Effect of Rock Strength Criterion On The Predicted Onset of Sand Production

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Effect of Rock Strength Criterion on the Predicted Onset

of Sand Production
X. Yi1; P. P. Valkó2; and J. E. Russell3

Abstract: The variation of sanding onset prediction results with the selection of one or another rock strength criterion is investigated. In
this paper, four commonly used rock strength criteria in sanding onset prediction and wellbore stability studies are presented. There are
Mohr–Coulomb, Hoek–Brown, Drucker–Prager, and modified Lade criteria. In each of the criterion, there are two or more parameters
involved. In the literature, a two-step procedure is applied to determine the parameters in the rock strength criterion. First, the Mohr–
Coulomb parameters like cohesion So and internal friction angle ␾ f , are regressed from the laboratory test data. Then, the parameters in
other criteria are calculated using the regressed Mohr–Coulomb parameters. It is proposed that the best way to evaluate the parameters in
a specific rock strength criterion is to perform direct regression of the laboratory test data using that criterion. Using this methodology, it
is demonstrated that the effect of various rock strength criteria on sanding onset prediction is less dramatic than using the commonly used
method. With this methodology, the uncertainties of the effect of rock strength criterion on sanding onset prediction are also reduced.
Through this study, it is also demonstrated that a sanding onset prediction problem cannot be properly solved by adopting strength criteria
that are not influenced by the intermediate principal stress if laboratory test data indicate rock failure is dependent on intermediate
principal stress.
DOI: 10.1061/共ASCE兲1532-3641共2005兲5:1共66兲
CE Database subject headings: Sand; Triaxial tests; Parameters; Rocks.

Introduction gressed from the conventional triaxial test data. Then, the rock
material parameters in other criteria are calculated using the re-
Sand production from oil/gas wells is a common field phenom- gressed Mohr–Coulomb parameters So and ␾ f . Using this proce-
enon. It poses some disadvantages or even risks for oil/gas pro- dure, it is concluded that one rock strength criterion predicts a
duction. Therefore it is desirable to predict under what conditions less conservative critical mud weight than the others. It is not the
sanding may occur and then to choose the appropriate sand con- intention of this work to judge if this conclusion is right or wrong.
trol technique. However, a different methodology is proposed to evaluate the
Generally, in sanding onset prediction, the near well stress rock material parameters appearing in each rock strength crite-
state is simulated using poroelastic theory, then an appropriate rion.
rock strength criterion is introduced to predict when sanding may In this paper, four different rock strength criteria: Mohr–
occur 共Morita et al. 1989a,b; Weingarten and Perkins 1995兲. Coulomb 共Chen and Mizuno 1990兲, Hoek–Brown 共1980兲,
Therefore, rock strength criterion plays a key role in sanding pre- Drucker–Prager 共1952兲, and modified Lade 共Ewy 1999兲 are used
diction. In each rock strength criterion, there are some rock ma- in conjunction with an axisymmetrical poroelastic stress model to
terial parameters involved. In order to predict more accurately the predict the onset of sand production. Using conventional triaxial
sanding onset conditions, one needs to determine those material test data, regression for the rock material parameters appearing in
parameters based on laboratory test data of the rock. the rock strength criterion directly from the test data is proposed
Ewy 共1999兲 and McLean and Addis 共1990兲 studied the effect instead of using the regressed Mohr–Coulomb parameters So and
of rock strength criterion on wellbore stability. In their work, a ␾ f to calculate the rock material parameters indirectly. Then it is
two-step procedure is applied to obtain the rock material param- demonstrated how the sanding onset prediction results differ. In
eters in the rock strength criterion. First, the Mohr–Coulomb pa- addition, because two of the aforementioned rock strength criteria
rameters like cohesion So and internal friction angle ␾ f are re- are intermediate principal stress dependent criteria, using conven-
tional triaxial test data only may not be adequate. If a set of true
1
PhD, Baker Atlas, 2001 Rankin Rd., Houston, TX 77073. triaxial 共polyaxial兲 test data, which give the same Mohr-Coulomb
2
Harold Vance Dept. of Petroleum Engineering, Texas A&M Univ., parameters So and ␾ f as those given by the previous conventional
College Station, TX 77843-3116. triaxial test data, are used and direct regression is performed,
3
Harold Vance Dept. of Petroleum Engineering, Texas A&M Univ., quite different sanding onset prediction results are reached. Fig. 1
College Station, TX 77843-3116. illustrates the comparison scheme in this paper.
Note. Discussion open until August 1, 2005. Separate discussions In this paper, as a rule, all the stresses are indicated as effective
must be submitted for individual papers. To extend the closing date by
stresses and compressive stress is assumed positive. Pore fluid
one month, a written request must be filed with the ASCE Managing
Editor. The manuscript for this paper was submitted for review and pos- pressure is always positive. At first, four rock strength criteria and
sible publication on February 24, 2003; approved on September 10, 2003. a near wellbore poroelastic stress model are introduced. Then it is
This paper is part of the International Journal of Geomechanics, Vol. 5, demonstrated how different ways of processing the same set of
No. 1, March 1, 2005. ©ASCE, ISSN 1532-3641/2005/1-66–73/$25.00. data yield different sanding onset prediction results according to

