Journal of Petroleum Science and Engineering 35 (2002) 127 – 150

www.elsevier.com/locate/jpetscieng

Analytical, numerical and experimental investigations of transverse

fracture propagation from horizontal wells
M.M. Rahman, M.M. Hossain, D.G. Crosby, M.K. Rahman *, S.S. Rahman
School of Petroleum Engineering, The University of New South Wales, Sydney 2052, Australia
Received 16 July 2001; accepted 14 March 2002

Abstract

This paper presents results of a comprehensive study involving analytical, numerical and experimental investigations into
transverse fracture propagation from horizontal wells. The propagation of transverse hydraulic fractures from horizontal wells is
simulated and investigated in the laboratory using carefully designed experimental setups. Closed-form analytical theories for
Mode I (opening) stress intensity factors for idealized fracture geometries are reviewed, and a boundary element-based model is
used herein to investigate non-planar propagation of fractures. Using the mixed mode fracture propagation criterion of the
model, a reasonable agreement is found with respect to fracture geometry, net fracture pressures and fracture propagation paths
between the modeled fractures and the laboratory tested fractures. These results suggest that the propagation of multiple
fractures requires higher net pressures than a single fracture, the underlying reason of which is theoretically justified on the basis
of local stress distribution. D 2002 Elsevier Science B.V. All rights reserved.

Keywords: Reservoir stimulation; Hydraulic fracturing; Transverse fracture; Enhanced recovery

1. Introduction treatments. A number of studies carried out recently

compare the production characteristics between frac-
The productivity of vertical wells in low-perme- tured horizontal wells and fractured/unfractured verti-
ability tight-gas formations is commonly enhanced cal wells (Economides et al., 1989; Mukherjee and
through the application of hydraulic fracturing. While Economides, 1991). Deimbacher et al. (1993) reported
the productive benefit from hydraulically fractured that a horizontal well intersected by four evenly
horizontal wells is potentially large, it has not been spaced transverse hydraulic fractures over a drainage
pursued widely because of difficulties associated with area of 260 ha possessed a production index of 75%
complex fracture evolution, particularly multiple frac- greater than that of an unfractured horizontal well, and
turing. Transversely fractured wells have the ability to 200% greater than that of a fractured vertical well and
greatly increase production rates by virtue of the fact a longitudinally fractured horizontal well. The reduced
that any number of fractures may be widely distributed contact areas between horizontal wellbores and trans-
along the length of horizontal wells through separate verse fractures, however, introduce additional ‘choke
skin effects’ (McLennon et al., 1989) that hinder
productivity and increase injection pressures. The
*
Corresponding author. Fax: +61-2-9385-5182. effect is fortunately negligible in low-permeability
E-mail address: mk.rahman@unsw.edu.au (M.K. Rahman). formations.

0920-4105/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 0 - 4 1 0 5 ( 0 2 ) 0 0 2 3 6 - X
128 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150

The geometry of fractures initiating from the hor-

izontal well is dependent on the in situ stress which is
the most important reservoir parameter that affects
hydraulic-fracture design and treatment. Hydraulic
fractures from a wellbore, regardless of their origin,
always attempt to propagate in planes orthogonal to
the minimum in situ stress, what is called the ‘pre-
ferred fracture plane’. Thus, while hydraulic-fracture
propagation planes are fixed, the horizontal wellbores
may be drilled in any direction depending on other
considerations. The location of fracture initiation from
the wellbores depends on many factors such as the
presence of natural fractures, perforation orientation
and near-wellbore stress concentration. When the Fig. 2. Fracture twisting.
location of fracture initiation does not coincide with
the preferred direction of fracture propagation, various Twisting fractures are similar to turning fractures
propagation complexities such as fracture turning and with the exception that reorientation occurs in two
twisting occur (Weng, 1993). planes, as shown in Fig. 2. Due to the nature of near-
Turning hydraulic fractures (sometimes referred to wellbore stress concentrations, fractures are more likely
as ‘S’-shaped or reoriented fractures) propagate some to initiate longitudinally regardless of wellbore orien-
distance from the wellbore before reorienting towards tations or the nature of in situ stress regimes. However,
another plane, as illustrated in Fig. 1. There are many where horizontal wellbores are oriented perpendicular
causes of fracture turning, most of which relate to the to preferred fracture planes, initial longitudinal frac-
distribution of near-wellbore stresses at fracture ini- tures reorient as they propagate from the wellbores in
tiation (Akgun et al., 1997). The near-wellbore seg- response to the far-field stresses such that they even-
ments of turning fractures are often subjected to greater tually assume transverse configurations. Just as with
stresses than those acting on the main fractures leading fracture turning, fracture twisting produces severe
to near-wellbore width restriction which eventually pinching in near-wellbore regions. The degree of
results in a mechanical flow restriction. This fracture restriction in the case of twisting is, however, far more
pinching often leads to excessively high treating pres- acute and its effect on fracture treatments and produc-
sures and may, in extreme cases, lead to proppant tions are more severe. For example, a parametric study
bridging and premature, near-wellbore screen-outs. by Deimbacher et al. (1993) of the influence of a
fracture, which initiates longitudinally and then reor-
ients transversely by twisting on the productivity dem-
onstrated that fracture widths at the point of twisting
were up to 83% smaller than those initiating trans-
versely from the horizontal wellbore. Twisted fractures
are thought to commonly afflict field-scale fractured
horizontal wells (Weng, 1993) and are frequently ob-
served in laboratory experiments (Weijers et al., 1992).
Turning and twisting of fractures may also occur
during the propagation of multiple fractures, where the
process itself changes the local stress distribution and
influences the fractures to change their propagation
direction. Evidence supporting the ubiquity of multiple
fractures is largely circumstantial, usually inferred
through the application of treatment pressure theory.
Fig. 1. Fracture turning. Medlin and Fitch (1983) first promoted the concept of
 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150 129

