Energies 15 01967

energies
Article
Mathematical Modeling of Hydraulic Fracture Formation and
Cleaning Processes
Nickolay Smirnov 1, *, Kairui Li 2 , Evgeniya Skryleva 1 , Dmitriy Pestov 1 , Anastasia Shamina 1 , Chengzhi Qi 2
and Alexey Kiselev 1
1 Scientific Research Institute for System Analysis of the Russian Academy of Sciences,
117218 Moscow, Russia; jennyne@yandex.ru (E.S.); dmitr-ey94@mail.ru (D.P.);
anashamina90@mail.ru (A.S.); akis2006@yandex.ru (A.K.)
2 School of Civil and Transportation Engineering, Beijing University of Civil Engineering and Architecture,
Beijing 100044, China; kerryli1989@yandex.ru (K.L.); qichengzhi65@163.com (C.Q.)
* Correspondence: mech.math.msu@inbox.ru
Abstract: The effectiveness of the hydraulic fracturing procedure is crucially dependent on the stage
of fracture planning and design. Forecasting fracture behavior in rock formations characterized by
non-uniform toughness is a serious challenge. In the present paper, a planar-3D model considering the
rock’s non-uniform fracture toughness has been developed for the uneven propagation of a hydraulic
fracture. The series of numerical experiments were designed to study the effect of inhomogenous
fracture toughness. The results show that the fracture toughness contract significantly controls the
overall direction of fracture propagation, and a combination of toughness contrast and the proportion
between the pay zone and barrier zone determine the fracture profile: from almost circular with or
without a pair of narrow wedges when the proportion is small to almost rectangular otherwise. This

paper also discusses the process of cleaning a fracture from hydraulic fracturing fluid by oil. Using
Citation: Smirnov, N.; Li, K.; numerical modeling on the basis of the constructed mathematical model, a relationship is established
Skryleva, E.; Pestov, D.; Shamina, A.; between the quality of hydraulic fracture cleaning and the geometrical parameters of the fracture and
Qi, C.; Kiselev, A. Mathematical the region filled with the hydraulic fracturing fluid. The results of numerical experiments show that
Modeling of Hydraulic Fracture while fracturing fluid is more viscous than oil, the length of the fracture has a greater influence on the
Formation and Cleaning Processes. cleaning process than the viscosity of the fracturing fluid.
Energies 2022, 15, 1967. https://
doi.org/10.3390/en15061967 Keywords: non-uniform fracture toughness; hydraulic fracture; planar-3D model; comparative
Academic Editors: Vamegh Rasouli, analysis; seepage flows; fracture cleanup
Fred Aminzadeh, Uchenna Odi and
Ali Rezaei
Received: 21 January 2022

1. Introduction
Accepted: 2 March 2022
Published: 8 March 2022
With the increasing demand for fossil fuels (oil and gas) and the depletion of conven-
tional oil and gas reservoirs, unconventional reservoirs have attracted great attention and
Publisher’s Note: MDPI stays neutral
now occupy a large proportion of the hydrocarbon supply [1,2]. Previously, unconventional
with regard to jurisdictional claims in
reservoirs were difficult for exploitation due to their extremely low permeability and con-
published maps and institutional affil-
ductivity [3,4]. Hydraulic fracturing is an efficient technique for improving unconventional
iations.
reservoirs’ permeability and conductivity via hydraulic fractures [5]. In the hydraulic
fracturing design fracture geometry (length, width and height) is a critical target, which is
determined by two “controllable” variables: hydraulic fluid viscosity and wellbore pump-
Copyright: © 2022 by the authors.
ing rate, and four “natural” variables: in situ stress, reservoir’s stiffness (Young’s modulus
Licensee MDPI, Basel, Switzerland. and Poisson’s ratio), fluid loss (porosity and permeability) and fracture toughness [6].
This article is an open access article Hydraulic fluid viscosity and wellbore pumping rate are controlled artificially [7–9], so
distributed under the terms and they can be easily studied and their working mechanism is relatively clear. The remaining
conditions of the Creative Commons four “natural” variables are mainly controlled by geological conditions, which have strong
Attribution (CC BY) license (https:// randomness and complexity. Therefore, the attention of researchers was mainly focused on
creativecommons.org/licenses/by/ studying the effects of these four “natural” variables on fracture geometry and propagation.
4.0/). The sustainability of the hydraulic fracturing procedure is crucially dependent on the
Energies 2022, 15, 1967. https://doi.org/10.3390/en15061967 https://www.mdpi.com/journal/energies

Energies 2022, 15, 1967 2 of 35
stage of fracture planning and design. Forecasting fracture behavior in rock formations
characterized by non-uniform toughness is a serious challenge.
Sedimentary rocks around the reservoir usually exhibit layered homogeneity due to
the deposition and compaction processes [10,11]. Different layers have their own stiffness,
porosity and permeability, fracture toughness as well as in situ stress. This situation causes
significant contrasts of the four “natural” variables between layers, especially between the
pay zone and the adjacent layers. In the early stages of fracturing studies, Warpinski [12,13],
conducted experiments to determine which of the two parameters (stiffness contrast and
in situ stress contrast) is predominant for controlling hydraulic fracture containment.
Experiments demonstrated that the in situ stress contrast is the most important factor
controlling fracture height, while the stiffness contrast has little effect. This conclusion
is consistent with numerical and semi-analytical results in the paper [14]. In contrast
to conclusions in [12–14], papers [6,15] gave quantitative descriptions of the effects of
formation stiffness (Young’s modulus and Poisson’s ratio) contrast on hydraulic fracture
propagation. In [16,17], pseudo-3D models were presented for hydraulic fracture growth
in a layered formation with contrasts in both stiffness and in situ stress. In addition, some
attention was paid to studying the effects of fluid loss (porosity and permeability) on
hydraulic fracture propagation. Warpinski et al. [13] presented a qualitative conclusion that
permeability and pore pressure have an important effect on hydraulic fracture containment.
Hanson et al. [18] developed theoretical and numerical models and indicated that the
increase in the pore pressure tends to reduce the tendency for crack extension, and the
fracture growth could be impeded in areas where the permeability is large. In papers [19,20],
the models for hydraulic fracture propagation in an inhomogeneous poroelastic medium
were proposed for demonstrating the influence of reservoir permeability on the dynamics
of fracture propagation. Gao et al. [21] established a 2D fluid–solid coupled model to
study the hydraulic fracture propagation in a non-uniform pore pressure field. The results
indicated that the distributions of in situ stress and pore pressure are gradually altered
with the fracture propagation, and these alterations inversely affect subsequent fracture
growth. Furthermore, different models and algorithms were proposed for the combined
effects of the four “natural” variables on hydraulic fracture propagation in a layered
medium [17,22–25].
In the above descriptions of the effects of the four “natural” variables (in situ stress,
stiffness, fluid-loss and fracture toughness) on hydraulic fracture propagation in a multi-
layered medium there is a common feature: ignoring or weakening the influence of fracture
toughness parameter, because most researchers think that fracture toughness has a neg-
ligible effect on hydraulic fracturing, and it shows a relatively small variation range in
unconventional reservoirs [26]. In fact, relatively accurate distributions of fracture tough-
ness in the layered media of reservoirs are very rare, because the measurement methods for
fracture toughness in conditions of fracture growing under complex geological conditions
are still immature [27–30]. Thiercelin et al. [31,32] performed experiments and numerical
studies on the influence of fracture toughness in a homogeneous reservoir as well as on the
influence of fracture toughness contrast in a layered medium. The results show that fracture
toughness has a significant effect on fracture propagation not only in the homogeneous
formation, but also in the layered medium. When the contrast in fracture toughness is
sufficiently high, the arrest or diversion of fracture may be obtained. It was also pointed
out in [32], that fracture toughness is a critical mechanical property influencing hydraulic
fracture propagation, especially in the cases where the in situ stress contrast is small, the
fluid is of low viscosity, and the fracture is relatively small.
Since fracture toughness is not the main factor in the above four “natural” variables,
and the data of fracture toughness in the layered medium are difficult to measure, there are
very few studies of the effect of fracture toughness contrast. In order to better understand
the mechanism of the effect of non-uniform fracture toughness on hydraulic fracturing,
especially in the early stage when the fracture is small, in this paper we established a
plane-3D model considering an inhomogeneous fracture toughness for hydraulic fracture
Energies 2022, 15, x FOR PEER REVIEW 3 of 36
Energies 2022, 15, 1967 3 of 35
a plane-3D model considering an inhomogeneous fracture toughness for hydraulic frac-

