Sie sind auf Seite 1von 10

ARMA 17-0008

A new model for failure and yield envelopes of anisotropic porous sandstone
Louis, L.
New England Research, White River Junction, VT, USA
Baud, P.
Institut de Physique du Globe de Strasbourg (UMR 7516 CNRS), EOST, Université de Strasbourg, Strasbourg, France
Wong, T.-f.
Earth System Science Programme, Faculty of Science, The Chinese University of Hong Kong, Hong Kong

Copyright 2017 ARMA, American Rock Mechanics Association


This paper was prepared for presentation at the 51st US Rock Mechanics / Geomechanics Symposium held in San Francisco, California, USA, 25-
28 June 2017. This paper was selected for presentation at the symposium by an ARMA Technical Program Committee based on a technical and
critical review of the paper by a minimum of two technical reviewers. The material, as presented, does not necessarily reflect any position of ARMA,
its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent
of ARMA is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 200 words; illustrations may not be copied. The
abstract must contain conspicuous acknowledgement of where and by whom the paper was presented.

ABSTRACT: The objective of this paper is to provide an overview of a new data-based approach to understanding and modeling
strength anisotropy in porous sandstone with potential application to a wider range of geomaterials. The rationale for the contribution
is the interest in quantifying the impact of mechanical anisotropy in practical industrial applications such as the prediction of
mechanical behavior of hydrocarbon reservoirs during depletion or injection. The onset of yielding can have profound impact on
flow and elastic properties, and it is still unclear to what extent anisotropy may affect those predictions. After recalling an extensive
mechanical data set obtained previously on samples of Rothbach sandstone, we define our modeling strategy using other experimental
and theoretical work from the literature. This results in the necessity of determining what an appropriate isotropic yield envelope
might be. We propose an approach whereby isotropic and anisotropic models are defined and parametrized sequentially. We obtain
a consistent framework where the microstructural controls on strength, including anisotropy, may be understood better than
previously, allowing the revisiting of legacy data sets as well as the designing of better informed geomechanical testing programs in
core analysis.

and its bedding plane does not appear to constitute a plane


1. INTRODUCTION of weakness when favorably oriented with respect to the
Mechanical anisotropy is known to adversely affect maximum compressive stress. Samples cored at 45
strength predictions, especially when associated with the degrees to the bedding exhibit intermediate strength with
presence of weak layers [1-3], which can result in respect to the ones cored parallel and perpendicular to it
unexpected borehole failure. In homogeneous materials, [4].
anisotropy has been showed to directly affect the shape of The objective of this work is to provide a framework
yield surfaces [4-7] and can also cause the strain field to capable of handling anisotropic data sets. In order to do
be oblique to the macroscopic stress tensor. In this latter so, we use the concept of modified stress tensor detailed
case, an important question to be answered is to what in Tobita (1988) [17] and used by Oda (1993) [7] on
extent mechanical anisotropy should be incorporated into anisotropically consolidated sand. This approach
geomechanical models, and how does it ultimately proposes to use a contact tensor (considered as the source
compare with work that has been done on the effect of of the observed mechanical anisotropy) to rewrite the
true triaxial states of stress [8]. stress tensor that is applied to an otherwise isotropic
material. Though this modification of the stress tensor is
The present paper builds on an extended petrophysical
straightforward, the problem then hinges on the choice of
and mechanical data set that was acquired over the years
an appropriate isotropic model.
in the Rothbach sandstone, a Triassic 20% porosity cross-
Instead of resorting to popular yield loci geometries
bedded fluvial sandstone belonging to the Buntsandstein
derived from plasticity models such as cam clay and yield
Formation in Northern Europe [4; 9-16]. The Rothbach
cap [18-20], we focus on determining the stress
sandstone is anisotropic with respect to key physical
conditions for failure in a simple 2D granular lattice. This
properties such as permeability, acoustic velocities and
results in the definition of a new set of yield surfaces for
mechanical strength. As far as mechanical strength is
an isotropic aggregate. This approach allows us to extract
concerned, this sandstone presents the particularity of
from the available data a stress sensitivity curve for the
being stronger when loaded perpendicular to the bedding,
effective grain contact area, and also interpret the corner 68% quartz, 16% Feldspar, 12% clays and about 3%
that is observed at the junction between brittle and oxides and mica.
compactive yield surfaces in terms of a ratio between the
Despite a clearly visible layering, the Rothbach sandstone
idealized shear and compactive strengths of the
presents a larger strength as well as static and dynamic
intergranular contacts.
stiffness perpendicular to the bedding compared to
As a result, we demonstrate that: (1) The geometry of the
parallel to it. This has been proposed to originate from the
yield surfaces of an isotropic material strongly depends
substantial anisotropy that is observed in the
on the ratio between microstructural shear and
microstructure and more specifically in the intergranular
compactive strengths, as well as on the sensitivity of the
contacts length and orientation statistics (Fig. 1c). This is
intergranular contacts to confining stress; (2) Anisotropy
also supported by SEM and cathodoluminescence
essentially affects the yield surfaces by introducing a
observations that have suggested a relatively intense
multiplier in the principal stress space compared to the
compaction history (Fig. 1d-e).
reference isotropic case; (3) At relatively high confining
stress, anisotropy results in a drastic change in the shape
of the compactive yield surface due to a rotation of the
180
principal shortening direction. Perpendicular
160 Oblique

