An incrementally non-linear model for clays with directional stiffness and a small strain emphasis

In response to construction activities and loads from permanent structures, soil generally is subjected to a variety of loading modes varying both in time and location. It also has been increasingly appreciated that the strains around well-designed foundations, excavations and tunnels are mostly small, with soil responses at this strain level generally being non-linear and anisotropic. To make accurate prediction of the performance of a geo-system, it is highly desirable to understand soil behavior at small strains along multiple loading directions, and accordingly to incorporate these responses in an appropriate constitutive model implemented in a finite element analysis.; This dissertation presents a model based on a series of stress probe tests with small strain measurements performed on compressible Chicago glacial clays. The proposed model is formulated in an original constitutive framework, in which the tangent stiffness matrix is constructed in accordance with the mechanical nature of frictional materials and the tangent moduli therein are described explicitly. The stiffness description includes evolution relations with regard to length of stress path, and directionality relations in terms of stress path direction. The former relations provide distinctive definitions for small-strain and large-strain behaviors, and distinguish soil responses in shearing and compression. The latter relations make this model incrementally non-linear and thus capable of modeling inelastic behavior.; A new algorithm based on a classical substepping scheme is developed to numerically integrate this model. A consistent tangent matrix is derived for the proposed model with the upgraded substepping scheme. The code is written in FORTRAN and implemented in FEM via UMAT of ABAQUS. The model is exercised in a variety of applications ranging from oedometer, triaxial and biaxial test simulations to a C-class prediction for a well-instrumented excavation. The computed results indicate that this model is successful in reproducing soil responses in both laboratory and field situations.