multiple fracturing as a means of explaining the exces- etc. Analytical methods are mostly based on the stress
sive treatment pressures during massive hydraulic- function suggested by Westergaard (1939) and the
fracture treatments in low permeability sands of the complex stress function suggested by Mushkelishvili
Mesa Verde Formation. Numerous studies (Davidson (1977). The application of these methods is limited to
et al., 1993; Stadulis, 1995; Hainey et al., 1995) have a few simple loading situations. Finite element and
documented examples where hydraulic-fracture treat- boundary element methods are the widely used
ments displayed injection pressures which were sig- numerical techniques for computing stress intensity
nificantly higher than those predicted by conventional factors for problems with very complex geometry, as
fracture models. These studies concluded that the well as loading. For the problem considered herein,
presence of severe multiple fracturing was the primary analytical formulas for calculating the stress intensity
cause of these treatment pressure anomalies. factor for opening mode are available. These formulas
While hydraulic-fracture complexity is an acknowl- are essentially based on idealized shapes of fractures
edged impediment to the widespread application of and load distribution.
hydraulically fractured horizontal wells, there exists a
lack of practical design tools and guidelines with 2.1. Stress intensity factors for single fractures
which the engineers could predict the severity of
complex fracturing. To minimize the shortcoming, a A single, two-dimensional line fracture subjected
unique combination of experimental, analytical and to uniform internal fracture pressure ( Pf) penetrating a
numerical investigations was performed (Crosby, plate of infinite width acted upon by confining stress
1999). A companion paper has reported the findings (r3) is a commonly employed limiting fracture geom-
associated with transverse hydraulic-fracture initiation etry. Irwin (1957) derived an analytical solution for
from horizontal wells (Crosby et al., 2002). This paper the Mode I stress intensity factor at the tip of such a
presents the major findings associated with propaga- fracture as
tion of hydraulic fractures from horizontal wells. pffiffiffiffiffiffiffiffi
Numerical studies were carried out for this purpose KIs ¼ ðPf  r3 Þ pLf ð1Þ
using sophisticated numerical software, FRANC-3D
where Lf is the fracture half-length.
and HYFRANC-3D. Results show that two or more
A three-dimensional radial fracture in an infinite
multiple transverse fractures may propagate but they
medium, also subjected to uniform internal fracture
involve different treatment pressure behaviours com-
pressure and confining stress, is another limiting
pared to a single fracture.
fracture geometry. Sneddon (1946) proposed that the
stress intensity factor for such three-dimensional frac-
tures differed from those of two-dimensional cases
2. Fracture propagation theories
only by a factor of 2/p. That is, the Mode I stress
intensity factor for a radial fracture is provided by the
Theories for mixed mode fracture propagation are
following expression
well documented in the literature of linear-elastic frac-
ture mechanics (Erdogan and Sih, 1963; Sih, 1974; 2 pffiffiffiffiffiffiffiffi
KIs ¼ ðPf  r3 Þ pRf ð2Þ
Hussain et al., 1974). These theories are implemented p
in terms of ‘stress intensity factors’ or ‘strain energy
release rate’. Three basic modes of crack extension are where Rf is the fracture radius.
distinguished in the literature of linear-elastic fracture Tada (1985) found the following analytical solu-
mechanics—opening mode, shearing mode and tearing tion for the opening mode stress intensity factor at the
mode—and three different stress intensity factors KI, tip of a radial fracture intersected at its center by a
KII and KIII are defined, respectively. wellbore
The major theoretical methods for evaluating stress qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2ðPf  r3 Þ
intensity factors in a triaxial loading system include KIs ¼ pffiffiffiffiffiffiffiffi R2f  rw2 ð3Þ
Rf p
boundary collocation, conformal mapping, analytical
and numerical techniques, the use of Green function, where rw is the wellbore radius.
130 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150

2.2. Stress intensity factors for multiple fractures 2.3. Boundary element model for fracture propagation

Similar to single fractures, analytical solutions for the Obviously, equations presented in the foregoing
stress intensity factors for multiple fractures can be section are valid only for idealized fracture geometries
derived where fracture geometries are relatively simple. and loadings. For general three-dimensional problems
For example, for two-dimensional line fractures, Isida subjected to arbitrary loads, the displacement correla-
(1973) developed the following solution for stress inten- tion technique is found to be an efficient yet accurate
sity factor at the tips of multiple fractures as a function of method of calculating stress intensity factors for all the
spacing three modes (Ingraffea and Saouma, 1985). In this
pffiffiffiffiffiffiffiffi technique, displacements computed at nodes near the
KIm ¼ ðPf  r3 Þ pLf *FIðSÞ ð4Þ crack tip are correlated with the theoretical values
derived based on assumed displacement functions.
where KIm is the stress intensity factor at the tips of Based on this principle, functions of stress intensity
multiple fractures; FI(S) = 1  0.293S(1  (1  S)4); factors in terms of nodal displacements are derived by
S = Lf/(Lf + s) and s is the orthogonal half-spacing Ingraffea and Manu (1980) for isotropic materials using
between two multiple fractures. Eq. (4) is based on series quarter-point elements and by Saouma and Sikiotis
expansion of complex potentials. Using the ‘body force (1986) for anisotropic materials using singular isopara-
method’, Isida et al. (1985) derived a lengthy cumber- metric elements. Nodal displacements are usually esti-
some analytical equation for stress intensity factors at the mated either by finite element or boundary element
tips of elliptical multiple fractures. analysis.
It is noted that as the spacing between multiple The numerical fracture propagation tool used in
fractures approaches zero, the stress intensity
pffiffiffi factors this study was developed by the Cornell University
at their individual tips are reduced by 1= 2 (0.707) Fracture Group. The non-coupled version is called
compared with that of equivalent single fractures. This FRANC-3D (Fracture Analysis Code) which is a
gives the following relationship UNIX-based modeling system that analyses the
KIs response of linear-elastic materials, containing three-
KIm ¼ pffiffiffi : ð5Þ dimensional fractures of arbitrary geometries, to
2 applied stresses. Nodal displacements in the discretized
As the spacing between multiple fractures rock model are computed by a Boundary Element
pffiffiffi increases, System (BES) which is an efficient method for analyz-
KIm normalized by KIs varies from 1= 2 to 1. This
relationship is for uniformly pressurized parallel and ing crack/boundary problems, as only the cracks them-
symmetric fractures. For closely spaced N number of selves (as opposed to the surrounding media in the case
multiple fractures, the opening mode stress intensity of a finite element method) are discretized (Crouch and
factor is suggested by Crosby (1999) to be Starfield, 1983). Stress intensity factors are estimated
using the ‘displacement correlation technique’ (Ingraf-
KIs fea and Manu, 1980; Saouma and Sikiotis, 1986; Boon
KIm ¼ pffiffiffiffi : ð6Þ et al., 1987), which is capable of correctly reflecting the
N
effect of any number of fractures. Thus, there is no need
This
pffiffiffiffi is equivalent to increasing fracture toughness for explicit formulations for multiple fractures like
by N which qualitatively justifies the approach analytical solutions represented by Eqs. (4) – (6).
adopted by the petroleum industry in modeling N The modified version of FRANC-3D is called
multiple fractures using conventional commercial hy- HYFRANC-3D, which allows true, coupled simula-
draulic-fracture models such as FRACPRO (Resource tion of fluid-driven fractures. Such simulation requires
Engineering Systems) and FRACANAL (SIMTECH the highly nonlinear coupling between internal pres-
Consulting Services) by adjusting the elastic modulus. sure distribution arising from fluid flow and deforma-
Nolte (1987) first suggested these adjustments in tion of the fracture. Such coupling is a particular
single fracture models to use for analysis of multiple problem at the fracture tip, which is subject to large
fractures. fluid pressure gradients. HYFRANC-3D makes use of
 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150 131

