You are on page 1of 18

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/327822076

A Study of Proppant Transport With Fluid Flow in a Hydraulic Fracture

Article in SPE Drilling & Completion · September 2018


DOI: 10.2118/174973-PA

CITATIONS READS

21 1,001

3 authors:

Christopher A.J. Blyton Deepen Gala


University of Texas at Austin University of Texas at Austin
6 PUBLICATIONS 138 CITATIONS 16 PUBLICATIONS 238 CITATIONS

SEE PROFILE SEE PROFILE

Mukul Sharma
University of Texas at Austin
554 PUBLICATIONS 13,735 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Negotiations Conflict Managment and Business Development Research View project

New P3D Modeling of Hydraulic Fracture Propagation and Closure View project

All content following this page was uploaded by Deepen Gala on 22 September 2018.

The user has requested enhancement of the downloaded file.


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 1 Total Pages: 17

A Study of Proppant Transport With Fluid


Flow in a Hydraulic Fracture
Christopher A. J. Blyton, Deepen P. Gala, and Mukul M. Sharma, University of Texas at Austin

Summary
The effective placement of proppant in a fracture has a dominant effect on well productivity. Existing hydraulic-fracture models
simplify proppant-transport calculations to varying degrees. A common assumption applied is that the average proppant velocity caused
by flow is equal to the average carrier-fluid velocity, while the settling-velocity calculation uses Stokes’ law. To more accurately deter-
mine the placement of proppant in a fracture, it is necessary to account for many effects not included in previous assumptions.
In this study, the motion of particles flowing with a fluid between fracture walls is simulated with a coupled computational-fluid-
dynamics/discrete-element method (CFD/DEM) code that uses both particle dynamics and CFD calculations to account for both
particles and fluid. These simulations (presented in metric units) determine individual particle trajectories as particle-to-particle and
particle-to-wall collisions occur, and include the effect of fluid flow. The results show that the ratio of proppant diameter to fracture
width governs the relative average velocity of proppant and fluid.
A proppant-transport model developed from the results of the direct numerical simulations and existing correlations for particle-
settling velocity has been incorporated into a fully 3D hydraulic-fracturing simulator. This simulator couples fracture geomechanics
with fluid-flow and proppant-transport considerations to enable the fracture geometry and proppant distribution in the main hydraulic
fracture to be determined. For two typical shale-reservoir cases, the proppant placement and width distribution have been determined,
allowing comparison at the hydraulic-fracture scale, including effects observed at the particle scale. This allows for optimization of the
treatment to a specific application, and the results are presented in oilfield units, considered more familiar to our readers.

Introduction
Overview. During hydraulic-fracture stimulations, it is common to inject a proppant-laden slurry consisting of a low-viscosity
Newtonian fluid and well-sorted sand. Alternative fluids and proppants, including gels and ceramics, are less commonly used, at least
for treatments pumped in the USA between 1947 and 2010 according to Gallegos and Varela (2015), who analyzed fluid and proppant
types reported in this region and period. Proppant is included in a stimulation treatment to provide significant fracture conductivity after
fracture closure.
Numerical simulation of the fracturing process at hydraulic-fracture scale is often used to provide a prediction of effective fracture
length, while analysis of pressure-buildup tests and production history matching have been used to infer the actual effective fracture
length after a stimulation treatment. Several researchers have inferred effective fracture lengths from production data that are much
shorter than those predicted. Lee and Holditch (1981) and Sharma et al. (2005) examined pressure-buildup and production data from
fractured low-permeability gas wells using several techniques. They consistently demonstrated that the inferred effective fracture
lengths were considerably shorter than those predicted by fracturing models.
Several reasons have been offered to explain the differences in the fracture lengths predicted by hydraulic-fracturing models and
those inferred from production and pressure-transient data. Higher leakoff caused by induced unpropped fractures, as examined by
Sharma and Manchanda (2015); fracture complexity induced by natural fractures; and planes of weakness and proppant retardation
have all been suggested as possible explanations; see Malhotra et al. (2014). This paper presents a study of the effect of proppant flow
with fluid on proppant placement in fractures.
The width of a hydraulic fracture during pumping is small, and, in some cases, may be only a small multiple of the proppant diameter.
One specific example of proppant transport between fracture walls of small spacing, relative to the particle size, is in a complex fracture
network consisting of dilated natural fractures connected to the main hydraulic fracture. However, one should note that such complexity
of fracture geometry is commonly not included in numerical simulations at the hydraulic-fracture scale, including for this study.
This research reduces the study of proppant transport at the granular scale to that of two-phase fluid-and-solid transport between two
parallel plates. The motion of discrete particles, moving in a fluid domain with appropriate boundary conditions applied, has been simu-
lated. Transport of proppant results from the macroscopic pressure gradient driving fluid flow. The relative average velocity of proppant
and fluid and the proppant-settling velocity are key to more accurately determining the proppant distribution within a hydraulic fracture.
Because these simulations are numerically very expensive, direct coupling to a geomechanical fracture simulator is not practical. How-
ever, using a correlation developed in this study, the results from these simulations have been incorporated into a fracture simulator,
allowing for more-accurate proppant-transport modeling at the fracture scale in an efficient manner with respect to computational
expense. The proppant placement and width distribution of a main hydraulic fracture have been calculated for two typical cases—with
and without inclusion of the particle-scale effects determined in the granular-scale study—to allow for comparison in this application.

Literature Review. Solid-Sphere Transport With Fluid Flow. The velocity of a single rigid sphere, transported by various fluid
flows, is a problem that has been studied by several researchers, including Ho and Leal (1974), Ganatos et al. (1980), and, more
recently, Staben et al. (2003) and Jones (2004). Ho and Leal (1974) used the method of reflections in an early and approximate treat-
ment of two problems: a sphere transported in Couette flow and in Poiseuille flow. Their results are accurate for small diameter/slot-
width ratios, but increasingly inaccurate when this ratio is large (i.e., significant errors arise when their approach is applied to a narrow
slot). Ganatos et al. (1980) used a numerical boundary-collocation technique to examine four cases—the translation of a sphere without
rotation, rotation of a sphere without translation, Poiseuille flow past a rigidly held sphere, and Couette flow past a rigidly held sphere.

Copyright V
C 2018 Society of Petroleum Engineers

This paper (SPE 174973) was accepted for presentation at the SPE Annual Technical Conference and Exhibition, Houston, 28–30 September 2015, and revised for publication. Original
manuscript received for review 18 September 2017. Revised manuscript received for review 26 March 2018. Paper peer approved 4 April 2018.

2018 SPE Drilling & Completion 1

ID: jaganm Time: 17:16 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 2 Total Pages: 17

At small Reynolds number (Re), the solutions can be summed. A limitation of that approach is that the sphere must be at least 10% of
its radius from the wall.
Staben et al. (2003) used a boundary integral algorithm to examine the motion of a sphere in Poiseuille flow. This approach directly
incorporates the effects of wall interactions into the stress tensor, allowing for highly accurate results for larger particle-diameter/slot-
width ratios and for a particle close to the wall. This overcomes the limitations of the solutions provided by Ho and Leal (1974) and
Ganatos et al. (1980). Staben et al. (2003) extended the results for a single-particle translation velocity to a multiparticle case using the
assumption that the concentration is uniform and dilute. Jones (2004) applied a Fourier-transform technique to the same problem inves-
tigated by Staben et al. (2003). Resembling Staben et al. (2003), this approach overcomes the restrictions of the solutions by Ho and
Leal (1974) and Ganatos et al. (1980).
All the studies discussed previously apply to a single particle moving in a Newtonian fluid flowing in a creeping-flow regime. A
proppant slurry is usually prepared at nondiluted concentrations, and proppant settles in the fracture before closure, which may result in
the largest concentration at the bottom of the fracture. On the other hand, the ramping of proppant concentration during pumping a
crosslinked-gel-based slurry might result in larger concentration closer to the depth where the fracture connects to the well, rather than
at its bottom. The carrier fluid used is often pumped at injection rates that produce Re that may be in the laminar regime but are much
greater than that for creeping flow. As a result, the solutions currently available in the literature cannot be directly applied to the more
complex case investigated in this research, where larger concentration and Re prevail.
Solid-Sphere Settling in Fluid. Stokes (1851) published an analytical solution for the settling velocity of a single rigid sphere
falling through an infinite volume of a Newtonian fluid at a very small particle Re. This solution is often termed Stokes’ law and is
given by
d2 ðqp  qf Þg
uStokes ¼ ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð1Þ
18lf