ture propagation. In this model, the three-dimensional deformation of the fracture is cou-
propagation.
pled In this model, the
with the two-dimensional three-dimensional
flow of hydraulic fluid, deformation
and the two-of the fracturepropaga-
directional is coupled
with the two-dimensional flow of hydraulic fluid, and the two- directional propagation
tion in the fracture plane is described in the framework of linear elastic fracture mechanics
inin the
the fracture plane
cylindrical is described
coordinate system.inBesides,
the framework of linear elastic
a very small-time step isfracture
appliedmechanics
for captur-in
the cylindrical coordinate system. Besides, a very small-time step is applied
ing the uneven propagation on different locations of the fracture front. Furthermore, for capturing
the uneven propagation on different locations of the fracture front. Furthermore,
many numerical experiments have been carried out for quantitative analysis of the effect many
numerical experiments have been carried out
of fracture toughness on hydraulic fracture propagation.for quantitative analysis of the effect of
fracture toughness on hydraulic fracture propagation.
2. Mathematical Model for Hydraulically Driven Fracture Propagation
2. Mathematical Model for Hydraulically Driven Fracture Propagation
Consider a vertical hydraulic fracture across a horizontal wellbore in a layered me-
Consider a vertical hydraulic fracture across a horizontal wellbore in a layered medium,
dium, as illustrated in Figure
as illustrated in Figure 1. For1.clarity
For clarity of discussion
of discussion on theon the effect
effect of inhomogeneity
of inhomogeneity of
of fracture
fracture toughness, the medium is assumed here to be impermeable and
toughness, the medium is assumed here to be impermeable and other properties of the other properties
ofmedium,
the medium, Young’s
Young’s modulus
modulus 𝐸, Poisson’s
E, Poisson’s ν and𝜈inand
ratio ratio situinstress
situ stress 𝜎 are assumed
σ0 are assumed uniform
uniform and homogeneous, except the fracture toughness. Hydraulic fluid is
and homogeneous, except the fracture toughness. Hydraulic fluid is injected at a constant injected at a
constant rate 𝑄 and is assumed to follow a Newtonian rheology
rate Q and is assumed to follow a Newtonian rheology with a viscosity µ. with a viscosity 𝜇.
0
Figure
Figure AA
1. 1. planar-3D
planar-3D hydraulic
hydraulic fracture
fracture across
across a horizontal
a horizontal wellbore
wellbore in in a medium
a medium with
with inhomoge-
inhomoge-
neous fracture
neous fracturetoughness.
toughness.
2.1. Governing Equations

2.1. Governing Equations
For establishing the planar-3D model for hydraulic fracture propagation in an im-
For establishing the planar-3D model for hydraulic fracture propagation in an imper-
permeable medium we consider: the three-dimensional deformation of fracture, the two-
meable medium we consider: the three-dimensional deformation of fracture, the two-di-
dimensional flow of hydraulic fluid in the fracture, and the two directional propagations
mensional flow of hydraulic fluid in the fracture, and the two directional propagations in
in the inhomogeneous medium. This corresponds to the planar 3D approach described in
the inhomogeneous medium. This corresponds to the planar 3D approach described in
the introductory section, in which the three-dimensional problem of hydraulically driven
the introductory section, in which the three-dimensional problem of hydraulically driven
fracture propagation in a non-uniform medium is treated under the assumption that the
fracture propagation in a non-uniform medium is treated under the assumption that the
fracture propagates within a definite plane and is symmetrical in respect to that plane.
fracture propagates within a definite plane and is symmetrical in respect to that plane.
This is an assumption, which is supported by the majority of test cases, nevertheless, it is
not always the case. This assumption provides the possibility for a serious simplification
Energies 2022, 15, 1967 4 of 35
of the mathematical model regarding the fracture propagation and fluid flow within the
frame of a 2-D approximation thus reducing the dimension of the problem and reducing
essentially the computer simulation time. On the other hand, the stress-strained state of
the medium surrounding the fracture is treated in a full scale 3-D statement. Given the
elastic properties, toughness and confining stress of the medium, the fluid properties, the
injection rate and the initial fracture, the solution of fracture propagation is composed of
the fluid pressure p f ( x, y, t), the fracture width, namely fracture opening w( x, y, t) and the
projection area Ω(t) of the fracture on the plane xoy.
2.1.1. Elastic Deformation

For the planar fracture propagation, namely mode I fracture, the fluid pressure in the
fracture can be expressed by the following integral equation [33]
E0 w( x 0 , y0 , t)dΩ( x 0 , y0 )
Z
p( x, y, t) = p f ( x, y, t) − σ0 = − 3/2
, (1)
8π Ω(t)
[( x 0 − x )2 + (y0 − y)2 ]
where p( x, y, t) is the net pressure in the fracture, E0 = E/ 1 − ν2 is the plane strain

elastic modulus.
2.1.2. Fluid Flow

According to the lubrication theory and Poiseuille’s law, the generalized continuity
equation without considering the fluid loss is
∂w → → w2
+ div w u = 0, u = − ∇ p, (2)
∂t 12µ
→
where u is the fluid velocity inside the fracture, ∇ p is the net pressure gradient.
2.1.3. Boundary Conditions

Since the fluid is injected into the fracture with a constant rate Q0 , the fluid flux at the
wellbore is formulated as
Q0
q 0 = w0 u 0 = , (3)
2πr0
where w0 and u0 are the fracture width and fluid velocity at the wellbore, respectively, and
r0 is the wellbore radius.
According to the linear elastic fracture mechanics and the assumption of zero fluid
lag, the fracture width and the fluid flux at the fracture front equal zero, namely
w(sc , t) = 0, q(sc , t) = 0, (4)
where sc is the boundary curve of the projection area Ω(t) of the fracture on the plane xoy.
The additional boundary condition at the fracture front is given by the typical fracture
propagation criterion:
K I (sc , t) < K IC (sc ) (5)
where K I (sc , t) is the stress intensity factor with respect to parameters time and fracture
front coordinates, K IC (sc ) is the fracture toughness with respect to fracture front coordinates
because of inhomogeneity.
2.1.4. Initial Conditions

It is generally known that fracture initiation is a complex process [34,35], which is
not the focus of our model. To simplify the model, we make an assumption that at the
initial time of simulation (t0 = 0) the fracture has a small penny-like shape with a radius
R0 and is opened by fluid pressure with a uniform distribution p0 (Figure 2), therefore, the
cylindrical coordinate system rθz is chosen for better describing the penny-shape fracture
time of simulation (𝑡 = 0) the fracture has a small penny-like shape with a radius 𝑅 and
Energies 2022, 15, 1967 is opened by fluid pressure with a uniform distribution 𝑝 (Figure 2), therefore, 5 of the
35 cy-
lindrical coordinate system 𝑟𝜃𝑧 is chosen for better describing the penny-shape fracture
evolution, especially the two-dimensional fluid flow. In the cylindrical coordinate system,
for evolution,
formulating the layered
especially homogeneousfluid
the two-dimensional fracture
flow.toughness of formation
In the cylindrical we assume
coordinate system, that
for formulating
fracture toughnesstheislayered homogeneous
only related to the angular toughness of𝜃,
fracture coordinate formation
namely we assume that
fracture toughness is only related to the angular coordinate θ, namely
𝐾 = 𝐾 (𝜃) (6)
K IC = K IC (θ ) (6)
and 𝐾 (𝜃)=𝐾 (𝜃+𝜋) (0 ≤ 𝜃 ≤ 𝜋). For example, in Figure 2 the thick red line drawn across
theand
point
K ICsource
(θ ) = Krepresents a≤
IC ( θ + π ) (0 homogeneous rock layer
θ ≤ π). For example, with a2uniform
in Figure the thickfracture toughness:
red line drawn
(0)=𝐾the
𝐾 across (𝜋).
point source represents a homogeneous rock layer with a uniform fracture
toughness: K IC (0) = K IC (π ).
Figure
Figure 2. The
2. The initial
initial conditionsof
conditions of the
the model
modeland
androck
rockmechanical
mechanicalproperties.
properties.
2.2. Model in Cylindrical Coordinate System
2.2. Model in Cylindrical Coordinate System
The governing Equation (1) of fracture deformation can be easily transformed into
theThe governing
cylindrical Equation
coordinate (1) of
system fracture
with the helpdeformation can bebetween
of the relationship easily transformed
two systems: into
thexcylindrical
= r cos θ, ycoordinate
= r sin θ. It system
should bewith
notedthethat
help
weof the
can relationship
solve between
this equation two systems:
in the Cartesian
𝑥 =coordinate
𝑟 cos𝜃, 𝑦system
= 𝑟 sin𝜃.
xoy It should
first, be noted
and then thatthe
transform wefinal
can results
solve this
into equation in the
the cylindrical Cartesian
system
coordinate system 𝑥𝑜𝑦 first, and then transform the final results into the cylindrical sys-
instead of directly substituting the transformation relationship into Equation (1) for solving
temit instead
in the system roθ. The
of directly purpose ofthe
substituting this is to simplify the
transformation calculationinto
relationship greatly, because
Equation (1) for
Equation (1) is a special hyper singular integral expression, which can be calculated easier
solving it in the system 𝑟𝑜𝜃. The purpose of this is to simplify the calculation greatly,
in the Cartesian system xoy than in the polar system roθ.
because Equation (1) is a special hyper singular integral expression, which can be calcu-
lated easier in the Cartesian system 𝑥𝑜𝑦 than in the polar system 𝑟𝑜𝜃.
The governing Equation (2) can be transformed into the polar system 𝑟𝑜𝜃 as
Energies 2022, 15, 1967 6 of 35
The governing Equation (2) can be transformed into the polar system roθ as
1 1 ∂ w3 ∂p