2. MECHANICAL DATA 140 Parallel

Differential stress (MPa)


120

100
The mechanical data set used here was presented in its
entirety in Louis et al. (2009) [4]. Parts of it had been 80
previously obtained by Wong et al (1997) [10], Baud et
60
al. (2004; 2005) [13-14] and Louis et al. (2007) [16]. The
sandstone tested, which is named Rothbach after a quarry 40
in Alsace (France) is also known as the “Vosgian 20
sandstone” and is found in numerous outcrops on either
side of the Rhine Graben in Northern Europe. 0
0 50 100 150 200 250
Effective mean stress (MPa)

c. Fig. 2. Peak stress and onset of inelastic compaction in samples


a.
of Rothbach sandstone cored perpendicular, oblique (45
degrees) and parallel to the bedding (Louis et al. [4]).

The failure data showed in Figure 2 consist of strength test


results obtained over a wide range of effective pressure
d. conditions (from 5 to 250 MPa) in series of plugs cored
e. perpendicular, oblique (45 degrees) and parallel to the
b.
bedding plane. Contrary to what is often seen in foliated
rocks [1-3], the bedding surface does not constitute a
plane of weakness and the plugs cored at 45 degrees fail
at intermediate level of differential stress between the
ones for parallel and perpendicular samples. In their
Fig. 1. a. Strasbourg Cathedral, Strasbourg, France. b. Block of study, Louis et al. [4] proceeded to investigate in greater
Vosges sandstone with laminations. c. Intergranular contacts depth the respective roles microstructural anisotropy and
length (grey line) and orientation (black line) statistics from thin heterogeneity on strength and spatial distribution of
section analysis. d-e. SEM and cathodoluminescence image of damage. When analyzing the compaction behavior of the
a grain contact. three sets of samples in the differential stress vs. effective
mean stress space (later referred to as the P:Q space), (1)
The Vosgian sandstone has long been used as building the yield envelopes were all fitted individually as would
material (Fig. 1a). In Figure 1b, a picture of the block have been done for data coming from different materials,
tested by Louis et al. (2003) [12] exhibits mm to cm thick and (2) elliptical caps were used for the initial yield loci
cross lamination with slight compositional and grain size as well as at prescribed levels of volumetric strain. While
variations. Its porosity is about 20% and David et al. this approach was considered satisfactory in terms of fit
(1994) [9] provides a modal mineralogical composition of to the experimental data, this unique data set was not
analyzed through a single theoretical anisotropic Note that in the compactive regime (region B), if one
framework. assumed that the strength and size of the grain contacts
were fixed, and neglecting porosity changes, the axial
The present paper is an attempt to establish this
stress should be solely responsible for failure. As a result,
framework.
the compaction yield curve should be perfectly horizontal
in the principal stresses space (and show a slope of -1.5 in
the P:Q space). As it only would seem to constitute an
3. MODELING APPROACH
extreme case, this means that another critical element is
needed, and that is the stress sensitivity of the grain
contacts, which causes contacts surface area to increase as
3.1. Two stress-sensitive strength envelopes the stresses rise. This element must thus be incorporated
into the model as a function of the applied stresses.
We first replot the data of Figure 2 in the principal stress From the observation of the data obtained on the plugs
space (axial and radial). In this space, the hydrostatic parallel to bedding, we retain that two distinct
loading path lies of the 1:1 line. Focusing on the data mechanisms must be honored (i.e. two yield envelopes, as
obtained on the plugs cored parallel to the bedding - in the series of models introduced by DiMaggio and
which are here the weaker ones - we can separate the yield Sandler [20]), and also that the contact surface area must
envelope into two regions. be present in the model as a stress sensitive parameter. It
300 can be easily shown that ignoring stress sensitivity would
produce two straight lines for yield surfaces, the one for
250 compaction being horizontal in the principal stresses
space and at a slope of -1.5 in the P:Q space.
200
B
sAx (MPa)

