Journal of Petroleum Science and Engineering (xxxx) xxxxxxxx

Contents lists available at ScienceDirect

Journal of Petroleum Science and Engineering

journal homepage: www.elsevier.com/locate/petrol

Dynamic conductivity of proppant-lled fractures

M. Mehdi Mollanouri Shamsi, Shahram Farhadi Nia, Kristian Jessen

University of Southern California, United States

A R T I C L E I N F O

A BS T RAC T

Keywords:
Hydraulic fractures
Conductivity
Proppant
Dierential stress
Lattice Boltzmann model
Discrete element method

A common challenge in matching the production from hydraulically fractured wells is related to changes in
conductivity of propped fractures due to changes in the stress conguration. In this paper, we investigate the
dynamic nature of fracture conductivity, by evaluating the permeability and width of a propped fracture
segment at variable dierential stress conditions. We combine the Discrete Element Method (DEM) and the
Lattice Boltzmann Model (LBM) to simulate the evolution of a fracture segment lled with proppant and to
evaluate the related fracture conductivity at various stress states. An ensemble of spherical proppant particles is
generated and compacted under a specied conning stress via DEM to generate an initial propped fracture
segment. A representative elementary volume (REV) of the proppant pack is then extracted from the DEM
representation and used for uid ow simulations. LBM is used to calculate the detailed single-phase ow eld,
at the pore-scale, and allows for calculation of permeability. In-situ conditions are then simulated by applying
conning and dierential stress on the REV via DEM. The dierential stress is gradually increased on the REV,
to simulate a reduction in the pore pressure due to production, until the formation of shear band(s) is observed
and fracture failure begins. Permeability of the proppant pack, from LBM simulations, combined with the
fracture aperture, from DEM simulations at dierent dierential stress levels, allow us to determine the
dynamic conductivity of the propped fracture segment. From a range of proppant-size distributions, we
demonstrate that a well-graded proppant pack outperforms a poorly graded pack in maintaining the fracture
conductivity over a broader range of dierential stress conditions. Furthermore, we demonstrate that a wellgraded proppant pack keeps the fracture segment open over a larger range of dierential stress states.

1. Introduction
To achieve economic production rates from low permeability
formations (e.g. gas shales), the subsurface rocks are often fractured.
To induce fractures, large volumes of water are commonly pumped into
the subsurface. The induced fractures extend until the rate of uid loss,
into the formation, exceeds the pumping rate. Once the pumping is
stopped, the pressure can decrease below the fracture closure pressure,
which is the uid pressure needed to initiate the opening of a fracture,
and the fracture may close rapidly and result in a loss of conductivity in
the stimulated volume. To avoid this, a propping agent (sand or
ceramic particles) is added to the injected uid to ensure that the ow
paths remain open after the uid injection stops (Economides and
Nolte, 2000; Reinicke et al., 2010). The proppants create a high
permeability pack that enables continued uid ow over an extended
period of time. During production, uids ow from the rock matrix and
related microfractures into the larger induced fractures, which are lled
with proppant, to arrive at the wellbore.
A major challenge in the design of hydraulic stimulation relates to

the optimization of pumping sequence and additive selection that

allows for a good proppant placement. The eective fracture length (the
length which is lled with proppants), is signicantly smaller than the
length of the induced fracture network (Liu et al., 2007). As the
induced fractures provide the main pathways for uid ow from the
subsurface to the well, it is desired to keep the fractures as conductive
as possible throughout production operations. Several laboratory
studies have shown that porosity and conductivity of induced fractures
decline considerably during production (e.g. Palisch et al., 2007;
Weaver and Rickman, 2010). It has also been demonstrated that
conductivities observed in the eld may be considerably dierent from
the laboratory measurements (Zho et al., 2012).
The use of slickwater (a water-based uid and proppant combination with a low viscosity) in well stimulation has become popular due to
an observed increase in the stimulated area/volume and a reduced risk
of fracture uid damage to the formation (King, 2010). Experimental
studies have been performed to investigate the deposition of proppant
particles in fracture slots (Kern et al., 1959; Barree and Conway, 1995;
Brannon et al., 2006; Sahai et al., 2014). In particular, Alotaibi and

Corresponding author.
E-mail address: jessen@usc.edu (K. Jessen).

http://dx.doi.org/10.1016/j.petrol.2016.12.030
Received 16 October 2016; Received in revised form 9 December 2016; Accepted 20 December 2016
0920-4105/ 2016 Elsevier B.V. All rights reserved.

Please cite this article as: Jessen, K., Journal of Petroleum Science and Engineering (2016), http://dx.doi.org/10.1016/j.petrol.2016.12.030

Journal of Petroleum Science and Engineering (xxxx) xxxxxxxx

M.M.M. Shamsi et al.

Nomenclature

kaf
i
Mi
Ii

N
S

Lu
U

Reservoir rock permeability, D

angular acceleration of a particle, m/s2
proppant pack permeability, D
acceleration of a particle, m/s2
Velocity of a particle, m/s
sum of forces acting on each particle, N
mass of a particle, kg
gravitational acceleration, m/s2
Inter particles cohesion, N/m2
Damping coecient
Permeability from Darcy, D
viscosity, cp
length of the sample, mm
porosity of the sample
Kozeny-Carman constant
Eective stress, N/m2
pore pressure, N/m2
fracture permeability, D
Initial conductivity of the fracture, m2.cm
Fracture width, mm
internal friction angle, degree
Normal stress, N/m2

kf
i
Kp
ui
ui
fi
mi
gi
c
D
k

C
eff
pp
k
ki
w

p
S0
kKC
c

t
Cfp

where k is the fracture permeability. By integrating Eq. (3) and

substituting for eff we nd

Miskimins (2015) conducted an experimental study to investigate the

height of the dunes created by the injected proppant and demonstrated
that dunes in slickwater injection develop at a height exceeding 96% of
the fracture slot height. They also demonstrated that proppants are
able to migrate into both secondary fractures (smaller fractures
emerging from the major induced fractures) and tertiary fractures
(smaller fractures emerging from the secondary fractures).

k = ki exp c pp + c pp ,

i

k = ki exp ppf ppi .