where uStokes is the Stokes’ settling velocity (m/s); d is the particle diameter (m); qp and qf are the particle density (kg/m3) and fluid den-
sity (kg/m3), respectively; g is gravitational acceleration (m/s2); and lf is the fluid dynamic viscosity (Pas). This solution arises from
equating the buoyancy and drag forces upon a particle at terminal velocity. Commonly used carrier-fluid viscosity, proppant size, and
density difference to the fluid place the particle Re outside the creeping-flow regime for which Stokes’ law is valid. For larger particle
Re, inertial effects are significant, and empirical friction-factor solutions are often used for the drag force. One such friction-factor solu-
tion is available in Bird et al. (2007).
The term “hindered settling” is often used to describe the reduction in settling velocity, compared with that of a single particle,
observed when multiple particles settle. The reduction is commonly quantified by the modification of a single-particle-settling solution
that was based on an empirical function of volumetric particle concentration. Richardson and Zaki (1954) developed one such correla-
tion from physical experiments in which a concentrated suspension settled in a closed tube. Peker and Helvaci (2008) present a compre-
hensive review of similar correlations.
As noted previously, Ganatos et al. (1980) provide a rigorous solution to a sphere translating between two parallel walls. The solu-
tion form is that of a dimensionless force coefficient, which quantifies the reduction in settling velocity resulting from the presence of
walls, compared with that predicted by Stokes’ law for an infinite volume of fluid. This solution is applicable to small particle Re.
Liu (2006) proposed an approximate solution to proppant-settling velocity in a hydraulic fracture. This involves modifying uStokes
by using independent functions to account for inertial, concentration, and wall effects. The inertial correction used is provided in an
explicit form, which removes the need for interations required by friction-factor solutions. The concentration correction used is a poly-
nomial fitted to the empirical correlations of Richardson and Zaki (1954), Maude and Whitmore (1958), Daneshy (1978), and Nolte
(1988). Finally, the wall-spacing correction formulated by Liu (2006) is based on the solution of Lorentz (1907), available in Happel
and Brenner (1965), for a sphere translating near a wall. The corrected settling velocity (m/s), usettling, is given by
" # "    2 #
al0:57
f 2 3 d d
usettling ¼ uStokes 0:29 0:29 0:86
ð1  4:8c þ 8:8c  5:9c Þ 1  1:563 þ 0:563 ; . . . . . . . . . . . . . . . . ð2Þ
qf ðqp  qf Þ d W W

where a is a unit-conversion constant, c is the volumetric particle concentration (volume fraction), and W is the width of the hydraulic
fracture (m).

CFD/DEM Simulation Approach


Application of CFD/DEM to Proppant Transport. A numerical solution to the average velocity of solid spheres, transported by fluid
between two parallel plates at several intermediate concentrations, is presented relative to the average fluid velocity. Appropriate
boundary conditions are applied such that the resultant flow field is fully developed and the concentration of solid spheres, representing
proppant particles, remains constant throughout the simulation.
A CFD/DEM approach was used. CFD provides numerical solutions to the fundamental equations governing the flow of fluids, and
finds application to problems where analytical solutions are not available. A necessary preprocessing step required by this approach is
discretization of the simulation domain by creation of a “mesh” comprising many gridblocks, over which the governing equations are
solved by the finite-volume method. The DEM solves for the motion of individual particles resulting from the forces acting upon them.
These forces arise from interactions between each other and walls, both cases where collisions take place in the case of particle trans-
port. Another set of forces that acts upon particles during proppant transport is caused by the motion of the surrounding fluid.
This research uses a “resolved” CFD/DEM approach, which refers to the use of a CFD mesh smaller than the particles and calculat-
ing the fluid force acting on each particle directly. In contrast, many studies use an “unresolved” CFD/DEM approach, whereby the
CFD mesh is composed of gridblocks larger than the particles and a correlation is used for the drag force acting upon each particle. The
“unresolved” approach is not appropriate for this specific application because the spacing between the parallel plates is only slightly
larger than the particles in some cases, leading to a complexity to the fluid flow around particles that would not be adequately repro-
duced by a coarse mesh.

Simulation Approach. CFD Simulation. The open-source CFD library OpenFoam (Open Source Field Operation and Manipulation)
was used to solve the fluid-flow field between particles; see Weller et al. (1998). This involves the solution of the incompressible
Navier-Stokes equations for a Newtonian fluid.

2 2018 SPE Drilling & Completion

ID: jaganm Time: 17:16 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 3 Total Pages: 17

The continuity equation is

r  uf ¼ 0; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð3Þ

where uf is the fluid-velocity vector (m/s). The momentum equation is


@uf
qf þ qf uf  ruf ¼ rp þ lf r2 uf ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð4Þ
@t
where t is time (seconds) and p is the pressure (Pa). These equations were solved numerically on an orthogonal mesh by a finite-
volume method.
DEM Simulation. The open-source DEM package LIGGGHTs was used to solve for the particle motion; see Kloss et al. (2012).
This method applies Newton’s second law to the calculation of individual particle trajectories that result from the sum of forces and tor-
ques acting on each. The forces on particles caused by the surrounding fluid flow and collisions with other particles and walls are
included. The equations of motion may be expressed as

d2 xi
mi ¼ Fi total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð5aÞ
dt2
d2 hi
Ii ¼ Mi total ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð5bÞ
dt2
where mi is the mass of particle i (kg), xi ¼ (xi, yi, zi) is the coordinates of its center of gravity (m), and Fi total is the total force acting
upon the particle (N). Considering rotation, Ii is the moment of inertia (kgm2), hi is the angular position (radians), and Mi total is the
total moment acting on the particle (Nm).
A nonlinear force-displacement law for the normal component of contact between elastic spheres developed by Hertz (1882) and a
solution for the tangential component from Mindlin and Derieswicz (1953) were applied to particle collisions. If the tangential compo-
nent of the interparticle force exceeds the Coulomb frictional force, sliding is assumed to occur. Combining normal and tangential com-
ponents gives

Fij ¼ ðkn dnij  cn Dup nij Þ þ ðkt dtij  ct Dup tij Þ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð6Þ

where Fij is the interparticle force (N), k is the spring coefficient (N/m), d is the overlap (m), c is the damping coefficient (Ns/m), and
Dup is the relative speed (m/s). Symbols nij and tij are used to denote unit vectors in the normal and tangential directions considering a
Langrangian coordinate system. The subscripts n and t are used to denote the normal and tangential directions. Finally, subscript ij is
applied to the interaction of particle i with particle j. For more detail on the calculation of the spring and damping coefficients from the
particle material properties, see Kloss et al. (2012). The same contact-force calculation applies when a particle collides with a wall;
however, in this case, the wall is assumed to have infinite mass and size.
CFD and DEM Coupling. The CFD and DEM representations of the physical system simulated are coupled. For illustration pur-
poses only, a simple 2D domain with one immersed solid body is shown in Fig. 1.

Γs Ωs

Fig. 1—The 2D CFD/DEM simulation domain shown for illustration of the coupling method.

An individual solid body with domain Xs is shown immersed in a fluid. The surface of the solid is denoted Cs. The coupling consists
of two parts. First, from the DEM solution, the position and velocity of each particle are determined and sent to the CFD solver. This
information is used to provide an additional boundary condition to the fluid domain for each immersed particle, given by

uf ¼ ui on Cs : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð7Þ

The second part of the coupling involves determining the fluid force acting on each particle from the CFD solution and sending this
information to the DEM solver.
One should note that the form of the Navier-Stokes equations solved does not include a body-force term, from which the hydrostatic
pressure gradient arises in the case of gravitational force applying. For this specific application, a study of the average particle velocity
and average fluid velocity for flow between two parallel plates, gravitational force is neglected. With this assumption, there is no buoy-
ancy contribution to the fluid force exerted on each particle.
The fluid force acting on each particle is calculated with a method proposed by Shirgaonkar et al. (2009). Hager et al. (2014) provide
details of the specific implementation that has been used in this research, for which the governing equations are solved over a 3D
domain with multiple immersed bodies.

2018 SPE Drilling & Completion 3

ID: jaganm Time: 17:16 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 4 Total Pages: 17

The traction on the immersed solid body, tCs , is given by

tCs ¼r  n^ on Cs ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð8aÞ

where r is the total stress tensor of the fluid field and n^ is the outward normal unit vector on the surface of the immersed solid body.
The total force exerted by the fluid on the immersed solid body can be obtained by integration of Eq. 8a,
ð ð
tCs dCs ¼ r  n ^ dCs : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð8bÞ
Cs Cs

The divergence theorem allows an equivalent expression to be written as


ð ð
tCs dCs ¼ r  rdXs ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð8cÞ
Cs Xs

which transforms the integral across the surface of the sphere to a volumetric integral over the solid domain. For an incompressible
Newtonian fluid,
n h io
r  r ¼ rp þ r  lf ruf þðruf ÞT : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð8dÞ

Substitution and simplification of the previous result produce


ð ð
 
tCs dCs ¼ rp þ lf r2 uf dXs : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð8eÞ
Cs Xs

Initial and Boundary Conditions. Simulation Domain. The domain used for simulation is illustrated in Fig. 2.
The length of the domain (m), L, is 9d and the height of the domain, H, is 6d. The initial and boundary conditions for the CFD and
DEM domains are specified separately.

Top boundary

d
Inlet-CFD Outlet-CFD
H
boundary boundary

z
x

y Bottom boundary

Inlet-DEM Outlet-DEM
boundary boundary

Back boundary

Poiseuille-flow upx
inlet-boundary W
condition

Front boundary

2d 2d
L

Fig. 2—The 3D CFD/DEM simulation domain, including boundary condition at the CFD inlet.