3.2. Anisotropy and oblique compaction


150

A Perpendicular
100
Another important feature to notice in the data of figure 3
Oblique
is the peculiar behavior of the yield curve for plugs cored
Parallel perpendicular to bedding (red curve). In the compaction
50
domain at high stresses, there is a decrease in the axial
stress needed for failure down to the hydrostatic grain
0 crushing pressure also referred to as P*. What is the
0 50 100 150 200 250 300 microstructural mechanism that allows the initiation of
sRad (MPa)
compaction at lower total axial stress under higher
Fig. 3. Mechanical data of Figure 2 plotted in the principal confinement? The answer lies in anisotropy and similarly
stresses reference. shaped envelopes are commonplace in normally
consolidated soils (Wood [21]).
In region A, as the radial (i.e. confining) stress increases,
it is necessary to apply an increasing amount of
differential stress in order to attain failure. In other words,
in that region, the radial stress is a hindrance to failure.
On the other hand, in region B, as the radial stress
increases, progressively lower amounts of differential
stress are necessary to attain failure. In this case we 90 75 60 45 30 15 0

consider that the radial stress is promoting failure. This


observation is consistent with the concept that, Fig. 4. Experiment of Allirot et al. (1977) [21]. Samples of
microstructurally, distinct failure mechanisms are at play diatomite with different coring orientations deform
with a domain dominated by shear failure at grain anisotropically and mainly perpendicularly to the bedding in
contacts, which is logically hindered by an increase of hydrostatic stress conditions.
confining stress (results in larger normal stress on
contacts), and a domain dominated by grain fracturing and A convincing illustration of a possible mechanism at play
crushing, in which the confining stress plays a promoting that involves anisotropy can be found in the work of
role for failure since confining and differential stresses Allirot et al. (1977) [22]. Their experiment consisted of
add up to increase normal stress at contacts, which is in deforming under hydrostatic stress up to 40 MPa
this case responsible for the initiation of compaction. cylindrical samples of highly porous diatomite that had
been cored in different directions with respect to the
bedding. Figure 4 summarizes the results obtained at the Note that Oda (1993) used this concept along with a
highest pressure for all orientations starting from nearly modified Cam Clay plastic model (elliptical yield locus)
identical size cylinders. The diatomite exhibiting a much to successfully map the yield envelopes of anisotropically
weaker strength perpendicular to the bedding compared consolidated Aio sand from Yasufuku [6]. More recently,
to parallel to it, regardless of the plug orientation, Oda’s results have been showed to compare favorably to
isotropic stress loading resulted in a strongly anisotropic the more elaborate MIT-E3 plastic model (Akl and Louis
strain field with the largest compactive distortion always [23]).
observed perpendicular to the bedding.
Here, the concept of modified stress will be used in its
These observations materialize the fact that compactive simplest form where we will assume collinearity between
deformation is the result of the interplay between the the grain fabric and macroscopic stress tensor, and reduce
macroscopic stress state and intrinsic sample anisotropy. our problem to finding a unique coefficient responsible
To that regard, the experiment of Allirot et al. [22] for the behavior of the perpendicular samples relative to
isolated the effect of intrinsic anisotropy. the parallel ones. The complexity that arises for the case
of a microstructural fabric oblique to the macroscopic
Let’s now consider an anisotropic sample being loaded
stress tensor is beyond the scope of this study and will be
triaxially along its stronger direction. The transverse
addressed in a separate work.
orientation being weaker, it is expected that lateral
compaction may initiate at lower stress than it would
vertically. If the axial and confining stresses are close
4. ISOTROPIC MODEL
enough, then the anisotropy can result in meeting the
failure conditions first along a direction oblique to the
sample axis. In this case, the sample may compact at a
4.1. General scheme
lower macroscopic axial stress – i.e. the yield envelope
goes down in the principal stresses space. As the
confining pressure increases and the macroscopic stress We find that the behavior observed macroscopically for
state approaches hydrostatic conditions and P*, the the samples cored parallel to bedding in Figure 3 can be
apparent strengths of samples cored in any orientation reproduced using a simple 2D analog of three contacting
converge since they are all reflecting the onset of grain spheres. This conceptual model is showed in Figure 5.
crushing along the weaker direction. We propose this Looking at the spheres arrangement, forces can be
mechanism to be the reason for the shape of the virtually applied in order to simulate (1) dilatant behavior
compactive yield envelope in the vicinity of the by reaching the tensile strength ST* of the lower vertical
hydrostatic grain crushing pressure P* for samples that contact, (2) shear failure by reaching the shear strength SS*
were not cored along their weakest direction. on the two oblique contacts or (3) compaction by attaining
the grain crushing stress SN* on these same contacts.

3.3. Modified stress tensor


sAx
As we are proposing that, in addition to stress sensitivity
at intergranular contacts, intrinsic anisotropy also
2R
contributes to the shape of the yield envelopes, a
framework is needed to incorporate this anisotropy into a
sRad sS sRad
model. To that effect, the concept of modified stress used
by Oda [7] and detailed in Tobita [17] is going to be used.
sN 2r
The modified stress concept consists of assuming that sT
intrinsic anisotropy arises from a fabric at the level of
intergranular contacts, which is precisely the case for the
Rothbach sandstone. A fabric tensor is determined and
used to recompute the stresses applied to an otherwise
isotropic material. Introducing a fabric tensor F of trace 1
sAx
reducing to I/3 in the case of isotropy, a modified stress
tensor T can be written from the macroscopic stress tensor Fig. 5. Scheme used for the analog aggregate.
s as follows:
1 1 Note that the star symbol ‘*’ is being used throughout this
Tij  Fik s kj (1) paper to signify that the stress considered is associated
3
with material yielding. As sketched in Figure 5, the
transmission of forces between the macroscopic and Hertz model for normal compression of two identical
microscopic scales is not explicitly resolved and is spheres (see for instance Mavko et al., 1998 [24]).
parametrized in the model by using a multiplier that
a. b.
combines porosity and a loosely defined structural 300 300

