Professional Documents
Culture Documents
Summary more, testing even the simplest of stress models requires extensive
This paper presents an evaluation of two fundamentally different data processing.
stress models: an elastic model, which is based on linear transverse
isotropic elasticity, and a failure model, which is based on the concept Elastic and Failure Stress Models
that rocks are in an equilibrium statc of shear failure. The models are Elastic Model. The stress/strain response of a fluid-saturated po-
evaluated by using physical parameters measured on core, pore pres-
rous material assumed to be linear elastic is 20 ,21
sure, and in-situ stress data from the Oas Research Institute (ORI)
Staged Field Experiments (SFE's) in east Texas. It is shown that the
elastic and failure models provide satisfactory predictions for most of
::::
E = C
-(:::: ::::)
0 - Kp p ...............•.•••..••....•.... (I)
the lithologies encountered. However, the failure model is more accu-
rate for predicting stress in soft shales. An example of stress predic- or, in suffix notation,
tions based on log-derived elasticity parameters that gives stress es-
Eij = Cijkl(Okl - KktPP)' . • . . . . . . . . . . . . . . . . . . . . . . . . . . (2)
timations comparable to core-based predictions is also shown.
The most commonly invoked stress model is based on linear elas-
Introduction ticity and assumes that rock is a semi-infinite isotropic poroelastic
The elastic uniaxial strain model is often used in the petroleum in- medium subjected to gravitational loading and zero horizontal strain.
dustry to predict the minimum stress as a function of depth and rock The vertical stress is therefore a result of the overburden and is a prin-
lithology. 1-4 Its popularity certainly comes from its simplicity: it re- cipal stress. The other two principal stresses lie in the horizontal
quires only parameters that can be estimated from wireline measure- plane. The influence of uniform horizontal principal strains or aniso-
ments or core data. tropy also could be considered. In the following section, we will ana-
It is, however, difficult to accept the idea that such a model can lyze the case of isotropic and transversely isotropic material.
apply to rock that has undergone compaction, cementation, heating,
cooling, and tectonic deformations spanning millions of years. In Isotropy. For an isotropic material,
some situations, stress models that assume that the rock is close to
shear failure S,6 or under lithostatic stress 7 appear physically more 1 J.l -~
0 0 0
reasonable. In other situations, more-complex models, which in- E E E
clude effects of inelastic creep, 8.9 load history,8,9 anisotropy, 10 ther- J.l I J.l
mal stresses, 8 and topography 11,12 may have to be considered.
-E E E
0 0 0
A dependence of stress on E is obtained, which means that the Mohr-Coulomb Failure Model. A stress model based on the
greater the E the greater the horizontal stress. This model gi ves a bet- Mohr-Coulomb failure criterion assumes that the maximum in-situ
ter qualitative description of the state of stress measured in areas in shear stress is governed by the shear strength of the formation. 5,6
which compressive tectonic stresses are suspected. This model can The Mohr-Coulomb criterion for shear failure in a plane is 22
account for situations where sandstones are under higher horizontal
stress than adjacent shales. The overburden stress is a principal r = c + tan(a)(oll - pp) . ......................... (13)
stress but not necessarily the maximum principal stress. Ifthe strains
The Mohr-Coulomb failure criterion can be rewritten to give 01 at
are equal to zero, this model reduces to the uniaxial strain model.