Production from a fractured well causes a pressure decline in the

fracture that, in turn, increases the eective stress on the fracture.
Increased eective stress can result in proppant reorientation, embedment and crushing that reduces the fracture conductivity and aects
the well productivity. The stress dependence of the conductivity of a
propped fracture is controlled primarily by the type, size and concentration of the proppants (Palisch et al., 2007). In order to simulate the
high initial production rates and the rapid decline at later times that
are commonly observed in production from unconventional wells, the
dynamic nature of the fracture conductivity must be represented
accurately. The mechanical properties of a porous material are a
function of the eective stress (Pedrosa, 1986), and can be expressed
by the eective stress law:
(1)

(2)

At reservoir conditions, the closure stress is constant (assuming no

tectonic activities), and the pore pressure controls the eective stress
and consequently the conductivity. A permeability modulus ( ) is often
used to express stress dependent fracture conductivity (Yilmaz et al.,
1991):

1 k
,
k eff

(5)

In commercial simulators, the pressure/stress dependence of

fracture conductivity is often implemented in tabular form (Okouma
et al., 2011; De Souza et al., 2012). This approach, however, assumes
that the conductivity is solely a function of the eective stress and
neglects the eect of other sources of conductivity loss. A study by
Palisch et al. (2007) indicates that estimates of in-situ fracture
conductivity can be a low as 5% of the related laboratory observations.
Separate measurements by Davies and Kuiper (1988) show that
fracturing uid residue may reduce the proppant conductivity by
90% and, in addition, that two-phase ow eects may further reduce
the eective fracture conductivity signicantly (by a factor of about 5).
Furthermore, laboratory-scale measurements by Weaver et al. (2007,
2010) indicate that geochemical interaction between proppants and the
formation water at high stress and temperature conditions accelerates
the proppant dissolution process and reduces permeability, as much as
90% in less than a year.
Rate transient analysis has been used to characterize changes in
hydraulic fracture and matrix permeabilities (De Souza et al., 2012;
Clarkson, 2013). Clarkson (2013) used, in the context of rate-transient
analysis, modied pseudo-pressure and pseudo-time to incorporate
stress-dependent matrix permeability and approximated changes in the
fracture conductivity via a time-dependent skin eect. These authors
show that the traditional linear and square-root-time plots become
non-linear, due to loss of conductivity during production, unless
modied pseudo-variables are used. By ignoring this eect, one may
interpret the start of non-linearity as an indication of pseudo-steady
state and underestimate the reserves (stimulated volume).

where c is the closure stress, is the Biot's poroelastic constant and pp

is the pore pressure. Biot's constant is unity for unconsolidated
materials, such as a proppant pack residing inside a fracture segment.
Therefore, Eq. (1) simplies to:

eff = c pp .

(4)

where subscripts i and f refer to initial and owing conditions,

respectively. If the closure stress is assumed to be constant, Eq. (4)
further reduces to:

1.1. Dynamic conductivity of fractures

eff = c pp ,

Altered fracture face permeability, D

angular velocity of a particle, m/s
total moment acting on a particle, N.m
moment of inertia for a particle, kg.m2
viscous energy adsorption
normal contact stiness, N/m
Tangential contact stiness, N/m
Inter particle friction coecient
Density of particles, kg/m3
lattice unit
sum of the x-direction components of velocity along a
constant x-plane, m/s
pressure drop along the length, N/m2
surface area per unit volume of the solid phase
Kozeny-Carman Permeability, D
closure stress, N/m2
Biot's poroelastic constant
permeability modulus
Time, sec
Fracture proppant pack conductivity, m2 cm
Proppant segment deformation
Shear strength, N/m2
Cohesion, N/m2

(3)
2

Journal of Petroleum Science and Engineering (xxxx) xxxxxxxx

M.M.M. Shamsi et al.

a given proppant segment is then subjected to dierential stress

conditions. The dierential stress is gradually increased until shear
bands form and the fracture segment starts closing. LBM is used to
calculate a detailed ow eld at the pore-scale and the permeability is
calculated using the velocity and pressure distribution. The permeability of the propped segment is calculated at a range of dierential
stress states. Finally, the conductivity is evaluated from fracture width
and permeability as a function of the dierential stress. To validate the
simulation work ow, we compare our results with previous theoretical
and experimental studies.

Several mechanisms contribute to dynamic changes in the performance of a propped fracture including but not limited to physical,
biological, chemical, and thermal interactions between formation,
proppant and reservoir uids, that can signicantly inuence the
performance of a fractured well (Reinicke et al., 2010). Numerous
eorts including laboratory, eld and theoretical studies have aimed to
delineate fracture damage mechanisms (Bishop, 1997; Lynn et al.,
1998; Civan, 2000; Fredd et al., 2000; Moghadasi et al., 2002; Behr
et al., 2006; Nasr-El-Din, 2003; Wen et al., 2006; Reinicke et al., 2010).
Mechanisms that contribute to the loss of conductivity include
proppant embedment, reorganization and crushing, fracturing uid
damage to the formation, multi-phase ow eects, nes migration and
proppant diagenesis (Weaver et al., 2007; Weaver and Rickman, 2010).
Despite these good eorts, there is, to our best knowledge, currently no
rigorous conductivity model or correlation available for production
modeling. As a consequence, reservoir engineers often have to resort to
simple approximations e.g. that fracture conductivity declines exponentially with time

k = ki exp [t ].

2. Modeling approach
2.1. Conceptual model
After completion of hydraulic fracturing, including placement and
settlement of proppants, the stress eld on the proppants begins to
change. We assume that a proppant pack in a fracture segment exists at
uniform conning stress/pressure prior to the onset of production.
Fig. 1a illustrates a segment of a stimulated well. As shown in the
Fig. 1b, and discussed previously, we assume that the fracture is lled
with proppants (dune exceeds 95% of fracture height). Prior to
production, the proppant pack is exposed to a conning stress equals
to as shown in Fig. 1c. As production starts, the pore pressure
decreases while the stress from other sides does not change signicantly. The reduction of the pore pressure in the fracture increases the
closure load on the proppant pack (Daneshy, 2005): The stress load
from the fracture faces increase as shown by (dierential stress) in
Fig. 1d. During production, increases gradually, and causes shear
stress on (and related shear deformation of) the proppant pack. This
shear stress is the main reason for particle to particle sliding and
dislodging. Experimental studies by Reinicke et al. (2010) showed that
at 5 MPa of dierential stress, nes (smaller particles) are generated.
By increasing the dierential stress, the porosity of proppant pack
decreases and a shear band may form as shown as a red band in
Fig. 1d. Fig. 1e illustrates the resultant shear strain of a proppant-lled
fracture segment. A further increase in the dierential stress, beyond

(6)

Here, is an adjustable parameter and ki is the initial conductivity

of the fracture. The parameters of Eq. (6) can be estimated via history
matching of production data. As these parameters are primarily
dependent on the type, size and concentration of the proppant, one
may attempt to nd average values from history matching of data from
wells that are stimulated similarly and apply these values to new wells.
In summary, appropriate modeling of in-situ fracture conductivity is
essential for accurate simulation of production. However, the dependence of fracture conductivity on a range of parameters complicates the
in-situ conductivity modeling.
In the following sections, we present an integrated 3D DEM-LBM
simulation workow to study the dynamics of propped fracture
segments. To investigate the impact of proppant selection, three
dierent grain-size distributions, including well-graded, uniformlygraded and poorly-graded proppant packs are investigated. Initial
propped segments are generated and compacted via DEM under
uniform conning stress. To simulate in-situ conditions of a fracture,

Fig. 1. Mechanism of fracture closing and formation of shear bands: (a) a stimulated well, (b) a fracture lled with proppant prior to production, (c) in-situ condition of proppant pack
before production, (d) a proppant pack under dierential stress and (e) a fracture segment lled with proppants during production.