∂w 1 1 ∂ ∂p
− 0 rw3 − 0 = 0, (7)
∂t µ r ∂r ∂r µ r ∂θ r ∂θ
where µ0 = 12µ. In this equation the radial and angular fluid flows in the fracture are
considered:
w2 ∂p w2 1 ∂p
ur = − 0 , uθ = − 0 . (8)
µ ∂r µ r ∂r
The boundary conditions at the wellbore and at the fracture front are rewritten as
w03 ∂p Q0
q0 = − = , r = r0 , (9)
µ0 ∂r 2πr0
3
q = − wµ0 ∂r = 0, r = R(θ )
∂p
, (10)
w ( R ( θ ), t ) = 0
where R(θ ) is the fracture radius with respect to θ, because the fracture has different radii
at different radial directions.
3. Numerical Scheme
The closed equation system of this model is composed of Equations (1), (7), (9) and
(10). The unknowns net pressure p(r, θ, t) and fracture opening w(r, θ, t) are coupled by
governing Equations (1) and (7) in so complex form that only the implicit scheme can
be used to solve this system. Furthermore, it should be noted that Neumann boundary
conditions (Equations (9) and (10)) require a special consideration [36]: the boundary
pressure at the fracture front p( R, θ, t) as an estimated known variable has to enter into the
system for a solution, and then the accurate numerical solution of p( R, θ, t) will be obtained
by iteration according to the law of conservation of fluid volume:
Z
Q0 t = wdΩ. (11)
Ω(t)
3.1. Fracture Discretization

The fracture plane is discretized with a fixed polar grid using triangular and trape-
zoidal elements (Figure 3). ∆r and ∆θ are the grid step lengths in the r and θ directions
respectively. The main physical parameters (location coordinates, fracture opening and
fluid pressure) are evaluated at the element centers of gravity. Besides, the elements with
j
the same angular coordinate θ j have the same fracture toughness K IC , just as the elements
passed by the red dotted line in Figure 3. The boundary elements marked with black points
j
contain an extra parameter, namely the stress intensity factor K I . If at a certain moment
j j
there are some boundary elements satisfying the condition K I ≥ K IC , the fracture front
where these boundary elements are located would propagate one element forward, see the
new elements with red dotted curves at directions θ j+1 and π + θ j+1 .
Energies 2022, 15, 1967 7 of 35
Figure 3. The discretized polar grid for the fracture. Black points–boundary grid nodes of the fracture
Figure
front, red3.points–internal
The discretizedpolar
grid grid
nodes, for
blue the fracture. of
points–nodes Black points–boundary
gravity gridand
center of elements nodes of the frac-
the green
tureisfront,
point red points–internal
the wellbore. ri , θ j and grid
ri , θ jnodes,
+1 are blue points–nodes
the elements’ of gravity
coordinates. center“iof” elements
Subscripts and the
and “j ” are
green point is the wellbore. (𝑟 , 𝜃 ) and (𝑟 , 𝜃 ) are the elements’ coordinates.
the element number in the r and θ directions, in which j = 1, 2, 3, · · · , n, and n = 2π/∆θ. Subscripts “𝑖” and
“𝑗” are the element number in the 𝑟 and 𝜃 directions, in which𝑗 = 1, 2, 3, ⋯ , 𝑛, and 𝑛 = 2𝜋⁄𝛥𝜃.
3.2. Discretization of Equation System
3.2.As
Discretization of Equation
already mentioned System 2.2, Equation (1) is solved in the Cartesian system
in Section
first, and
As then the obtained
already mentionedresults are transformed
in Section into the
2.2, Equation (1) polar system.
is solved ThisCartesian
in the can greatly
system
simplify the calculation process, so Equation (1) is discretized as
first, and then the obtained results are transformed into the polar system. This can greatly
simplify the calculation process, 0soi=Equation
m j=n (1) is discretized as
E
∑ ∑

p( x, y) = I x, y, xi , y j w xi , y j , (12)
8π
𝑝(𝑥, 𝑦) = i=1 j=1 𝐼(𝑥, 𝑦, 𝑥 , 𝑦 )𝑤(𝑥 , 𝑦 ), (12)

where
where ( x,(𝑥,
y) 𝑦)
are are
coordinates of any
coordinates of fracture element
any fracture in Cartesian
element xi , y j are
system,system,
in Cartesian (𝑥 coordi-
, 𝑦 ) are co-
nates of each fracture element in the Cartesian system. m is the element number in the r
ordinates of each fracture element in the Cartesian system. 𝑚 is the element number in
direction of the farthest fracture element away from the wellbore, n = 2π/∆θ. The stiffness
the 𝑟 direction of the farthest fracture element away from the wellbore, 𝑛 = 2𝜋⁄𝛥𝜃. The
coefficient I x, y, xi , y j is formulated as
stiffness coefficient 𝐼(𝑥, 𝑦, 𝑥 , 𝑦 ) is formulated as

A xi , 𝐴(𝑥
yj , 𝑦 )
𝐼(𝑥,
I x, y, xi ,𝑦, , 𝑦 −) = −
yj𝑥 = 2 3/2 ⁄
(13)
(13)
x )2−+𝑥)y j +
[( xi −(𝑥 − (𝑦
y −] 𝑦)
Energies 2022, 15, 1967 8 of 35

where A xi , y j is the area of fracture element xi , y j . The coordinates of fracture elements
in the two kinds of coordinate systems can be transformed into each other with the help of
the following relationship:
x = r cos θ, y = r sin θ, xi = ri cos θ j , y j = ri sin θ j . (14)
According to the finite volume method, Equation (7) for fluid flow can be discretized
as follows
∆t 1 1
∆w = rw3 ∂r − rw3 ∂p
∂p
µ0 ri ∆ri ∂r +
i +1/2,j,k i −1/2,j,k
3 (15)
∆t 1 1 3
w ∂p
µ0 ri ∆θ j r ∂θ i,j+1/2,k − wr ∂p
∂θ ,
i,j−1/2,k
where ∆w = wi,j,k+1 − wi,j,k represents the increment of fracture opening of the element

ri , θ j at the moment t = tk+1 . Subscript “k” is the time step number.
Boundary conditions (9) and (10) are transformed as
Q0 µ 0

∂p
rw3 =− (16)
∂r 1−1/2,j 2π

rw3 ∂r
∂p
= 0,
m+1/2,j (17)
wm+1/2,j = 0,
where subscripts “1 − 1/2” and “m + 1/2” in the r direction represent the inner boundary
node at the wellbore and the outer boundary node on the fracture front, respectively. For
the utilization of Equation (17), the pressure at the fracture front pm+1/2,j should enter into
Equations (12), (15)–(17) as an estimated known parameter [36], and then be solved by
iteration according to Equation (11).
The stress intensity factor K I in the three-dimensional model can be solved by its
definition: the limit of the product of the stress near the fracture tip and the root of the
distance from the fracture tip. Bui [37], proposed an equivalent method for solving K I in
the three-dimensional model with a precondition that the discontinuous displacement near
the fracture tip is known. Its discrete form is
E0
r
π wm,j
KI θj = p < K IC θ j , (18)
4 2 R j − rm
where wm,j and rm are the fracture opening and radial coordinate of the nearest element
away from the fracture front in each direction θ = θ j respectively. R j is the fracture radius
in the direction θ = θ j .
3.3. Implicit Algorithm

The discretized model consists of Equations (12), (15)–(17) with the tip condition of
stress intensity factor (18). Firstly, substituting Equation (12) into (15)–(17), we obtain an
implicit nonlinear equation system for the unknowns wi,j,k+1 with the highest fourth power,
which can be solved by the Newtonian iteration method. Then, the obtained values of
wi,j,k+1 must satisfy the law of fluid volume conservation (11): if condition (11) is satisfied,
proceed into the second step; otherwise, further iteration is needed for wi,j,k+1 until the
condition of fluid volume conservation (11) is satisfied.

