Three-dimensional inverse analyses of a deep excavation in Chicago clays

Numerical models are commonly used to estimate excavation-induced ground movements. Two-dimensional (2D) plain strain assumption is typically used for the simulation of deep excavations which might not be suitable for excavations where three-dimensional (3D) effects dominate the ground response. This paper adapts an inverse analysis algorithm to learn soil behavior from field measurements using a 3D model representation of an excavation. The paper describes numerical issues related to this development including the generation of the 3D model mesh from laser scan images of the excavation. The inverse analysis to extract the soil behavior in 3D is presented. The model captures the measured wall deflections. Although settlements were not sufficiently measured, the predicted settlements around the excavation site reflected strong 3D effects and were consistent with empirical correlations.