Journal of Petroleum Science and Engineering (xxxx) xxxxxxxx

M.M.M. Shamsi et al.

granular materials. On the uid mechanics side, the Lattice Boltzmann

Method (LBM) has been demonstrated to be ecient in modeling uid
ow in porous materials: LBM utilizes a set of kinetic equations capable
of recovering Navier-Stokes equations for slightly compressible uids
at low Mach and Knudsen numbers (Chen et al., 1992).
During production from stimulated/fractured well the proppants
will compact and rearrange resulting in changes to the conductivity of
the propped fracture. The purpose of this work is to simulate these
change using DEM, and investigate their eect on the fracture
conductivity using LBM (e.g. Chen et al., 1992; Shan and Chen,
1993; Borujeni et al., 2013). Fig. 5 provides a schematic of a proppant
pack in a fracture segment as studied in this work. We assume that the
fracture walls are parallel (it's a common assumption for all experimental studies as well as numerical simulation).
For modeling purposes, it is assumed that the proppants are rigid
spherical particles residing in a cubic space and that the particles are
not crushed or deformed. The proppant particles are selected from a
specied diameter-size distribution and initialized as randomly distributed in space with no initial contact prior to the compaction. The
DEM is then used to compact a given proppant pack subject to a
specied stress state (uniform or dierential). Fluid ow simulation
using the LBM is then performed, at dierent stress states, to calculate
the permeability of a given proppant-pack conguration.
To model an assembly of proppant particles, 1000 spherical
particles are randomly distributed in a loose cubic packing without
any initial contacts (see Fig. 7a below). Particle diameters are selected
from a distribution function to represent the gradation of the proppant
(see details below). By applying a uniform conning pressure (e.g.
25 MPa), the particles are displaced and allowed to contact. The cubic
space is then divided into a set of sub-boxes to facilitate grain contact
detection. At this stage, a box or a group of boxes is assigned for every
particle (Shamsi and Mirghasemi, 2012). By applying boundary loads,
each particle (i) is then allowed to move based on Newton's second law
of motion:

the formation of shear bands, causes the shear band to extend and
initiate closing/failure of the propped segment.
The conductivity of a hydraulic fracture depends on two factors: the
proppant (size, concentration, and type) and the dimensions of the
fracture (Daneshy, 2007). The propped fracture conductivity (Cfp) is
calculated from the product of the fracture width (w) and the proppant
pack permeability (kfp)

Cfp = wk fp,

(7)

where, w is the fracture width and kfp is the permeability of the

proppant pack. In simulation of production from fractured wells, the
permeability of three zones must be represented accurately (Reinicke
et al., 2010). These include the reservoir rock permeability (kr), the
permeability of the altered fracture face (kaf) at the interface between
fracture and proppants, and the proppant pack permeability (kfp). In
this work, we focus exclusively on the permeability and conductivity of
the proppant. We assume that the fracture segment is lled with
proppant and we do not consider the eects of embedment, proppant
crushing, and cementation. Accordingly, we consider the permeability
of the fracture to be equal to the permeability of the proppant pack. At
each step of the compaction process (DEM simulation, see below), we
can then calculate the fracture width (w) from the axial strain ()

w = (1 )wi ,

(8)

where wi is the initial width of the proppant pack before dierential

stress is applied.
2.2. Proppant selection
Several factors must be considered when selecting proppants
including proppant size, grain-size distribution, weight, shape,
strength, resistance to cyclic stress changes and nes generation. In
this work, we are focusing on the proppant-size distribution and its
eect on dynamic conductivity of a propped fracture segment.
The conductivity of a propped fracture segment is given by the
product of the proppant permeability and the fracture aperture. In
terms of permeability, a uniform proppant pack with maximum
porosity and permeability is preferred. However, a uniform proppant
pack is easier to compact, compared to a well-graded proppant pack,
and can promote reductions in fracture aperture. On the other hand, a
well-graded proppant pack has a lower porosity/permeability but is
more dicult to compact and will keep a fracture segment open over a
larger range of dierential stress conditions. To study of the interplay
between permeability and fracture aperture, we consider 3 dierent
grains-size distributions (see Fig. 2) including: Well graded (denser
packing), poorly graded (less dense packing), and uniformly graded (all
grains have the same size least dense packing).
Propped fractures with an aperture greater than 0.25 (6.35 mm)
are very dicult to achieve due to limitations in uid pressure,
especially in the deeper subsurface where the default stress on the
rocks increases: In addition, there is not a single grain-size distribution
that will perform best for all well conditions (Montgomery and
Steanson, 1985). To provide a realistic representation of proppants
in a given segment, ve dierent particle sizes with diameters in the
range of 0.11.5 mm were selected for each ensemble (proppant
grade). The three dierent ensamples of proppants obtained from the
3 dierent size distributions are illustrated in Figs. 3 and 4.

ui + ui =

Fi
+ gi ,
mi

i + i =

Mi
.
Ii

(9)

(10)

Eqs. (9+10) describe the linear movement and rotation of particles,

respectively. ui and ui denote acceleration and velocity of a particle, Fi is
the sum of forces acting on each particle from contact (normal force
and shear force), mi is the mass of a particle and gi is the gravitational
acceleration. In Eq. (10), i and i represent angular acceleration and
velocity, respectively. Mi is total moment acting on a particle and Ii is
the moment of inertia. in both equations represents energy dissipation.
A centered nite dierence method is utilized to integrate the
motion equations (Cundall and Strack, 1979, 3DEC, 1988) by approx-

2.3. Numerical approach

To arrive at a better understanding of the proppant conductivity in
hydraulically fractured wells, under dynamic stress conditions, a
coupled uid-ow and proppant-pack-mechanics approach must be
utilized. To simulate the behavior of a proppant pack, the Discrete
Element Method (DEM), as discussed by Cundall and Strack (1979), is
very eective in modeling quasi-static and dynamic deformation of

Fig. 2. Grain-size distribution for proppants: Well graded, uniformly graded and poorly
graded.

Journal of Petroleum Science and Engineering (xxxx) xxxxxxxx

M.M.M. Shamsi et al.