Secondly, the stress intensity factors K I θ j in all directions θ can be calculated
(Equation (18)) using the obtained values of wi,j,k+1 . If all stress intensity factors satisfy the

condition K I θ j < K IC θ j , it means that the values of wi,j,k+1 are the final solutions of the
model at the moment t = tk+1 and the third step for solving pi,j,k+1 starts, otherwise, it

means that fracture front where stress intensity factors satisfy K I θ j ≥ K IC θ j propagates,
and the fracture planar area with the fracture elements is updated. Therefore, the entire
Energies 2022, 15, 1967 9 of 35
propagates, and the fracture planar area with the fracture elements is updated. Therefore,
the entire calculation area for the model is updated, and the calculation needs to return to
calculation
the first stepareafor for
newthe model
values ofis𝑤updated,
, , , untiland allthe calculation
stress intensityneeds 𝐾 (𝜃 ) to
to return
factors the first
satisfy the
step for new
condition values
(18). of wi,j,k
The final values of 𝑤all
+1 , until , ,
stress as intensity
well as thefactors
new K
fracture
I θ j satisfy
front the condition
satisfying the
(18). The final values
conditions (11) and (18) areof w +1 numerical solutions of the model at the moment 𝑡 = 𝑡
i,j,kthe as well as the new fracture front satisfying the conditions .
(11) and (18) are
Thirdly, thethe numerical
obtained solutionsvalues
numerical of the modelof 𝑤 , ,at thearemoment
inversely t = tsubstituted
k +1 . into
Thirdly,
Equation (12)the
forobtained numerical
the solutions of 𝑝 , values . Atoflast,wi,j,k
the are inversely
+1 calculation substituted
enters into theinto
nextEqua-
time
,
tion (12) for the solutions
step 𝑡 = 𝑡 , and solutions 𝑤i,j,k of p + . At
1 and 𝑝 last, the calculation
are regarded enters
as into
the the next
solutions time
of the step
pre-
, , , ,
= tk+time
tvious 2 , and solutions
step. w i,j,k + 1 and p i,j,k + 1 are regarded as the solutions of the previous
time step.
4. Numerical Experiments
4. Numerical Experiments
4.1. Experiment
4.1. Experiment Parameters
Parameters
For numerical
For numerical calculation
calculationthe themain
mainparameters
parametersare aregiven
givenasasfollows:
follows: Young’s
Young’s modu-
modulus
lus = 30 𝐺𝑃𝑎, Poisson’s ratio 𝜈 = 0.2,
= 30 GPa, Poisson’s ratio ν = 0.2, the injection rate at the wellbore Q0 = 10 𝑚
the injection rate at the wellbore 𝑄 = 10 − 4 m3⁄/s,
𝑠,
fluid viscosity 𝜇 = 10 − 𝑃𝑎
3 ⋅ 𝑠. The initial conditions of the fracture are
fluid viscosity µ = 10 Pa·s. The initial conditions of the fracture are as follows: the as follows: the in-
itial radius
initial radius𝑅 R=0 0.5 𝑚, the
= 0.5 initial
m, the pressure
initial pressure𝑝 =p0.1 𝑀𝑃𝑎. Figure 4 shows the fixed grid of
0 = 0.1 MPa. Figure 4 shows the fixed
∘ ∘⁄ ∘
fracture,
grid grid parameters
of fracture, Δ𝑟 = 0.02
grid parameters ∆r 𝑚= and
0.02 Δ𝜃 = 20∆θ
m and (𝑛==20360◦ (n =20360= ◦18).
/20◦The= symmet-
18). The
rical yellow areas
symmetrical yellowrepresent one layer,
areas represent onei.e., thei.e.,
layer, paythe
zonepaywith
zonea with
fracture toughness
a fracture in the
toughness
horizontal
in direction,
the horizontal the other
direction, the white areas represent
other white the other
areas represent the layers, i.e., thei.e.,
other layers, barrier layers
the barrier
with another
layers fracture
with another toughness.
fracture toughness.
Figure 4. The discrete grid of initial fracture and the locations of the pay zone and the barrier layers.
Figure 4. The discrete grid of initial fracture and the locations of the pay zone and the barrier layers.
In this paper we want to study the effects of the fracture toughness contrast of the
two different layers and the relative width of the pay zone, i.e., the proportion of the
yellow areas on fracture propagation in early time, so the fracture toughness contrast and
the proportion of the yellow areas are selected as the governing variables of numerical
In this paper we want to study the effects of the fracture toughness contrast of the
two different layers and the relative width of the pay zone, i.e., the proportion of the yel-
Energies 2022, 15, 1967 low areas on fracture propagation in early time, so the fracture toughness contrast and 10 ofthe
35
proportion of the yellow areas are selected as the governing variables of numerical exper-
iments. The crossover experiment strategy is adopted for quantitative analysis of the ef-
fects of these The
experiments. two crossover
governingexperiment
variables. strategy is adopted for quantitative analysis of the
In Figure 4 the proportion
effects of these two governing variables. of the yellow areas involving only one pair of radial grids
is 𝜉 In ⁄9. According
= 1Figure 4 the proportion of thediscretization
to this grid in Figureonly
yellow areas involving 4 weone choose
pair of3 radial
valuesgrids
of theis
proportion: 𝜉 = 1⁄ 9 , 3⁄ 9 and 5 ⁄ 9 corresponding to 1, 3 and 5
ξ = 1/9. According to this grid discretization in Figure 4 we choose 3 values of the proportion:pairs of radial grids as
ξshown
= 1/9,in3/9Figure
and 5. 5/9The values of fracture
corresponding toughness
to 1, 3 and 5 pairs in
of the two
radial layers
grids for numerical
as shown in Figureex- 5.
periments
The values ofare obtained
fracture from papers
toughness in the two [31,32]:
layersthe fracture toughness
for numerical experiments ofare
barrier
obtainedlayers
from is
.
fixed at 𝐾 = 2 𝑀𝑃𝑎 ⋅ 𝑚 , the fracture toughness of the pay zone
b
papers [31,32]: the fracture toughness of barrier layers is fixed at K IC = 2 MPa·m , the fracturechanges from
0.5 the min-
. .
imum value
toughness of the pay 1 𝑀𝑃𝑎zone⋅ 𝑚 changestofrom the maximumvalue
the minimum value
1 MPa 4 ·𝑀𝑃𝑎 ⋅ 𝑚the
m0.5 to : 𝐾 =
maximum
.
1, 1.25, 1.5, 1.75,
0.5 2.0,
p 2.5, 3.0, 3.5, 4.0 𝑀𝑃𝑎 ⋅ 𝑚 for obtaining
value 4 MPa·m : K IC = 1, 1.25, 1.5, 1.75, 2.0, 2.5, 3.0, 3.5, 4.0 MPa·m for obtaining 9 different0.5 fracture tough- 9
different fracture𝜁toughness
ness contrasts: ⁄
= 𝐾 𝐾 contrasts: = 0.5, 0.625,
ζ=K
p
0.75,/K0.875,
b = 1.0,0.625,
0.5, 1.25,0.75,
1.5, 0.875,
1.75, 2.0,
1.0, respectively.
1.25, 1.5, 1.75,
IC IC
Therefore,
2.0, according
respectively. Therefore,to the crossover
according experiment
to the crossover strategy
experiment westrategy
conductwe 3 groups
conduct of exper-
3 groups
iments based on the values of 𝜉, and each group of experiments
of experiments based on the values of ξ, and each group of experiments includes 9 experiments includes 9 experiments
correspondingtoto9 9different
corresponding different values
values ofThe
of ζ. 𝜁. The detailed
detailed information
information of experiments
of experiments is ex-
is exhibited
hibited
in Table 1.in Table 1.
Figure 5. Three
Figure 5. Three kinds
kinds of
of proportion
proportion of
of the
the yellow
yellow areas
areas (the
(the pay
pay zone)
zone) corresponding
corresponding to
to three
three groups
groups
of
of crossover
crossoverexperiments.
experiments.
Table1.1.The
Table The27
27subgroups
subgroupsof
ofvalues
valuesof
ofgoverning variablesξ 𝜉and
governingvariables and 𝜁 for
ζ for crossover
crossover experiments.
experiments.
Experiment
Experiment Group
Group 1 1 Experiment
Experiment Group
Group 2 2 Experiment
Experiment Group
Group 33
Number
Number Ξ = 1/9 1/9
Ξ = Number
Number Ξ
Ξ = 3/9= 3/9 Number
Number Ξ = 5/9
Ξ = 5/9
11 ξ =ξ1/9,
= 1/9,
ζ =ζ0.5
= 0.5 10 10 ξ = 3/9,
ξ = 3/9, ζ = 0.5
ζ = 0.5 19 19 ξ ξ= =1/9,
1/9,ζ ζ= =0.5
0.5
2
2 ξ = 1/9, ζ = 0.625
ξ = 1/9, ζ = 0.625 11
11 ξ = 3/9, ζ = 0.625
ξ = 3/9, ζ = 0.625 20
20 ξ = 1/9, ζ = 0.625
ξ = 1/9, ζ = 0.625
3 ξ = 1/9, ζ = 0.75 12 ξ = 3/9, ζ = 0.75 21 ξ = 1/9, ζ = 0.75
3 ξ = 1/9, ζ = 0.75 12 ξ = 3/9, ζ = 0.75 21 ξ = 1/9, ζ = 0.75
4 ξ = 1/9, ζ = 0.875 13 ξ = 3/9, ζ = 0.875 22 ξ = 1/9, ζ = 0.875
4 ξ = 1/9, ζ = 0.875 13 ξ = 3/9, ζ = 0.875 22 ξ = 1/9, ζ = 0.875
5 ξ = 1/9, ζ = 1.0 14 ξ = 3/9, ζ = 1.0 23 ξ = 1/9, ζ = 1.0
56 ξ =ξ1/9, ζ =ζ 1.0
= 1/9, = 1.25 14 15 ξ = 3/9, ζ = 1.0
ξ = 3/9, ζ = 1.25 23 24 ξξ==1/9,
1/9,ζζ==1.01.25
67 ξ =ξ 1/9, ζ = 1.25
= 1/9, ζ = 1.5 15 16 ξ = 3/9, ζ = 1.25
ξ = 3/9, ζ = 1.5 24 25 ξ= 1/9, ζ = 1.25
ξ = 1/9, ζ = 1.5
78 ξ =ξ1/9,
= 1/9,
ζ =ζ 1.5
= 1.75 16 17 ξ = 3/9,
ξ = 3/9, ζ = 1.75
ζ = 1.5 25 26 ξξ==1/9,
1/9,ζζ==1.51.75
89 ξ =ξ = 1/9,
1/9, ζ = ζ1.75
= 2.0 17 18 ξ = ζ3/9,
ξ = 3/9, ζ = 2.0
= 1.75 26 27 ξ= ξ 1/9,
= 1/9,
ζ =ζ 1.75
= 2.0
9 ξ = 1/9, ζ = 2.0 18 ξ = 3/9, ζ = 2.0 27 ξ = 1/9, ζ = 2.0
The inhomogeneity of rock fracture toughness leads to the uneven propagation of frac-
ture, i.e., different front points of the fracture may grow forward with different velocities,
moreover, at each moment there are different front points, which satisfy the propagation
condition and grow forward. In principle, the smaller the time step ∆t, the more accurately
the simulation captures this uneven feature of fracture propagation. However, an exces-
Energies 2022, 15, 1967 11 of 35
sively small-time step leads to non-convergence of model calculation, and an overly large
time step makes it difficult to capture the uneven feature of fracture propagation. Therefore,
in our experiments, the time step is assumed as = 0.01 s, and the total simulation time of
each experiment is set as T = 20 s for the early time evolution of fracture propagation, i.e.,
the time step number: N = 2000.
4.2. Experiment Results