failure in term of 03:
Transverse Isotropy. Transverse isotropy characterizes a rock that is 01 - Pp = auf + N a (03 - pp), ..................... (14)
isotropic in the bedding plane but differs in the direction normal to
this plane. Transverse isotropy is actually a more-representative mod-
el for shales and unfractured laminated rocks than the conventional
65r---~--~--~--~--~--~--~--'
isotropic approximation. With the assumption that direction three is
the direction normal to the plane of isotropy, the tensor Cis 60
1 /1 /1' ~ 55
E -E E 0 0 0
/1 1 /1' ~ 50
0 0 0
E E E' ~ 45 o
::: "3
c= -~ _ /1' J... 0 0
~40
0 ........ (8)
E' E' E'
0 0 0 G' 0 0
0 0 0 0 G' 0
2(1 + /1)
0 0 0 0 0
E
Tensor 1C is given by
E /1' [ 1 * 2653 m
°Hmin - XPP = E'l -/1 °v - x(1 - ~)pp
x 26.54 m
E E/1
+ 1 _/12€Hmin + I _/12€Hmax .................... (II)
EEl 2657 III
and 0Hmax -
E /1'
XPp = E'l _ /1 [0v - xO - ~)pp 1 Fig. 1-Scatterplot showing E as a function of mean pressure in
SFE 2 sandstones between 2649 and 2657 m [8,691 and 8,717ft];
E E/1 core depths (in meters [feet]) are +=2651 [8,698],0=2652 [8,701],
+ -1--2€Hmax + -1--2€lImin' .............•..... (12)
*=2653 [8,704], x =2654 [8,707], and 0=2657 [8,717].
-/1 -/1
where Na = tan2(% +~) .......................... (IS) most of the reservoir sandstones and bounding shales are naturally
fractured. Vertical-extension fractures are prevalent in the sand-
stones, while fractures dipping < 45° are more common in mud-
and 0ul = 2cfN;, . ................................ (16) stones. Extension fractures exhibit a range of quartz mineralization
If the critical shear stress is controlled by slip along preexisting from almost no quartz to fully mineralized. Low-angle fractures in
surfaces, c can be assumed to be negligible. However, a residual co- the mudstones occasionally show slickensides, which indicates that
hesion may still exist. a is usually measured by using peak strength shear movement has occurred. More complete geologic descrip-
data as a function of confining pressure (a =ap ) obtained during tions are given in the references cited above. 26-27
triaxial testing. One can also use the angle of friction that character-
izes the frictional properties of two surfaces that are touching. This In-Situ Data
angle can be measured by using residual strength data as a function Stress State. We have good estimates of the vertical stress and the
of confining pressure (a =ar ) obtained during triaxial testing once least horizontal stress in the SFE wells. By integrating the bulk den-
the sample has failed. Using a r rather than ap in a failure stress mod- sity log, we estimated the vertical stress to be equal to 0.024 MPa/m
el should be more consistent with the assumption that the minimum [1.06 psi/ft]. We also have good evidence that the overburden is a
stress is controlled by friction along preexisting planes. principal stress from the analysis of strain relaxation cracks 27 or
Finally, if the formation is in extension, the vertical stress is the Borehole televiewer (BHTV) images. 28
maximum principal stress and Eq. 14 becomes The minimum stress was measured with small-volume hydraulic
fractures. The test procedure for stress using the microhydraulic
0Hmin - Pp = ~ (ov - pp), ....................... (17) fracturing technique in a cased-hole situation 29 essentially consists
a
of perforating a 0.6-m [2-ft] interval, isolating the interval with
in which the effect of cohesion has been neglected. straddle packers, and fracturing the rock with low-viscosity fluids.