Fig. 3. 3D proppant packs: (a) well graded, (b) poorly graded, and (c) uniformly graded.

imating the translational and rotational velocities at time t as

t

t +
1 t t
xi = xi 2 + xi 2 ,

i =

1 t 2t
i

i = i

xi (t +

t
t +
i 2 ,

i =

1
t

(13)

t
t
i 2 .

(14)

By inserting Eqs. (11)(14) in Eqs. (9) and (10) and solving for
t
velocities at time + 2 , translational and rotational velocities can be
evaluated from:
t
t +
2

xi

t
t +
2

F (t )
t
= D1xi 2 + i + git D2 ,

mi

t 1
2) .

In Eqs. (15) and (16), D1=1

and D2 = (1 +
The
changes in translation and rotation for each sphere is nally evaluated
from

xi =

t
xi (t + 2 )t ,

(19)

mmin
,
kmax

(20)

where kmax is the maximum of the normal or shear contact stiness,

mmin is the mass of the smallest particle and frac is a coecient set to
0.05 in these calculations to ensure numerical stability. Ng (1989)
investigated the eect of dierent parameters such as time step, sand
density and damping coecient on the behavior of semi-static granular
materials. Ng (1989) introduced the Unbalanced Force Index (UFI) to
control stability (and maintain static conditions) throughout the DEM
calculations

(16)
t
2

= xi (t )+xi .

t = frac

(15)

M (t )
t
= D1i 2 + i t D2.

mi

t )
2

To calculate contact forces, normal stiness (N) and tangential

stiness (S) are utilized in the force-displacement equations (Eqs. (9)
and (10)). To consider the roughness of a grain surface, a friction
coecient () is used and the damping coecient (D) is dened to
dissipate kinetic energy and ensure that a static equilibrium condition
is obtained for a given particle assembly (Abedi and Mirghasemi,
2011). The particles are traced forward in time via time steps that are
selected suciently small to ensure a stable numerical solution
(Shamsi and Mirghasemi, 2012). The procedure is repeated with the
updated particle positions in subsequent time steps via the forcedisplacement relations. To ensure that the scheme is numerically
stable, the time-step is evaluated from

(12)

t

t
1 t + 2t
xi
xi 2 ,

t + t
2
i

(18)

and the new position of particle i is calculated from

(11)

while the acceleration terms are approximated by

xi =

(t + t )
2 t ,

AverageInertiaForces
UFI=
=
AverageContactForces

(17)

InertiaForcesofSpheres
TotalNumberofSpheres
ContactForces
TotalNumberofContacts

Fig. 4. 2D schematics of proppants in a fracture segment: (a) well graded, (b) poorly graded and (c) uniformly graded.

,
(21)

Journal of Petroleum Science and Engineering (xxxx) xxxxxxxx

M.M.M. Shamsi et al.

Fig. 5. (a) A fracture segment lled with proppant, (b) Magnied view of a proppant-pack in fracture segment, (c) Numerical representation (approximation) of the proppant-pack.

and demonstrated that UFI values less than 0.01 result in stable
calculations: We followed this guideline in our calculations.
Fig. 6 shows an example of compaction modeling using our DEM
simulation setup with input parameters reported in Table 1. The
parameters were selected to mimic the behavior of sand particles and
to guarantee the numerical stability of the calculations (Shamsi and
Mirghasemi, 2012). Fig. 6a shows the initial arrangement of the
proppant (no initial contact between the particles) prior to compaction.
The particles are then subjected to a hydrostatic strain rate equal to
1.0107 per time step. This value was selected to reach UFI < 0.01
(see discussion above) and maintain a stable numerical scheme. Fig. 6b
shows the proppant pack at 25 MPa of conning stress.
In Table 1, the specic weight of the particles is higher than for
normal sand. This choice is intentional to decrease the CPU time of our
simulations. This use of larger values for the density is known as
density scaling as discussed by Thornton (2000) and Cui et al. (2007),

Table 1
Particle parameters for DEM simulation (Shamsi and Mirghasemi, 2012).
Parameters

values

Normal contact stiness (N)

Shear contact stiness (S)
Inter particle friction coecient ( )
Inter particles cohesion (c)
Density of particles ( )
Damping coecient (D)

21010 (N/m)
21010 (N/m)
0.5
0.0
2104 (Kg/m3)
5105

and does not aect the nal static results.

The conductivity of a proppant pack, at a given stress state, was
calculated using the Lattice Boltzmann Method (LBM). To calculate the
permeability in a given direction (e.g. x-direction), the inlet and outlet
faces are left open while all the other faces are closed with no-ow

Fig. 6. DEM simulation (a) initially particle ensemble; (b) particles compacted at 25 MPa of conning pressure.

Journal of Petroleum Science and Engineering (xxxx) xxxxxxxx

M.M.M. Shamsi et al.

permeability was measured at constant conning pressures of 5, 10,

15, 20 and 25 MPa. Fig. 12 compares the results from the experimental
work (with 0% coal nes) with our numerical calculations.
As observed from Fig. 12, the permeability obtained from numerical
calculations, after the compaction stage (no dierential stress), are
larger but in reasonable agreement with the experimental observations:
The numerical calculations reported in Fig. 12 do not consider the
actual shape of proppants that are assumed to be spherical. Therefore,
permeability from the experimental test is lower due to better
compaction of proppants and lower porosity.
In a second validation, we consider the Kozeny-Carman correlation.
The Kozeny-Carman correlation, originally developed for proppant
packs, relates the permeability of a porous medium to the specic
surface area and the porosity (Wyllie and Gregory, 1955)

boundaries to approximate the experimental conguration used for

permeability measurements. Fig. 7 shows the LBM simulation setup:
All four sides orthogonal to the ow direction (here the x-direction)
have no-ow boundaries, dened via solid points.
A periodic boundary condition with a body force is then applied in
the ow direction. Permeability has the dimension of length squared,
and in the LBM approach the length dimension is denoted Lu (lattice
unit) which is, often, the resolution of the input image e.g:

k (D ) = K (Lu 2 ) (resolutioninmicron )2 .

