Site response analysis using one-dimensional equivalent-linear method and Bayesian filtering

Abstract

Site response analysis is crucial to define the seismic hazard and distribution of damage during earthquakes. The equivalent-linear (EQL) is a numerical method widely investigated and used for site response analysis. Because several sources of uncertainty are involved in this type of analysis, parameters defining the numerical models need to be identified from in-situ measurements. In this paper, a Bayesian inference method to estimate the expected values and covariance matrix of the model parameters is presented. The methodology uses data from downhole arrays recorded during earthquakes. Two numerical applications show the good performance and prediction capabilities of the proposed approach.