Usually < 0.4 m3 [100 gal] is injected into the formation at a flow
Field Cases rate on the order of 0.0005 m 3/s [0.13 ft 3/sec). The downhole pres-
sure is recorded during pumping and shut-in to determine the mini-
We have evaluated elastic and failure models using data obtained
mum stress from the pressure records. The minimum stress in the
from SFE's 1,2,15,23 and 3. 24 These wells have been drilled in the
SFE's has been obtained from an estimation of the closure stress
East Texas basin. instead of the instantaneous shut-in pressure (lSIP).30 On the basis
Stress measurements in SFE's I and 2 are in the Cretaceous Travis of fracture behavior considerations in a permeable medium during
Peak formation. Travis Peak is composed of fine- to very-fine- shut-in, the closure stress is assumed to be a better measure of the
grained sandstones, muddy sandstones, and sandy mudstones. Po- minimum stress than is the lSI? The closure stress is taken as the
rosity ranges from 3% to 17% but is typically < 8%. Matrix perme- point on the pressure-decline curve where the pressure deviates
ability is less than = 0.00 1 ~m [ = 1 md] because of abundant quartz from a linear relationship to the square root of shut-in time. 31 -34 The
cement. 25 SFE 3 is < 1.61 km [< 1 mile] from SFE 1, but stress stress-state values, which are given in Table 1, are published
measurements are from the Jurassic Taylor sandstone and Bossier data. 23 ,24 Pore pressure is one of the most difficult parameters to ob-
shale. Grain size, porosity, and permeability are similar to those in tain for low-permeability formations. Pore-pressure measurements
the Travis Peak formation. with either repeat formation tester or pressure buildup were made in
Laubach26 and Laubach and Monson 27 analyzed fractures in the SFE's. These techniques work well in reservoir rocks with
cores from SFE 2 and other East Texas basin wells. They found that permeability on the order of tens of millidarcies or more but are
SFE 1 1878 to 1880 [6,159 to 6166] Mudstone 43.9 33.7 43.6± 1.2
1887 to 1880 [6,190 to 6,193] Sandstone 48.6 42.6 117.2±15.6
1897 [6,224] Mudstone 51.9 57.4 112.4
1897 [6,226] Mudstone 51.5 58.5 257.2
2262 to 2269 [7,420 to 7,444] Sandstone 51.3 46.4 58.6±9.6
2275 [7,461] Sandstone 56.9 50.6 169.7
2281 to 2283 [7,483 to 7,489] Mudstone 35.5 24.1 64.8± 1
SFE2 2513 [8,242] Sandstone 196.9±25.1
2533 [8,307] Mudstone 32.7 27.0 64.1
2652 to 2653 [8,699 to 8,702] Sandstone 55.7 52.5 49.9 200.7±10.1
2898 [9,506] Sandstone 56.5 56.0 51.6 260
2920 [9,571] Siltstone 40.2 38.8 73.9± 13.3
3012 [9,881] Sandstone 59.8 55.7 260
3030 [9,940] Sandstone 321.4
SFE3 2835 [9,300] Sandstone 48.5 47.2 44.2 152.0 ± 19.8
2856 [9,367] Shale 29.0 16.8 49.1±11.78
~
55
c... 55
:::E 50
.S ~'" 50
"0 45 c::
.- 45
~
C,)
:e 40 "2
~
'"'" 35
~
t:s
i:~
'"
til
30 ~ 30
e'"
::s 25 '"
e § 25
'8 E
::; 20 '2 20
._
:::E
15 15
10 IO~~--~~--~~--~~--~~~
10 15 20 25 30 35 40 45 50 55 60 10 15 20 25 30 35 40 45 50 55 6()
Minimum stress measured in MPa Minimum stress measured in MPa
Fig. 2-ScaHerplot showing stress prediction from the elastic Fig. 3-ScaHerplot of stress prediction from the elastic uniaxial
uniaxial strain model, assuming isotropy and lC = 1. strain model, assuming transverse isotropy, lC = 1, and ~ = o.
much less accurate when permeability is on the order of :s 0.0001 tained at 30 MPa [4350 psi] confining pressure. This approach gives
,Um [:s 0.1 md]. The problem in determining pore pressure is most values of E and E' that are close to the un fractured rock values.
pronounced in mudstones and shales. In this study, we estimated These should be considered as lower bounds to the actual unfrac-
pore pressure in shales and mudstones by interpolation of measure- tured values. The,U' values reported are the values measured at the
ments made in adjacent but more-permeable layers. This approach same stress level as E. Finally, we found that,U is difficult to measure
is valid if the thickness of the impermeable layers is relatively small, with accuracy because it has to be measured exactly in the bedding
as observed in most of the Travis Peak formation, and if the perme- plane. The values in Table 2 are given only for samples in which the
able layers have not been depleted. Unfortunately, some of the sand- results were repeatable. Notice that the mudstones and siltstones
stones penetrated by SFE 1 have been significantly depleted owing
usually exhibit a higher JI than the sandstones. An exception is the
to production from adjacent wells. Therefore, for the low-perme-
SFE 3 shale, the only sample that exhibits significant anisotropy.
ability mudstones in SFE 1, we have indicated two values of pore
Table 2 also lists values of the Biot coefficient, lC, which have
pressure: the first obtained by interpolation of the pore pressure
from the value measured in adjacent layers and the second by inter- been published by the Gas Research InstY The determination of 1(
polation of the pore pressure from apparently non depleted layers. has been made within I m [3.3 ft] of the depth indicated in Table 2.
