The objective of this dissertation is to effectively estimate the reliable unsaturated shear strength and soil water characteristic curve, which are required for the slope stability and seepage analyses of unsaturated soil slopes.
A simple but reliable method that could more effectively predict the unsaturated shear strength with respect to the matric suction was proposed. The nonlinear equation, containing only one more empirical parameter ($C_max$) for the purpose of representing the unsaturated shear strength, was formulated as a hyperbolic type. Validation of the proposed equation was shown with regard to the published data and test results obtained in this study. It has been found that the proposed hyperbolic formulation could well reproduce the apparent cohesion of unsaturated soils, nonlinearly increasing with the increase in the matric suction.
An interrelationship between basic soil properties that can be easily measured and the ultimate increment of apparent cohesion ($C_max$) could be acquired using the ANN(artificial neural network), based on the collected data as well as data obtained from this study. We were also able to understand the degree of effects of the input parameters used in the neural network on the output by conducting a parametric study. Furthermore, the network model resulted in relatively reliable predictions of $C_max$, even though new input parameter data were given in the trained network. In virtue of this work, it is expected that the problem of limited use of unsaturated shear strength can be solved and this nonlinear unsaturated shear strength modeling can be easily adopted in existing geotechnical analysis methods by simple modifications.
In order to reasonably obtain the soil water characteristic curve(SWCC) for the domestic weathered granite soils in experiments, the test apparatus and test procedure were modified. One of differences in the test process used in this study was to measure the volume change of the specimen ...