Liquefaction potential of soils: A fully probabilistic approach

Abstract

This research is aimed at improving existing methods for evaluating liquefaction potential of soils using probabilistic approach. The research deals with several closely related studies. First, a Cone Penetration Test (CPT)-based model for calculating liquefaction resistance is developed using artificial neural network approach, taking advantage of a recently updated seismic loading model and a comprehensive data set of CPT-based liquefaction/no-liquefaction case histories. Second, the liquefaction resistance model developed is incorporated into the formulation of liquefaction potential index, and the various issues related to the liquefaction potential index are examined using probability concept. Third, general applicability of the commonly used index properties-based criteria for liquefaction susceptibility is investigated. Fourth, a methodology for mapping liquefaction-induced ground failure is developed. As an example application, liquefaction-induced ground failure potential for Charleston quadrangle, South Carolina is mapped using seismic parameters specified for the area for a hazard level corresponding to 475-year return period. Finally, a new framework that integrates the conditional probability of liquefaction with the variation of earthquake sources and the underlying model uncertainty, as reflected in the U.S. Geological Survey Seismic Hazard Maps, for a fully probabilistic analysis of soil liquefaction in a given time exposure is developed and demonstrated with an example.