Modelling effects of structure in soft natural clays
Geological processes give natural clays a different structure to that of clays that are reconstituted in the laboratory. In soft clays, this structure often breaks down under loading; this is called destructuration. This project aimed to develop a model to predict destructuration in soft natural clays. An understanding of the main characteristics of the behaviour of these clays detained was from data reported in the literature. Existing frameworks that describe the behaviour of these clays were reviewed, and basic concepts proposed to model structured soils. The Sensitivity framework (Cotecchia & Chandler, 2000) uses sensitivity as a parameter that can represent structure in both volumetric and stress space before significant destructuration takes place. Study of the behaviour of three soft clays with low to medium sensitivities; Sibari, Bothkennar and Pisa clays, demonstrated that sensitivity changes in parallel with destructuration during both volumetric compression and undrained shearing such that there is a single expression that directly relates change in normalised sensitivity to change in damage strain, where the increment of damage strain is the magnitude of the vector of plastic strain increment. This destructuration law was used to extend an existing model, the Three-surface kinematic hardening (3-SKH) model which was developed by Stallebrass & Taylor (1997) for reconstituted clays. The new model requires only three new parameters to represent structure and its degradation that can each be derived from data from a single isotropic compression test. They are: the initial sensitivity, which represents the initial degree of structure in the natural clay; the ultimate sensitivity, which represents the stable elements of structure in the clay; and the parameter k, which controls the rate of destructuration with plastic strains. The other parameters used are the same as in the 3-SKH model and are derived from data from tests on the corresponding reconstituted clay. The model was evaluated against data from tests on Bothkennar and Pisa clay. Qualitatively, the model could predict the important features of behaviour observed in these clays. Quantitatively, results of analyses showed that determining initial sensitivity in a consistent way by using the Sensitivity framework leads to predicted values of undrained shear strength within 10 to 20% of the experimental values. Typically destructuration was correctly predicted in analyses simulating volumetric compression, but it was over-predicted by about 15 to 25% in analyses simulating undrained tests. This could be improved in some cases by using an ultimate sensitivity greater than unity in analyses simulating tests on specimens that are likely to have stable elements of structure arising from fabric. Structural anisotropy seemed to influence the behaviour of Pisa clay, and a model including structural anisotropy may improve predictions on such soils. The main limitation of the current research is the difficulty in determining the initial stress state and sensitivity to be used in the analyses; improvement of this should be the prime aim of further work.