(22)

The LBM simulation results in a detailed ow eld (pressure and

velocity) that is then used to calculate the permeability (k) from Darcy's
equation

k=

U x
.
,
p

(23)

kKC =

where U is the sum of the x-direction components of velocity along the

yz-plane, which due to mass conservation for incompressible systems is
the same for all planes perpendicular to the x-axis. is the dynamic
viscosity of the uid (specied via the relaxation time), x is the sample
length, and p is the pressure drop along the length.
To ensure that we use an appropriate REV size, we calculated the
permeability of (DEM) sub-cubes with side lengths in the range of 3
7 mm: The permeability was calculated at two dierent conning stress
states including 1 MPa and 5 MPa. Fig. 8 demonstrates that the
calculated permeability remains almost constant for subvolumes larger
than 5.5 mm5.5 mm5.5 mm for both stress states (1 MPa and
5 MPa). Accordingly, we assume that a REV size of 555 mm3 is
suciently accurate for this study.
To decrease the computational cost, while maintaining a REV, we
selected 505050 voxel (see e.g. Lenoir et al., 2010) sub-samples for
our LBM simulation as demonstrated in Fig. 9.
Samples of this size result in the same permeability, independent of
the location in the larger (DEM) domain, and was found also to equal
the permeability as calculated for the whole DEM domain. Table 2
reports permeability in the dierent areas of the larger (DEM) domain:
We observe a good agreement across subsamples. Based on this
observation, we assume that one sub-element is sucient to represent
the overall DEM domain in the context of permeability/conductivity
calculations during the compaction process.
Fig. 10 provides a owchart that summarizes the combined DEM/
LBM approach for calculation of conductivity of a given proppant pack.

3
CS02(1)2

(24)

is the porosity of the sample, S0 is the surface area per unit volume of
the solid phase, and C is the Kozeny-Carman constant. For equal-sized
spheres, the specic surface area can be expressed through the grain
diameter, So=3/r, where r is the radius of the spheres (Jin et al., 2012).
The constant C is approximately equal to 5 for ow through unconsolidated sand (Wyllie and Gregory, 1955). For the proppants in this
study, and S0 were calculated from the DEM results at dierent
conning pressures, while C was set equal to 5. Fig. 13 compares the
permeability of proppant packs at dierent conning pressures, as
calculated from DEM/LBM and predicted using the Kozeny-Carman
correlation. We observe a good agreement between numerical calculations and the Kozeny-Carman correlation over the full range of
porosities.
3.2. Validation Compaction under dierential stress
To keep a fracture open, at increasing dierential stress levels, we
need to inject proppant with high shear strength that also provides for
a high porosity (permeability) of the proppant pack. To this end, we
need to consider the internal friction angle that is a mechanical
property of proppants related to shear forces. () and cohesion c
(MPa) are estimated by the MohrCoulomb shear strength criterion
(Terzaghi, 1936):

= c + tan.

(25)

The shear strength of a proppant is greatly inuenced by the

particle-size distribution (Zelasko et al., 1975). To calculate for a
given grain-size distribution, we have several options including a direct
shear test or a triaxial test. In this study, we simulated the triaxial test.
In a triaxial test, the mobilized friction angle of a cohesion-free
proppant can be determined by Eq. (26) as a function of the major
principal stress:

3. Results and discussion

3.1. Validation Compaction under uniform stress
Proppant packs were generated from the 3 distributions (see. Fig. 3)
and compressed via DEM calculations. Fig. 11 reports the porosity of
the 3 packs over a range of conning pressures (uniform stress). As
expected, by increasing the conning pressure the porosity decreases.
We also observe (as discussed above) that the porosity of the wellgraded sample is lower than that of the uniformly- and poorly-graded
proppant packs.
To validate our numerical calculations, we compare our results with
experimental work and/or analytical results. We start by comparing
our results with the experimental observations from Zou et al. (2013),
who conducted an experimental study to simulate coal-nes migration
through a proppant-pack. The experimental apparatus used in their
study is based on the API-standard fracture-conductivity instrument.
In their work, the conductivity cell was paved with high-strength
proppant with particle sizes ranging from 0.85 to 2 mm. They
considered dierent particle mixtures (sand+coal-nes) including 0
10% (by weight) of coal nes. Two steel plates were used to simulate
fracture surfaces. The conning pressure was then increased and

sin =

1 3
,
1 + 3

(26)

The major principal stress was determined from the average stress
tensor within an assembly (Shamsi and Mirghasemi, 2012) and the
internal friction angle was evaluated from Mohr-Coulomb theory. The

Fig. 7. LBM setup for simulation of ow in the x-direction.

Journal of Petroleum Science and Engineering (xxxx) xxxxxxxx

M.M.M. Shamsi et al.

Table 2
LBM permeability in different parts of main sample (DEM simulation sample).
Sub-volume Position
(x,y,z)

(40,40,40)

(70,70,70)

(100,100,100)

(130,130,130)

Permeability (Darcy)

2.24102

2.12102

2.18102

2.21102

3.3. Simulation of dynamic proppant conductivity

In our rst calculation example, we consider a poorly-graded
proppant pack in a fracture segment. The pack was initialized at a
uniform stress of 20 MPa and the dierential stress was then gradually
increased. A sub-volume of the proppant pack (REV) was then
extracted for ow simulations (permeability) at discrete values of the
axial strain. Fig. 15 reports the DEM-LBM result for a poorly-graded
proppant pack in terms of dierential stress, axial strain and conductivity.
We observe from Fig. 15 that the propped segment can resist a
dierential stress of ~24.2 MPa. Beyond this magnitude of the dierential stress, a shear band is formed in the proppant pack. The
maximum dierential stress that a proppant pack can endure before
shear band(s) start to form depends on several factors including the
initial conning stress and, more signicantly, the grain-size distribution and the related mechanical properties. We also observe that an
initial increase in dierential stress causes the conductivity to decline.
The main reason for this reduction is the compaction of proppant:
reduction of both the permeability and the width of the propped
segment. As the dierential stress is further increased, the proppant
segment reaches a maximum compaction and at this point the
permeability levels o. Usually close to the maximum dierential
stress level, a shear band starts to form, causing the porosity to change.
At this point, the porosity/permeability increase due to sliding,
relocation and movement of compacted particles inside the shear
band. However, the formation of shear band causes a reduction in
the width of the propped segment: The increase in permeability is more
signicant than the reduction of the segment width, and the conductivity of the propped segment increases. By further increasing the
dierential stress, the proppant pack (and shear band) starts to
collapse while the proppants behave like a uid and a fracture failure
will occur (occurs at ~35% in our calculations).
In a second calculation example, we consider a fracture segment
lled with uniformly-graded proppants. Fig. 16 reports the change in
conductivity as a function of axial strain and dierential stress for the
uniformly-graded proppant pack.
From Fig. 16, we observe a higher initial conductivity for the
uniformly-graded proppant segment, as compared to Fig. 15 (poorlygraded pack), due to the higher permeability of the uniformly-graded
proppant pack. However, as the dierential stress increases, a sudden

Fig. 8. Impact of subvolume size on calculated permeability for well graded proppant at
a conning stress of 1 MPa and 5 MPa.

results from the simulated triaxial tests are reported in Table 3.

