Conditional Probabilistic Analysis of Cement-Treated Soil Column Strength

Abstract

The quality of cement-treated soil columns is usually assessed by examining the strength of cored samples. This paper presents a finite-element analysis approach with random field theory for assessing the quality of a cement-treated soil column from which cored samples are retrieved. Monte Carlo simulations with conditional distribution (conditional simulation) were conducted to generate spatial dependence random fields with the data known at some locations. Simulations without the known data (unconditional simulation) were also performed for comparison. Using the random field samples generated from the Monte Carlo simulations, finite-element analysis was performed to simulate the compression behavior of a full-scale column in which material properties vary with a presence of spatial autocorrelation. Finite-element analysis with conditional simulation predicts the overall strength of a targeted cement-treated soil column appropriately, and the overall strength variability obtained with the conditional simulation is lower than that obtained with the unconditional simulation. The results suggest that the conditional simulation offers advantages over the unconditional simulation when assessing the probability of failure of a cement-treated soil column with large autocorrelation distance.