According to the above parameters, 27 numerical experiments are simulated based
on this model and algorithm. Figures 6–8 show the results of the distribution of fluid
pressure in the fracture at the moment = 20 s, Figure 9 shows comparison of fluid pres-
sures along radial zones θ = 0 and θ = π/2 for different numerical experiments, and
Figures 10–12 are the results of the distribution of fracture opening at the moment = 20 s,
in which Figures 6 and 10 exhibit the results of experiments of Group 1 with ξ = 1/9
corresponding to the case of Figure 5a; Figures 7 and 11 are related to experiments of Group
2 with ξ = 3/9 corresponding to the case of Figure 5b; Figures 8 and 12 are related to exper-
iments of Group 3 with ξ = 5/9 corresponding to the case of Figure 5c. Figures 9 and 13
show 6 examples (Experiments 1, 9, 10, 18, 19 and 27) of distribution of fluid pressure and
fracture opening, respectively, along the radial zones θ = 0 and θ = π/2, which are the
further supplements to Figures 6–8 and 10–12.
Distributions of
Figure6.6.Distributions
Figure of fluid
fluidpressure
pressureofof
Group
Group1 numerical experiments
1 numerical (ξ = (𝜉
experiments 1/9)
= 1with 9 different
⁄9) with 9 differ-
fracture toughness contrasts at the moment = 20
ent fracture toughness contrasts at the moment = 20 s.s.
Energies 2022, 15, 1967 12 of 35
Figure 6. Distributions of fluid pressure of Group 1 numerical experiments (𝜉 = 1⁄9) with 9 dif
ent fracture toughness contrasts at the moment = 20 s.