66 / INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MARCH 2005


␴1 + ␴3 ␴1 − ␴3
F= sin ␾ f + So cos ␾ f − 共3兲
2 2
Failure occurs when F 艋 0.

Hoek–Brown Criterion „1980…


Hoek and Brown 共1980兲 studied the published experimental re-
sults of a wide variety of rocks and proposed the following
strength criterion:

␴1 ␴3
=
Co Co
+ 冑 s+m
␴3
Co
共4兲

Fig. 1. Comparison scheme in this paper where m and s = constants which depend on the properties of the
rock and on the extent to which it has been broken before being
subject to the stresses. Parameter

再 冎
the scheme illustrated in Fig. 1. Finally conclusions are drawn
and recommendations are made. =1 for intact rock
s
⬍1 for previously broken rock
It is assumed s = 1 in this study. Co = uniaxial compressive strength
Rock Strength Criterion
of the intact rock in the specimen. Co is related to the Mohr–
Coulomb parameters through the following expression:
In this section, four most commonly used rock strength criteria in
wellbore stability analysis and sand production prediction are pre- 2So cos ␾ f
sented. Co = 共5兲
1 − sin ␾ f
The failure function is
Mohr–Coulomb Criterion „Chen and Mizuno 1990…
Mohr–Coulomb failure criterion is the most commonly used
strength criterion for geomaterials. According to the Mohr–
F=
␴3 ␴1

Co Co
+ 冑 s+m
␴3
Co
共6兲

Coulomb criterion, the shear strength increases with increasing Failure occurs if F 艋 0.
normal stress on the failure plane. It can be represented by the Mohr–Coulomb and Hoek–Brown criteria are only dependent
following equation 共see also Fig. 2兲: on the maximum and minimum principal stresses. The effect of
␶ = So + ␴ tan ␾ f 共1兲 intermediate principal stress is not considered. The Mohr–
Coulomb criterion is linear with ␴3 while the Hoek–Brown crite-
where ␶ = shear stress; ␴ = normal stress; So = cohesive strength; rion is nonlinear. Application of Hoek–Brown criterion in sand
and ␾ f = internal frictional angle. production prediction is considered in Wang and Wu 共2002兲.
The Mohr–Coulomb criterion can be written in terms of prin-
cipal stresses, which results in
Drucker–Prager Criterion „1952…
␴1 − ␴3 ␴1 + ␴3
= sin ␾ f + So cos ␾ f 共2兲 An approximation to the Mohr–Coulomb criterion was presented
2 2 by Drucker and Prager 共1952兲 as an extended Mohr–Coulomb
in which ␴1 = maximum principal stress and ␴3 = minimum prin- rule that employs the Von Mises criterion often used for ductile
cipal stress. metals. It has the form

␣I1 + k = 冑J2
The failure function F is
共7兲
where
I1 = ␴1 + ␴2 + ␴3 共8兲
which is the first invariant of the stress tensor
1
J2 = 6 关共␴1 − ␴2兲2 + 共␴2 − ␴3兲2 + 共␴1 − ␴3兲2兴 共9兲
which is the second invariant of the stress deviator tensor.
The failure function is

F = ␣I1 + k − 冑J2 共10兲


Failure occurs if F 艋 0.
The material constants in Drucker–Prager criterion can be de-
termined by matching two particular points with those of the
Mohr–Coulomb criterion, and thus the two constants, ␣ and k,
can be expressed in terms of the Mohr–Coulomb parameters So
Fig. 2. Mohr–Coulomb strength criterion and ␾ f .

