proposed approach, called the “Equivalent-Nonlinear” method (EQNL), defines a soil transfer function evolving in time, depending on the shear stress and strain demands. The EQNL method approximates the response of a nonlinear system as an incrementally changing viscoelastic system and could represent a valuable tool for nonlinear deconvolution. This article shows the analytical formulation and the first set of validations of the EQNL approach, with detailed comparisons with the EQL and NL methods and vertical array data. These comparisons show the potentialities of the EQNL approach to reproduce the results of the nonlinear dynamic analysis. The EQNL approach has been implemented in MATLAB, and the source code is provided as supplementary material for this article. A more comprehensive validation is underway, aiming to better characterize the limitations and the capabilities of the method.