CFD Initial and Boundary Conditions. Fluid flows across the domain in the x-coordinate direction resulting from a velocity
boundary condition at the CFD inlet, appropriate for fully developed flow between two infinite parallel plates, and a constant-pressure
boundary condition at the CFD outlet. This approach eliminates a developing flow region and allows a smaller domain to be used. The
CFD front and back boundaries, which represent fracture walls, use a velocity boundary condition. In the case of zero leakoff, these use
the no-slip condition commonly assumed at a fluid-to-solid interface. When leakoff is present, fluid leaves the domain at a constant
velocity perpendicular to the front boundary and the back boundary. The CFD top and bottom boundaries use a cyclic boundary condi-
tion, which represents a domain that is infinitely large in height. This is achieved by setting the flux and pressure at one cyclic boundary
to those at the paired cyclic boundary. To minimize the duration of the transient response of the system, the initial pressure and velocity
fields used are for fully developed flow between two infinite parallel plates. Further details on the CFD initial and boundary conditions
applied are provided in Appendix A.

4 2018 SPE Drilling & Completion

ID: jaganm Time: 17:16 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 5 Total Pages: 17

DEM Initial and Boundary Conditions. Particles move from the DEM inlet to the DEM outlet resulting from fluid flow in the
x-coordinate direction. A periodic boundary condition is applied between the two, whereby particles leave the domain at the DEM out-
let and re-enter at the DEM inlet with the same position on the y–z-plane and the same velocity. The DEM inlet and DEM outlet are
positioned 2d within the CFD inlet and CFD outlet, respectively, as illustrated in Fig. 2. This setup is a practical consideration whereby
a numerical artifact concerning the use of periodic DEM boundaries in coupled CFD/DEM simulations is minimized; see a detailed dis-
cussion of this subject in Blyton (2016). Walls are placed on the DEM front and back faces, representing the fracture walls. The DEM
top and bottom faces also use a periodic boundary condition, although it is unlikely that particles will cross between these because the
pressure gradient driving flow is in the x-coordinate direction only.
Initial particle positions are selected with each coordinate value from a uniform distribution bounded by limits of one particle radius
inside the simulation-domain boundary. Positions of a sufficient number of nonoverlapping particles to provide the desired volumetric
concentration are determined. Because no particles are added after the initial seed and the DEM boundaries are periodic in the x- and
z-coordinate directions and with walls in the y-coordinate direction, the concentration remains constant throughout the simulation. The
initial velocity of each particle is equal to the average initial fluid velocity to minimize the duration of the transient response of the system.
CFD and DEM Numerical Considerations and Data Exchange. The timestep required for numerical stability of the DEM solver
is smaller than that of the CFD solver. There are two criteria necessary for DEM-simulation stability: use of a timestep smaller than
both the Raleigh and Hertz times. Both criteria are functions of particle size and elastic moduli, while the former also depends on
density and the latter includes a relative velocity term. Precisely how much smaller than the two criteria the timestep needs to be for
stability is a function of coordination number; see O’Sullivan and Bray (2004). For most simulations, with a particle diameter of
0.0004 m, the DEM timestep used was 3108 seconds. Simulations with a larger particle size, a diameter of 0.001 m, enabled use of a
larger timestep of 7108 seconds.
For fully developed fluid flow between two infinite parallel plates, in the absence of immersed particles, zero pressure gradient and
large velocity gradients are present in the coordinate direction normal to the two plates. A linear-pressure gradient and zero-velocity
gradient exist parallel to flow. As a result, accurate numerical simulation of this type of flow by the finite-volume method requires less
mesh refinement in the x- and z-coordinate directions compared with the y-coordinate direction, using the coordinate system shown in
Fig. 2. For this research, which, by necessity, includes details of the flow around each particle, the width of the gridblocks in the x- and
z-coordinate directions is half the particle diameter. In the y-coordinate direction, a gridblock width of half a particle diameter is the
largest used. This was applied to large-slot-width cases, which exhibit smaller velocity gradients than do small-slot-width cases at the
same Re. For the smaller-slot-width cases, a gridblock width less than half a particle diameter is used. Further, dynamic refinement of
the CFD mesh has been applied for greater resolution of the fluid flow around each particle, while incurring less numerical expense in
regions of the domain where a fine mesh is not required. Two steps, each of which splits a gridblock into two in each of the three Carte-
sian coordinate directions, were applied near each particle.
The Courant number, a measure of the magnitude of the local fluid velocity relative to the size of the gridblock for a specific time-
step, is the relevant criterion for CFD-simulation stability. Considering a fluid-flow-only simulation (not a coupled CFD/DEM simula-
tion), a Courant number of more than 1.0 produces numerical instability. In the case of the coupled CFD and DEM simulation approach
used in this research, a coupling interval is required. This determines a specific number of smaller DEM timesteps taken before a larger
CFD timestep is made. In the absence of a superior method of selecting an appropriate coupling interval, a trial-and-error approach has
been adopted. A result of the smaller DEM timestep and the necessary coupling interval selected is that the CFD timestep used produces
a Courant number much smaller than unity for all simulations.
In brief, there are three key steps involved in the data exchange for coupling the DEM and CFD representations of the system:
1. The CFD solver is used to calculate the fluid pressure and velocity fields, assuming that only the fluid phase is present in the
domain (i.e., the presence of solid particles is neglected for the purposes of the initial solution).
2. The DEM timesteps prescribed by the selected coupling interval are completed, using the individual fluid-to-particle forces deter-
mined resulting from the last CFD solution. The resultant state of particle velocities is “imposed” on the CFD gridblocks where
they are “covered” by the particles, as determined from the position of each. However, when the velocity field for the fluid
domain is modified in this way, the continuity equation is no longer satisfied. A correction is made to ensure that the conservation-
of-mass requirement is once again met; see Hager et al. (2011) for details.
3. Finally, the force exerted on each particle by the fluid is calculated and sent to the DEM solver.
These three steps are repeated for the duration of the simulation.
The coupled approach is explicit in terms of the force that the fluid exerts upon each particle. During Step 2, the fluid-to-particle force
applied to each particle is taken from the last CFD solution. Even if only one DEM timestep is taken, which is not practical for the appli-
cations considered in this research because of the considerable computational expense of the method, the coupling does not include itera-
tion between the CFD and DEM solutions. Using a coupling interval greater than 1.0 may be considered to increase the degree to which
the solution is explicit, because the DEM solution evolves further without an update of the individual fluid-to-particle forces.
The computational expense of the method and constraints upon run times necessitate a relatively large coupling interval to allow
production of sufficient simulation data for nondiluted cases. For example, the simulation of a system with many particles using a cou-
pling interval of 1.0 would not reach steady state before the run-time limit set for the hardware used. Confirmation that the accuracy of
the calculations is maintained with a relatively large coupling interval was investigated using less-computationally-expensive single-
particle cases in slot Poiseuille flow. The smallest slot width considered requires a particle-diameter/slot-width ratio of 0.95. This geom-
etry was used to make two comparisons of the calculated translational velocity of a single particle in the center of a slot at Re of 1.0 and
1,000, which place the flow in the creeping and laminar flow regimes, respectively. The results determined from comparing simulations
using coupling intervals of 1.0 and 20 agreed within approximately 2%. For most of the simulations presented, the coupling interval
used was 20. The exception is cases with a particle-diameter/slot-width ratio of 0.1, for which a coupling interval of 40 was used. This
is by necessity, given the relatively large number of particles simulated in these cases, but acceptable, given the smaller fluid and parti-
cle velocities apparent. The level of agreement (approximately 2%) is considered acceptable for this application.

Verification. CFD/DEM results for the velocity of a single sphere translating in Poiseuille flow at small Re were compared with those
of Staben et al. (2003). The slot-flow Re used in this study is defined on the basis of a characteristic length of double the width of the
fracture. This is appropriate for infinite parallel plates that are based on consideration of the hydraulic diameter and given by
2Whufx iinlet qf
Re ¼ ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð9Þ
lf

2018 SPE Drilling & Completion 5

ID: jaganm Time: 17:16 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 6 Total Pages: 17

where <ufx>inlet is the cross-slot average fluid velocity at the inlet. Although Staben et al. (2003) did not state the range of Re for which
their results are applicable, their simulations were confined to the creeping-flow regime. For comparison purposes, the CFD/DEM simu-
lations were conducted at a Re of 1.0. For normalization, the average fluid velocity at the inlet was used to reduce the CFD/DEM parti-
cle velocity. The inlet fluid velocity represents the undisturbed fluid velocity used by Staben et al. (2003). The comparison was made at
10 different ratios of particle diameter to slot width and two different particle y-coordinate locations in the slot, as shown in Fig. 3.

1.2 Center (Staben et al. 2003)


Wall (Staben et al. 2003)

upx /<ufx >inlet


Center (CFD/DEM)

0.8 Wall (CFD/DEM)

0.4

0
0 0.2 0.4 0.6 0.8 1
d/W

Fig. 3—The upx /<ufx>inlet determined by CFD/DEM simulation and provided by Staben et al. (2003) vs. d/W for a single particle
transported at two locations in the slot, at the center and at the wall.

For fully developed flow between two parallel plates, in the absence of particles, the maximum fluid velocity is 1.5 times greater
than the average fluid velocity. For d/W equal to 0.1, a particle in the middle of a slot translates at very slightly less than the local undis-
turbed (particle-free) fluid velocity, from the results of Staben et al. (2003). As a result, the particle velocity normalized by the average
fluid velocity presented in Fig. 3 is approximately 1.5 times in this case. Across the range of d/W and particle position in the slot, the
comparison between the CFD/DEM results and those from Staben et al. (2003) demonstrates good agreement.
The results compared previously are for a single sphere translating in slot Poiseuille flow. For practical application to proppant trans-
port, larger concentrations of particles are transported at larger Re. The method is general in that there is no theoretical limit to particle
concentration and there is no simplification to the Navier-Stokes equations solved (i.e., the effects of fluid inertia are included). As a
result, the method is suitable for this application. Of course, how well the idealized numerical model reflects the complexity of the
actual physical system depends on the accuracy of the assumptions invoked. Two key assumptions—perfectly spherical particles and
smooth walls—are clearly considerable simplifications of irregularly shaped proppant grains and rough fracture walls. Modeling each
of these in more detail requires a significantly expanded study and is beyond the scope of this research.