Reinicke et al., 2010 conducted an experimental study to evaluate
the conductivity of two dierent proppant packs, including high
strength proppants (HSP), composed of 20/40 mesh sintered bauxite,
and intermediate strength proppants (ISP), made from 20/40 mesh
fused ceramics (proppant particle diameter: 0.40.8 mm). They used a
constant conning stress of 10 MPa and investigated varying dierential stress states. In their study, the internal friction angel of
proppants is 45.7 (higher than for our current numerical study).
Given the larger internal friction angel of sand used in the experiments,
the conductivity could be measured for 6, 21, 36 and 51 MPa of
dierential stress. In our DEM simulation, because we consider
proppants to be spherical, the maximum internal friction angle is
33.6 for well graded proppant packs. As a result, the proppant pack in
our calculations will fail at lower dierential stress states (for 10 MPa
of conning pressure) and, accordingly, our current simulations do not
cover 20 MPa and 35 MPa of dierential stress at a conning stress of
10 MPa. Fig. 14 reports the comparison of the experimental results of
Reinicke (2010) to the numerical results from this work (prior to the
formation of shear bands).
From Fig. 14, we observe that the conductivity evaluated from
DEM/LBM calculations with well-graded and uniformly-graded proppants is slightly higher than what is reported from the experimental
study by Reinicke (2010). The agreement is very reasonable, however,
considering that we assume that particles are spherical and rigid (no
breakage and nes generation). Both breakage and nes formation
cause additional conductivity reduction. Overall, our numerical results
are in good agreement with the experimental study, in particular for
the trend of the conductivity reduction with increasing dierential
stress.

(b)

(a)

Fig. 9. (a) Compacted particles after DEM simulation, (b) sub-volume for LBM calculations.

Journal of Petroleum Science and Engineering (xxxx) xxxxxxxx

M.M.M. Shamsi et al.

Fig. 10. Flow chart of numerical approach to calculate dynamic conductivity of a fracture segment.

Fig. 13. Comparison of permeability from DEM/LBM simulation and Kozeny-Carman's

equation as a function of conning pressure.

Fig. 11. Porosity versus conning pressure for dierent proppant packs.

Table 3
Internal friction angle (), for dierent grain size distributions.
Proppant pack

Well graded

Poorly Graded

Uniformly graded

()

33.6

27.3

20.1

Fig. 12. Permeability from LBM-DEM calculations (current study) and data (Zou et al.,
2013).

drop in conductivity is observed. Conductivity decreases before reaching the highest dierential stress. The main reason for the sudden
reduction, relative to for the poorly-graded proppant, is the lack of
interlocking between grains to resist the shear forces. Uniformly graded
proppants slide more easily and the fracture closing starts at a lower
value of the dierential stress. We also observe from Fig. 16 that the
conductivity remains constant over a broader range of axial strain (10

Fig. 14. Comparison of conductivity versus dierential stress at 10 MPa of conning

pressure (a) experiments by Reinicke (2010) on HSP and ISP proppants pack; (b)
Current DEM/LBM calculations.

Journal of Petroleum Science and Engineering (xxxx) xxxxxxxx

M.M.M. Shamsi et al.

mechanical properties (i.e. internal friction angle of proppants) inuence the conductivity of a fracture segment. Fig. 17 also demonstrates
that the conductivity of the uniformly-graded proppant-pack is constant till failure occurs, while for the well-graded proppant pack, the
conductivity increases after the formation of the shear band (~30%
higher than for the uniformly-graded proppant pack).
For uniformly-graded proppant pack, ecient interlocking of
particles during shear-band formation does not occur and results in
sliding of particles. As a result, there is not a large dierence between
permeability of the proppant pack inside and outside the shear band.
However, the calculated permeability, inside the shear band, for wellgraded and poorly-graded proppant packs is about 22.5 times larger
than outside the shear band. This result is also in agreement with the
observations of Sun et al. (2013). Here, we consider a sub-volume of
the proppant pack including the shear band. This sub-volume is not a
representative of whole fracture width, however, because the fracture
width is very small and the process of shear band formation is very
slow, we assume that the ow has sucient time to align with the shear
band (even during its formation). Accordingly, the fracture conductivities calculated from the presented work ow should be representative
of relevant propped fracture segments.

Fig. 15. Conductivity (and dierential stress) of a poorly-graded proppant pack at

=20 MPa as a function of axial strain.

4. Summary and conclusions

In the previous sections, we have presented an ecient workow
based on DEM and LBM to promote the understanding and description
dynamic fracture conductivity in the context of unconventional oil and
gas production. The proposed workow provides an eective approach
for analyzing production dynamics related to variable proppant conductivity, and provides a path for optimizing the proppant size
distribution based on expected variations in the stress eld.
Furthermore, the workow provides an approach for generating
conductivity input tables for reservoir simulation and history matching
applications.
To investigate the impact of proppant selection, three dierent
grain-size distributions, including well-graded, uniformly-graded and
poorly-graded packs were generated and compacted via DEM at
variable conning stress. By considering changes in the eective stress
on a proppant pack, we have calculated the fracture width and
permeability to arrive at a relationship between the stress state and
the fracture conductivity. Dierential stress, induced by the walls of a
fracture, causes the formations of a shear band in the proppant pack.
The shear band will gradually extend and rupture in responds to
additional increases in the dierential stress. Following this event,
facture failure occurs and the fracture segment can close. Based on our
example calculations and discussion, we arrive at the following
conclusions:

Fig. 16. Conductivity (and dierential stress) of a uniformly-graded proppant pack at

=20 MPa as a function of axial strain and dierential stress.

30%) until fracture failure occurs. This observation suggests that

uniformly-graded proppants pack cannot resist high dierential stress
due to sliding between particles. Moreover, it demonstrates that the
pore space of the uniformly-graded particle assembly remains nearly
isotropic even after undergoing signicant shear as discussed by Sun
et al. (2013).
The conductivity of the uniformly-graded proppant pack is higher
than for the poorly-graded proppant pack at low dierential stress (i.e.
early time of production). As the dierential stress increases, permeability of the uniformly-graded proppants decreases due to sliding of
particles. On the other hand, due to the interlocking of grains and also
dilation inside the shear band in the poorly-graded proppant pack, the
conductivity increases signicantly. Furthermore, fracture closing
occurs sooner for the uniformly-graded pack than for the poorlygraded proppant pack. The main reason for this is the higher shear
strength of the well-graded proppant pack. Table 3 conrms the role of
during increasing dierential stress (production): A higher internal
friction for proppant helps to keep the fracture open at increasing
dierential stress conditions (production time) and provides for a
higher conductivity (increased production).
Fig. 17 reports the eect of grain-size distribution on the dynamic
conductivity for the 3 dierent proppant packs considered in this study
at a conning stress of 20 MPa. For moderate axial strain ( < 10%),
the conductivity of the uniform proppant pack is signicantly higher
than for the poorly-graded and the well-graded proppant packs (e.g.
70% higher than the well-graded proppant pack). At increasing
dierential stress (~10% axial strain), the conductivity of all proppant
packs becomes similar. As the shear band is formed at 1520% of axial
strain, the mechanical properties of the proppant packs signicantly
aect the conductivity. As a result, the combination of porosity and

1. Our calculation results demonstrate that the combination of poros-

Fig. 17. Inuence of grain-size-distribution on dynamic conductivity for dierent

proppant packs at = 20MPa .