intensification factor . We will see later that the 250 250

Differentail Stress Q (MPa)


characteristic parameters extracted from the mechanical

Axial Stress (MPa)


Dilatancy
200 200
Shear
data are in fact algebraically insensitive to this scaling Compaction
150 150
factor.
100 Dilatancy 100
The equation obtained for the tensile, shear and normal Shear
Compaction
50 50
stresses applied on the intergranular contacts write as Hydrostatic Line

follows: 0
0 50 100 150 200 250 300
0
0 50 100 150 200 250 300
Radial Stess (MPa) Mean Stress P (MPa)

Fig. 6. Graphical expression of the isotropic model of Figure 5


3  R2
sN   s Ax (2) through equations (2) to (4). (a) In the principal stresses
4 1 r2 reference. (b) In the P:Q space.

1  R2
sS   (s Ax  s Rad ) (3) The features of the yield envelopes in Figure 6
4 1 r2 satisfactorily honor what is traditionally observed in
mechanical data sets. The successive failure hindrance
3  R2 3 1
sT   ( s Rad  s Ax ) (4) and failure promotion effects of the two yield functions is
4 1 r 2
2 2 clearly represented. The following sections are focused on
analyzing these envelopes in greater depth. Note that the
modeling of the onset of dilatancy is going to be ignored
where  is the intensification factor which honors the from here and on as it does not reflect ultimate
effect of the structural arrangement of the aggregate,  the macroscopic failure in the data analyzed.
porosity, R the grain radius, r the radius of the
intergranular contacts, and sAx and sRad the macroscopic
axial and radial (i.e. confining) stresses. As mentioned 4.2. Envelopes intersect point
earlier, the radius of the intergranular contact is stress
sensitive. It is straightforward to rewrite equations (2) to At the transition between the two failure envelopes, which
(4) to isolate the axial stress term and make them suitable would correspond to the location where the slope of the
for plotting in the principal stresses space. Also, it is best fitting curve is equal to 1 in Figure 6a or to zero in
assumed that there is only one microscopic strength Figure 6b (the role of the confining stress changes from
criterion for each mechanism and that this criterion is hindrance to promotion of failure), both equations are
satisfied all along each respective yield curve. The true. This means that at this specific location, a ratio
equations for the envelopes in the principal stresses space which we name  can be defined between the normal and
write: shear contact strength. This ratio writes:

sN
*
s *Ax
4 1 r *
2   3  (8)
s*  s (5) sS
*
s *Ax  s Rad
*

3  R
Ax 2 N

This ratio is independent of the contact radius, hence of


1 r2 * the contact law that is being used. This coefficient locates
s* 4 s  s Rad (6)
Ax
 R2 S on a yield envelope the stress conditions for which it
becomes easier to yield in compaction than in shear, and
8 1 r2 * thus represents the brittle-ductile transition within the
s *
 3s Rad  s (7)
3  R framework of our model.
Ax 2 T