Figure 7. Distributions of fluid pressure of Group 2 numerical experiments (ξ = 3/9) with 9 different
Figure 7. Distributions of fluid pressure of Group 2 numerical experiments (𝜉 = 3⁄9) with 9 dif
fracture toughnessent
contrasts
fractureat the moment
toughness = 20 at
contrasts s. the moment = 20 s.
Figure 8. Distributions of fluid pressure of Group 3 numerical experiments (ξ = 5/9) with 9 different
Figure 8. Distributions of fluid pressure of Group 3 numerical experiments (𝜉 = 5⁄9) with 9 differ-
fracture toughness
ent fracture contrasts
toughness at the
contrasts at moment = 20=s.20 s.
the moment
p (Pa)
p (Pa)
The results of the fracture opening in Figures 10–12 indicate that the distribution of
Energies 2022, 15, 1967 fracture opening is similar to the crack profile. The distributions of fracture opening 13 along
of 35
Figure 8. Distributions of fluid pressure of Group 3 numerical experiments (𝜉 = 5⁄9) with 9 differ-
all radial zones are the same under the uniform
ent fracture toughness contrasts at the moment = 20 s. propagation, see Figures 10 e, 11e and 12e.
When the fracture has a very irregular contour, for example, Figures 10a and 12i, where
the radius of the radial zone 𝜃 = 0 is much larger (or much smaller) than the radius of
the radial zone 𝜃 = 𝜋⁄2, the distributions of fracture opening along these two radial zones
are very different from each other, see Figure 13a,f, because for all radial zones boundary
p (Pa)
p (Pa)
conditions of the fracture opening at the fracture front (𝑤 = 0) and at the wellbore (𝑤 )
are the same (see the start and the end of all curves of fracture opening in Figure 13), so
the average fracture opening gradient of the longer radial zone with a larger radius is
smaller than the average fracture opening gradient of the shorter radial zone with a
smaller radius. It means that because the radial zone with a less fracture toughness prop-
agates faster than radial zones with a larger fracture toughness, the fracture opening gra-
dient of the shorter radial zone near the fracture front deeply increases, which leads to the
larger and larger stress intensity factor according to Equation (18) (see the red curves in
p (Pa)
p (Pa)
Figure 13b,d,f). When the stress intensity factor of the shorter radial zone reaches the limit
value (𝐾 ), it propagates forward and the radius increases. After that its stress intensity
factor rapidly decreases, as shown by the black curves in Figure 13a,c,e. The mechanism
of uneven fracture propagation is that the zones with less fracture toughness drive the
zones with larger fracture toughness to propagate because of the compatibility of fracture
opening on each radial zone. This is the reason why the radial zone with smaller fracture
toughness cannot always propagate without the growth of the radial zone with larger
p (Pa)
p (Pa)
fracture toughness, except under some special conditions. This feature is clearly seen in
Figure 13a. Although the radial zone 𝜃 = 0 with the smaller fracture toughness propa-
gated very fast from the initial radius 0.5 𝑚 to 3.9 𝑚, the radial zone 𝜃 = 𝜋⁄2 with a
larger fracture toughness did not keep still, and also propagated from 0.5 𝑚 to 1.3 𝑚, just
the velocity was smaller. It is opposite for the radial zones 𝜃 = 0 and 𝜃 = 𝜋⁄2 in Figure
13b:
Figure
Figurethe9.radial zone 𝜃of=fluid
Comparison
Comparison 0 propagated slower,
pressures along while
radial the
zones θ𝜃== 00and
radial zone
and 𝜃𝜋=
θ 𝜃==π/2 𝜋of
⁄2of ⁄experiments
2experiments
propagated
1,1,9,
9,
10,10,
18,18,
faster. 19,19,
27.27.
Figure 10.
Figure Distributionsof
10. Distributions of fracture
fracture opening
opening of
of Group
Group 11 numerical
numerical experiments
experiments (𝜉 =11/9)
(ξ = with 99
⁄9) with
different fracture
different fracture toughness
toughness contrasts
contrasts at
at the moment =
the moment = 20
20 s.
s.
Energies 2022, 15, 1967
Energies 2022, 15, x FOR PEER REVIEW 14 of 35 1
Figure 11.of Distributions of fracture opening of Group 2 numerical experiments (𝜉 = 3⁄9) w

Figure 11.
Figure Distributions
11. Distributions of fracture opening
fracture opening of
of Group
Group 22 numerical experiments(ξ(𝜉==3/9)
numericalexperiments 3⁄9)with
with99
different fracturedifferent
toughnessfracture toughness
contrasts at the contrasts=at20
moment the
s. moment = 20 s.
different fracture toughness contrasts at the moment = 20 s.
Figure 12. Distributions

Figure 12.of Distributions
fracture opening of Group
of fracture 3 numerical
opening experiments
of Group (ξ experiments
3 numerical 9 = 5⁄9) w
= 5/9) with (𝜉
different fracturedifferent
toughnessfracture toughness
contrasts contrasts=at20
at the moment s. moment = 20 s.
the
Figure 12. Distributions of fracture opening of Group 3 numerical experiments (𝜉 = 5⁄9) with 9
different fracture toughness contrasts at the moment = 20 s.
Energies
Energies 2022,
2022, 15,
15, x1967
FOR PEER REVIEW 16
15 of
of 36
35
w (m)
w (m)
w (m)
w (m)
w (m)
w (m)
Figure
Figure 13.
13. Comparison
Comparison of
of fracture
fracture openings
openings along
along radial zones θ𝜃 =
radial zones = 00 and 𝜋⁄2ofofexperiments
andθ 𝜃==π/2 experiments1,
1, 9, 10, 18, 19, 27.
9, 10, 18, 19, 27.
4.3. Validation
Except forof Figures
Model and
6e,Algorithm
7e and 8e, there are several distinct weak pressure zones ap-
Savitski
pearing near and Detournay
the fracture front[38]
in studied the propagation
the remaining graphs of of a penny-shaped
Figures 6–8: the redhydraulic
curve in
Figure 9a shows the specific distribution of weak pressure zone
fracture in the solid medium with a uniform fracture toughness, namely the experiment on the radial zone θ=0
near
5, the23fracture
14 or in Tablefront.
1. TheItslarge-toughness
feature is a sharp decreasesolution
asymptotic in the pressure gradient
to the fracture nearwas
radius the
fracture front.
obtained [38]: The reason is that the fracture front located at these weak pressure zones
propagated at the previous time step, which led to stress release and the emergence of
new spaces near these fracture tips. 𝑅(𝑡) The new= 𝐿 (𝑡)𝛾 (𝑡),
spaces were initially liquid-free with zero (19)
pressure,
where theand large
length pressure
scaling gradients𝐿 between
parameter and thethe propagating fracture
dimensionless zones and 𝛾 are for-
the surrounding
radius
non-propagating
mulated: areas were generated at these tips, i.e., these weak pressure zones. As
the hydraulic fluid flows into these new spaces, ⁄
the pressure gradients near the fracture
𝑄 𝐸 𝑡
front become smaller and smaller, as shown in Figure 9b–f. These features can only be
𝐿 = , 𝛾 = 𝛾 + 𝑀𝛾 , (20)
observed in the uneven crack propagation. 𝐾 For the case of uniform crack propagation, as
shown in Figures 6e,⁄ 7e and 8e, near the fracture front there is no distinct weak pressure
where 𝐾 = 4(2the ⁄𝜋) 𝐾 , 𝛾 = 0.8546, 𝛾 = −0.7349 and the dimensionless viscosity
zone, because whole
⁄
fracture tips at different locations propagate at the same time,
𝑀and ⁄𝐾 𝑡 to .form distinct weak pressure area. It should be noted that these
= it𝑄is𝐸impossible
weakThe comparison
pressure of theaffect
areas only resultstheofpressure
the large-toughness asymptotic
distribution near solution
the fracture to and
front the frac-
lead
ture radius
to the andflow
circular numerical
at theseexperiments
locations. Away 5, 14 orfrom23 with the uniform
the fracture front,rock toughness
for example, in this
near the
paper are exhibited in Figure 14. It shows that the two results generally
wellbore, only the radial pressure gradient appears due to fluid injection, there is almost no coincide with each
other
pressureverygradient
well, with
alongonly
theahoop
smalldirection,
difference asat the beginning
shown in Figuresand 6–8.the end
This of thecan
feature time
be
more clearly
interval. observed
The reasons areinthat
Figure 9: pressures
in this model, the offracture
differenthas
radial (θ = 0of
zonesradius
an initial and𝑅 θ==0.5 𝑚
π/2)
nearthe
and theinitial
coordinate origin
pressure are almostisthe
distribution notsame.
enough to propagate the fracture, so at the be-
ginningTheofresults of the fracture
the simulation, opening radius
the fracture in Figures 10–12unchanged.
remains indicate thatFurthermore,
the distribution of frac-
it should
turenoted
be openingthatisthe
similar to the crack
asymptotic profile.
solution in The
[38]distributions of fracture
is obtained with opening along
the assumption thatallthe
ra-
dial zones are the same under the uniform propagation, see Figure
fracture is always in dynamic propagation equilibrium; however, in our model, the nu- 10e, Figures 11e and 12e.
merical solution is obtained with the assumption that the fracture propagates radially one
Energies 2022, 15, 1967 16 of 35
When the fracture has a very irregular contour, for example, Figures 10a and 12i, where the
radius of the radial zone θ = 0 is much larger (or much smaller) than the radius of the radial
zone θ = π/2, the distributions of fracture opening along these two radial zones are very
different from each other, see Figure 13a,f, because for all radial zones boundary conditions
of the fracture opening at the fracture front (w = 0) and at the wellbore (w0 ) are the same
(see the start and the end of all curves of fracture opening in Figure 13), so the average
fracture opening gradient of the longer radial zone with a larger radius is smaller than the
average fracture opening gradient of the shorter radial zone with a smaller radius. It means
that because the radial zone with a less fracture toughness propagates faster than radial
zones with a larger fracture toughness, the fracture opening gradient of the shorter radial
zone near the fracture front deeply increases, which leads to the larger and larger stress
intensity factor according to Equation (18) (see the red curves in Figure 13b,d,f). When the
stress intensity factor of the shorter radial zone reaches the limit value (K IC ), it propagates
forward and the radius increases. After that its stress intensity factor rapidly decreases, as
shown by the black curves in Figure 13a,c,e. The mechanism of uneven fracture propagation
is that the zones with less fracture toughness drive the zones with larger fracture toughness
to propagate because of the compatibility of fracture opening on each radial zone. This is
the reason why the radial zone with smaller fracture toughness cannot always propagate
without the growth of the radial zone with larger fracture toughness, except under some
special conditions. This feature is clearly seen in Figure 13a. Although the radial zone θ = 0
with the smaller fracture toughness propagated very fast from the initial radius 0.5 m to
3.9 m, the radial zone θ = π/2 with a larger fracture toughness did not keep still, and also
propagated from 0.5 m to 1.3 m, just the velocity was smaller. It is opposite for the radial
zones θ = 0 and θ = π/2 in Figure 13b: the radial zone θ = 0 propagated slower, while the
radial zone θ = π/2 propagated faster.
4.3. Validation of Model and Algorithm