10

Journal of Petroleum Science and Engineering (xxxx) xxxxxxxx

M.M.M. Shamsi et al.

2.

3.

4.

5.

Jin, G., Patzek, T.W., Silin, D.B., 2012. Modeling the impact of rock formation history on
the evolution of absolute permeability. J. Pet. Sci. Eng. 100, 153161.
Kern, L.R., Perkins, T.K., Wyant, R.E., 1959. The Mechanics of Sand Movement in
Fracturing, paper SPE-1108-G presented at the Annual Fall Meeting of Society of
Petroleum Engineers, Houston, TX, 58 October. Doi: http://dx.doi.org/10.2118/
1108-G.
King, G.E., 2010. Thirty years of gas shale fracturing: what have we learned, Paper SPE133456-MS, presented at the SPE Annual Technical Conference and Exhibition,
Florence, Italy, 1922 September. SPE-133456-MS. Doi: http://dx.doi.org/10.
2118/133456-MS.
Lenoir, N., Andrade, J.E., Sun, W.C., Rudnicki, J.W., 2010. In: Aishibli, K.A., Reed, A.H.
(Eds.), In situ permeability measurements inside compaction bands using X-ray CT
and lattice Boltzmann calculations, in Advances in Computed Tomography for
Geomaterials: GeoX2010. John Wiley, Hoboken, N. J, 279286.
Liu, Y., Gadde, P.B., Sharma, M.M., 2007. Proppant placement using reverse-Hybris
Fracs. SPE Prod. Oper., 348356.
Lynn, J.D., Hisham, A., Nasr-El-Din, H.A., 1998. Evaluation of formation damage due to
frac stimulation of a Saudi Arabian clastic reservoir. J. Pet. Sci. Eng. 21, 179201.
Moghadasi, J., Jamialahmadi, M., Muller-Steinhagen, H., Sharif, A., Izadpanah, M.R.,
2002. Formation damage in iranian oil elds. SPE73781. In: Proceedings of the SPE
International Symposium and Exhibition on Formation Damage Control, Louisiana,
USA
Mollanouri Shamsi, M.M., Mirghasemi, A.A., 2012. Numerical simulation of 3D semireal-shaped granular particle assembly. Powder Technol. 221, 431446.
Montgomery, C.T., Steanson, R.E., 1985. Proppant selection: the Key to Successful
fracture stimulation (Revised). J. Pet. Tech. 37 (12), 21632172. http://dx.doi.org/
10.2118/12616-PA.
Nasr-El-Din, H.A., 2003. New mechanisms of formation damage: lab studies and case
histories, Paper SPE-82253-MS, presented at SPE European Formation Damage
Conference, 1314 May, The Hague, Netherlands. Doi: http://dx.doi.org/10.2118/
82253-MS.
Ng, T.T., 1989. Numerical simulation of granular soil under monotonic and cyclic
loading: a particulate mechanics approach (Ph.D. thesis). submitted to Rensselaer
Polytechnic Institute, Troy, New York.
Okouma, V., Guillot, F., Sarfare, M., San, V., Ilk D., Blasingame, T.A., 2011. Estimated
ultimate recovery (EUR) as a function of production practices in the Haynesville
Shale, Paper SPE 147623 presented at the SPE Annual Technical Conference and
Exhibition, Denver, Colorado, 30 October2 November. http://dx.doi.org/10.2118/
147623-MS
Palisch, T., Duenckel, R., Bazan, L., Heidt, H.J., Turk, G., 2007. Determining realistic
fracture conductivity and understanding its impact on well performance- Theory and
Field Examples, SPE 106301, Hydraulic FracturingTechnology Conference, College
Station, Texas, USA, 29-31January. Doi: http://dx.doi.org/10.2118/106301-MS.
Pedrosa, O.A., 1986. Pressure transient response in stress-sensitive formations, Paper
SPE 15115 presented at the SPE California Regional Meeting, Oakland, California,
USA, April 24.
Reinicke, A., 2010. Mechanical and Hydraulic Aspects of RockProppant Systems:
laboratory Experiments and Modelling Approaches(Doctoral Thesis). University at
Potsdam, Germany.
Reinicke, A., Rybacki, E., Stanchits, S., Huenges, E., Dresen, G., 2010. Hydraulic
fracturing stimulation techniques and formation damage mechanisms-implications
from laboratory testing of tight sandstone-Proppant systems. Chem. der ErdeGeochem. 70, 107117.
Sahai, R., Miskimins, J.L., Olson, K.E., 2014. Laboratory results of proppant transport in
complex fracture systems, presented at the SPE Hydraulic Fracturing Technology
Conference, Woodlands, Texas, SPE-168579-MS. Doi: http://dx.doi.org/10.2118/
168579-MS.
Shan, X., Chen, H., 1993. Lattice Boltzmann model for simulating ows with multiple
phases and components. Phys. Rev. E 47 (3), 18151819.
Sun, W., Kuhn, M.R., Rudnicki, J.W., 2013. A Multiscale DEM-LBM analysis on
permeability evolutions Inside a dilatant shear band. J. Acta Geotech. 8 (5), 465480
. http://dx.doi.org/10.1007/s11440-013-0210-2.
Terzaghi, K., 1936. The Shear Resistance of Saturated Soils and the Angles between the
Planes of Shear. Proceeding of 1st International Conference Soil Mechanics and
Foundation Engineering. Boston, Vol. 1, pp. 5456.
Thornton, C., 2000. Numerical simulations of deviatoric shear deformation of granular
media. Geotechnique 50 (1), 4353.
Weaver, J., Parker, M., Batenburg, D.V., Nguyen, P., 2007. Fracture-Related Diagenesis
May Impact Conductivity, SPE Journal, 272281. In: Presented at International
Symposium and Exhibition on Formation Damage Control, Lafayette, L.A., USA.
Weaver, J.D., Rickman, R.D., 2010. Productivity Impact from Geochemical Degradation
of Hydraulic Fractures. SPE130641-MS. Presented at the SPE Deep Gas Conference
and Exhibition, Manama, Bahrain.
Wen, Q., Zhang, S., Wang, L., Liu, Y., Li, X., 2006. The Eect of Proppant Embedment
upon the Long-term conductivity of fractures. J. Pet. Sci. Eng. 55, 221227.
Wyllie, M.R.J., Gregory, A.R., 1955. Fluid ow through Unconsolidated porous
aggregates: Eect of porosity and particle shape on Kozeny-Carman constants. Ind.
Eng. Chem. 47 (7), 13791388.
Yilmaz, O., Nur, A., Nolen-Hoeksema, R., 1991. Pore pressure proles in fractured and
compliant rocks. Paper SPE 22232 available from SPE, Richardson, Texas
Zelasko, J.S., Krizek, R.J., Edil, T.B., 1975. Shear Behavior of Sands as a Function of
Grain Characteristics. Istanbul Conference on Soil Mechanics and Foundation
Engineering, 1, pp. 5564.
Zho, D., Zhang, G., Ruan, M., HeA., Wei, D., 2012. Comparison of Fracture
Conductivities from Field and Lab, presented at the International Petroleum
Technology Conference. Paper IPTC 14706, Bangkok, Thailand, February 9th.
Zou, Y.S., Zhang, S.C., Zhang, J., 2013. Experimental method to simulate coal nes
migration and coal nes aggregation prevention in the hydraulic fracture. Transp.
Porous Med 101, 1734.

