ABSTRACT
Deep excavations are commonly used in densely populated areas with the aim of cost and space optimization. Soft soil deposits present in Bogotá are inherently variable, so a combination of numerical and probabilistic approaches are suitable toassess the safety conditions of the excavations to prevent damages to neighboring buildings. In the analyses, the soil constitutive parameters can be treated as random fields to deal with uncertainty due to spatial variability. Geostatistical simulation techniques are useful to generate random fields of soil properties using available measures. Two issues related with random field simulations are the number of required realizations to obtain the final estimate and the horizontal scale of fluctuation because measurements in that direction are usually scarcer. Constitutive model Hardening Soil Small Strain was considered to represent soil behavior and parameters were obtained from laboratory and field test results, mostly SCPTu. Descriptive and spatial statistical analyses were performed on obtained sets of residuals after detrending. Sequential Gaussian co-simulations were used to obtain conditional random fields for selected constitutive parameters considering different number of realizations for each simulation and different anisotropy ratios for the scales of fluctuation. Random fields were mapped to numerical models and 3D finite element analyses were performed on an idealized deep excavation model obtaining the system response in terms of damage potential indexes.