The ratio  also provides a valuable characteristic


parameter for the material being tested. This ratio should
For illustrative purposes, Figure 6 provides a result be essentially controlled by the degree of cementation and
obtained both in the principal stresses space and in the the predisposition of the grains to internally develop
P:Q space using these equations. For the stress sensitivity cracks and fail.  should be low in the case of cemented
of the intergranular contact radius, we used the classical weak grains and high in the case of strong non cohesive
grains. In the first case, the observed yield envelope 4.4. Application to the Rothbach parallel data set
should be mainly one of compaction with a relatively
small brittle domain, and in the other case, mainly one of
The Rothbach data set is anisotropic (as would likely be
brittle failure, i.e. a nonlinear Coulomb envelope, up until
the compaction domain is attained. most sandstones if investigated as thoroughly), but the
results obtained on samples cored parallel to the bedding
 is a key parameter and constitutes a direct output of our provide an example of a material that fails along the main
model as long as the transition from shear dominated to loading direction under any stress conditions. For
compaction dominated yielding is covered by the data set. simplicity, we will seek the modeling of the equivalent
isotropic material, with implications that have no impact
on the present demonstration. A complete analysis, which
4.3. Data analysis and stress sensitivity will be conducted elsewhere, needs to approach the data
sets obtained along all orientations together.
Starting from a given data set, the equations for two yield The first step of the envelope retrieval is to identify in the
envelopes need to be retrieved. This is going to be data set the location at which the transition between the
achieved through curve fitting. Let us define the function two failure regimes occurs in order to properly distribute
 extracted from equation (5) or (6) as follows: the data points. This has already been done in Figure 3
where the data point between regions A and B is
1 r2 * considered as belonging to both regimes. In practice, this
4 sS (9)
 R2 point must be located where the data trend transitions
from a slope greater than 1 to a slope smaller than 1.
The values of at each failure point can be obtained by
writing   s *Ax  s Rad for the points on the shear Based on the finding of the intersect point, the parameter
 can be calculated. This parameter represents the ratio of
dominated yield envelope and   3s *Ax /  on the the contact compaction strength vs. contact shear
compaction dominated yield envelope. These data can strength. In the Rothbach parallel data set, we obtain a
now be fitted as a function of radial (confining) stress. value of   S N* / S S*  3.7 .
Once a function has been obtained, the expression for
  f (s Rad ) can be redistributed in the two failure Then, the function can be retrieved as outlined in 4.3.,
the illustration of which is provided in Figure 7. Here,
regimes following s *Ax  (s Rad )  s Rad for the ‘shear’ since only the grain contact area is considered sensitive to
envelope and s *Ax  (s Rad ) / 3 for the ‘compactive’ stress, it is also responsible for the nonlinear relationship.
envelope.
140
The function   f (s Rad ) not only represents the core
120
behavior of the material studied. Within the frame of our
model, aside from stresses, the only parameter that is 100
allowed to vary as a function of the confining stress is the
 (MPa)

intergranular contact radius r, assuming that porosity 80


variations in the elastic domain are comparatively Shear
60
negligible. Hence,  is also a function that describes the Comp
evolution of the intergranular contact surface area with 40 Hertz:   63.5  1.43s Rad 2 / 3
confining stress. It is easy to verify from the equations just Power:   38 .6  s Rad 0.203
Ln:   22 . 6  16 . 6  Ln (1  s Rad )
provided that if were made insensitive to stress 20

(constant intergranular contact radius), the two yield 0


envelope would be two straight lines in either principal 0 50 100 150 200 250 300
stresses or P:Q spaces. Confining Stress (MPa)

The example of Figure 6 used a hertzian contact law for


illustrative purposes. Without formulating any such Fig. 7. Sensitivity of to the confining stress in the Rothbach
theoretical assumption, our approach allows studying the parallel data set.
relationship that emerges from the data.
In the following section, all the elements mentioned here A significant assumption that is made in our model and
are used to analyze the Rothbach parallel data set. needs to be mentioned here is that we consider the
intergranular contact surface area to be only sensitive to
the minimum stress, i.e. the hydrostatic component of the
stress field. And it is assumed that the addition of a  An intergranular contact strength ratio  that also
deviatoric stress is mostly handled at the microscopic acts as a relative scaling coefficient for the brittle
scale by shear motion on contacts of constant surface area. and compaction domains.
This assumption does not affect the process of finding a 
 A stress sensitive function  that comprises two
function since it is data-based and we only suggest what
fitting parameters recognized as the UCS strength
it might represent. But it allows us to compare the newly
found relationship to the classical hertzian result. In and the stress sensitivity factor .
Figure 7, several trends are compared. The one that is  A value for the hydrostatic grain crushing
called ‘Hertz’ is a power law function of the form pressure P* if not already available. P* can be
  0  A(s Rad
*
) n that honors an exponent of n=2/3 to easily obtained by solving for sAx=sRad in
the confining stress. Then the exponent n is allowed to equation (11).
deviate from 2/3 (‘Power’). Finally, a natural logarithm Therefore, within the frame of our model, the three
relationship of the form   0  BLn(1  s Rad*
) is also parameters UCS,  and  are able to fully define the brittle
proposed, with the virtue of locking-in the nonlinearity, and compaction yield envelopes of an isotropic sandstone.
thus reducing the number of fitting parameters to two.
Note that  in the relationships used actually represents
the UCS strength. The stress sensitivity factor  has a critical importance in
shaping the yield envelopes. That includes the slope of the
Picking the natural logarithm among the relationships compaction line at the hydrostatic crushing pressure P*.
obtained for , we now fit the failure envelopes of the This element is worth mentioning since it has been a
Rothbach parallel data set. The result is showed in Figure matter of discussions in recent studies (Gueguen and
8. And the final equations obtained in the principal Fortin, 2013 [25]; Baud et al., 2015 [26]). Though
stresses space are as follows: anisotropy is going to add complexity to that picture, as
far as the envelope of an isotropic material is concerned,
we see here that the slope of the compaction line in the
(Shear) s *Ax  22.6  s Rad  16.6Ln(1  s Rad ) (10) P:Q space can theoretically vary between 0 and -1.5, the
latter reflecting the absence of stress sensitivity
3.7
(Comp.) s *Ax  [22.6  16.6 Ln(1  s Rad )] (11) represented by a straight horizontal line in the principal
3 stresses space (material fails at a fixed axial stress
regardless of the confining pressure).
a. b.
300 300