ity/permeability and mechanical properties of the proppants such as

internal friction angle of proppants jointly determine the conductivity of a propped fracture segment.
Although a uniformly-graded proppant pack provides for a higher
permeability than a well-graded proppant pack, the lack of interlocking between grains during shear combined with grain sliding,
prevents the fracture from remaining open. Accordingly, the fracture
can close at a lower level of dierential stress.
In a well-graded proppant pack, the interlocking of grains provides
for a very strong bond during shearing. Accordingly, a propped
fracture segment will remain open for larger values of the dierential
stress. In addition, the permeability inside the shear band increases
in some intervals of the axial strain.
Our simulation results show that the porosity in the proppant pack
will not remain homogenous after the formation of a shear band. By
selecting a proppant pack that is able to keep the shear band open
for a longer period without failing, will increase production and
maintain production for a longer period of time.
Our results indicate that a well-graded proppant pack will provide
for a higher conductivity and promote production over a longer
period of time as compared to a uniformly-graded proppant pack.

References
3DEC: 3-Dimensional Distinct Element Code, 1988. Theory and Background, Itasca
Consulting Group, Inc., Minneapolis, Minnesota.
Abedi, S., Mirghasemi, A.A., 2011. Particle shape consideration in numerical simulation
of assemblies of irregularly shaped particles. Particuology 9 (4), 387397.
Alotaibi, M.A., Miskimins, J.L., 2015. Slickwater Proppant Transport in Complex
Fractures: New Experimental Findings: Scalable Correlation. Paper SPE-174828-MS
presented at SPE Annual Technical Conference and Exhibition. Houston, Texas,
USA, 2830 September. Doi: http://dx.doi.org/10.2118/174828-MS.
Barree, R.D., Conway, M.W., 1995. Experimental and Numerical Modeling of Convective
Proppant Transport. Paper SPE-28564-MS presented at SPE Annual Technical
Conference and Exhibition, New Orleans, Louisiana, USA, 2528 September. Doi:
http://dx.doi.org/10.2118/28564-MS.
Behr, A., Mtchedlishvili, G., Friedel, T., Haefner, F., 2006. Consideration of damaged
zone in a tight gas reservoir model with a hydraulically fractured well. SPE Prod. Op.;
Op. J. 21, 206211. http://dx.doi.org/10.2118/82298-PA.
Bishop, S.R., 1997. The Experimental Investigation of Formation Damage due to the
Induced Flocculation of Clays within a Sandstone Pore Structure by High Salinity
Brine. SPE paper38156, Presented at SPE European Formation Damage Conference,
The Hague, Netherlands.
Borujeni, A.T., Lane, N.M., Thompson, K., Tyagi, M., 2013. Eects of image resolution
and Numerical resolution on computed permeability of Consolidated packing using
LB And FEM pore-scale simulations. Comput. Fluids 88, 753763. http://
dx.doi.org/10.1016/j.compuid.2013.05.019.
Brannon, H.D., Wood, W.D., Wheeler, R.S., 2006. Large Scale Laboratory Investigation
of the Eects of Proppant and Fracturing Fluid Properties on Transport. Paper SPE98005-MS presented at the International Symposium and Exhibition on Formation
Damage Control, Lafayette, Louisiana, 1517 February. Doi: http://dx.doi.org/10.
2118/98005-MS.
Chen, H., Chen, S., Matthaeus, W.H., 1992. Recovery of The NavierStokes equations
using a lattice-gas Boltzmann method. Phys. Rev. A. 45, 53395342.
Civan, F., 2000. Reservoir Formation Damage. Gulf Publishing Company, Houston, TX,
USA.
Clarkson, C.R., 2013. Production data analysis of unconventional gas wells: review of
theory and best practices. Int. J. Coal Geol. 109110, 101146.
Cui, L., OSullivan, C., ONeill, S., 2007. An analysis of the triaxial apparatus using a
mixed boundary three-dimensional discrete element model. Geotechnique 57 (10),
831844.
Cundall, P.A., Strack, O.L., 1979. A discrete Numerical model for granular assemblies.
Geotechnique 29 (1), 4765.
Daneshy, A.A., 2005. Proppant distribution and Flowback in O-balance hydraulic
fractures. SPE J. Prod. Facil. 20, p41p47. http://dx.doi.org/10.2118/89889-PA.
Daneshy, A.A., 2007. Pressure variation Inside the hydraulic fracture and its Impact on
fracture propagation, conductivity, and screen-out". SPE J. Prod. Oper., p107p111.
http://dx.doi.org/10.2118/95355-PA.
Davies, D.R., Kuiper, T.O.H., 1988. Fracture Conductivity in Hydraulic Fracture
Stimulation. J. Petrol. Technol. 40 (5).
Diaz de Souza, O.C., Sharp, A.J., Martinez, R.C., Foster, R.A., Reeves Simpson, M.,
Piekenbrock, E.J., Abou-Sayed, I., 2012. Integrated Unconventional Shale Gas
Modeling: a Worked Example from the Haynesville Shale, de Soto Parish, North
Louisiana, SPE 154692, Americas Unconventional Resources Conference,
Pittsburgh, PA, 5e7June.
Economides, M.J., Nolte, K.G., 2000. Reservoir Stimulation third edition. Wiley and Sons
Ltd., West Sussex, England.
Fredd, C.N., McConnell, S.B., Boney, C.L., England, K.W., 2000. Experimental study of
hydraulic fracture conductivity demonstrates the benets of using proppants. Paper
SPE 60326, SPE Rocky Mountain Regional/Low-Permeability Reservoirs
Symposium and Exhibition, Mar. 1215, 2000, Denver, Colorado.

11