Analysis of Deep Excavations in Clay

 

 This dissertation describes the application of non‑linear finite element analyses for predicting and interpreting ground deformations associated with deep, braced excavations in clays.

 

 The first part of this research reviews algorithms for the implementation of elasto-plastic constitutive equations in non‑linear finite element analyses. Robust and efficient numerical convergence characteristics are achieved for the Modified Cam Clay soil model through the formulation of an implicit numerical integration scheme, together with a consistent tangent stiffness matrix. A second soil model, MIT‑E3, has shown excellent capabilities for predicting complex aspects of the measured behavior of soft clays including a) anisotropy stress‑strain‑strength, b) small strain non‑linearity and c) hysteretic and path dependent behavior for overconsolidated clays. The numerical implementation of MIT‑E3 in the ABAQUS finite element code uses explicit integration with a quasi‑consistent tangent stiffness matrix.

 

 Part two presents an extensive program of numerical experiments to evaluate the mechanisms controlling the performance of excavations in clay. The analyses focus on the short-term behavior for deep excavations in Boston Blue Clay, supported by structural diaphragm walls and braced internally using techniques of top down construction. The principal parameters considered in the study include the stress history and modeling of clay behavior, wall length and support spacing. The results show that the excavation sequence induces small strains in the surrounding soil, and imposes complex loading on soil elements, characterized by principal stress rotation and stress reversals. These loading paths generate earth pressures, acting over the excavated height of the diaphragm wall, which can exceed the initial lateral stresses. Improved modeling of this behavior using MIT‑E3 leads to more realistic predictions of wall deflections, surface settlements and failure mechanisms in the soil than can obtained using simple elasto-plastic models. Wall length has negligible effect on pre‑failure deformations but affects the mechanisms of failure. Summary charts describe the expected wall deflections and ground movements as functions of the support conditions and excavation depth. Further predictions show that excavation response for clay deposits with variable stress history can be interpreted from results obtained using simplified soil profiles.

 

 The third part describes a detailed simulation of the real time construction sequence for the seven‑story, underground garage at Post Office Square in Boston. Model predictions are evaluated through comparisons with extensive field data, including wall deflections, soil deformations, surface settlements and piezometric elevations. Differences between predicted and measured wall movements are attributed primarily to post‑construction shrinkage of the roof and floor system, which brace the structure, while predicted settlements are affected by modeling of piezometric elevations in the underlying rock. The results demonstrate those reliable and consistent predictions of soil deformations and transient groundwater flow can be achieved by advanced methods of analysis, without recourse to parametric iteration. The study also emphasizes the need for adequate characterization of engineering properties for all pertinent soils in the profile.