Читать книгу Instabilities Modeling in Geomechanics - Jean Sulem - Страница 15

To provide an experimental illustration of a local instability, under the assumption of the absence of localization (diffuse plastic strain), is almost impossible. Figure 1.3(a) shows a series of triaxial test results with increasing confining stress values, from Paterson (1958) for Wombeyan marble. All but those with the highest confining stress eventually exhibit unstable behavior, but for pressure values lower than 35 MPa the material exhibits localized instability, either as a vertical spalling (1) or single (2) or conjugate (3) shear bands, whereas for larger confining stress the behavior is qualified as ductile (4,5) for 70 and 100 MPa (3,4) (Figure 1.3(b)).

Figure 1.3. Triaxial compression of Wombeyan marble. (a) Axial stress–strain curves; (b) localization and diffused damage modes, from Patterson (1958)

Notably, for a higher level of confining stress in these tests (70 and 100 MPa), the deformation pattern, with a substantial inelastic component, can be viewed as uniform or non-localized. The stress–strain curves do not suggest an unstable or non-unique behavior. Nevertheless, the pattern formation of microcracks brings some concerns about the homogeneity of the strain and stress distribution across the sample, which in reality is not homogeneous. However, rigorous or at least somewhat codified measuring and understanding of what is “sufficiently homogeneous” are conspicuously missing.

In contrast, for all tests below 35 MPa of confining stress, one or more of stability criteria are failed, but the deformation is invariably localized at a certain point.

Acoustic emission recording techniques allow us to monitor sound-emitting micro-fracturing, which initiates long before the loss of stress–strain curve linearity, even before noticeable dilatancy onset, and long before approaching the peak stress in triaxial conditions for marble (Figure 1.4; Hallbauer et al. 1973). Indeed, what could qualify as an onset of non-uniqueness and instability coincides rather with the coalescence of microcracks into a shear band or macrocrack. Notably, the crack pattern evolution is a gradual process and does not suggest any threshold behavior or values. However, that is not always the case, as seen in comparisons of uniaxial failure in salt-rock, granite and marble (see Figure 1.7 from Zhang et al. (2015)).

Figure 1.4. Axial and lateral stress measured on a set of argillaceous quartzite with the corresponding evolution of the distribution of microcracking (Hallbauer et al. 1973)

For sand, the situation is quite similar. Unstable and non-unique behavior is seen in triaxial tests at low confining stress, and is invariably associated with a localized deformation into a shear band (Figure 1.5; Vardoulakis et al. 1978).

Figure 1.5. Biaxial compression of sand with visible localized shear band (Vardoulakis et al. 1978)

Homogeneous behavior at higher confining stress is rarely seen in triaxial strain, and is usually associated with material stability and uniqueness of response to drained tests. There is no established database to support the claim that there may be unstable sand behavior with a uniform strain across the sample, at least in drained triaxial tests (Drescher 2016). Karner et al. (2008) report sound emission attributed to intense intergranular friction at low mean stress, but at higher stress (and sometimes elevated temperature), no strain localization is mentioned, while post-test observations indicate grain breaking and comminution. The associated stress–strain curve implies stable behavior at high confinement (Figure 1.6).

Figure 1.6. (a) Low and (b) high confining stress compression of a quartz sand: isotropic effective stress versus porosity and acoustic emission decreasing after yielding at high confinement; (c) deviatore stress–strain curves showing stable behavior at 24°C at high confining pressure (Karner et al. 2008). For a color version of this figure, see www.iste.co.uk/stefanou/instabilities.zip

Figure 1.7. Comparison of uniaxial compression of rock salt, granite and marble, with a different intensity of acoustic emission at different stages of loading (Zhang et al. 2015). For a color version of this figure, see www.iste.co.uk/stefanou/instabilities.zip

Figure 1.8. Evolution of the distribution of acoustic emission during uniaxial compression of salt rock, granite and marble. For a color version of this figure, see www.iste.co.uk/stefanou/instabilities.zip

A separate issue arises in undrained tests on sands, during which the volumetric strain is imposed to be constant. In such tests, the material exhibits unstable behavior; however, there is no indication of localization; in other words the deformation appears to be homogeneous or diffuse. The term “diffuse failure” has been adopted for this type of behavior (Daouadji et al. 2011). Figure 1.9 shows the corresponding stress–strain curve and the effective stress path for such a test on Hostun sand. Rightfully, Daouadji et al. indicate a restriction of the undrained or constant volume conditions, clearly imposing a peculiar deformation pattern.

Figure 1.9. Unstable behavior during an undrained test of Hostun sand (Daouadji et al., 2011). For a color version of this figure, see www.iste.co.uk/stefanou/instabilities.zip

Also interesting is the instability developing in sand during constant deviatoric stress drained tests. In these tests, failure coincides with the changing sign of the volumetric strain (or at a maximum of attained dilatancy) (see Figure 1.10). However, in such tests the condition of static admissibility of the local stress rate at the softening regime may be violated. As will be seen later in the paper, experimentation with thermal pressurization of clay at constant stress deviator undrained heating test leads to a similar response (Hueckel and Pellegrini 1991). However, in that test both the localized and the diffuse strains developed at failure.

There are several observations to be made concerning laboratory experiments on small triaxial samples. The underlying assumption for such experiments is that all the fields: stress, strain and plastic strain (and possibly microcracking) are uniform across the sample until a possible appearance of localization. This unfortunately is not necessarily true, as shown in Figure 1.4 (Hallbauer et al. 1973). Similarly, the evolution of local porosity monitored via CAT in sand suggests an early loss of uniformity prior to shear banding (Desrues et al. 1996). Stress is obviously not measurable directly in such experiments. It is usually considered as an average resulting from the measured force assumed as uniformly distributed across the area. A number of reasons for the non-uniformity are quoted, such as axial symmetry of specimens and a development of roughly planar shear bands, or stress concentration at piston boundaries, with a different intensity in sand, clay or rock. Remedies to the experimental techniques have been sought, by introducing truly triaxial testing (Muir Wood 1973), biaxial strain testing (e.g. Vardoulakis 1978), etc. Drescher and De Josselin de Jong (1972) conducted a series of tests in which they subjected a 2D photosensitive granular medium to shear under constant vertical load between two rigid smooth arms rotating around a pin with a controlled rate, causing a displacement of the medium between the arms with a globally unstable or stable force response, depending on the direction of the medium displacement, and dilative or contractile volumetric strain. These and many other subsequent similar experiments have shown that contact stresses lead to the formation of chains of compressed column forces within the medium, separated by lightly or completely unloaded grains. In addition, said compressed columns undergo periodic unstable buckling and rebuilding of such columns, so that the entire process – while apparently monotonic at the force– displacement level – is unstable, non-homogeneous and non-monotonic at the level of grain structures, which form vortices and other patterns (Kozicki and Tejchman 2016). Similar conclusions were derived from much later studies of discrete element methods (Iwashita and Oda 1998). This prompted the investigation of even small- scale experiments as boundary value problems and the treating of their stability and uniqueness through global criteria. Finally, there is a question of validity beyond the point of loss of stability or uniqueness of the stress–strain curves obtained in a single experiment. Indeed, each such curve in a non-uniqueness range is just one of an infinity of possible responses, as they go through a singularity point of H = 0.

Figure 1.10. Stress–strain curves in a q = cons. drained test with instability (Daouadji et al. 2011). For a color version of this figure, see www.iste.co.uk/stefanou/instabilities.zip

Some additional insight may be expected from multiscale analyses using discrete element method (DEM) computations, together with progress in a rigorous approach to interscale data interpretation.