250 250
5. ANISOTROPIC MODEL
Deviatoric Stress Q (MPa)
Axial Stress (MPa)

200 200

Shear Shear
150 150
Comp Comp

100 100 Using the isotropic model defined previously and


50
recalling the concept of modified stress used in Oda [7],
50
we are now in a position to model the mechanical
0 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
anisotropy of the Rothbach sandstone. For simplicity, we
Confining Stress (MPa) Mean Stress P (MPa) will only focus on the parallel and perpendicular data sets.
The oblique data set was very useful in showing that the
Fig. 8. Shear and compactive yield envelopes in the Rothbach
bedding plane does not constitute a plane of weakness,
parallel data set. a. Principal stresses reference. b. P:Q space.
thus allowing the use of a homogeneous anisotropic
approach. Note also the cross-bedded nature of the
Note that the multiplier in front of the natural logarithm sandstone and associated potential challenge in obtaining
represents the stress sensitivity factor for the rock studied a true 45 degrees orientation. Another working hypothesis
and should be identified and reported as an output of the here is that the principal axes of the internal fabric are
model. We thus propose to identify this stress sensitivity perfectly aligned with the macroscopic stress reference in
both parallel and perpendicular data sets.
factor by the letter .
Equation (1) taken from Oda [7] can be explicitly written
In summary, the simple scheme of Figure 5 allows us to
as follows:
use the strength data of a given granular material to
retrieve:
 T11 T12 T13  1 / F11 0 0  s 11 s 12 s 13  controlled by sample strength anisotropy. In these
  1   (12)
 T21 T22 T23    0 1 / F22 0  s 21 s 33 s 23  circumstances, we actually have T33>T11, what the yield
T  3
 31 T32 T33   0 0 1 / F33  s 31 s 32 s 33  envelope must reflect by essentially mirroring itself
across some line of symmetry since sRad or PConfining is now
The modified tensor Tij is non-symmetric in general and
the stress that is responsible for failure.
the obliquity of the stress and fabric references adds some
complexity that is not addressed in this paper. As we That line of symmetry in fact constitutes the ‘true’
assume here a perfect alignment between the fabric and hydrostatic stress state, which is not on the 1:1 line but
the stress tensor references, the non-diagonal stress terms rather oblique to it since a little more stress needs to be
are equal to zero and we can simply rewrite the effect of applied in the axial direction to achieve T11=T33.
anisotropy as: Another practical consequence of this understanding of
1 1 1 the effect of anisotropy is the fact that under traditional
T11  s 11 ; T22  s 22 ; T33  s 33 (13) hydrostatic loading (sAx=sRad), the value of P* should be
3F11 3F22 3F33
about the same regardless of plug coring angle since
Let’s identify better the applied stresses by writing failure always initiates along the same (weaker) direction.
s11=sAx and s22=s33=sRad=Pconfining. Equation (13) shows
a.
that the fabric is responsible for some amount of stress 300
anisotropy in the transverse plane whereby the applied
stresses s22 and s33 are equal but not the ones seen by the 250
isotropic aggregate T22 and T33.

Axial Stress (MPa)


200
As our model is two-dimensional, the workflow consists
of retaining the two stresses leading the maximum 150
differential stress, which can be named Tmax and Tmin, and
use them in place of sAx and sRad in equations (10) and 100
Parr Shear
(11). Parr Comp
50 Perp
Because of the microstructural fabric present in the 'True' Perp Hydro Line
Rothbach sandstone, for a given state of applied stress, 0
0 50 100 150 200 250 300
Tmax will be different among parallel and perpendicular
Confining Stress (MPa)
samples. And for most yield conditions, Tmax is also going b.
to be aligned with sAx. 300

For simplicity, we propose here to consider that both 250 Parr Shear
parallel and perpendicular samples respond in the same
Differential Stress Q (MPa)

Parr Comp
manner to the applied radial stress as far as contact surface 200 Perp
area goes, and focus on the difference in axial stress at 'True' Perp Hydro Line

failure (height of the envelopes in the principal stresses 150


space) which is governed by the ratio between F11 and F33.
Concretely, this equates to retrieving the multiplier that 100