CFD/DEM Simulation Results


CFD/DEM Simulation-Processing Procedure. Calculation of Average Particle Velocity in the Direction of Fluid Flow. Of key
concern to this study is the relative average velocity of proppant and fluid. The average particle velocity in the direction of fluid flow is
not specified directly by the boundary conditions imposed. Rather, each individual particle is transported, resulting from the fluid forces
acting. As a result, it is necessary to determine the average particle velocity in the direction of fluid flow for each specific
simulation case.
Cumulative-particle mass vs. time-simulation data was recorded by a counter in the y-z-plane, at an x-coordinate location in the cen-
ter of the domain. An example of the simulation data recorded for the base case with d/W of 0.1 is shown in Fig. 4.

×10–5 ×10–6
5 5
Mass Mass
Least-squares regression Least-squares regression
4 4
Cumulative Mass (kg)

3 3

2 2

1 1

0 0
0 0.02 0.04 0.06 0.08 0.1 0 0.002 0.004 0.006 0.008 0.01
Time (seconds) Time (seconds)

Fig. 4—Left: cumulative mass vs. time for base case with d/W of 0.1. Right: cumulative mass vs. time for base case with d/W of 0.1,
subset of simulation data from 0 to 0.01 seconds shown.

6 2018 SPE Drilling & Completion

ID: jaganm Time: 17:16 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 7 Total Pages: 17

During 0.1 seconds of simulation time, there are 548 particle counts recorded by the counter, whereas there are only 86 particles
present in the domain for this specific simulation case. This is a result of the periodic boundary condition applied to the DEM domain
between the DEM outlet and the DEM inlet. The right panel of Fig. 4 shows the discrete nature of the recorded data, the smallest incre-
ment in cumulative mass results from the registration of the passing of a single particle. The left panel shows a least-squares regression
fit to 0.1 seconds of simulation data, providing an adequate measure of the average gradient, which represents the average particle-mass
_ see Blyton (2016) for more detail. The average particle velocity in the x-coordinate-direction, <upx>, was determined
rate (kg/s), m;
from the average particle-mass rate as
m_
hupx i ¼ ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð10Þ
cAqp

where A is the cross-sectional area of the simulation domain in the y–z-coordinate plane (m2), given by the product of W and H.
Calculation of Average Fluid Velocity. The cross-slot average fluid velocity at the inlet to the simulation domain, at an x-coordinate
location of zero, is specified by means of the fully developed boundary condition imposed on the CFD inlet, as detailed in Appendix
A and shown in Fig. 2. The periodic DEM inlet, where particles re-enter the domain after leaving through the DEM outlet, is at an
x-coordinate location of 2d. As a result, there is a region of particle-free fluid between the x-coordinate locations of zero and 2d. In this
region, the average fluid velocity remains constant with x-coordinate location, by conservation-of-mass considerations. Because particles
leave the domain at the DEM outlet, there is a second region of particle-free fluid between the x-coordinate locations of 7d and 9d, where
the average fluid velocity is equal to the cross-slot average fluid velocity at the inlet to the simulation domain and constant with
x-coordinate location.
However, between the x-coordinate locations of 2d and 7d, particles are present at a volumetric concentration c. Particles are trans-
ported by fluid flow in this region, resulting in the average fluid velocity not being specified directly by the boundary conditions
imposed, as is the case in the two particle-free regions. The discontinuous concentration distribution is illustrated in Fig. 5.

0.8
c (volume fraction)

0.6

0.4

0.2

0
0 2 4 6 8
x/d

Fig. 5—The c vs. dimensionless x-coordinate location, x/d.

The particles are transported with an average particle velocity in the direction of fluid flow <upx>, determined from the average
particle-mass rate recorded, with the detail of the procedure provided previously. Conservation of total flux is maintained throughout
the domain. Thus, for any cross section in the y–z-plane, total flux in the x-coordinate direction is

chupx i þ ð1  cÞhufx i ¼ b; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð11aÞ

where b is a constant. Because this conservation equation applies at all x-coordinate locations in the domain, application at the CFD
inlet allows for solution of b, because at this location c is equal to zero and <ufx> is specified,

b ¼ hufx iinlet : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð11bÞ

Substitution and rearrangement allow for solution of <ufx> in the region between x-coordinate locations of 2d and 7d, where both
particles and fluid are transported as
hufx iinlet  chupx i
hufx i ¼ : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð11cÞ
ð1  cÞ
That is, although the cross-slot average fluid velocity at the CFD inlet is specified, both the average particle velocity and the average
fluid velocity in the region where both phases flow are not specified, and arise from the solution. These two velocities, and the relation-
ship between them, are significant in proppant transport driven by fluid flow.

CFD/DEM Simulation Cases. The model was run with each independent variable of interest set to values appropriate for typical
hydraulic-fracturing applications, included in Table 1.
A base case was defined, and one independent variable changed at a time, while leaving all other independent variables constant.
This allowed identification of the main effects with an efficient number of runs. The base case included all values of d/W, d of
0.0004 m, c of 0.15, qp of 2650 kg/m3, qf of 1000 kg/m3, lf of 0.001 Pasn, Re of 1,000, and leakoff of 0 m/s.

2018 SPE Drilling & Completion 7

ID: jaganm Time: 17:16 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 8 Total Pages: 17

Independent Variable Value (SI units) Application


Particle-diameter/slot-width ratio, d/W 0.1, 0.2, 0.3, 0. 4, 0.5, 0.6, 0.7, 0 .8 , 0.9, 0.95
Particle diameter, d 0.0004, 0.001 (m) 40/60-, 16/30-US-mesh proppant
Volumetric-solids concentration, c 0.05, 0.1, 0.15, 0.2, 0.25 (volume fraction)
Particle density, ρp
3
2650 kg/m Silica sand
Fluid density, ρf
3
1000 kg/m Aqueous fracturing fluids
n
Fluid viscosity, µf 0.0002, 0.001 Pa·s 0.2- and 1-cp slickwater
Slot-flow Reynolds number, Res 1,000, 2,000
Carter leakoff coefficient after 10 minutes of
0.5
injection. Shale (0 ft/min ), high-permeability
0, 1.606 × 10 m/s
–6
Leakoff 0.5
reservoir (0.001 ft/min ). For more details on
Carter leakoff, see Howard and Fast (1957).

Table 1—Values of independent variables used in the CFD/DEM simulation study.

Average Proppant-Phase Velocity in the Direction of Fluid Flow. The results are presented in a dimensionless format. The average
proppant velocity in the x-coordinate direction, normalized by the average fluid velocity in the x-coordinate direction, is termed the
dimensionless proppant velocity. This quantity is presented vs. the particle-diameter/slot-width ratio for each simulation case. The
results of several simulation cases, with several different concentrations of particles and particle-diameter/slot-width ratios, are pre-
sented in Fig. 6.

1.12
c = 0.05
c = 0.10
1.15 c = 0.15
c = 0.20
c = 0.25
<upx>/<ufx >

1.1

1.05

0.95
0 0.2 0.4 0.6 0.8 1
d/W

Fig. 6—The < upx>/<ufx > vs. d/W for several values of c.

The results shown in Fig. 6 demonstrate a peak in the dimensionless proppant velocity at a particle-diameter/slot-width ratio of 0.8.
Because the results are discrete in nature, it is not possible to determine precisely the particle-diameter/slot-width ratio for which the
dimensionless proppant velocity is at a maximum. However, from the results available, it is apparent that this occurs at approximately
0.8 particle-diameter/slot-width ratio. There is no systematic dependence on concentration, though some scatter is evident. The results
of several simulation cases with different Re, viscosity, particle diameter, and fluid leakoff are presented in Fig. 7.
All the results shown in Fig. 7 demonstrate the same trend observed in Fig. 6 (i.e., a peak in the dimensionless proppant velocity
at a particle-diameter/slot-width ratio of 0.8). Again, although scatter is evident, there is no dependency on Re, lf, d, or fluid
leakoff observed.
A continuous piecewise function has been fitted to the CFD/DEM simulation results for use in a hydraulic-fracturing simulator. At
particle-diameter/slot-width ratios equal to or smaller than 0.4, a linear function with a constant value of 1.0 has been used. For d/W
greater than 0.4 and less than or equal to 0.95, a cubic polynomial curve was fitted for the dependency upon d/W. For d/W equal to 1.0
(i.e., for a particle diameter the same size as the slot width), it is expected that the particle would jam and the dimensionless velocity
would be zero as a result. Simulations were not conducted for d/W greater than 0.95, and therefore the exact behavior of the dimension-
less velocity in this range is not known. As such, the simplest trend possible has been fitted, a linear function for d/W. The complete cor-
relation fitted is given by
8
> d
>
>  0:4 1 if
>
> W
  >
> 3  2  
hupx i d < d d d d
¼h ¼ 4:83 þ 8:40  4:24 þ 1:66 if 0:4 <  0:95
hufx i W >
> W W W W
>
>    
>
> d d
>
: 21:48 þ 21:48 if > 0:95:                                 ð12Þ
W W

8 2018 SPE Drilling & Completion

ID: jaganm Time: 17:16 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 9 Total Pages: 17

1.2 1.2
Res = 1,000 μ = 0.001 Pa⋅s
1.15 1.15 μ = 0.0002 Pa⋅s
Res = 2,000

<upx>/<ufx>
1.1 1.1

1.05 1.05

1 1

0.95 0.95
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

1.2 1.2
d = 0.0004 m Fluid loss = 0 m/s
1.15 d = 0.001 m 1.15 Fluid loss = 1.6×10–6 m/s
<upx>/<ufx>

1.1 1.1

1.05 1.05

1 1

0.95 0.95
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
d/W d/W

Fig. 7—The < upx >/<ufx > vs. d/W for different Re (top left), lf (top right), d (bottom left), and leakoff (or fluid loss) (bottom right).