Savitski and Detournay [38] studied the propagation of a penny-shaped hydraulic
fracture in the solid medium with a uniform fracture toughness, namely the experiment
5, 14 or 23 in Table 1. The large-toughness asymptotic solution to the fracture radius was
obtained [38]:
R ( t ) = L k ( t ) γk ( t ) , (19)
where the length scaling parameter Lk and the dimensionless fracture radius γk are formulated:
2
!1/5
Q20 E0 t2
Lk = , γk = γk0 + Mγk1 , (20)
K0 2
where K 0 = 4(2/π )1/2 K IC , γk0 = 0.8546, γk1 = −0.7349 and the dimensionless viscosity
1/5
13 18
M = Q20 E0 /K 0 t2 .
The comparison of the results of the large-toughness asymptotic solution to the fracture
radius and numerical experiments 5, 14 or 23 with the uniform rock toughness in this paper
are exhibited in Figure 14. It shows that the two results generally coincide with each other
very well, with only a small difference at the beginning and the end of the time interval.
The reasons are that in this model, the fracture has an initial radius of R0 = 0.5 m and the
initial pressure distribution is not enough to propagate the fracture, so at the beginning of
the simulation, the fracture radius remains unchanged. Furthermore, it should be noted
that the asymptotic solution in [38] is obtained with the assumption that the fracture is
always in dynamic propagation equilibrium; however, in our model, the numerical solution
is obtained with the assumption that the fracture propagates radially one element at a time,
so the difference between the asymptotic solution and the numerical solution increases
with time, as shown in Figure 14 at the end of the time interval.
solution increases with time, as shown in Figure 14 at the end of the time interval.
The reason for not choosing the assumption of dynamic propagation equilibrium for
our model is that this assumption would involve one additional unknown for each bound-
ary grid element into the model, namely the propagating length at each time step. How-
Energies 2022, 15, 1967 ever, there is no corresponding linearly independent equation introduced into 17 the model
of 35
for solving.
2
the numerical solution in this paper
1.8
the asymptotic solution in [38]
1.6
1.4
1.2
0.8
0.6
0.4
0.2
0
0 2 4 6 8 10 12 14 16 18 20
t (s)
Figure 14.14.
Figure Comparison
Comparisonofofthe
thefracture radiusofofthe
fracture radius thelarge-toughness
large-toughness asymptotic
asymptotic solution
solution and the
and the
numerical solution under the uniform propagation in this paper.
numerical solution under the uniform propagation in this paper.
4.4. Analysis of Results

The reason of Numerical
for not Experiments
choosing the assumption of dynamic propagation equilibrium
for our model is that this assumption would involve one additional unknown for each
Figures 15–17 present the comparison of the results of fracture profiles affected by
boundary grid element into the model, namely the propagating length at each time step.
the fracture toughness contrast 𝜁, in which the proportion of pay zones is fixed as 𝜉 =
However, there is no corresponding linearly independent equation introduced into the
1⁄9model
, 3⁄9 for
and 5⁄9, respectively. Figure 18 shows the comparison of fracture profiles af-
solving.
fected by the proportion of pay zones 𝜉, in subgraphs (a–i) of which the fracture tough-
ness4.4. Analysisisoffixed
contrast as of𝜁Numerical
Results = 0.5, 0.625, 0.75, 0.875, 1.0, 1.25, 1.5, 1.75 and 2.0, respec-
Experiments
tively. The black
Figures dotted
15–17 curves
present in Figures 15–18
the comparison represent
of the results the initial
of fracture fracture
profiles profile.
affected by theFrom
fracture toughness contrast in which the proportion of pay zones is fixed
the results in Figures 15–18 it is found that the fracture profile is significantly affected
ζ, as ξ = 1/9, by
3/9 and 5/9, respectively. Figure 18 shows the comparison of fracture
the fracture toughness contrast and the proportion of the pay zones, namely variables profiles affected by 𝜁
the
and 𝜉. proportion of pay zones ξ, in subgraphs (a–i) of which the fracture toughness contrast
is fixed as ζ = 0.5, 0.625, 0.75, 0.875, 1.0, 1.25, 1.5, 1.75 and 2.0, respectively. The black
The fracture profile is generally a circle in Figure 15 under 𝜉 = 1⁄9. When the frac-
dotted curves in Figures 15–18 represent the initial fracture profile. From the results in
tureFigures
toughness it is found𝜁 that
15–18contrast < 1,thethefracture
fracture profile
profile is like a circle
is significantly withbya the
affected pairfracture
of obverse
wedges
toughness contrast and the proportion of the pay zones, namely variables ζ and ξ. red curve
at the symmetric pay zone with the smaller fracture toughness (see the
in Figure 15). As the fracture toughness contrast disappears (𝜁 → 1), the pair of wedges
also gradually disappears, see the curves of 𝜁 = 0.625, 0.75 and 0.875. The green dotted
curve in Figure 15 represents the standard circular fracture profile under the uniform
propagation with no fracture toughness contrast (𝜁 = 1), which is the same as in Figures
ness (here we define a pair of reverse wedges with their tips facing each other, conversely
it is a pair of obverse wedges). When 𝜁 = 1.25, the reverse wedges are not obvious, be-
cause the length of reverse wedges, namely, the retracted length of radial zones (𝜃 = 0
and 𝜃 = 𝜋) with the larger toughness is relatively small. With the increase of fracture
Energies 2022, 15, 1967
toughness contrast (𝜁 = 1.5, 1.75 and 2.0), the shape of reverse wedges becomes 18 more
of 35
obvious, i.e., the retracted length increases with the increase of fracture toughness contrast
under 𝜁 > 1.
y (m)

Figure
Figure 15. Fractureprofiles
15. Fracture profilesonon
thethe plane
plane 𝑥𝑜𝑦
xoy of of Group
Group 1 experiments
1 experiments under ξ under 𝜉 the
= 1/9 at ⁄9 at the
= 1moment = mo-
20 s.
ment = 20 s.
The fracture profiles under the proportion of pay zone 𝜉 = 3⁄9 and 5⁄9 are very
similar to each other, but significantly different from the profiles under 𝜉 = 1⁄9, see Fig-
ures 16 and 17. When 𝜁 < 1, as the proportion of pay zone increases from 1⁄9 to 3⁄9,
and even to 5⁄9, the fracture profile consisting of a circle with a pair of obverse wedges
transforms into a horizontal rectangle with its long side along 𝑥-axis. As the fracture
toughness contrast disappears (𝜁 → 1), the difference between the long and short sides of
this horizontal rectangle also disappears, and the fracture profile approaches a circle (the
green dotted curve). With the increase of fracture toughness contrast over unity, the frac-
ture profile is transformed into a vertical rectangle with the long side along 𝑦-axis. The
results of fracture profiles under 𝜉 = 3⁄9 , 5⁄9 in Figures 16 and 17 show that when the
y (m)
fracture toughness of the pay zone is larger than that of the barrier layers, the hydraulic
fracture easily and mainly propagates toward the barrier layers, and the fracture along
the pay zone is relatively small, which deviates from the goal of hydraulic fracturing.
Figure 16. Fracture