brings the data points from the parallel samples onto the
50
ones of the perpendicular samples. Algebraically, this
multiplier is simply appended to equations (10) and (11).
0
0 50 100 150 200 250 300
Before introducing the figure displaying our results, we
Effective Mean Stress P (MPa)
need to address the other fundamental expression of
anisotropy in yield envelopes, which is the decrease of
Fig. 9. Result of the modeling of anisotropic yield envelopes for
axial yield stress in the vicinity of the hydrostatic crushing
the perpendicular and parallel Rothbach data sets. a. Principal
pressure P*. stresses space. b. P:Q space.
The assumption of collinearity between the fabric and
applied stress tensor allows the description of the
anisotropy effect to be provided in a very simple form. Figure 9 presents the final result of the modeling of the
Looking at the factors used for T11 and T33, if s11 and s33 anisotropic yield envelopes in the Rothbach sandstone.
are close to one another, the fabric may cause T33 to The envelope for the perpendicular samples is a scaled up
actually become greater than T11, leading the sample to version (along sAx) of the one for the parallel samples up
respond to the radial stress for failure as opposed to the until the modified lateral stress becomes greater than the
axial one. This is what the results of Allirot et al. [22] in modified vertical stress. There, the scheme is rotated to
Figure 4 illustrates: a preferential strain direction reflect a lateral compaction. Such singularity at the
junction between the two compaction curves is a direct redrawn upon some prescribed changes in porosity and
effect of the approach taken whereby the maximum and fabric, the evolution of which must be closely tied to the
minimum stress directions are instantaneously swapped. kinematics of the deformation.
Practical cases where perfect collinearity between fabric
Lastly, in order to take full advantage of the work, it is
and stress references should not be expected, not to
also necessary to define a practical engineering workflow
mention the fact that real grain size and contact statistics
to apply to a data set. The most important aspect of it is
are also expected to have a strong smoothing effect.
that only one coring direction should be available and it is
The envelopes for the perpendicular samples were unknown a priori how that direction is oriented relative to
obtained using a ratio of vertical to horizontal contacting maximum and minimum strength angles. Therefore, if
surface of F11/F22=1.19, which corresponds to a any anisotropy arises in the data set in the form of a
mechanical anisotropy of 17%. This number is consistent decrease of the axial strength towards P*, it is only an
with the cement radius ratio obtained by Louis et al. [12] indication of apparent anisotropy between the loading
from unconfined P-wave velocity data using a slightly direction and the minimal strength direction.
altered version of the cemented spheres model of Dvorkin
The proposed workflow is as follows:
and Nur (1996) [27].
In the discussion that follows, we recapitulate the
findings of this paper including some implications for Replot strength data in the axial and radial stress space
rock properties modeling, then provide a practical
engineering workflow that can be applied to a data set
Select data points up to the maximum axial stress and
obtained on plugs cored along one direction only. not past it (exclude decrease towards P*)

Identify the transition between regimes by sliding a 1:1


6. DISCUSSION AND CONCLUSION line onto the data set

Compute  using the stress conditions at the transition


This paper presented a new approach to modeling strength
envelopes in anisotropic porous sandstone, taking
advantage of a unique data set that comprised fully probed Rewrite the stresses to fit a unique  function using:
failure stresses for plugs cored in three different   UCS   . Ln(1  s Rad )
orientations with respect to the bedding. Though a
strategy was already available for honoring anisotropy
Write the equations for the yield envelopes as in (10)
(the modified stress tensor), observation of the data set and (11)
obtained parallel to the bedding plane led to defining a
new isotropic model that attempts to better relate the
Assuming P* is available as an experimental data
macroscopic strength to the microstructure. point, determine the multiplier to append to equation
(11) in order to intersect P*. This multiplier is the
Taking into account both the isotropic and anisotropic apparent anisotropy ratio
parts of our model, the yield envelopes of plugs cored in
various direction can be fully defined using:
 The three parameters UCS, , and . We believe such workflow should allow revisiting
 A 3D fabric tensor (of which the anisotropy ratio numerous existing mechanical data sets as well as
determined earlier is just one component). designing more sensible geomechanical testing programs
in core analysis.
A more thorough derivation including a general approach
to anisotropy for oblique fabric and stress tensor
references will be provided elsewhere. As proposed here, 7. REFERENCES
our model directly challenges the assumption of
normality during plastic deformation since anisotropy is
postulated to occasion compaction increment that are 1. Donath, F.A.. 1972. Effects of cohesion and
granularity on deformational behaviour of anisotropic
oblique to the loading axis. In fact, it appears that the
rock. In Studies in mineralogy and Precambrian
yielding of both consolidated porous rocks and soils geology, Geol Soc Am Mem 135, ed. B.R. Doe and
(sands and clays) could potentially be approached as D.K. Smith, 95-128.
presented in this study. Our model also comprises the
basic elements to defining a path towards plastic 2. McLamore, R., Gray, K.E.. 1967. The mechanical
modeling in the sense that envelopes can be easily behaviour of anisotropic sedimentary rocks. J Eng
Ind-T ASME 89: 62–73.
3. Rawling, G.C., Baud, P., Wong, T.-f.. 2002. 15. Louis, L., David, C., Metz, V., Robion, P., Menendez,
Dilatancy, brittle strength, and anisotropy of foliated B., Kissel, C.. 2005. Microstructural control on the
rocks: Experimental deformation and anisotropy of elastic and transport properties in
micromechanical modeling. Journal of Geophysical undeformed sandstones, Int. J. Rock Mech. 42(7–8):
Research. 107(B10): ETG8-1. 911–923.