Hydraulic-Fracture-Simulation Approach
Overview of Incorporation of CFD/DEM Correlations. The correlation developed for proppant transport with fluid flow can be
incorporated into any hydraulic-fracturing simulator. This section provides a brief discussion of how this has been performed in a 3D
planar hydraulic-fracturing simulator that couples fracture mechanics with fluid and proppant transport; see Ribeiro and Sharma (2013).
The constitutive equations shown next are based on the appropriate assumptions for an incompressible, isothermal, single-phase, and
single-component carrier fluid.

Hydraulic-Fracture Mathematical-Problem Definition. The fracture is assumed to be planar, vertical, and symmetric with respect
to the wellbore (biwing) and propagating in a purely elastic medium. Both height and length growth are permitted, with the capability
to include horizontal layers with different elastic moduli and minimum horizontal stress. The fracture width is negligible compared
with fracture height and length, and thus slurry flow through fracture is modeled as 2D in the x–z-plane between parallel porous walls.
The illustration of the fracture domain in Fig. 8 shows the portion of the boundary where the injection of slurry is imposed, the per-
forated interval, @Xperf, forming an important boundary condition. To model a fracture in a horizontal well, the perforated interval is
small. The slurry flows over the entire fracture domain with vector components in the x-coordinate direction and z-coordinate direction.
A typical distribution of proppant is shown, whereby settling has produced a large concentration in the bottom portion of the fracture.
Hence, the propped-fracture length is often much smaller than the created fracture length.

z
Slurry Phase

∂Ωfront
Fluid Phase
Ω
y
∂Ωperf
x
Proppant

Fig. 8—Hydraulic-fracture-simulation domain, X, fracture front ›Xfront and the perforated interval along the wellbore ›Xperf,
adapted from Ribiero and Sharma (2013).

Fracture Mechanics and Propagation Criteria. The opening equation for a tensile Mode-I planar fracture in an isotropic, homogene-
ous, 3D elastic medium is a boundary integral equation. A review of literature relevant to Mode-I fractures is available in Zoback
(2007). The equation is provided in Kossecka (1971) as

2018 SPE Drilling & Completion 9

ID: jaganm Time: 17:16 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 10 Total Pages: 17

ð     
G @ 1 @W @ 1 @W
pðx; zÞ þ rhmin ðx; zÞ ¼ þ dX; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð13aÞ
4pð1  Þ @x r @x0 @z r @z0
X

where
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r¼ ðx0  xÞ2 þ ðz0  zÞ2 : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð13bÞ

Eq. 13a relates the net fluid pressure in the fracture (Pa), p þ rhmin, to the fracture width in the y-coordinate direction (m), W, both of
which vary across the fracture domain. The net fluid pressure in the fracture varies across the domain caused by variation in both –p
and rhmin, the former resulting from the solution to a calculation coupled to slurry flow (detailed next) and the latter which is an input
to the calculation and is often assumed to increase linearly with depth across a specific interval. G is the shear modulus (Pa), and  is
the Poisson’s ratio of the reservoir rock.
The stress-intensity factor (Pam0.5), KI, is given by
sffiffiffiffiffiffiffiffiffiffiffi
G 2p
KI ¼ W ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð14Þ
4ð1  Þ rtip

where rtip is the distance to the fracture tip (m). The fracture propagation Dd (m) is governed by the difference between the stress-
intensity factor and rock toughness, as established by Mastrojannis et al. (1979),
   
KI  KIC
Dd ¼ max Ddmax ; 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð15Þ
KIC
where KIC is the rock toughness (Pam0.5).

Fluid and Proppant Transport. The overall conservation of mass of fluid and proppant is given by

@ðqs WÞ !
þ r  ðqs Whus iÞ þ qf qL ¼ 0; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð16Þ
@t
!
where qs and hus i are the slurry density (kg/m3) and average slurry velocity (m/s), respectively, and qL is the fluid leakoff (m/s) calcu-
lated from Carter’s leakoff coefficient. More detail covering Carter’s leakoff model is available in Howard and Fast (1957). Slurry prop-
erties apply to the carrier fluid and proppant together; that is, a continuum assumption is used. The slurry velocity is related to the
pressure gradient by the solution to the incompressible Navier-Stokes equation for a power-law fluid flowing between two infinite
plates. Using correlations for the rheological properties of a proppant-laden slurry developed by Shah (1993) that was based on experi-
ments, the transport of slurry can be more accurately modeled. Shah (1993) determined that both the flow-consistency index (Pasn) and
the flow-behavior index were functions of proppant concentration. Incorporation of the Navier-Stokes equation result in the previous
conservation-of-mass statement forms the pressure equation. The pressure equation is solved, coupled to the fracture-opening equation,
to yield the average slurry velocity, pressure, and width across the fracture domain.
The slurry density and slurry-mass flux are given by

qs ¼ ð1  cÞqf þ cqp : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð17Þ

and
! ! !
qs hus i ¼ ð1  cÞqf huf i þ cqp hup i: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð18Þ

The relative average velocity of the two phases, proppant and fluid, resulting from convection is given by the CFD/DEM-derived
correlation h(d/W). These each may have vector components in x- and z-coordinate directions. There is an additional vector component
in the z-coordinate direction for the proppant phase because of settling, given by the model suggested by Liu (2006). Rearrangement of
!
the definition of slurry-mass flux, h(d/W), and usettling allows solution of the average proppant velocity. With hup i known, the continuity
equation for proppant can be solved to determine proppant distribution across the fracture domain. The continuity equation for proppant
is given by
@ðqp cWÞ !
þ r  qp cWhup i ¼ 0: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð19Þ
@t

Hydraulic-Fracture Simulation Results


Overview of Hydraulic-Fracture Simulation Results. The fracturing simulator described in the previous section, including the CFD/
DEM-derived correlation for proppant transport with flow, can be used to design fracture treatments. For a given application, defined
by the in-situ stress profile, rock elastic moduli and toughness, and leakoff properties of the reservoir, the treatment can be optimized.
For two typical cases, the effect of several factors affecting proppant transport has been evaluated separately. That is, for each case, a
comparison has been made between the proppant distribution predicted by Stokes settling alone, Stokes settling and the CFD/DEM-
derived correlation for flow, corrected Stokes settling alone, and, finally, corrected Stokes settling and the CFD/DEM-derived correla-
tion for flow.

Hydraulic-Fracture Simulation Cases. A large barrier-stress contrast limits height growth, and was used in all hydraulic-fracture-
simulation cases. The minimum-horizontal-stress profile and the proppant schedule used are shown in Fig. 9.

10 2018 SPE Drilling & Completion

ID: jaganm Time: 17:17 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 11 Total Pages: 17

σhmin (psi)
5,000 6,000 7,000 8,000
9,750

9,800

9,850 0.71 psi/ft 2.00

Proppant Loading (ppa)


9,900 1.75
1.50
9,950 1.25

Depth (ft)
0.61 psi/ft 1.00
10,000
0.75
10,050 0.50
0.25
10,100
0.00
0 10 20 30 40 50
10,150 0.71 psi/ft Time (minutes)
10,200

10,250

Fig. 9—Minimum-horizontal-stress profile vs. depth (left) and proppant loading vs. time (right).

Table 2 includes all necessary inputs required by the hydraulic-fracture simulator described previously. When only one value for an
independent variable is included, it was used for all simulations.

Independent V ari able Value (SI uni ts) Value (field units)
Shear modulus, G 11.03, GPa 1.6, MMpsi
(Young’s modulus) 27.58, GPa 4, MMpsi
Poisson’s ratio, ν 0.25
0.5 0.5
Rock toughness, KIC 1.099, MPa·m 1,000, psi-in.
Height 60.96, m 200, ft
Proppant diameter, d 0.0003, m 40/70-US-mesh proppant
Proppant density, ρp
3 3
2650, kg/m 165, lbm/ft
Fluid density, ρf
3 3
1000, kg/m 62.4, lbm/ft
Fluid viscosity, μf 0.001, Pa·s 1-cp slickwater
3
Injection rate (biwing, hence half into single fracture) 0.03975, 0.06625, m /s 16.7, 25, bbl/min
0.5
3.935 × 10 , m/s
–6 0.5
Leakoff, CL 0.0001, ft/min

Table 2—Values of independent variables used in the hydraulic-fracture simulation study. MM 5 million.