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MARCH 2005 / 67


Modified Lade Criterion „Ewy 1999…
The modified Lade criterion was proposed by Ewy 共1999兲 based
on Lade criterion 共Lade 1977兲. Application of this criterion in
sand production is also shown by Ewy et al. 共2001兲. The criterion
can be written as

I⬙13
= 27 + ␩ 共17兲
I3⬙

where

I⬙1 = 共␴1 + S1兲 + 共␴2 + S1兲 + 共␴3 + S1兲 共18兲

Fig. 3. Failure envelopes projected on ␲ plane for Mohr–Coulomb I3⬙ = 共␴1 + S1兲共␴2 + S1兲共␴3 + S1兲 共19兲
criterion, circumscribed Drucker–Prager criterion, inscribed
Drucker–Prager criterion, and Drucker–Prager middle circle and

In three dimensional 共3D兲 stress matching, if the points are So


S1 = 共20兲
selected in such a way that the failure surface circumscribes the tan ␾ f
Mohr–Coulomb hexagonal pyramid 共see Fig. 3兲, the material con-
stants are 共Chen and Mizuno 1990; McLean and Addis 1990兲
4 tan2 ␾ f 共9 − 7 sin ␾ f 兲
2 sin ␾ f ␩= 共21兲
␣= 共11兲 1 − sin ␾ f
冑3共3 − sin ␾ f 兲
The failure function is
6So cos ␾ f
k= 共12兲 I1⬙3
冑3共3 − sin ␾ f 兲 F = 27 + ␩ − 共22兲
I⬙3
The matching points may also be selected in such a way that the
failure surface touches the other three apexes 共see Fig. 3兲. In this Failure occurs if F 艋 0.
case, the material constants are 共Chen and Mizuno 1990; McLean
and Addis 1990兲
2 sin ␾ f Near Wellbore Poroelastic Stress Model
␣= 共13兲
冑3共3 + sin ␾ f 兲
Radial Fluid Flow Model
6So cos ␾ f If single-phase fluid flow is assumed, the material balance equa-
k= 共14兲
冑3共3 + sin ␾ f 兲 tion for the fluid in an axisymmetrical reservoir is 共Aziz and
Settari 1979兲
If the Drucker–Prager and Mohr–Coulomb criteria are expected

冉 冊 冉 冊
to give identical limit loads for the plane strain case, the material
constants can be determined as 共Chen and Mizuno 1990; McLean 1 ⳵ kr ⳵ p ⳵ kz ⳵ p ␸共c f + cr兲 ⳵ p
r + = 共23兲
and Addis 1990兲 r ⳵r ␮ f B f ⳵r ⳵z ␮ f B f ⳵z Bf ⳵t
tan ␾ f where kr and kz = permeability in radial direction and vertical di-
␣= 共15兲
冑9 + 12 tan2 ␾ f rection, respectively; ␮ f = reservoir fluid viscosity; B f = reservoir
fluid formation volume factor; ␾ = formation prorsity; C f = fluid
3So compressibility; Cr = rock compressibility; p = pressure; t
k= 共16兲
冑9 + 12 tan2 ␾ f = production time; and r and z = coordinates in the r and z direc-
tions.
Under this condition, the Drucker–Prager failure surface inscribes Eq. 共23兲 can be solved with appropriate boundary conditions
the Mohr–Coulomb failure surface 共see Fig. 3兲. and initial condition by finite difference or finite element method.