This approach (see Table I) shows that stresses predicted for SFE Although not reported, the lC coefficient must certainly be pressure
I at 1879 and 1904 m [6,165 and 6,247 ft] are inaccurate because sensitive in the Travis Peak formation because it is in theory a linear
of the large uncertainty in the value of the pore pressure. function of the compressibility of the matrix. The stiffness matrix
measured during the lC determination 37 is in agreement with the val-
Laboratory Data ues in Table 2 and suggests that lC has been determined at a stress
Elastic parameters E', E,,U', and,U, were measured by triaxial tests on level similar to that of the elastic parameters.
SFE 2 and SFE 3 cores to provide stiffness tensor values for use in
stress prediction. Samples were loaded at a constant axial strain rate Strength Determination. Triaxial strength tests were made on pre-
of 2 x 10-5 seconds-I under drained conditions. Samples were tested served samples as a function of confining pressure. Sample pres-
"as preserved," and deformation was measured with either cantilever ervation and loading-rate conditions are the same as those used in
displacement gauges or strain gauges bonded directly to the samples. the elastic parameter testing.
Test results show that the elastic parameters are not constant but Loading was continued after peak strength so that a residual
are function of the stress state. An example of nonlinear behavior is strength could be measured on some sandstones. Tests were can'ied
demonstrated in Fig. 1, which shows the tangent E' determined dur- out to a confining pressure of up to 30 MPa [4350 psi]. Higher con-
ing loading as a function of the mean pressure [1/3(01 + 203)]. Six fining pressures were generally not tested because the sandstone
sandstones from SFE 2 obtained from depths between 2649.9 and strength exceeded the strength limit of the triaxial cells used. Table
2656.6 m [8,694 and 8,716 ft] are shown. The data scatter, which is 3 gives uniaxial strength Oitl and three representations of the angle
on the order of ± 10 GPa [1.45 106 psi], is typical of SFE's sand- of friction: ap ' a~ a r . These friction angles were determined by lin-
stones within a 6-m [J9.7-ft] interval. The values of E' do not be- ear regression by using the peak strengths, ap and a;', and residual
come constant until very high confining pressures are reached. For strength, ar , as a function of confining pressure. ap was obtained by
example, the sandstone obtained at 2653 m [8,704 ft] appears to using all the peak-strength data; a;' was obtained without using
reach its linear elastic regime only at about 50 MPa [7250 psi].
uniaxial strength data to see whether eliminating samples that failed
Although we believe that the nonlinear elastic behavior of these
by axial splitting gave better strength predictions.
rocks is caused by coring-induced microcracks,35.36 it was not pos-
Table 3 shows that most of the sandstones are extremely strong and
sible to prove the hypothesis with in-situ measurements. Uncertain-
ty about the origin of the microcracks introduces uncertainty in de- have a compressive strength of =200 MPa [=2900 psiJ. Generally,
termining elastic parameters from laboratory data. To overcome we find that the peak strength and friction angle of Travis Peak sand-
some of the uncertainties, we chose to determine our parameters stones increase slightly with depth. It is also clear that the friction angle
from the linear portion of the stress/axial-strain curves. Linearity is is lithology dependent, the lowest friction angles being found in shales
observed at high confining pressures once all the microcracks are and in some mudstones. We also find that the textural term of mudstone
closed. The effects of this approach on the stress predictions will be as it is used in core description rarely carries reliable information about
discussed where appropriate. Table 2 gives the elastic parameters rock strength or elastic properties. Some mudstones are as stiff and as
determined from the linear portion of the deformation curves ob- strong as the sandstones, while others are compliant and weak.