Each of the two typical shale-reservoir cases examined specifies the same overall injection rate of 100 bbl/min. The difference
between the two cases is that the overall injection rate is evenly split between four or six clusters, resulting in 25 or 16.7 bbl/min
per biwing fracture, respectively. The 40/70-US-mesh proppant was included as per the schedule shown in Fig. 9. Slickwater is well-
characterized as a Newtonian fluid with a viscosity of 1 cp. The value used for Carter’s leakoff coefficient was 0.0001 ft/min0.5 in
all cases.

Hydraulic-Fracture Proppant Distribution. Final Proppant Distribution With Injection at 25 bbl/min. The proppant distribution
present at the completion of the pumping process for the case with an injection rate of 25 bbl/min into one biwing fracture is shown in
Fig. 10. One should note that the “effective-fracture length” (or “propped-fracture length”) has not been explicitly defined for this por-
tion of the discussion. There are several options by which such a quantity could be defined, but each requires the specification of a cut-
off value related to proppant distribution, which can only be selected in an arbitrary manner. Similar to effective-fracture length,
“effective-fracture height” has not been explicitly defined, because this would also require the specification of an arbitrarily selected
cutoff value related to proppant distribution. Rather, the proppant solid mass per unit area (lbm/ft2) presented in Fig. 10 uses a consist-
ent scale for all panels, allowing the comparison of any contour of interest. The comparisons detailed next have been made on the basis
of examining several contours, for which the trends between the results produced by the different proppant-transport models are found
to be consistent.
First, after comparing the top-left and top-right panels, it is apparent that the effective-fracture length is smaller if the Stokes settling
velocity is assumed vs. the Stokes settling velocity corrected with the correlation proposed by Liu (2006). The Liu (2006) correlation
reduces the settling velocity, compared with that calculated with Stokes’ law, and as a result, proppant is carried farther into the frac-
ture, producing a relatively longer effective-fracture length. The longer effective-fracture length is accompanied by smaller effective-
fracture height. Second, comparing the bottom-left and bottom-right panels, both of which include the CFD/DEM correlation for
dimensionless proppant velocity, the same trends for effective-fracture length and effective-fracture height are apparent. That is,
whether or not the CFD/DEM correlation is applied, the effective-fracture length is longer and the effective-fracture height is smaller if
the Liu (2006) correlation is assumed, compared with Stokes’ law for the settling-velocity calculation.

2018 SPE Drilling & Completion 11

ID: jaganm Time: 17:17 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 12 Total Pages: 17

9,500 9,500

9,600 9,600
9,700 Solid Mass per
Area (lbm/ft2)
9,700 Solid Mass per

True Vertical Depth (ft)

True Vertical Depth (ft)


Area (lbm/ft2)
9,800 3.000 9,800 3.000
1.750 1.750
1.021 1.021
9,900 0.596 9,900 0.596
0.348 0.348
0.203 0.203
10,000 0.118
0.069 10,000 0.118
0.069
0.040 0.040
10,100 0.024 0.024
0.014 10,100 0.014
Stokes 0.008 Corrected Stokes 0.008
10,200 10,200
10,300 10,300

10,400 10,400
10,500 10,500
0 100 200 300 400 500 600 700 800 900 1,000 0 100 200 300 400 500 600 700 800 900 1,000
Length (ft) Length (ft)
9,500 9,500

9,600 9,600
Solid Mass per
Area (lbm/ft2)
9,700 9,700
True Vertical Depth (ft)

True Vertical Depth (ft)


Solid Mass per 3.000
Area (lbm/ft2) 1.750
9,800 9,800 1.021
3.000 0.596
9,900 1.750 9,900 0.348
1.021 0.203
0.596 0.118
10,000 0.348 10,000 0.069
0.203 0.040
0.118 0.024
10,100 0.069 10,100
0.014
Stokes and CFD/DEM 0.040 Corrected Stokes and CFD/DEM 0.008
10,200 0.024 10,200
0.014
0.008
10,300 10,300

10,400 10,400
10,500 10,500
0 100 200 300 400 500 600 700 800 900 1,000 0 100 200 300 400 500 600 700 800 900 1,000
Length (ft) Length (ft)

Fig. 10—Final proppant distribution after injection at 25 bbl/min per biwing fracture. Top left: Stokes settling. Top right: Corrected
Stokes settling. Bottom left: Stokes settling and CFD/DEM correlation h(d/W). Bottom right: Corrected Stokes settling and CFD/
DEM correlation h(d/W).

Third, comparing the top-left and bottom-left panels, one finds negligible difference in the effective-fracture length and effective-
fracture height resulting from the inclusion of the CFD/DEM-derived correlation for dimensionless proppant velocity, while assuming
Stokes settling in both cases. Finally, assuming the Liu (2006) correlation for the settling velocity in place of Stokes settling, there is
negligible difference in the effective-fracture length and effective-fracture height observed with and without inclusion of the CFD/
DEM-derived correlation, seen in the top-right and bottom-right panels. That is, the CFD/DEM-derived correlation for dimensionless
proppant velocity has negligible effect on the final proppant distribution.
Fig. 10 presents contours of the proppant solid mass per unit area (lbm/ft2), and implicitly provides information on the fracture-
width distribution. The proppant solid mass per unit area can be calculated as the product of fracture width, proppant concentration, and
proppant density. Because the proppant density is identical between all simulations presented, it follows that a relatively larger proppant
mass per unit area results from an increase in the product of fracture width and proppant concentration. Although fracture width could
be presented by means of a contour plot, doing so does not convey any information regarding the proppant distribution, the focus of this
study. In this portion of the discussion, effective-fracture width has been defined as the product of fracture width, proppant concentra-
tion, and a factor of the inverse of the maximum proppant concentration. The maximum-allowable proppant concentration is 0.52, from
maximum packing considerations, and proppant concentration of this value results in the effective-fracture width equaling the fracture
width. Conversely, zero proppant concentration produces zero effective-fracture width. This definition allows an average with height to
be calculated and plotted vs. length and comparisons to be made between the results assuming different proppant-transport models. The
results have been presented in this manner in Fig. 11.
The average width of the effective fracture vs. length shown in Fig. 11 indicates that shorter lengths are predicted when Stokes set-
tling velocity is assumed, compared with application of the correlation proposed by Liu (2006). There is negligible difference between
effective-fracture lengths predicted assuming the CFD/DEM-derived correlation or that the dimensionless proppant velocity takes a
value of 1.0 across the entire range of the ratio of particle diameter to slot width.
One should note that the hydraulic-fracture simulator models only the main hydraulic fracture, following common engineering prac-
tice. Possible dilation of natural fractures is accounted for only implicitly through the Carter leakoff coefficient, which accounts for fluid
loss to the formation but does not account for the possibility of proppant transport in any natural-fracture network. In the main hydraulic
fracture, the particle-diameter/slot-width ratio is small over most of the domain. That is, the slot width is considerably larger than the
particle diameter. Because the CFD/DEM-derived correlation deviates from the common assumption of equal average proppant and
average fluid velocity, or a dimensionless proppant velocity of 1.0, only for particle-diameter/slot-width ratios greater than 0.4, it fol-
lows that behavior of the function in this range is unimportant for most of the hydraulic-fracture domain. The peak in dimensionless
proppant velocity at a particle-diameter/slot-width ratio of 0.8 is of greater importance for proppant transported in a dilated natural frac-
ture, where the slot width is often much smaller.

12 2018 SPE Drilling & Completion

ID: jaganm Time: 17:17 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 13 Total Pages: 17

0.1

Propped Width (in.)


0.01

0.001
0 100 200 300 400 500 600 700 800
Distance Along Fracture Length (ft)
Stokes Corrected Stokes
Stokes and CFD/DEM Corrected Stokes and CFD/DEM

Fig. 11—Average effective-fracture width vs. length with injection at 25 bbl/min.

Final Proppant Distribution With Injection at 16.7 bbl/min. One biwing fracture was simulated with an injection rate of
16.7 bbl/min. The proppant distribution present at the completion of the pumping process is shown in Fig. 12. Comparing all four
panels, there is negligible difference in the effective-fracture length and effective-fracture height predicted. That is, assuming Stokes
settling velocity for proppant or the Liu (2006) correlation and assuming the CFD/DEM-derived correlation for dimensionless proppant
velocity or that it is equal to 1.0 for all particle-diameter/slot-width ratios, the same results are produced when the injection rate is
smaller. This contrasts with the larger-injection-rate case. With an injection rate of 25 bbl/min, the proppant is carried farther into the
fracture, producing a longer effective-fracture length and smaller effective-fracture height, assuming the smaller settling velocity
proposed by Liu (2006).

