The main objective of this study regards the analysis of results of a series of compressive strength tests on granite rocks carried out on a press with servo-controlled loading in order to characterize their strain softening behavior. We pay particular attention to the evolution of failure criteria from peak to residual strength that obviously presents plastic strain dependence. A graphical methodology was used to link strength with plastic strain all along the process of plastic yield and for different levels of confining pressure. It was possible to fit a mathematical expression, which is able to reflect this relation and which also allows obtaining evolving strength criteria (cohesion and friction values). We have also accounted for the post-failure dilation behavior of these rocks, studied elsewhere. The derived model has been implemented in a numerical code, and some simple simulations are carried out. These models have been able to reproduce the initial experimental results reasonably well.
The relevance of a thorough knowledge of the complete stress-strain curve in rock has frequently been highlighted in the rock mechanics field, and most particularly, in its earlier breakthroughs (Hudson et al., 1971). The aim of the present research is to study the actual behavior of rocks and propose strain softening and dilation models able to better represent the observed stress-strain response of rock samples.
It is widely acknowledged that rocks, and also rock masses, suffer a strain softening or strength weakening process after achieving a peak stress. This is a complex behavior, so even simple strain softening models need a good number of parameters to reproduce its most relevant features.
Simple strain-softening models for rocks require at least elastic parameters (E and ν), peak and residual strength envelopes (typically two parameters each, as for instance cpeak, ϕpeak, cres and ϕres) and two post failure parameters (for instance drop modulus M and dilatancy φ). These 8 parameters serve to recover the main features of this type of behavior (see upper graphs of Fig. 1). However, one must admit that this representation is still not very similar to the actual stress-strain response of rock tests, as it can be derived from the comparison of this representation to actual test results, as shown in the lower graphs of Figure 1. In this paper we investigate a possible approach to provide a more accurate model to simulate strain-softening behavior of granitic rocks as observed in the lab.
To study this behavior we have performed in our lab 3 series of around thirty unconfined and confined triaxial tests (σ3 in the range 0–12MPa) in three different granitic rocks named Amarelo País, Blanco Mera and Vilachán, respectively (Fig. 2). To do this, a servo controlled press was modified to control the axial stress in triaxial tests and also to measure the volume of hydraulic fluid displaced from Hoek's cell, so that the volumetric strain in the sample can be computed.