4. Louis, L., Baud, P., Wong, T.-f.. 2009. Microstructural 16. Louis, L., Wong, T.-f., Baud, P., 2007. Imaging strain
inhomogeneity and mechanical anisotropy associated localization by X-ray radiography and digital image
with bedding in Rothbach sandstone. Pure appl. correlation: Deformation bands in Rothbach
Geophys.. 166: 1063-1087. sandstone. J. Struct. Geol. 29 (1): 129–140.

5. Baud, P., Meredith, P., Townend, E., 2012. 17. Tobita, T.. 1988. Contact tensor in constitutive model
Permeability evolution during triaxial compaction of for granular materials. In Studies Applied Mechanics,
an anisotropic porous sandstone. Journal of ed. Elsevier, London, 263-270.
Geophysical Research. 117: B05203.
18. Roscoe, K.H., and Burland, J.B.. 1968. On the
6. Yasufuku, N., Murata, H., Hyodo, M.. 1991. Yield generalized stress-strain behaviour of ‘wet’ clay. In
characteristics of anisotropically consolidated sand Engineering Plasticity, Eds J. Heyman and F.A.
under low and high stresses. Soils and Foundations Leckie, Cambridge: Cambridge University Press. 535-
31(1): 95-109. 609.

7. Oda, M.. 1993. Inherent and induced anisotropy in 19. Schofield, A.N., and Wroth, C.P.. 1968. Critical state
plasticity theory of granular soils. Mechanics of soil mechanics. London: McGraw-Hill.
Materials 16(1-2): 35-45.
20. DiMaggio, F.L. and I.S. Sandler. 1971. Material
8. Ma, X., and Haimson, B.C.. 2016. Failure model for granular soils. J. Eng. Mech. Div. Am. Soc.
characteristics of two porous sandstones subjected to Civ. Eng.. 97: 935-950.
true triaxial stresses, J. Geophys. Res. Solid Earth,
121, 6477–6498, doi:10.1002/2016JB012979. 21. Wood, D.M.. 1990. Soil behaviour and critical state
mechanics. New-York: Cambridge University Press.
9. David, C., Wong, T.-f., Zhu., W., Zhang, J.. 1994.
Laboratory measurement of compaction induced 22. Allirot, P., J.P. Boehler, A. Sawczuk. 1977.
permeability change in porous rocks: Implication for Irreversible deformation of an anisotropic rock under
the generation and maintenance of pore pressure hydrostatic pressure. Int. J. Rock Mech. Min. Sci. &
excess in the crust. Pure Appl. Geophys. 143: 425– Geomech. Abstr. 14: 77-83.
456.
23. Akl, S.A. and L. Louis. 2013. Using asymmetric yield
10. Wong T.-f., C. David, W. Zhu. 1997. The transition surfaces to model material anisotropy. In Proceedings
from brittle faulting to cataclastic flow in porous of the 47th US Rock Mechanics / Geomechanics
sandstones: Mechanical deformation. J. Geophys. Symposium, San Francisco, 23-26 June 2013,
Research. 102: 3009-3025. ARMA13-149.

11. Besuelle, P., Baud, P., Wong, T.-f.. 2003. Failure 24. Mavko, G., Mukerji, T., Dvorkin, J.. 1998. The rock
mode and spatial distribution of damage in Rothbach physics handbook: Tools for seismic analysis of
sandstone in the brittle-ductile transition. Pure Appl. porous media. Cambridge: Cambridge University
Geophys. 160 (5–6): 851–868. Press.

12. Louis, L., C. David, P. Robion. 2003. Comparison of 25. Gueguen, Y. and Fortin, J.. 2013. Elastic envelope of
the anisotropic behavior of undeformed sandstones porous rock. Geophysical Research letters.
under dry and saturated conditions. Tectonophysics DOI:10.1002/grl.50676.
370: 193-212.
26. Baud, P., Reuschlé, T., Ji, Y., Cheung, C., Wong, T.-
13. Baud, P., Klein, E.,Wong, T.-f.. 2004. Compaction f.. 2015. Mechanical compaction and strain
localization in porous sandstones: spatial evolution of localization in Bleurswiller sandstone. J. Geophys.
damage and acoustic emission activity. J. Struct. Geol. Res. Solid Earth, 120, doi:10.1002/2015JB012192.
26 (4): 603–624.
27. Dvorkin, J. and A. Nur. 1996. Elasticity of high
14. Baud, P., L. Louis, C. David, G.C. Rawling, T.-f. porosity sandstones: theory for two North Sea data
Wong. 2005. Effects of bedding and foliation on sets. Geophysics 61(5): 1363-1370.
mechanical anisotropy, damage evolution and failure
mode. Geol. Soc. London Special Publ.. 245: 223-249.

Das könnte Ihnen auch gefallen