‘Linear Elastic Hydraulic Fracturing’ (LEHF) rock/ with a negligible permeability (8e  5 mD). In order to
fluid coupled solutions in near-tip regions (Desroches provide sufficient contrast between the induced
and Carter, 1996). The capability of a dual-porosity hydraulic fracture and the fracture test material, finely
poroelastic model to analyze such coupled fluid flow ground (less than 45 Am) iron oxide was added to the
and deformation problems is also recently investi- fracture fluid at a concentration of 10% by weight of
gated by simulating laboratory experiments with frac- fracture fluid. The red colour of the iron oxide com-
tured rock specimens (Bai et al., 1999). bined with the existing pink colour of Mobilgear SHC
Both FRANC-3D and HYFRANC-3D have the 6800 to produce a bright red fracture fluid that con-
capacity to propagate hydraulic fractures of complex trasted well with the white fracture test material. Table 1
shape. They have options to simulate fracture prop- summarizes the laboratory test material properties and
agation using any of the criteria: (1) maximum tan- other test parameters compared to a set of their field-
gential stress (Erdogan and Sih, 1963), (2) maximum scale values. Injection rates and fluid viscosity were,
energy release rate (Hussain et al., 1974), (3) mini- however, varied for different configurations.
mum strain energy density (Sih, 1974) and (4) planar The synthetic fracture test material was cast into
propagation (fracture extension coplanar with initial blocks measuring 400  400  400 mm. Thin circu-
fracture). In the numerical examples presented herein,
the maximum tangential stress criterion was employed
for mixed-mode propagation of fractures. Table 1
A set of field-scale fracturing parameters and their laboratory scaled
dimensions

3. Hydraulic-fracture experiments at laboratory Parameter Field-scale Laboratory

dimension scale dimension

A series of laboratory scale hydraulic-fracture ex- Wellbore dimensions

Wellbore radius, 0.091 0.006
periments was carried out in an attempt to investigate rw (m)
the initiation and propagation of, and interaction Reservoir properties
between multiple transverse fractures (Crosby, 1999). Permeability, 0.1 8e  5
In order to produce field-representative results, dimen- K (mD)
sional analysis was used to derive the appropriate scale Geomechanical properties
Elastic modulus, 50 000 7000
factors (de Pater et al., 1992). The laboratory test E (MPa)
dimensions were scaled upon those typically encoun- Poisson’s ratio, v 0.25 0.2
tered during the exploitation of low permeability gas Fracture toughness,
pffiffiffiffi
3 0.27
resources in Central Australia. Length dimensions were KIc (MPa m)
scaled according to wellbore radius (rw). Extremely low In situ stresses
Vertical in situ stress, 64 7
injection rates were employed in order to achieve stable rv (MPa)
fracture initiation and growth. This necessitated the use Minimum in situ 47 5
of an extremely high viscosity fracture fluid and a low horizontal stress,
leak-off fracture medium. Correct scaling is also rh (MPa)
required so that the fracture medium possesses a low Maximum in situ 57 6
horizontal stress,
modulus (E) and toughness (KIc). A high viscosity rH (MPa)
industrial lubricant, Mobilgear SHC 6800 was Hydraulic-fracture
employed as a fracture fluid. This fluid is Newtonian treatment data
in nature. The fracture test material was a portland Leak-off, 4.8e  5 3.8e  8
pffiffi
cement-based material composed of 70% silica flour, kl (m= s)
Injection rate, 0.072 3e  7  1e  6
30% portland cement. The low cement content ensured Q (m3/s)
that the fracture test material possessed a low elastic Injection period, 27 1
modulus
pffiffiffiffi (7000 MPa) and low toughness (0.27 MPa t (min)
m). In addition, the high proportion of silica flour and Fracture fluid 0.5 30 (at 23 jC)
inclusion of acrylic emulsion provided the test material viscosity, l (Pa s)
132 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150

Fig. 3. Rock block configuration for laboratory tests.

lar plastic disks were positioned inside the blocks hydraulic fractures initiated. Two separate block tests
during casting (Fig. 3). When intersected by a well- were performed. The first contained a single wellbore
bore, these disks constrained the location at which of 12 mm in diameter drilled through a single plastic

Fig. 4. ‘Single completion’ fracture test configuration.

 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150 133

Fig. 5. ‘Dual completion’ fracture test configuration.

disk, as shown in Fig. 4. The surfaces of the plastic In another test configuration described in Fig. 5,
disc were roughened in order to enhance cohesion two plastic discs separated by a distance of 30– 50 mm
with the cement. were intersected by two hydraulically isolated well-

Fig. 6. Polyaxial cell configuration for fracture tests.

134 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150

bores, each possessing a radius of 6 mm. Stainless steel computer-controlled linear displacement pump, while
injection tubes of 3-mm internal radius were grouted the dual completion tests used two (one for each
into the wellbores, leaving a small ‘open hole’ section wellbore), shown in schematic form in Fig. 7. In
immediately adjacent to the plastic disks. The absence addition, transducers independently measured the ini-
of a sand fraction left the synthetic fracture medium tiation and propagation pressures of each multiple
susceptible to shrinkage cracking. However, this was fracture. During injection testing, both wellbores were
minimized by storing the blocks under conditions of pressurized at similar rates. A primary fracture even-
100% humidity during curing and until immediately tually initiated from one of the two wellbores. Imme-
prior to testing. In this block, grease was applied to the diately after the primary fracture initiated, the
surfaces of the plastic disks in order to decrease wellbore from which it emanated was shut-in. Injec-
cohesion, simulating the presence of uncemented nat- tion into the primary fracture recommenced only after
ural fractures or pre-broken-down perforations. the secondary fracture (from the opposing wellbore)
The blocks were placed in a polyaxial cell, as was initiated. Injection into both fractures was
illustrated in Fig. 6. Through the use of water-filled allowed to proceed at a constant rate for a period of
flat-jacks, the polyaxial cell exerted stresses on all approximately 2 min, whereupon they were both shut-
faces of each block. The magnitudes of the applied in. Thus, the experiments allowed an approximate
stresses (Table 1) were such that the wellbores were injection time of 120 s for fracture propagation.
oriented in the direction of the minor principal stress, Although the injection rate plays a significant role in
rz. This orientation between wellbore and stresses fracture initiation and early growth, the injection rate
induced the formation of transverse hydraulic frac- was not varied in any of the test configurations to
tures. The single fracture test employed only one study this effect.

Fig. 7. Schematic of the injection and data recording system for laboratory fracture tests.
 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150 135

Fig. 8. Injection pressure record during multiple fracturing test, DC-2.