Table 1. Assumed Rock Strength Test Data Points


Case A—Conventional triaxial test Case B—True triaxial test

␴1 ␴2 ␴3 ␴1 ␴2 ␴3
Experiment 共MPa兲 共MPa兲 共MPa兲 Experiment 共MPa兲 共MPa兲 共MPa兲
Uniaxial 16.72 0 0 Uniaxial 16.72 0 0
Conven. triaxial 31.20 4.83 4.83 True Triaxial 20.86 3.45 1.38

68 / INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MARCH 2005


Table 2. Strength Criterion Parameters for Case A—Derived from Regressed Mohr–Coulomb So and ␾ f
k S1 So ␾f Co
Strength criterion ␣ 共MPa兲 共MPa兲 ␩ 共MPa兲 共deg兲 共MPa兲 m
M-C — — — — 4.83 30 — —
H-Ba — — — — — — 16.72 5.15
D-P cirm. circle 0.23094 5.79 — — — — — —
D-P ins. circle 0.16013 4.02 — — — — — —
D-P middle circle 0.16496 4.14 — — — — — —
Modified Lade — — 8.36 14.6667 — — — —
a
m and Co are directly calculated without using Mohr–Coulomb So and ␾ f .

Axisymmetrical Poroelastic Stress Model ria, then only conventional triaxial tests 共␴1 ⬎ ␴2 = ␴3兲 are neces-
sary. Otherwise, true triaxial tests 共␴1 ⬎ ␴2 ⬎ ␴3兲 are needed.
If the in situ horizontal stress is isotropic and all the formation
However, in reality, it is not known which criterion the formation
properties and loading are axisymmetrical, an axisymmetric po-
rock satisfies until proven by laboratory tests. True triaxial test
roelastic stress model can be established to describe the near well-
data can always be used to determine which criterion best de-
bore stress state.
scribes the formation rock strength. However, for many reasons,
The stress model in cylindrical coordinates is established in r
and z directions using the concept of momentum equilibrium 共Ti- conventional triaxial test data are commonly used but true triaxial
moshenko and Goodier 1969兲 and poroelasticity 共Biot 1941兲, test data are rarely available for petroleum engineering usage.
which is Besides, the procedure used to process the conventional triaxial
test data is frequently regressing for the Mohr–Coulomb criterion
⳵␴r ⳵␶rz ␴r − ␴␪ ⳵p parameters like cohesion So and internal friction angle ␾ f through
+ + + f r + ␣b = 0 共24兲
⳵r ⳵z r ⳵r the ␴1, ␴3 plot, and then deriving the parameters in other strength
criteria through their relation with the Mohr–Coulomb parameters
⳵␴z ⳵␶rz ␶rz ⳵p 关see Eqs. 共11兲–共16兲, 共20兲, and 共21兲兴. Even if there are relations
+ + + f z + ␣b = 0 共25兲 between Mohr–Coulomb parameters and the parameters in the
⳵z ⳵r r ⳵z
other criteria, the relations are not based directly on laboratory
where ␴r = effective radial stress; ␴␪ = effective tangential stress; test data. Direct regression of the test data for each different rock
␴z = effective vertical stress; ␶rz = shear stress in z direction; f r strength criterion is recommended.
= body force per unit volume in radial direction; f z = body force To support the recommendation, in this section, the following
per unit volume in vertical direction; and ␣b = Biot’s constant. are demonstrated: 共1兲 Using conventional triaxial test data, the
Combining the radial fluid flow model and the above poroelas- rock strength criteria parameters obtained from the regressed
tic stress model with appropriate boundary conditions, the stress Mohr–Coulomb parameters and from direct regression of the test
state can be determined using the finite element method for a data are different. Consequently, the approach used in processing
given well production condition as a function of time. laboratory test data does matter. 共2兲 Using true triaxial test data
which give the same regressed Mohr–Coulomb parameters as be-
fore, it is demonstrated that the rock strength criteria parameters
Determination of Rock Strength Criterion obtained from direct regression of the test data are different from
Parameters those obtained using conventional triaxial test data.
It is unfortunate that laboratory test data are not available and
From the above listed rock strength criteria, it is noticed that there therefore the regression process cannot be performed 共interested
are two parameters involved in each criterion 共in Hoek–Brown, s reader should refer to Colmenares and Zoback 2002b, a兲. Two
has been assumed to be unity兲. Laboratory tests are needed to assumed strength data points are used in two cases 共see Table 1兲,
determine those parameters for a specific rock. Among the listed to illustrate our ideas. As a result, it is important to mention that,
rock strength criteria, Mohr–Coulomb and Hoek–Brown criteria in this paper, all the rock strength criterion parameters are calcu-
are intermediate principal stress independent 共␴2 independent兲 lated from assumed data points instead of regressed from actual
while Drucker–Prager and modified Lade criteria are intermediate test data points. In Case A, one uniaxial compressive test data
principal stress dependent 共␴2 dependent兲. If it is known that the point and one conventional triaxial test data point are used. In
formation rock satisfies any of the ␴2 independent strength crite- Case B, one uniaxial compressive test data point and one true

