Effect of uncertain spatial variability of soils on nonlinear seismic site response analysis

In this paper, recently developed stochastic elastic-plastic finite element method (SEPFEM) is used to propagate the uncertain spatial variability of soil properties at a site in northern California through the governing wave equation in quantifying its effect on non-linear seismic site response analysis. The SEPFEM computational framework utilizes Karhunen-Loeve and polynomial chaos expansions for discretizing (in stochastic and spatial spaces) the soil properties and response variable (displacement) random fields, respectively. Fokker-Planck- Kolmogorov approach is used to probabilistically integrate the elastic-plastic constitutive equations at the material points. The spatial variability parameters - correlation length and variance - of the soil properties along with their mean values for the site are quantified by statistically analyzing in-situ test data using maximum likelihood technique. The computed site response is quantified in terms of evolutionary probability density function of surface acceleration. A parametric study is also performed by varying the correlation length and variance of soil properties to investigate the effects of soil spatial variability on the computed response.