Based on measured low permeabilities, the leak-off fractures for test configuration DC-3 is illustrated in
during fracture propagation was assumed to be negli- Fig. 9 with corrections for compressibility of the
gible. Injection period vs. pressure curves for all the injection system and fracture fluid. The figure shows
test configurations broadly supported this assumption a strong decline in the measured pressure after shut-in.
with very slow pressure decline after shut-in (e.g. Fig. The laboratory configuration was such that upon
8 for test DC-2), except test configuration DC-3. The completion of injection, valves were shut in the injec-
injection pressure recording for primary and multiple tion line close to the top of the test block (although not

Fig. 9. Injection pressure record during multiple fracturing test, DC-3.

136 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150

shown in Fig. 7) to prevent fluid passing back to the modeled for efficient computation. By this way, it was
sump. Therefore, the post-shut-in pressure decline possible to restrain the lateral movements on both x –z
must be due to either fracture extension or leak-off, and y –z planes. Movements along z-direction on the
or both. The authors believe that it was due to the top and bottom faces of the model were restrained to
combined effect of both and the leak-off rate was impose plane strain condition. The block, in which the
significantly high due to some cavity or lack of fracture was modeled, was subjected to stresses: rz = 5
material compaction in this test block. MPa; rx = 6 MPa; ry = 7 MPa. The wellbore was
oriented in the direction of rz. The fracture possessed
a uniform internal pressure of 20 MPa. Fig. 10
4. Fracture propagation analysis illustrates the configuration of the entire model for
numerical analysis.
4.1. Single fracture analysis Fig. 11 illustrates the Mode I stress intensity factors
estimated by FRANC-3D as functions of fracture radii
A comparison between the analytical and numer- and compares them with the analytical solution of
ical (FRANC-3D) estimates of stress intensity factors Tada (1985) (Eq. (3)). The figure shows a reasonable
for a single, uniformly pressured fracture was carried agreement between the analytical and numerical sol-
out in order to confirm the validity of the numerical utions, and thus provides confidence in the validity of
model. The fracture geometry, block dimensions and the numerical model results. Accuracy of the numeri-
material properties were defined based on those of the cally derived stress intensity factors was found to
laboratory hydraulic-fracture tests. improve with an increasing number of fracture tip
Radial fractures of varying diameters were mod- boundary elements. This, however, severely increased
eled. These fractures intersected a wellbore (0.01 m in run-times. In addition, the adoption of square-shaped
diameter) in a transverse fashion. Using the benefit of (as opposed to triangular or rectangular) fracture tip
symmetry, only a quarter of the block and fracture was boundary elements also increased accuracy.

Fig. 10. Discretized block and single, radial fracture intersecting the wellbore transversely.
 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150 137

Fig. 11. Comparison between numerical (FRANC-3D) and analytical estimates of Mode I stress intensity factors for a single fracture described
in Fig. 10.

4.2. Parallel multiple fractures analysis and were parallel. Fig. 12 shows the typical multiple
parallel fracture mesh scheme. Loading conditions
Numerical analyses using FRANC-3D were per- were similar to those applied in the single fracture
formed on a pair of closely spaced radial multiple case and each multiple fracture was equally pressur-
fractures. The multiple fractures were of equal size ized. Only a quarter of the block was modeled based

Fig. 12. FRANC-3D multiple parallel fracture mesh scheme.

138 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150

on the same arguments presented in the previous 4.3. Deviating multiple fracture analysis
section. It was not possible to model a half-space of
the two fractures for computational efficiency because The two closely spaced radial multiple fractures
the effect of one fracture on the other could not be modeled in Fig. 12 were allowed to propagate in
represented by a suitable boundary condition. increments not greater than 0.005 m to investigate
Fig. 13 illustrates normalized multiple fracture their propagation behavior and its effects on the stress
stress intensity factors for radial fractures of varying intensity factors. The block and fractures were
separations, as derived from the FRANC-3D model. stressed in a similar manner to the parallel fracture
Plotted on the same figure are the Isida (1973) case described earlier. The fractures were initially
analytical solution for line fractures (Eq. (4)) and separated by 0.01 m and had equal radii of 0.02 m.
the Isida et al. (1985) numerical solution for radial Fig. 14 illustrates the three-dimensional geometry of
fractures. Note that as for the line fracture analytical the propagated fractures, as determined by FRANC-
solution, the normalized stress intensity factors 3D. Fig. 15 describes the path of fracture propagation
derived
pffiffiffi from the FRANC-3D model approach 1= projected onto the z – x plane of symmetry. Despite an
2 (0.707) as the space between two fractures initial irregularity, which resulted from problems
decreases. This definitely supports the relationship associated with numerical convergence, the multiple
defined by Eq. (5) for two parallel fractures. The fractures linearly deviated away from one another.
numerical results of this study support the work by Fig. 16 displays the normalized stress intensity
Isida et al. (1985) to the close-spaced asymptote and factor at the tip of each diverging multiple fractures,
confirm that the degree of decrease in the stress as determined by FRANC-3D. Plotted on the same
intensity factors at the tips of multiple fractures at figure are the results from solutions for line fractures
close-spaced asymptotes is independent of fracture (Eq. (4); Isida, 1972) and for radial fractures (Isida et
geometry. al., 1985). The FRANC-3D solution indicates that as

Fig. 13. Normalized stress intensity factors for multiple parallel fractures as estimated by FRANC-3D and solutions for line fractures (Eq. (4))
and radial fractures (Eq. (5)).
 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150 139

Fig. 14. Diverging multiple radial fractures, as simulated by FRANC-3D.

the fractures diverge (decreasing s/Rf), the normalized al. (1985) trend lines. As the multiple fractures diverge,
stress intensity factors at their tips approach 1.0. This the stress intensity factors at their tips approach that
differs significantly from the Isida (1973) and Isida et of a single fracture. Note that one FRANC-3D data

Fig. 15. Propagation paths of multiple radial fractures obtained from FRANC-3D analysis.
140 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150

Fig. 16. Normalized stress intensity factors for deviating multiple fractures as functions of radii.