Table 3. Strength Criterion Parameters for Case A—Calculated Directly from Two Assumed Test Data Points 共Compare with Table 2兲
k S1 So ␾f Co
Strength criterion ␣ 共MPa兲 共MPa兲 ␩ 共MPa兲 共deg兲 共MPa兲 m
M-C — — — — 4.83 30 — —
H-Ba — — — — — — 16.72 5.15
D-P cirm. circle 0.23094 5.79 — — — — — —
D-P ins. circle 0.23094 5.79 — — — — — —
D-P middle circle 0.23094 5.79 — — — — — —
Modified Lade — — 8.36 14.6667 — — — —
a
m and Co are directly calculated without using Mohr–Coulomb So and ␾ f .

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MARCH 2005 / 69


Table 4. Strength Criterion Parameters for Case B—Derived from Regressed Mohr–Coulombb So and ␾ f
k S1 So ␾f Co
Strength criterion ␣ 共MPa兲 共MPa兲 ␩ 共MPa兲 共deg兲 共MPa兲 m
M-C — — — — 4.83 30 — —
H-Ba — — — — — — 16.72 4.33
D-P cirm. circle 0.23094 5.79 — — — — — —
D-P ins. circle 0.16013 4.02 — — — — — —
D-P middle circle 0.16496 4.14 — — — — — —
Modified Lade — — 8.36 14.6667 — — — —
a
m and Co are directly calculated without using Mohr–Coulomb So and ␾ f .

Table 5. Strength Criterion Parameters for Case B—Calculated Directly from Two Assumed Test Data Points 共Compare with Table 4兲
k S1 So ␾f Co
Strength criterion ␣ 共MPa兲 共MPa兲 ␩ 共MPa兲 共deg兲 共MPa兲 m
M-C — — — — 4.83 30 — —
H-Ba — — — — — — 16.72 4.33
D-P cirm. circle 0.11662 7.70 — — — — — —
D-P ins. circle 0.11662 7.70 — — — — — —
D-P middle circle 0.11662 7.70 — — — — — —
Modified Lade — — 21.02 3.4527 — — — —
a
m and Co are directly calculated without using Mohr–Coulomb So and ␾ f .

Table 6. Well, Reservoir, and Production Parameters triaxial test data point are used. These data are assumed such that
Parameter Value they provide same regressed Mohr–Coulomb parameters So and
␾ f in order to compare the results obtained in Cases A and B.
Wellboe radius 共m兲 0.1524
For Case A, if the Mohr–Coulomb parameters So and ␾ f are
Drainage radius 共m兲 609.60
calculated first, and then use So and ␾ f to calculate rock strength
Reservoir thickness 共m兲 6.096
criterion parameters in other criteria using Eqs. 共11兲–共16兲, 共20兲,
Total production time 共years兲 12
and 共21兲, Table 2 is obtained. If the parameters in each strength
Gas specific gravity 共fraction兲 0.7
criterion are calculated directly from the assumed test data with-
Reservoir temperature 共°C兲 42.22
out using Eqs. 共11兲–共16兲, 共20兲, and 共21兲, Table 3 is obtained.
Initial reservoir pressure 共MPa兲 19.31
Similarly, Tables 4 and 5 are obtained for Case B.
Permeability 共m2兲 9.87e − 15
From the results in Tables 2–5, it is noticed that the procedure
Porosity 共fraction兲 0.12
used in processing laboratory test data is important for evaluating
Initial production rate 共m2 / day兲 35,396
rock strength parameters. It is strongly recommended that direct
Minimum break horsepower 共MPa兲 3.45
regression based on laboratory test data be used to evaluate rock
Young modulus 共GPa兲 9.6
strength parameters.
Poisson’s ratio 共fraction兲 0.3
In the following section, it is shown how the sanding onset
Biot’s constant 共fraction兲 1.0
prediction results are affected by the methodology of obtaining
Horizontal stress 共MPa兲 21.10
the parameters in rock strength criterion.
Vertical stress 共MPa兲 23.44

