Physical and numerical modelization of the behaviour of rockfill dams during overtopping scenarios

Abstract

In recent years the technology on embankments dams has developed sensibly due to the advances in soil mechanic, and in all related sciences. Nevertheless their vulnerability to overtopping is still an open theme of discussion. Such an extreme phenomenon can lead to catastrophic consequences with loss of stability and failure of the entire structure. In spite of that, the consequences of an overtopping are still difficult to analyze and to prevent optimally. The possibility to define a numerical instrument to provide support for analyzing the failure of a dam is an important step ahead for organizing the intervention measures and for optimizing the economic plan. This is the principal reason why the Spanish ministry of Science and Innovation is funding the XPRES project, between the Polytechnic University of Madrid (UPM), the Centre for Hydrographical Studies of CEDEX and the International Centre for Numerical Methods in Engineering (CIMNE). The characterization of the failure process is dealt with by combining new numerical models, developed by CIMNE, with experimental tests on physical models of different sizes and scales carried out by UPM and CEDEX. In the present paper the work carried out by CIMNE is presented. The numerical model is based on combination of Lagrangian Particle Finite Element Method (PFEM) and Eulerian level set techniques in order to optimize the coupling between the dynamic effect of water inside and outside the dam, with the structural deformation and collapse induced on the down stream shoulder. The breaching formation and the consequent beginning of the failure are analyzed with particular interest. The specific features of the adopted method make it appropriate to treat the rockfill material and its large deformations and shape changes. A variable viscosity law and an energetic erosion criterion are used to take into account mass sliding and erosion acting in combined or alternative way.