point in the figure does not appear to follow the trend dary fracture of fixed radii (Rfm) upon the stress
set by the other points. This anomaly is associated with intensity factor of a primary fracture of increasing
the same numerical convergence problem that resulted radius (Rfs). The results demonstrate that the close
in the nonuniform initial fracture propagation path. proximity of secondary fractures has a negligible
impact upon the stress intensity factors of primary
4.4. Asymmetric multiple fractures analysis fractures, unless both fractures are of similar length.
For example, with reference to Fig. 17, if the radius of
Propagation of multiple fractures with equal radii the primary fracture is 1.2 times the secondary frac-
is very rare in practice. Usually, single, primary ture (Rfs/Rfm = 1.2), the stress intensity factor of the
fractures dominate propagation at the expense of primary fracture differs from that of a single fracture
much smaller secondary fractures. To study this by only about 14%. As expected, at large multiple
fracture configuration, a FRANC-3D numerical anal- fracture spacings (greater than 1.0 m), the presence of
ysis similar to that described above for multiple multiple fractures has a negligible impact upon the
parallel fractures was conducted. Only the primary stress intensity factors.
fracture was permitted to propagate in the presence
of a static smaller secondary fracture. The secondary
fracture radius was fixed at 0.02 m, while the 5. Comparison of laboratory results and numerical
primary fracture was propagated from an initial results
radius of 0.02 – 0.1 m. Separation between the pri-
mary and secondary fracture planes was thus main- While the technique for modeling the propagation
tained at 0.01 m by planar propagation of both of multiple fractures has been incorporated into a
fractures. number of commercial models, no laboratory study
The stress intensity factor of the primary fracture has yet been carried out to test its validity. As such,
normalized by that of a single fracture is plotted in one of the key aims of the experimental programme
Fig. 17. This figure shows the influence of a secon- undertaken in this study was to propagate a pair of
 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150 141

Fig. 17. Influence of secondary fractures upon the stress intensity factors of primary fractures with varying radius.

closely spaced multiple fractures, and compare the 5.1. Plotting of laboratory propagation pressures
propagation pressures of each individual fracture
branch with those predicted by the dynamic fracture Results of hydraulic-fracture propagation tests in
model. the laboratory are plotted in Fig. 18. For this plotting,

Fig. 18. Dimensionless net fracture pressures as functions of propagation period for all the laboratory hydraulic fractures.
142 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150

the net fracture pressures and the injection periods are showing that the final measured fracture radii closely
normalized as Pn/P* and t/s*, similar to that adopted match with fracture radii estimated from theoretical
by de Pater et al. (1993), where time – radius relationship. For two-dimensional radial
  1=4 fractures, Crockett et al. (1986) proposed relationships
E (Appendix A) for fracture radius, width and net
P* ¼ lQ ð7Þ
rw pressure as functions of injection rate, time and fluid
parameters. As mentioned earlier, all the fractures
 3=4  1=4 were allowed to propagate for 120 s after their
rw3 l
s* ¼ ð8Þ initiation. Measured fracture radii are compared with
Q E
that estimated by Eq. (A1) for this propagation time in
where P* is the characteristic net fracture pressure; s* Table 2 which shows acceptably close match for all
is the characteristic injection time; Q is the fracture the configurations.
fluid injection rate per fracture wing (as measured at The plotting of raw data created relatively scattered
pumps); l is the effective channel flow fluid viscosity curves due to
variations in wellbore fluid pressure
(equivalent to 12 times Newtonian fluid l); with time DP Dt . In order to provide better contrast
w
 viscosity,

E between the various experimental results, the mean
and E is the crack opening modulus 4ð12 Þ .
laboratory fracture propagation pressure trends for
It was apparent that compliance of the laboratory different test configurations are plotted in Fig. 18.
injection systems played a significant role during Propagation of the laboratory hydraulic fractures was
fracture propagation. A significant proportion of this taken to commence when the injection pressure deriv-
compliance was expected to be due to fluid compres- atives with respect to time deviated from their initially
sibility. Using estimates of injection system compli- constant values. Thus, Fig. 18 in fact shows fracture
ance, the laboratory injection rates were modified propagation time.
through the use of the following expression
Vsys DPw 5.2. Single fracture propagation pressures from tests
Q* ¼ Q þ ð9Þ
Csys Dt and radial model

where Q* is the net fracture fluid injection rate, per The fracture propagation pressures of the single
fracture wing, adjusted for fluid decompression; Vsys fracture laboratory experiments (DC-4 and -6), shown
is the total volume of fluids within the injection in Fig. 19, compare closely with those estimated by the
system; Csys is the average compliance of the entire single fracture radial model (Eq. (A1)). This provides
injection system and DPw/Dt is the rate of fluid confidence in the experimental procedure. It appears
pressure variation with respect to time. that material toughness and artificial toughness (due to
The comparison of normalized measured pressures the action of the confining stress on the large fluid lag
for different fracture configurations may lead to erro- zone) had a negligible effect on propagation pressures.
neous conclusions about the relative value of the net This is expected, considering the low ratio of confining
pressure if the high pressure at an early time happens
to be compared with a low pressure at a later time for
another configuration. The best way to avoid such an
Table 2
erroneous pressure comparison was to compare pres- Comparison of laboratory created and estimated fracture radii
sures with respect to fracture radius. The experimental
Test Injection Fluid Measured Estimated
procedure followed in this work, however, precluded rate (cm3/s) viscosity radius (mm) radius (mm)
measurement of fracture radius as function of injec- (Pa s)
tion time. Only the final fracture radii were measured DC-2 0.012 33 99 87
through dissection of the block. Therefore, it was not DC-3 0.011 81 93 78
possible to plot pressure as a function of the growth of DC-4 0.024 53 99 105
fracture radius. An indirect way is to show the DC-6 0.005 31 83 67
pressure behavior as a function of injection time by DC-8 0.005 33 64 65
 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150 143

Fig. 19. Comparison between the dimensionless net fracture pressures of the single laboratory hydraulic fractures, DC-4 and DC-6, and single-
fracture radial model.

stress to net fracture pressure (approximately 0.5). Significant discrepancies consistently exist be-
These results also seem to agree with Gardner’s tween the early time net fracture pressures measured
(1992) observations regarding the insensitivity of in the laboratory and those predicted by Eq. (A3) with
fracture widths and pressures to fluid lag. coefficients for two fractures, i.e. N = 2. This is prob-
ably because the radial model assumes that the frac-
5.3. Multiple fracture propagation pressures from tures possessed consistently symmetrical, radial and
tests and radial model planar geometry from a very early time. In the labo-
ratory, however, it was quite likely that very early
The fracture propagation pressures of the multiple fracture growth was characterized by highly asymmet-
fracture laboratory experiments (DC-3 and -8) were ric propagation, particularly during inflation of the
slightly higher than those of the single fractures. The region surrounding the plastic disks. These early time
multiple fractures in laboratory test DC-3 grew in pressure discrepancies are a common feature of labo-
opposite directions with respect to the wellbore axis. ratory scale hydraulic fractures (de Pater et al., 1992,
The multiple fractures also deviated from the oppos- 1993). It was also unfortunate that the laboratory
ing fracture planes. In addition, the lower (secondary) fractures could not be propagated for longer time
fracture displayed a slightly greater net pressure than periods in order to prevent propagation to the edges
the upper (primary) fracture, as demonstrated in Fig. of the blocks. The fracture propagation trends, how-
20. This slight difference in pressure is reasonable ever, are reasonably clear.
considering that the lower fracture initiated in the In the multiple fracture experiment DC-8, the upper
presence of the (already initiated) primary fracture. (secondary) fracture was forced to propagate for its
The initiation sequence probably contributed to the entirety in relatively close proximity to an adjacent,
greater deviation from the plane of preferred prop- lower (primary) fracture (Fig. 20). The primary frac-
agation experienced by the lower fracture, as illus- ture was initially propagated a short distance before
trated in Fig. 21. propagation of the secondary fracture was com-
144 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150

