Probabilistic Characterization of 3-D Spatial Variability of Soils: Methodology and Strategy

Abstract

The 3-D spatial variability of soils has significant impacts on the failure mechanism and reliability of geotechnical structures and deserves a quantitative characterization through site investigation. This study develops a probabilistic approach for characterizing the 3-D spatial variability of soils within the framework of maximum likelihood estimation, whose computational problem is addressed through a matrix decomposition technique. The sampling strategy to minimize the statistical uncertainty is explored systematically based on virtual site analysis. The empirical distance criterion and density criterion are proposed to control the statistical uncertainty to a practically acceptable low level.