Failure Models. The stress predictions using Eq. 17 and the value
a
of p and a;
are shown in Figs. 5 and 6, respectively. A better pre-
diction is obtained if one uses a; a
rather than p . Again, we see that
the model underpredicts the in-situ stress, but the sense and magni-
tude of stress contrast is well matched.
Discussion
One problem in the evaluation of stress models is related to prob-
15 lems of scale. In-situ stress data, as typically acquired in the petro-
1OL-----'---'---'--------'---'---'--------'---'---'------'
leum industry, represent the stress averaged over vertical distances
10 15 20 25 30 35 40 45 50 55 60 that span distinct lithological variations. Consequently, we are nev-
er sure whether the interval of cores tested is really representative
Minimum stress measured in MPa of the lithology that has been stress tested. Furthermore, in some
cases (especially in SFE 3) we have not obtained cores that corre-
Fig. 4-Scatterplot of stress prediction from the elastic uniaxial spond exactly to the depth of the stress test.
strain model, assuming transverse isotropy and from measured Heterogeneity is also typical of the rock tested. Within a few feet,
1(values.
rock properties may change significantly, even with apparently the
same lithology. However, we found that the variability affects the
Stress Prediction
elastic properties and cohesion more than it does the friction angle.
Previous work in the east Texas SFE's showed that the in-situ A typical example shown in Table 3 is the mudstone at 1897 m
stresses are correlated with the stress predicted by the elastic uniax- [6,224 ft] in SFE 1, which shows a large variation of cohesion with-
ial strain model (Eq. 5) using the log-derived Poisson 's.4,15,37 These out any variation of friction angle within an interval of 0.6 m [2 ft].
results suggest that the elastic strain terms in Eq. 6 are negligible. When evaluating models, one must also question the accuracy of
It also was shown that, when using the dynamic Poisson's ratio, the in-situ stress and pore-pressure measurements. Stress measurements
predictions were better if 1( = 1 (which is the upper bound) instead are uncertain when a closure pressure is not well defined or when
of if values typically measured in the laboratory were used. 4 ,37 multiple closures are observed. Variability of several MPa is not un-
Therefore, the following predictions have been made, neglecting common for stress interpretations. Vertical stresses are estimated
the horizontal strain and using either a value of 1( = I or the laborato- from the density log and not measured. Pore pressure accounts for
ry-derived values of 1( given in Table 2. roughly 75% of the predicted stress in SFE 2 sandstones. However,
accurate pore-pressure measurements in tight gas sandstones are dif-
Elastic Models. Fig. 2 shows the elastic prediction using the iso- ficult to achieve owing to their very low diffusivity. Pore pressure in
tropic model given by Eq. 5 and the assumption 1( = I. The predic- shale is only obtained by interpolating the pore pressure measured in
tions are systematically low. However, all but two points are within the adjacent, more-permeable layers. It is likely that pore-pressure es-
2 MPa [290 psi] of the measured stress. The predicted stress for the timates are within only a few MegaPascals (MPa) of the true value.
SFE 3 shale (the highest measured stress value) is about 15 MPa But the major problem concerns the evaluation of the elastic mod-
[2175 psi] too low. In Fig. 3, the influence of anisotropy is taken into els with the determination of elastic parameters for nonlinear rocks.
account (see Eq. 10), which results in a dramatic improvement of the We chose to use the elastic parameters of un fractured rock (intrinsic
stress prediction for the SFE 3 shale. In these predictions, it was as- elastic properties). These are determined from the apparently linear
sumed that!l =!l' when values for!l were not available and ~ = O. elastic regime observed at a confining pressure much higher than the
Finally, Fig. 4 confirms a result already mentioned 37 : a realistic val- estimate of the in-situ value. An alternativc solution would be to de-
60~~-~~--~--~~--~-~~--. 60.-~--~~--~~--~~--~~--.