Fig. 20. Comparison between the dimensionless net fracture pressures of the multiple laboratory hydraulic fractures, DC-3 and DC-8, and radial
fracture model.

menced. The primary fracture displayed, as expected, model. In addition, the secondary fracture pressure of
the net pressure characteristics more of a single frac- DC-8 declined rapidly with propagation, displaying a
ture. The secondary fracture displayed a much greater pressure decline trend approaching that of the DC-3
net propagation pressure, and while certainly larger primary fracture.
than that predicted by the single fracture radial model, Fig. 22 illustrates that the secondary DC-8 fracture
was smaller than that predicted by the multiple fracture deviated significantly from the preferred fracture

Fig. 21. Multiple fractures created during the dual completion test, DC-3.
 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150 145

Fig. 22. Multiple fractures created during the dual completion test, DC-8.

plane. This natural tendency to achieve a lower energy Experimental multiple fractures presented in this
state prevents the dual-fracture configuration adopted paper can be considered as the representative of the
in experiment DC-8 from adhering to the idealized near-wellbore condition because they were not propa-
multiple fracture radial model pressures. Such fracture gated far in the block. Other published laboratory
deviation would be expected in most multiple fracture results consistently show that secondary multiple frac-
environments. However, it would be expected that as tures rarely extend beyond the near-wellbore region, as
the number of multiple fractures increases, their they are squeezed shut by primary fractures that invar-
ability to deviate from each other becomes limited. iably develop (Narendran and Cleary, 1983; Behrmann
Thus, it is likely that the greater the number of and Elbel, 1991; Weijers et al., 1992). This is in contrast
multiple fracture strands, the more close their prop- to the widespread anecdotal conclusion that multiple
agation pressures are to those estimated by idealized fractures of roughly equal length do penetrate far into
multiple fracture models. formations in absence of dominating primary fractures.
From this comparative study of fracture propagation This conclusion is basically derived from bottom-hole
pressures, a significant observation is that the net treating pressure records. The results of the multiple
pressures in multiple fractures are consistently greater fracture experiments of this and above mentioned
than that expected of a single-radial fracture, despite studies suggest that the primary fractures dominate
the fact that the multiple fractures essentially develop and thus the conclusion of deeply penetrating multiple
as single fractures. This suggests that small, inflated fractures of equal length far into the formation may be
secondary multiple fractures may increase the propa- incorrect. Similar to the experimental findings in this
gation pressures of the primary fracture, though not to study, Germanovich et al. (1997) also suggested that
the same degree as closely spaced, symmetrical multi- from analytical fracture modeling, the close proximity
ple fractures. This may be most relevant where secon- of but partially opened multiple fractures, even at the
dary fractures are propped, pos-sibly as a result of the near-wellbore region, can increase the propagation
use of proppant slugs, early in treatments. pressures of ‘primary fractures’, though not to the same
146 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150

extent as for equally proportioned ‘far-field’ multiple and the highly complex nature of multiple fracture
fractures. iteration, each propagation step required between 2
days and 1 week. The entire propagated fracture set
5.4. Laboratory and numerical model propagation illustrated in Fig. 23 required over 1 month to run.
pressures and fracture geometries The authors, however, acknowledge that alternative
physical models might be developed which could be
In the previous section, non-planar propagation of computationally more efficient.
fractures was pointed to explain the discrepancy Fig. 24 compares the laboratory measured and
between the net fracture pressures measured in the numerically derived paths of the propagated multiple
laboratory and estimated by the radial model. To fractures. The most salient feature of this analysis is
investigate this further, the multiple fracture labora- the marked mutual deviation of the multiple fractures
tory test DC-8 was modeled using HYFRANC-3D, consistent with the work of Jeffrey et al. (1987) and
which has the capability to simulate non-planar frac- Cleary et al. (1993). In addition, the numerically
ture propagation. Fig. 23 illustrates the applied boun- derived fracture paths display a slightly greater devia-
dary conditions and fracture mesh. Modeling was tion, due likely to its rigid linear-elastic assumptions.
performed using material properties and injection In contrast, the laboratory fractures deviated to a
parameters similar to those employed during the lesser extent and to a declining degree with propaga-
laboratory test DC-8. Fracture propagation was per- tion, such that they approach a parallel relationship.
formed with discrete fracture increments. The max- Fig. 25 compares the normalized net fracture
imum tangential stress criterion was employed to propagation pressures for the laboratory fracture, the
propagate the fracture at each step. Fluid leak-off in analytical fracture model with two fractures and the
the model was set to zero. numerical model. Due to the characteristic divergence
Despite only a half of the block was discretized, of the multiple fractures, the analytical solution over-
each propagation step was extremely time consuming, estimates the net propagation pressures. In contrast,
even while running on a 500-MHz Dec Alpha. Due to the numerical and laboratory pressures display a
the high mesh density required to maintain stability greater correlation although the discrepancy between

Fig. 23. Divergent multiple hydraulic fracture propagation by HYFRANC-3D.

 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150 147

Fig. 24. Comparison of multiple fracture propagation at the laboratory and that simulated by HYFRANC-3D.

Fig. 25. Comparison of dimensionless net fracture pressures of the laboratory multiple fracture propagation (DC-8) and HYFRANC-3D
simulation.
148 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150