Fig. 4. Pressure change with time during gas production

70 / INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MARCH 2005


Application of Rock Strength Criterion in Sand
Production Prediction

Sand production is a very common field phenomenon. It may be


caused by a high production rate, which leads to near-wellbore
formation rock tensile failure 共Weingarten and Perkins 1995兲. It
may also be induced by the increase of near-wellbore effective
stress during the depletion of a reservoir, which causes near-
wellbore formation rock shear failure 共Morita et al. 1989a, b兲. In
this paper, it is assumed that sand production is caused by near-
wellbore formation rock shear failure.
A hypothetical vertical gas well with the parameters given in
Fig. 5. Wellbore surface principal effective stresses change with pro- Table 6 is considered. The well produces first at constant rate and
duction time assuming linear elasticity then at constant pressure after the bottomhole flowing pressure
reaches the allowed minimum value. The variation of average
reservoir pressure and bottomhole flowing pressure etc. with pro-
duction time is shown in Fig. 4. If the formation rock behaves
linear elastically, the variation of wellbore surface principal effec-
tive stresses with production time is calculated as shown in Fig. 5.
However, it is more likely that the formation rock behaves as an

Fig. 6. Predicted sanding onset results using different rock strength criteria for rock data in Cases A and B, with rock strength parameters from
Tables 2 and 4, respectively. Since Tables 2 and 4 are essentially the same, the sanding onset prediction results are exactly same for both cases.
Sanding onset prediction indicates that no sanding occurs if circumscribed Drucker–Prager criterion is used for both cases, data for this criterion
in this graph is at production time= 12 years.

Fig. 7. Predicted sanding onset results using different rock strength criteria for test data in Case A, with rock strength parameters from Table 3.
Predicted sanding onset results for three versions of Drucker–Prager criteria are same because parameters ␣ and k are same for different versions
of Drucker–Prager criterion. Since sanding onset prediction indicates that no sanding occurs if three Drucker–Prager criteria are used, data for
criteria in this graph are at production time= 12 years.

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MARCH 2005 / 71


Fig. 8. Predicted sanding onset results using different rock strength criteria for rock test data in Case B, with rock strength parameters from
Table 5

