The positive effect of artificial densification on the stress-strain performance of granular materials is acknowledged from very ancient times but its importance greatly increased after the development of powerful machineries for the construction of large earth and rockfill dams. It is, however, pointed out that the extensive exploitation of gravelly soil is rarely supported by a thorough analytical assessment of the compaction effects on the constitutive relationships of materials, as would be desirable considering the massive dimension of these works and the complexity of typical loading conditions. The present research aims to fill this gap by means of a detailed experimental investigation and a theoretical analysis on the stress-strain response of dense gravels under monotonic and cyclic loading. The experimental campaign consists of a large number of triaxial tests performed at different initial mean effective stresses, following different stress paths and sequences, on artificially reconstituted samples of large dimensions compacted at different initial densities. The great care placed in the accuracy of laboratory instrumentation enables a high repeatability of experimental results, which is necessary to provide a clear focus on non-linearity in the limited strain range of the pre-failure response of the gravel. Based on the curve-fitting method the ingredients of an elasto-plastic constitutive model have been defined to predict the response of gravel under monotonic and cyclic loading. Elastic stiffness is simulated with a model derived from the literature which assumes a dependency on soil density together with inherent and stress-induced anisotropy. Plastic strain development from different initial stress and volume states of gravel is simulated by a critical state, multiple yielding constitutive model. Hardening and flow rules for the latter have been obtained by modifying previously existing laws in order better to reproduce the observations under monotonic compression, extension and cyclic loading. Validation of the proposed model is finally provided by comparing simulations and experimental results in a variety of testing conditions.

No abstract