them is also considerable. Of particular note is the results, however, confirm that small, inflated secon-
late-time ‘flattening’ trend in numerically estimated dary multiple fractures increase the propagation pres-
pressure. It is uncertain whether this is a real feature or sure of the primary fracture. It is evident from literature
an artifact of the coarse resolution of the numerical survey and experimental results presented in this paper
pressure measurements. that secondary multiple fractures rarely extend beyond
the near-wellbore region because they are squeezed
shut by the primary fractures. These near-wellbore
6. Conclusions multiple fractures, however, significantly increase the
fracture propagation pressure. This effect becomes
Major findings of this study can be concluded as severe when closely spaced, symmetrical multiple
follows. fractures initiate and propagate even at the near-well-
(1) Reasonably good agreement is found between bore region.
the analytically and numerically estimated Mode I (5) The numerical model has simulated the exper-
stress intensity factors for the single fracture sce- imentally obtained diverging fracture propagation
nario. paths, which also deviated from the preferred fracture
(2) For two elliptical parallel fractures, the Mode I propagation direction. The numerically derived frac-
stress intensity factors estimated using the numerical ture paths display a slightly greater deviation, which
model and analytical methods are found to be in can be explained by the underlying assumption of
good agreement. It is also found that the degree to rigid, linear elasticity of materials.
which the stress intensity factor at the tip of multiple (6) Based on the comparison of net fracture pres-
fractures at a close-spaced asymptote reduces is sures and propagation paths, it is plausible to indicate
independent of fracture geometry. The stress inten- the existing weakness of the boundary element method
sity factor for two fractures reduces by a factor of in modeling highly nonlinear multiple fracture prob-
1/M2 compared to a single fracture in all the three lems such as the ones addressed in this paper. This
cases when the fracture space approaches a very study has, however, provided a high confidence in the
small value. great potential of such a numerical model, probably
(3) With increased divergence between two adja- with enhanced capability for the investigation of more
cent fractures, the stress intensity factor at each frac- complex hydraulic-fracturing problems.
ture tip approaches a value of that for a single fracture:
a fact that can also be supported intuitively. This effect
has been successfully modeled numerically, which has Acknowledgements
not been possible by using the ana-lytical formulas for
planar radial fractures. Consequently, the results for
The authors wish to acknowledge the Australian
deviating multiple fractures obtained by using the
Petroleum Cooperative Research Centre (APCRC) for
numerical model and the analytical formulas differ
supporting this work and CSIRO Petroleum for use of
significantly. Similarly, only the numerical model has
their laboratory facilities in Melbourne. The authors
successfully demonstrated that the close proximity of
also express their sincere appreciation to C.J. de Pater
secondary fractures has a negligible impact upon the
and the other anonymous reviewer for their con-
stress intensity factors of primary fractures unless both structive comments, which have improved the stand-
fractures are of approximately equal length.
ard of this paper.
(4) Experimentally obtained fracture propagation
pressures are compared with that estimated by the
numerical model and analytical formulas. It is found
that the analytical formulas overestimate the net prop- Appendix A
agation pressures significantly. In contrast, the numer-
ical and laboratory pressures display a greater cor- For a generalized two-dimensional radial fracture
relation although the discrepancy between them is still usually used in conventional, commercial hydraulic-
considerable. Experimental, analytical and numerical fracture models, the fracture parameters can be de-
 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150 149

fined through use of simple differential equations Vsys total volume of fluid injected in the system
derived by Crockett et al. (1986) as ww fracture width at the wellbore
r3 confining stress that tends to close the
"  3 # 19 fracture (closure stress)
3 c2 c4 E Q 4
rx stress applied on a face of the rock block
Rf ¼ t9 ðA1Þ
64 c1 c3 ð1  r2 Þl 2cr perpendicularly to the wellbore axis
ry stress applied on the orthogonal face of the
rock block perpendicularly to the wellbore
" 2  3 # 19 axis
64 c1 c3 ð1  r2 Þl Q 1 rz stress applied to the rock block parallel to the
ww ¼ t9 ðA2Þ
3 c2 c4 E 2cr wellbore axis
l fracturing fluid viscosity
l effective channel flow viscosity (12l)
  13 s* characteristic injection time
E 64 c1 c3 l 1
Pn ¼ t3 ðA3Þ
c1 4ð1  r2 Þ 3 c2 c4 E

where ww is the maximum hydraulic fracture width References

at the wellbore and Pn is the net fracture pressure
Akgun, F., Yang, Z., Crosby, D.G., Rahman, S.S., 1997. Factors
( Pf  r3). Other lumped coefficients for radial frac-
affecting hydraulic fracture initiation in high in-situ stress con-
tures are: cm = p/3; c1 = 2/p; c2 = 0.5; c3 = p/3 and ditions: a wellbore stress modelling approach. SPE 38631, SPE
c4 = 1.0. Coefficients by which single-radial fracture Annual Technical Conference and Exhibition, San Antonio,
parameters, such as width and net fracture pressure Texas, Oct. 5 – 8, 621 – 633.
(Eqs. (A1) and (A3)) are modified to arrive at corre- Bai, M., Meng, F., Elsworthy, D., Abousleiman, Y., Roegiers, J.-C.,
1999. Numerical modelling of coupled flow and deformation in
sponding multiple fracture values, are N2/3 and N  5/9,
fractured rock specimens. International Journal for Numerical
respectively. and Analytical Methods in Geomechanics 23, 141 – 160.
Behrmann, L.A., Elbel, J.L., 1991. Effect of perforations in fracture
Nomenclature initiation. SPE 20661. Journal of Petroleum Technology, 608 –
Csys average compliance of the injection system 615. May.
Boon, T.J., Wawrzynek, P.A., Ingraffea, A.R., 1987. Finite element
E Young’s modulus of rock
modelling of fracture propagation in orthotropic materials. En-
E crack opening modulus (E = E/4(1  m2)) gineering Fracture Mechanics 26 (2), 185 – 201.
KIs Mode I stress intensity factor for a single Cleary, M.P., Johnson, D.E., Kogsbol, H.H., Owens, K.A., Perry,
fracture K.F., de Pater, C.J., Stachel, A., Schmidt, H., Tambini, M.,
KIm Mode I stress intensity factor for multiple 1993. Field implementation of proppant slugs to avoid prema-
fractures ture screen-out of hydraulic fractures with adequate proppant
concentration. SPE 25892, SPE Rocky Mountain Regional/
Lf half-length of a line fracture Low Permeability Reservoirs Symposium, Denver, Colorado,
N number of multiple parallel fractures Apr. 12 – 14.
Pf average uniform pressure inside fractures Crockett, A.R., Okusu, N.M., Cleary, M.P., 1986. A complete in-
Pn fracture net pressure ( Pf  r3) tegrated model for design and real-time analysis of hydraulic
P* characteristic net fracture pressure fracturing operations. SPE 15069, 56th SPE California Regional
Meeting, Oakland, California, Apr. 2 – 4.
Q fracturing fluid injection rate Crosby, D.G., 1999. The initiation and propagation of, and interaction
Q* decompression adjusted injection rate per between, hydraulic fractures from horizontal wellbores. A disser-
fracture wing tation for PhD, The University of New South Wales, Sydney.
Rf radius of a radial fracture Crosby, D.G., Rahman, M.M., Rahman, M.K., Rahman, S.S., 2002.
rw wellbore radius Single and multiple transverse fracture initiation from horizontal
wells. Journal of Petroleum Science and Engineering, in press.
s half-spacing between two parallel multiple Crouch, S.L., Starfield, A.M., 1983. Boundary Element Methods in
fractures Solid Mechanics. George Allen and Unwin, London.
t fracturing fluid injection time Davidson, B.M., Saunders, B.F., Robinson, B.M., Holditch, S.A.,
150 M.M. Rahman et al. / Journal of Petroleum Science and Engineering 35 (2002) 127–150

