Empirical models are widely used in dam risk analysis because they can estimate the dam breach parameters easily, which enables a quantitative analysis of the breach process to be conducted. In this study, by applying a regression-based method to a database collected worldwide, a new empirical model has been developed that can predict the peak discharge, final breach average width, and failure time. The reservoir shape coefficient, the ratio of the water level when the dam is breached and the final breach depth, as well as the dam height, have been selected as the control variables of the model. Further, in the new model, the dam types (i.e., homogeneous dam, zoned-fill dam and core wall dam), as well as the failure modes (i.e., overtopping and piping) are considered. The advantage of this new model is that the formulas have been developed based on a larger database of dam breach cases and the choice of the input parameters. The model performance and comparison with the existing empirical models testify to the rationality of the new model.

Shallow plate-load tests were simulated using a two-dimensional discrete element procedure together with parameters calibrated from a series of biaxial compression experiments on aluminum rods. The deformation and force transmission characteristics within ground materials were investigated first, and then attentions were focused on studying the influences of the interparticle friction coefficient, the roughness of the loading plate, and the size of loading plate on the testing results, including the ultimate bearing capacity, the deformation behavior, and the failure mode. The well-known scale effect in plate-load tests was also analyzed from the perspectives of both continuum mechanics and discrete element method simulations. It was proven that the nonlinear strength behavior of the materials beneath the foundation is a possible source of the scale effect; however, this effect may be counterbalanced by the boundary restriction effect in laboratory experiments.

Large-scale triaxial compression and creep experiments were conducted with a saturated limestone rockfill material. The relationships between dilatancy ratio and stress ratio during loading and creeping were analyzed. It was found that the dilatancy ratio during both loading and creeping decreased nonlinearly when the stress ratio increased. In addition, the positive dilatancy ratio was considerably higher in creep than in vertical loading under the same stress state. Experimental data also showed that creep strains evolve linearly with the logarithm of time after a certain period, and both the final amounts of creep strains at the end of experiments and their rates are influenced by the stress states. A higher confining pressure and higher deviatoric stress lead to larger creep strains and higher strain rates. Empirical equations relating the final amounts of creep strains and their rates to the stress states are suggested.

A large-scale triaxial testing apparatus was upgraded by incorporating two pressure stabilization techniques so as to suppress the confining pressure fluctuation in dynamic experiments. Then it was used to study the residual strain behavior of a rockfill material subjected to different initial stress states and dynamic loading. It was found that the strain of the material under dynamic loading can be decomposed into a cyclic part and a residual part. In particular, the residual strains increase rapidly during the first few loading cycles. The strain accumulation rate, however, decreases gradually as the number of cycles increases. To simulate the increase of the residual (plastic) strains with the increasing number of loading cycles, a power function and an exponential function, taking into account the initial stress state, the cyclic shear strain amplitude, and the number of loading cycles, were proposed for the plastic shear strain and the plastic volumetric strain, respectively. The way to use ...