elastoplastic material. If a specific strength criterion is used, sand This is because the Biot’s constant is taken to be unity in our
production caused by the near wellbore area failure can be pre- study and the minimum principal stress ␴3 is zero.
dicted. In this section, it is shown how the sanding onset predic-
tion results vary with the methodology of obtaining the param-
eters in the rock strength criterion. Conclusions and Recommendations
For Case A, when only conventional laboratory measurements
are available, the rock strength criteria parameters using two dif-
The following are concluded and recommended from the above
ferent methodologies are calculated. Using parameters in Table 2,
work:
which is derived by first calculating the Mohr–Coulomb param-
1. Of the approaches considered, the best way to evaluate the
eters So and ␾ f , and then using So and ␾ f to calculate rock
rock strength criterion parameters is to perform direct regres-
strength criterion parameters in other criteria using Eqs.
sion of the laboratory test data. Using the regressed Mohr–
共11兲–共16兲, 共20兲, and 共21兲, the specific production time when sand
production occurs can be predicted, in addition to other data such Coulomb parameters to calculate the rock strength param-
as drawdown 共the difference between average reservoir pressure eters in other criteria is not recommended.
and bottom hole flowing pressure兲 and total drawdown 共the dif- 2. Using the methodology proposed in this paper, the uncertain-
ference between initial reservoir pressure and bottom hole flow- ties of the effect of rock strength criterion on sand production
ing pressure兲 at the onset of sand production. Those sanding pa- prediction are reduced.
rameters obtained by using different strength criteria are 3. Through this study, it is demonstrated that a sanding onset
illustrated in Fig. 6. Obviously, the variation of predicted onset of prediction problem cannot be properly solved by adopting
sanding with the applied strength criterion is too large. However, strength criteria that are not influenced by the intermediate
if parameters in Table 3 are used, which are calculated directly principal stress if laboratory test data indicate rock failure is
from the test data without using Eqs. 共11兲–共16兲, 共20兲, and 共21兲, a dependent on intermediate principal stress.
different set of sanding onset data are obtained 共see Fig. 7兲. Com-
parison of Figs. 6 and 7 indicates that the sanding onset prediction
differences for various versions of the Drucker–Prager criterion References
are eliminated if the rock strength criterion parameters are ob-
tained by direct regression of the laboratory test data. This is Aziz, K., and Settari, A. 共1979兲. Petroleum reservoir simulation, Applied
because in the Drucker–Prager criterion there are only two param- Science Publishers Ltd., London.
eters, ␣ and k, available for regressing. Therefore, using the same Biot, M. 共1941兲. “General theory of three-dimensional consolidation.” J.
test data, regression can only lead to the same set of ␣ and k even Appl. Phys., 12, 155–164.
for different versions of the Drucker–Prager criteria. Chen, W., and Mizuno, E. 共1990兲. Nonlinear analysis in soil mechanics—
In the same way, two different sanding onset prediction results Theory and implementation, Elsevier, New York.
are obtained for Case B. If the rock strength criterion parameters Colmenares, L., and Zoback, M. 共2002a兲. “A statistical evaluation of
in Table 4 are used, sanding prediction results are exactly the intact rock failure criteria constrained by ployaxial test data for five
same as those in Fig. 6. If the rock strength criterion parameters different rocks.” Int. J. Rock Mech. Min. Sci., 39, 695–729.
Colmenares, L., and Zoback, M. 共2002b兲. “Statistical evaluation of six
in Table 5 are used, different sanding prediction results are ob-
rock failure criteria constrained by polyaxial test data.” Proc., 38th
tained 共see Fig. 8兲. The sanding prediction results differ from each
U.S. Rock Mechanics Symp., Washington, D.C., 1251–1258.
strength criterion much less than that in Fig. 7 due to the fact that
Drucker, D., and Prager, W. 共1952兲. “Soil mechancis and plastic analysis
one true triaxial test data point in Case B and our recommended or limit design.” Q. Appl. Math., 10, 57–165.
methodology to process the test data points are used. Some dif- Ewy, R. T. 共1999兲. “Wellbore-stability predictions by use of a modified
ference still exists because the shape of each strength criterion lade criterion.” SPE Drill. Completion, 6, 85–91.
surface in the three-dimensional principal stress space is different. Ewy, R. T., Ray, P., Bovberg, C. A., Norman, P. D., and Goodman, H. E.
In Figs. 6–8, it is noticed that the Mohr–Coulomb and Hoek– 共2001兲. “Open hole stability and sanding predictions by 3D extrapo-
Brown criteria provide the same sanding onset prediction results. lation from hole collapse tests.” SPE Drill. Completion, 12, 243–251.

72 / INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MARCH 2005


Hoek, E., and Brown, E. 共1980兲. Underground excavations in rock, The alistic sand production prediction: numerical approach.” SPE Prod.
Institution of Mining and Metallurgy, London. Eng., 2, 15–24.
Lade, P. V. 共1977兲. “Elasto-plastic stress–strain theory for cohesionless Timoshenko, S., and Goodier, J. 共1969兲. Theory of elasticity, McGraw–
soil with curved yield surfaces.” Int. J. Solids Struct., 13, 1014–1035. Hill, New York.
McLean, M. R., and Addis, M. A. 共1990兲. “Wellbore stability: the effect Wang, Y., and Wu, B. 共2002兲. “Borehole collapse and sand production
of strength criteria on mud weight recommendations.” Proc., SPE evaluation: Experimental testing, analytical solutions and field appli-
Annual Technical Conf. and Exhibition, New Orleans, 9–17. cations.” Proc., 38th U.S. Rock Mechanics Symp., Washington, D.C.,
Morita, N., Whitfill, D. L., Fedde, O. P., Levik, T. H. 共1989a兲. “Paramet- 67–74.
ric study of sand production prediction: analytical approach.” SPE Weingarten, J. S., and Perkins, T. K. 共1995兲. “Prediction of sand produc-
Prod. Eng., 2, 25–33. tion in gas wells: method and Gulf of Mexico case studies.” J. Pet.
Morita, N., Whitfill, D. L., Massie, I., and Knudsen, T. W. 共1989b兲. “Re- Technol., 7, 596–600.

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MARCH 2005 / 73

You might also like