1993. Analysis of abnormally high fracture treating pressures Isida, M., 1973. Method of Laurent series expansion for internal
caused by complex fracture growth. SPE 26154, SPE Gas Tech- crack problems. In: Sih, G.C. (Ed.), Mechanics of Fracture I.
nology Symposium, Calgary, Alberta, Canada, Jun. 28 – 30. Noordhoff International Publishing, Leiden, pp. 56 – 130.
Deimbacher, F.X., Economides, M.J., Jensen, O.K., 1993. General- Isida, M., Hirota, K., Noguchi, H., Yoshida, T., 1985. Two parallel
ized performance of hydraulic fractures with complex geometry elliptical cracks in an infinite solid subjected to tension. Interna-
intersecting horizontal wells. SPE 25505, Production Operations tional Journal of Fracture 27, 31 – 48.
Symposium, Oklahoma City, Oklahoma, USA, 891 – 902. Mar. Jeffrey, R.G., Vandamme, L., Roegiers, R.G., 1987. Mechanical
21 – 23. interactions in branched or subparallel hydraulic fractures.
de Pater, C.J., Cleary, M.P., Quinn, T.S., Barr, D.T., Johnson, D.E., SPE 16422, SPE/DOE Low Permeability Reservoirs Symposi-
Weijers, L., 1992. Experimental verification of dimensional um, Denver, Colorado, May 18 – 19.
analysis for hydraulic fracturing. SPE 24994, European Petro- McLennon, J.D., Roegiers, J.-C., Economides, M.J., 1989. Ex-
leum Conference, Cannes, France, Nov. 16 – 18. tended reach and horizontal wells. In: Economides, M.J., Nolte,
de Pater, C.J., Weijers, L., Savic, M., Wolf, K.H.A.A., van den K.G. (Eds.), Reservoir Simulation, 2nd edn. Prentice-Hall, New
Hoek, P.J., Barr, D.T., 1993. Experimental study of nonlinear Jersey, pp. 19-01 – 19-07.
effects in hydraulic fracture propagation. SPE 25893, SPE Joint Medlin, W.L., Fitch, J.L., 1983. Abnormal treating pressures in
Rocky Mountain Regional Meeting/Low-Permeability Reser- MHF treatments. SPE 12108, Annual Technical Conference
voirs Symposium, Denver, Apr. 26 – 28. and Exhibition, San Francisco, CA, Oct. 5 – 8.
Desroches, J., Carter, B.J., 1996. 3 Dimensional modelling of a Mukherjee, H., Economides, M.J., 1991. A parametric comparison
hydraulic fracture. Proc. of 2nd North American Rock Mecha- of horizontal and vertical well performance. SPE 18303, SPE
nics Symposium (NARMS), Montreal, Quebeck, Canada, 995 – Formation Evaluation, 209 – 216. June.
1002. Mushkelishvili, N.I., 1977. Some Basic Problems of the Mathemat-
Economides, M.J., McLennan, J.D., Brown, E., Roegiers, J.-C., ical Theory of Elasticity. Noordhoff, Leyden.
1989. Part I: performance and stimulation of horizontal wells. Narendran, V.M., Cleary, M.P., 1983. Analysis of growth and inter-
World Oil 208, 41 – 45. June. action of multiple hydraulic fractures. SPE 12272, Reservoir
Erdogan, F., Sih, G.C., 1963. On the crack extension in plates under Stimulation Symposium, San Francisco, CA, Nov. 15 – 18.
plane loading and transverse shear. Journal of Basic Engineering Nolte, K.G., 1987. Discussion of influence of geologic discontinu-
85, 519 – 527. ities on hydraulic fracture propagation. Journal of Petroleum
Gardner, D.C., 1992. High fracturing pressures for shales and which Technology, 998. Aug.
tip effects may be responsible. SPE 24852, Annual Technical Saouma, V.E., Sikiotis, E.S., 1986. Stress intensity factors in aniso-
Conference and Exhibition, Washington, DC, Oct. 4 – 7. tropic bodies using singular isoparametric elements. Engineer-
Germanovich, L.N., Ring, L.M., Astakhov, D.K., Shlyapobersky, J., ing Fracture Mechanics 25 (1), 115 – 121.
Mayerhofer, M.J., 1997. Hydraulic fracture with multiple seg- Sih, G.C., 1974. Strain energy density factor applied to mixed mode
ments I. Observations and model formulation: II. Modeling. crack problems. International Journal of Fracture 10 (3), 305 –
International Journal of Rock Mechanics and Mining Sciences 321.
34 (3 – 4) Paper No. 097 and 098. Sneddon, I.N., 1946. The distribution of stress in the neighbourhood
Hainey, B.W., Weng, X., Stoisits, R.F., 1995. Mitigation of multiple of a crack in an elastic solid. Proceedings of the Royal Society
fractures from deviated wellbores. SPE 30482, Annual Techni- of London, 229 – 260.
cal Conference and Exhibition, Dallas, Texas, Oct. 22 – 25. Stadulis, J.M., 1995. Development of a completion design to control
Hussain, M.A., Pu, S.L., Underwood, J., 1974. Strain energy release screenouts caused by multiple near-wellbore fractures. SPE
rate for a crack under combined Mode I and Mode II. Fracture 29549, Rocky Mountain Regional/Low-Permeability Reservoirs
analysis. ASTM Special Technical Publication 560, 2 – 28. Symposium, Denver, CO.
Ingraffea, A.R., Manu, C., 1980. Stress-intensity factor computation Tada, H., 1985. The Stress Analysis of Cracks (Handbook), 2nd
in three dimensions with quarter-point elements. International edn. Paris Productions and Del Research, Saint Louis, MS.
Journal for Numerical Methods in Engineering 15, 1427 – 1445. Weijers, L., de Pater, C.J., Owens, K.A., Kogsboll, H.H., 1992.
Ingraffea, A.R., Saouma, V., 1985. Numerical modeling of discrete Geometry of hydraulic fractures induced from horizontal well-
crack propagation in reinforced and plain concrete. In: Sih, bores. SPE 25049, European Petroleum Conference, Cannes,
G.C., Ditommaso, A. (Eds.), Fracture Mechanics of Concrete. France, Nov. 16 – 18.
Martinus Nijhoff Publishers, Kluwer Academic Publishers Weng, X., 1993. Fracture initiation and propagation from deviated
Group, Dordrecht, The Netherlands: 171 – 225. wellbores. SPE 26597, SPE Annual Technical Conference and
Irwin, G.R., 1957. Analysis of stresses and strains near the end of a Exhibition, Houston, Texas, Oct. 3 – 6.
crack traversing a plate. Transactions of the ASME, Journal of Westergaard, H.M., 1939. Bearing pressures and cracks. Journal of
Applied Mechanics 24, 361 – 364. Applied Mechanics 6, A49 – A53.