55 55
«I
«I c..
50 ~ 50
~ .5
.S 45 '0 45
~
asu u
40 '6 40
t ....c..
~
~)I(
35 35
<J)
<J)
lIE '"
'"
~
l:l 30
~ 30
)I(
<J) '"e
8 25 ::I 25
::I )I( e
8 '2
'2 20 20
~
~ )I(
15
15
10 10
10 15 20 25 30 35 40 45 50 55 60 10 15 20 25 30 35 40 45 50 55 60
Minimum stress mea~ured in MPa Minimum stress measured in MPa
Fig. 5-Scatterplot of stress prediction from the failure model Fig. 6-Scatterplot of stress prediction from the failure model
witha=ap . with a~ .
0.8 0.8
0.7 0.7
v ~
E 0.6 .E 0.6
.;
~
00.5 ~ 0.5
:.:: lIE
0.4 0.4
lIE
0.3 0.3
0.2 0.2
0.1 0.1
0
o L-~ __~__~~__~__~~__~__~~
0 0.1 0.2 0.3 0.4 0.5 O.n 0.7 O.S 0.9 o 0.1 0.2 n.3 0.4 0.5 O.n 0.7 0.8 0.9
Ko ela<;tic Ko elastic (dynamic)
Fig. 8-Scatterplot of Ko for the failure model as a function of Ko Fig. 9-Scatterplot of Ko for the failure model as a function of Ko
for the elastic uniaxial strain model (transverse isotropy). forthe elastic uniaxial strain model (Poisson's ratio derived from
logs).
value of Ko with the elastic constants. This is exactly what it is found The similarity between the predictions of both models is caused
using the data in Tables 2 and 3 (Fig. 8). Actually, the correlation by an observed positive correlation between the coefficient of Earth
indicates a negative correlation between the friction angle and Pois- stress determined using the friction angle and the one determined
son's ratio for isotropic rocks; i.e., the higher the Poisson's ratio the using Poisson's ratio. This correlation explains why elastic models
lower the friction angle. A similar correlation is found if one uses may predict stress as well as failure models in tectonically relaxed
a Poisson's ratio derived from logs for the elastic Ko (Fig. 9). How- basins and why sonic logs provide satisfactory stress prediction.
ever, in this case, the correlation is weaker, especially if one includes
the Bossier shale found in SFE 3 (the rock with the highest value of Nomenclature
Ko failure) because of the assumption of isotropic behavior current- _ c = cohesion, m/Lt 2
- . 2
1y imposed by sonic logs. C. Cijkl = tensor of compliance, mILt
This correlation explains why elastic models may predict stress £= Young's modulus in the plane of isotropy, mlLt 2
as wcll as failure models in tectonically relaxed basins and why son- £' = Young's modulus normal to the plane of isotropy, mlLt2
ic logs provide satisfactory stress prediction, at least with the rock G' = shear modulus in any planes normal to the plane of
tested in the SFE wells. isotropy, mlLt2
Finally, although we have shown that very simple elastic and fail- hi = direction cosines of the i axis in relation to the axis of
ure models provide satisfactory stress predictions in east Texas, we symmetry
do not claim that these simple models include all effects presumed Ko = coefficient of Earth stress
to account for contemporary stresses in petroleum reservoirs. For Pp = pore pressure
example, we have not tried to reproduce the burial history of the ba- Na = passive coefficient
sin, and we have excluded thermal stress. The systematic underpre- a = angle of friction
dictions observed for both models could easily be accounted for by
ap, = angle of internal friction using peak strength
a p = angle of internal friction using peak strength under
inclusion of thermal stress or viscous deformations or by geophysi-
nonzero confining pressure
cally reasonable variations in pore pressure and vertical stress. The _a r = angle of friction using residual strength
model evaluations, however, indicate the sensitivity of stress pre- €, € ij = strain tensor
dictions to the different rock properties associated with the different fHmax = maximum horizontal strain
hypothetical deformation mechanisms. fHmin = minimum horizontal strain