9,700 9,700
Solid Mass per Solid Mass per
Area (lbm/ft2) Area (lbm/ft2)
9,800 9,800
True Vertical Depth (ft)

True Vertical Depth (ft)

3.000 3.000
1.750 1.750
1.021 1.021
9,900 0.596 9,900 0.596
0.348 0.348
0.203 0.203
10,000 0.118 10,000 0.118
0.069 0.069
0.040 0.040
0.024 0.024
10,100 0.014 10,100 0.014
Stokes 0.008 Corrected Stokes 0.008

10,200 10,200

10,300 10,300

0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800
Length (ft) Length (ft)

9,700 9,700

Solid Mass per Solid Mass per


9,800 Area (lbm/ft2)
9,800 Area (lbm/ft2)
True Vertical Depth (ft)

True Vertical Depth (ft)

3.000 3.000
9,900 1.750 9,900 1.750
1.021 1.021
0.596 0.596
0.348 0.348
10,000 0.203 10,000 0.203
0.118 0.118
0.069 0.069
10,100 0.040 10,100 0.040
0.024 0.024
Stokes and CFD/DEM 0.014 Corrected Stokes and CFD/DEM 0.014
0.008 0.008
10,200 10,200

10,300 10,300

0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800
Length (ft) Length (ft)

Fig. 12—Final proppant distribution after injection at 16.7 bbl/min per biwing fracture. Top left: Stokes settling. Top right:
Corrected Stokes settling. Bottom left: Stokes settling and CFD/DEM correlation h(d/W). Bottom right: Corrected Stokes settling
and CFD/DEM correlation h(d/W).

2018 SPE Drilling & Completion 13

ID: jaganm Time: 17:17 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 14 Total Pages: 17

Fig. 13 presents the average of the effective-fracture width over height vs. length using the same definition that was used in the cal-
culation of the results shown in Fig. 11.

0.1

Propped Width (in.)


0.01

0.001

0.0001
0 100 200 300 400 500 600
Distance Along Fracture Length (ft)
Corrected Stokes Stokes
Corrected Stokes and CFD/DEM Stokes and CFD/DEM

Fig. 13—Average effective fracture width vs. length with injection at 16.7 bbl/min.

There is negligible difference between the effective-fracture lengths predicted by either proppant settling velocity or dimensionless
proppant velocity considered.

Conclusions
The CFD/DEM simulation results demonstrate that the dimensionless proppant velocity is a function of the particle-diameter/slot-width
ratio only. Although simulation cases were completed with different volumetric particle concentrations, slot-flow Re, viscosities, parti-
cle sizes, and fluid leakoff values, only random scatter in the results was noted, and no systematic variation was identified. A peak in
the dimensionless proppant velocity of approximately 1.2 is observed at a particle-diameter/slot-width ratio of approximately 0.8.
The CFD/DEM results have been incorporated into a hydraulic-fracturing simulator using a simple correlation. Although the dimen-
sionless proppant velocity exhibits a peak at a particle-diameter/slot-width ratio of approximately 0.8, the main hydraulic-fracture
widths predicted are much larger than proppant particles, and the peak does not significantly change the effective-fracture lengths pre-
dicted. Proppant particles transported in dilated natural fractures, not modeled explicitly in this study, would be expected to be trans-
ported in the region where particle-diameter/slot-width ratio is large and the peak in the dimensionless proppant velocity
becomes significant.

Nomenclature
a ¼ unit-conversion constant
A ¼ cross-sectional area of simulation domain, m2
b ¼ constant
B ¼ half-slot width, m
c ¼ volumetric solids concentration, volume fraction
CL ¼ Carter leakoff coefficient, m/s0.5
d ¼ particle diameter, m
Fij ¼ interparticle force between particles i and j in Lagrangian coordinate system, N
Fi total ¼ total force acting on particle i, N
g ¼ gravitational acceleration, m/s2
G ¼ shear modulus, Pa
h ¼ function “h”
H ¼ height of simulation domain, m
Ii ¼ moment of inertia of particle i, kgm2
kn ¼ normal spring coefficient, N/m
kt ¼ tangential spring coefficient, N/m
KI ¼ Mode-I stress-intensity factor, Pam0.5
KIC ¼ rock toughness, Pam0.5
L ¼ length of simulation domain, m
m_ ¼ average particle mass rate in the x-coordinate direction, kg/s
mi ¼ mass of particle i, kg
Mi total ¼ total moment acting on particle i, Nm
nij ¼ normal unit vector between particles i and j in Lagrangian coordinate system
^ ¼ outward normal unit vector
n
p ¼ pressure, Pa
qL ¼ leakoff velocity, m/s
r ¼ distance, m
rtip ¼ distance to fracture tip, m
Re ¼ Reynolds number
t ¼ time, seconds

14 2018 SPE Drilling & Completion

ID: jaganm Time: 17:17 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 15 Total Pages: 17

tij ¼ unit tangential vector between particles i and j in Lagrangian coordinate system
tCs ¼ traction on Cs, N
uf ¼ fluid-velocity vector, m/s
ufx ¼ scalar x-component of fluid-velocity vector, m/s
ufx max ¼ maximum scalar x-component of fluid-velocity vector, m/s
ufy ¼ scalar y-component of fluid-velocity vector, m/s
ufz ¼ scalar z-component of fluid-velocity vector, m/s
<ufx> ¼ average fluid velocity in x-coordinate direction, m/s
<ufx>inlet ¼ average fluid velocity in x-coordinate direction at domain inlet, m/s
ui ¼ particle i velocity vector, m/s
!
hup i ¼ average proppant-phase velocity vector, m/s
<upx> ¼ average proppant velocity in x-coordinate direction, m/s
!
hus i ¼ average slurry velocity vector, m/s
uStokes ¼ Stokes’ settling velocity, m/s
usettling ¼ corrected Stokes’ settling velocity, m/s
W ¼ slot or hydraulic-fracture width, m
xi ¼ coordinates of center of gravity of particle i, m
xi ¼ x-coordinate of center of gravity of particle i, m
yi ¼ y-coordinate of center of gravity of particle i, m
zi ¼ z-coordinate of center of gravity of particle i, m
d ¼ overlap, m
cn ¼ normal damping coefficient, Ns/m
ct ¼ tangential damping coefficient, Ns/m
Dd ¼ fracture propagation, m
Dup ¼ relative speed, m/s
hi ¼ angular position of particle i, radians
lf ¼ fluid dynamic viscosity, Pas
 ¼ Poisson’s ratio
qs ¼ slurry density, kg/m3
qf ¼ fluid density, kg/m3
qp ¼ proppant density, kg/m3
r ¼ total stress tensor of the fluid, Pa
rh min ¼ minimum horizontal stress, Pa
Cs ¼ solid-domain surface
X ¼ fracture domain
Xs ¼ solid domain
dXfront ¼ fracture front
dXperf ¼ perforated interval

Acknowledgments
The authors acknowledge the Texas Advanced Computing Center at the University of Texas at Austin for providing high-performance
computing resources that have contributed to the research results reported within this paper.

References
Bird, R. B., Stewart, W. E., and Lightfoot, E. N. 2007. Transport Phenomena. New York: Wiley.
Blyton, C. A. J. 2016. Proppant Transport in Complex Fracture Networks. PhD dissertation, University of Texas at Austin.
Daneshy, A. A. 1978. Numerical Solution of Sand Transport in Hydraulic Fracturing. J Pet Technol 30 (1): 132–140. SPE-5636-PA. https://doi.org/
10.2118/5636-PA.
Gallegos, T. J. and Varela, B. A. 2015. Trends in Hydraulic Fracturing Distributions and Treatment Fluids, Additives, Proppants and Water Volumes
Applied to Wells Drilled in the United States From 1947 Through 2010—Data Analysis and Comparison to the Literature. United States Geological
Survey Scientific Investigations Report 2014–5131.
Ganatos, P., Pfeffer, R., and Weinbaum, S. 1980. A Strong Interaction Theory for the Creeping Motion of a Sphere Between Plane Parallel Boundaries.
Part 2: Parallel Motion. Journal of Fluid Mechanics 99 (4): 755–783. https://doi.org/10.1017/S0022112080000882.
Hager, A., Kloss, C., Pirker, S. et al. 2011. Efficient Realization of a Resolved CFD-DEM Method Within an Open Source Framework. Presented at the
Open Source CFD International Conference, Paris-Chantilly, France, 3–4 November.
Hager, A., Kloss, C., Pirker, S. et al. 2014. Parallel Resolved Open Source CFD-DEM: Method, Validation, and Application. Journal of Computational
Multiphase Flows 6 (1): 13–27. https://doi.org/10.1260/1757-482X.6.1.13.
Happel, J. and Brenner, H. 1965. Low Reynolds Number Hydrodynamics. New Jersey: Prentice-Hall.
Hertz, H. 1882. Ueber die Beruehrung Elastischer Koerper (On Contact Between Elastic Bodies). Available in English in Miscellaneous Papers. London:
MacMillan and Company Limited.
Ho, B. P. and Leal, L. G. 1974. Inertial Migration of Rigid Spheres in Two-Dimensional Unidirectional Flows. Journal of Fluid Mechanics 65 (2):
365–400. https://doi.org/10.1017/S0022112074001431.
Howard, G. C. and Fast, C. R. 1957. Optimum Fluid Characteristics for Fracture Extension. Drilling and Production Practice 24: 261–270. Conference
Paper API-57-261.
Jones, R. B. 2004. Spherical Particle in Poiseuille Flow Between Planar Walls. The Journal of Chemical Physics 121 (1): 483–500. https://doi.org/
10.1063/1.1738637.
Kloss, C., Goniva, C., Hager, A. et al. 2012. Models, Algorithms, and Validation for Opensource DEM and CFD-DEM. Progress in Computational Fluid
Dynamics, An International Journal 12 (2,3): 140–152. https://doi.org/10.1504/PCFD.2012.047457.
Kossecka, E. 1971. Defects as Surface Distribution of Double Forces. Archives of Mechanics 23: 481–494.

2018 SPE Drilling & Completion 15

ID: jaganm Time: 17:17 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 16 Total Pages: 17

Lee, W. J. and Holditch, S. A. 1981. Fracture Evaluation With Pressure Transient Testing in Low-Permeability Gas Reservoirs. J Pet Technol 33 (9):
1776–1792. SPE-9975-PA. https://doi.org/10.2118/9975-PA.
Liu, Y. 2006. Settling and Hydrodynamic Retardation of Proppants in Hydraulic Fractures. PhD dissertation, University of Texas at Austin.
Malhotra, S., Lehman, E. R., and Sharma, M. M. 2014. Proppant Placement Using Alternate-Slug Fracturing. SPE J. 19 (5): 974–985. SPE-163851-PA.
https://doi.org/10.2118/163851-PA.
Mastrojannis, E. N., Keer, L. M., and Mura, T. 1979. Stress Intensity Factor for a Plane Crack Under Normal Pressure. International Journal of Fracture
15 (3): 247–258. https://doi.org/10.1007/BF00033223.
Maude, A. D. and Whitmore, R. L. 1958. A Generalized Theory of Sedimentation. British Journal of Appl. Physics 9 (12): 477–482. https://doi.org/
10.1088/0508-3443/9/12/304.
Mindlin, R. D. and Derieswicz, H. 1953. Elastic Spheres in Contact Under Varying Oblique Forces. Journal of Applied Mechanics 20: 327–344.
Nolte, K. G. 1988. Fluid Flow Consideration in Hydraulic Fracturing. Presented at the SPE Eastern Regional Meeting, Charleston, West Virginia, 1–4
November. SPE-18537-MS. https://doi.org/10.2118/18537-MS.
O’Sullivan, C. and Bray, J. D. 2004. Selecting a Suitable Time Step for Discrete Element Simulations That Use the Central Difference Time Integration
Scheme. Eng. Comp. 21: 278–303. https://doi.org/10.1108/02644400410519794.
Peker, S. M. and Helvaci, S. S. 2008. Solid-Liquid Two Phase Flow. Amsterdam: Elsevier.
Ribeiro, L. H. and Sharma, M. M. 2013. A New 3D Compositional Model for Hydraulic Fracturing With Energized Fluids. SPE Prod & Oper 28 (3):
259–267. SPE-159812-PA. https://doi.org/10.2118/159812-PA.
Richardson, J. F. and Zaki, W. N. 1954. Sedimentation and Fluidization. Part 1. Trans. Inst. Chem. Eng. 32: 35–53. https://doi.org/10.1016/S0263-
8762(97)80006-8.
Shah, S. N. 1993. Rheological Characterization of Hydraulic Fracturing Slurries. SPE Prod & Fac 8 (2): 123–130. SPE-22839-PA. https://doi.org/
10.2118/22839-PA.
Sharma, M. M., Gadde, P., Sullivan, R. et al. 2005. Slick Water and Hybrid Fracturing Treatments: Some Lessons Learned. J Pet Technol 57 (3): 38–40.
SPE-0305-0038-JPT. https://doi.org/10.2118/0305-0038-JPT.
Sharma, M. M. and Manchanda, R. 2015. The Role of Induced Un-Propped (IU) Fractures in Unconventional Oil and Gas Wells. Presented at the SPE
Annual Technical Conference and Exhibition, Houston, 28–30 September. SPE-174946-MS. https://doi.org/10.2118/174946-MS.
Shirgaonkar, A. A., MacIver, M. A., and Patankar, N. A. 2009. A New Mathematical Formulation and Fast Algorithm for Fully Resolved Simulation of
Self-Propulsion. Journal of Computational Physics. 228 (7): 2366–2390. https://doi.org/10.1016/j.jcp.2008.12.006.
Staben, M. E., Zinchenko, A. Z., and Davis, R. H. 2003. Motion of a Particle Between Two Parallel Plane Walls in Low-Reynolds Number Poiseuille
Flow. Physics of Fluids 15 (6): 1711–1733. https://doi.org/10.1063/1.1568341.
Stokes, G. G. 1851. On the Effect of the Internal Friction of Fluids on the Motion of Pendulums. Trans. of the Cambridge Philosophical Society, Part II 9: 8–106.
Weller, H. G., Tabor, G., Jasak, H. et al. 1998. A Tensorial Approach to Computational Mechanics Using Object-Oriented Techniques. Computers in
Physics 12 (6): 620–631. https://doi.org/10.1063/1.168744.
Zoback, M. D. 2007. Reservoir Geomechanics. New York: Cambridge University Press.

Appendix A
The boundary conditions applied to the CFD domain are
• At the CFD inlet: fully developed slot Poiseuille flow,
(   )
absðy  BÞ 2 @p
ufx ðyÞ ¼ ufx max 1  ufy ¼ 0 ufz ¼ 0 ¼ 0; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-1aÞ
B @x

where ufx, ufy, and ufz (m/s) are the scalar components of the fluid-velocity vector and B is half the slot width. The expression for
ufx has been obtained analytically from conservation-of-momentum considerations for a Newtonian fluid.
• At the CFD outlet: constant pressure,
@ufx @ufy @ufz
¼0 ¼0 ¼ 0; p ¼ 0: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-1bÞ
@x @x @x
• At both the CFD front and CFD back boundaries: no slip condition for zero leakoff,
@p
ufx ¼ 0; ufy ¼ 0; ufz ¼ 0; ¼ 0: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-1cÞ
@x
• At both the CFD front and CFD back boundaries: constant leakoff to fracture walls,
@p
ufx ¼ 0 ufy jy¼0 ¼ qL ufy jy¼W ¼ qL ufz ¼ 0 ¼ 0: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-1dÞ
@x
• At the CFD top and CFD bottom boundaries: cyclic pairing,
ufx jz¼0 ¼ ufx jz¼6d ; ufy jz¼0 ¼ ufy jz¼6d ; ufz jz¼0 ¼ ufz jz¼6d ; pjz¼0 ¼ pjz¼6d : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-1eÞ

The cyclic pairing used for the top and bottom boundaries represents an infinitely tall slot. The initial conditions applied to the CFD
domain are appropriate for fully developed slot Poiseuille flow. As for the boundary conditions applied at the inlet, these have been
obtained analytically from conservation-of-momentum considerations and are
• Velocity distribution across the domain,
"  #
absðy  BÞ 2
ufx ðyÞ ¼ ufx max 1  ufy ¼ 0 ufz ¼ 0: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-2aÞ
B

• Pressure distribution across the domain,

18dlf ufx max x


pðxÞ ¼ 1 þ pjx¼9d : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-2bÞ
B2 qf 9d

16 2018 SPE Drilling & Completion

ID: jaganm Time: 17:17 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017


DC174973 DOI: 10.2118/174973-PA Date: 27-August-18 Stage: Page: 17 Total Pages: 17

Christopher A. J. Blyton is a geomechanics specialist at BP America. His current research interests include sand management,
hydraulic fracturing, and proppant transport. Blyton has authored or coauthored more than six journal articles and conference
proceedings. He holds a bachelor of engineering (with honors) degree in mechanical engineering from the University of
Wollongong, Australia, and MS and PhD degrees in petroleum engineering from University of Texas at Austin. Blyton is a member
of SPE.
Deepen P. Gala is a PhD-degree candidate at University of Texas at Austin. His research interests include hydraulic-fracturing
modeling and fracturing-fluid selection, coupling geomechanics with compositional reservoir simulation, minimizing damage
from interwell fracture interference, and designing gas huff ‘n’ puff IOR operations in tight oil reservoirs. Gala has authored or
coauthored more than five technical papers. He holds a BS degree in chemical engineering from the Institute of Chemical Tech-
nology, India. Gala is a member of SPE and has contributed to peer review for several SPE Journal papers.
Mukul M. Sharma is a professor and holds the “Tex” Moncrief Chair in the Department of Petroleum and Geosystems Engineering
at University of Texas at Austin where he has been for the past 33 years. He served as chair of the department from 2001 to 2005.
Sharma’s current research interests include hydraulic fracturing, oilfield water management, formation damage, and improved
oil recovery. He has published more than 400 journal articles and conference proceedings, and holds 21 patents. Sharma
founded Austin Geotech Services, an E&P consulting company, in 1996, and cofounded Layline Petroleum and Karsu Petroleum,
private E&P companies, in 2006. He holds a bachelor’s of technology degree in chemical engineering from the Indian Institute
of Technology and MS and PhD degrees in chemical and petroleum engineering from the University of Southern California.
Sharma is a member of the US National Academy of Engineering; he is the recipient of the 2017 SPE John Franklin Carll Award,
the 2009 SPE Anthony F. Lucas Gold Medal, the 2004 SPE Faculty Distinguished Achievement Award, the 2002 SPE Lester C. Uren
Award, and the 1998 SPE Formation Evaluation Award. He served as an SPE Distinguished Lecturer in 2002–2003, has served on
the editorial boards of many journals, and has taught and consulted for industry worldwide.

2018 SPE Drilling & Completion 17

View publication stats


ID: jaganm Time: 17:17 I Path: S:/DC##/Vol00000/180017/Comp/APPFile/SA-DC##180017

You might also like