Figure 16. Fractureprofiles
profilesonon
thethe plane
plane 𝑥𝑜𝑦
xoy of of Group
Group 2 experiments
2 experiments 3/9 at𝜉 the
under
under ξ = ⁄9 at the
= 3moment = mo-
20 s.
ment = 20 s.
Energies 2022, 15, 1967 19 of 35
Figure 16. Fracture profiles on the plane 𝑥𝑜𝑦 of Group 2 experiments under 𝜉 = 3⁄9 at the mo-
ment = 20 s.
y (m)
Figure 17.
Figure 17. Fracture
Fractureprofiles
profilesononthe
the
plane xoy𝑥𝑜𝑦
plane of Group
of Group 3 experiments
3 experiments underunder 𝜉 =at5⁄
ξ = 5/9 9 at
the the mo-
moment = 20 s.
ment = 20 s.
The fracture profile is generally a circle in Figure 15 under ξ = 1/9. When the fracture
toughness contrast ζ < 1, the fracture profile is like a circle with a pair of obverse wedges at
the symmetric pay zone with the smaller fracture toughness (see the red curve in Figure 15).
As the fracture toughness contrast disappears ( ζ → 1 ), the pair of wedges also gradually
disappears, see the curves of ζ = 0.625, 0.75 and 0.875. The green dotted curve in Figure 15
represents the standard circular fracture profile under the uniform propagation with no
fracture toughness contrast (ζ = 1), which is the same as in Figures 16 and 17. When the
fracture toughness contrast ζ > 1, the fracture profile is like a circle losing a pair of reverse
wedges at the symmetric pay zone with the larger fracture toughness (here we define a
pair of reverse wedges with their tips facing each other, conversely it is a pair of obverse
wedges). When ζ = 1.25, the reverse wedges are not obvious, because the length of reverse
wedges, namely, the retracted length of radial zones (θ = 0 and θ = π) with the larger
toughness is relatively small. With the increase of fracture toughness contrast (ζ = 1.5,
1.75 and 2.0), the shape of reverse wedges becomes more obvious, i.e., the retracted length
increases with the increase of fracture toughness contrast under ζ > 1.
The fracture profiles under the proportion of pay zone ξ = 3/9 and 5/9 are very
similar to each other, but significantly different from the profiles under ξ = 1/9, see
Figures 16 and 17. When ζ < 1, as the proportion of pay zone increases from 1/9 to 3/9,
and even to 5/9, the fracture profile consisting of a circle with a pair of obverse wedges
transforms into a horizontal rectangle with its long side along x-axis. As the fracture
toughness contrast disappears ( ζ → 1 ), the difference between the long and short sides
of this horizontal rectangle also disappears, and the fracture profile approaches a circle
(the green dotted curve). With the increase of fracture toughness contrast over unity, the
fracture profile is transformed into a vertical rectangle with the long side along y-axis. The
results of fracture profiles under ξ = 3/9, 5/9 in Figures 16 and 17 show that when the
fracture toughness of the pay zone is larger than that of the barrier layers, the hydraulic
fracture easily and mainly propagates toward the barrier layers, and the fracture along the
pay zone is relatively small, which deviates from the goal of hydraulic fracturing.
Energies 2022, 15, 1967 20 of 35
(a): =0.5 (b): =0.625 (c): =0.75

3
2 1.5
2
1
1
1 0.5
0 0 0
-0.5
-1
-1
-1
-2
-1.5
-2
-3
-2 0 2 -2 -1 0 1 2 -2 -1 0 1 2
x (m) x (m) x (m)
(d): =0.875 (e): =1.0 (f): =1.25

2
1.5 1.5 the initial profile
1 1 =1/9 1
=3/9
0.5 0.5
=5/9
0 0 0
-0.5 -0.5
-1
-1 -1
-1.5 -1.5
-2
-2 -1 0 1 2 -2 -1 0 1 2 -2 -1 0 1 2
x (m) x (m) x (m)
(g): =1.5 (h): =1.75 (i): =2.0

2
2 2
1 1 1
0 0 0
-1 -1
-1
-2 -2
-2
-2 -1 0 1 2 -2 0 2 -2 0 2
x (m) x (m) x (m)
Figure 18. Comparison of fracture profiles affected by different proportions of pay zone under the
same fracturetoughness
same fracture toughnesscontrast
contrastatatthe
themoment
moment==
2020 s. The
s. The legend
legend in the
in the subgraph
subgraph (e) is(e)also
is also ap-
applied
plied to other subgraphs
to other subgraphs (a–i). (a)–(i).
In
Bythis model, theitradii
comparison, of radial
is found thatzones 𝜃 = 0 and
the fracture 𝜃 = 𝜋⁄2 is
propagation aremore
the most representa-
sensitive to the
tive lengths to describe the fracture geometry because of symmetry,
formation with less than unity fracture toughness contrast: when the fracture toughness so here we take the
radii
of payofzoneradialis zone 𝜃 = more
less and 0 andthan 𝜋⁄2 of
𝜃 = that as the barrier
“half-length”
layers (𝑅 the) and
samethe “half-height”
proportion, the
(𝑅 ) of propagation
excess the fracture according
length ontoaccount
the traditional practice
of less than in most
unity fracture models of hydraulic
toughness contrast frac-is
turing.
larger thanThe ratio
the shortened 𝑅 is called
of 𝑅 andlength owingthe fracture
to more shape
than unity coefficient 𝜂 = 𝑅 ⁄𝑅 .contrast,
fracture toughness
p
Figures 19–21
for example, K IC =show
1 andthe MPa·m0.5of
3.0evolution 𝑅 , a𝑅proportion
with and 𝜂 with 𝑡 for
timeless
of 50% and the experiments
more than the
of Groups
fracture 1, 2 andof3 barrier
toughness corresponding
zones, the 𝜉 = 1⁄propagation
toexcess 9, 3⁄9 and lengths5⁄9, respectively.
of the radialItzoneis known
θ=0
from
comparedFigure 19 (𝜉
with 1⁄9) thatpropagation
the=uniform when 𝜁 < 1, are𝑅2.0,is 1.8,
always
1.4 mgreater than 𝑅 , but
corresponding to ξ with
= 1/9, the3/9in-
and 5/9,
crease of while the corresponding
the fracture shortened
toughness contrast from lengths
0.5 to 1,arethe
only 0.5, 0.7, between
difference 0.8 m. 𝑅 and 𝑅
It is clearly
decreases observed in
and approaches Figure
zero. When 𝜁 > 1,
18 that the𝑅fracture toughness
is in turn than 𝑅 between
greater contrast and the dif- the
pay zone
ference and the𝑅barrier
between and 𝑅 zones is the again,
appears core factor
evenofrapidly
determining
increases thewith
overall propagation
fracture tough-
direction
ness contrastof the fracture.
changing fromNo1 matter
to 2. The what proportion
fracture the pay zone
shape coefficient 𝜂 initially
occupies (ξ = rap-
changes 1/9,
3/9 or 5/9), when the fracture toughness of the pay zone is less
idly and then approaches one certain limit value for all fracture toughness contrasts. The than that of the barrier
zones (ζ < for
difference 𝜁 <fracture
1), the 1 and 𝜁in>total
1 ispropagates
that whenalong 𝜁 < 1,the 𝜂 pay
is anzone, see Figure
increasing 18a–d,ofontime
function the
contrary,
from unitwhen > 1, thelimit
to theζ certain fracture
value, while 𝜁 >generally
propagates 1, 𝜂 is aperpendicular
decreasing function to the pay zone,
of time see
from
Figureto18f–i.
unity The proportion
the limit of pay zone
value. According to theonly
dataplays a role 19,
in Figure in strengthening
it can be estimated or weakening
that the
the overall
limit valuesdirection of fracture
of the fracture shapepropagation,
coefficient for example,
𝜂 under 𝜉 =in1⁄subgraphs
9 are about (a–c) of 1.58,
2.55, Figure 18
1.27,
the horizontal propagation of the fracture is very significant under
1.11, 1, 0.84, 0.71, 0.61 and 0.53 for the corresponding fracture toughness contrasts 𝜁 = 0.5, ξ = 1/9, but only
concentrated
0.625, in very
0.75, 0.875, limited
1.0, 1.25, 1.5,zones,
1.75 and the2.0,
overall fracture profile is still a circle; when the
respectively.
Energies 2022, 15, 1967 21 of 35
proportion of pay zone is large enough (ξ = 3/9, 5/9), the fracture profiles are completely
transformed from a circle into a horizontal rectangular. Furthermore, when the variables
ζ and ξ are large enough, some fronts of the fracture cannot propagate, such as the blue
curve of ζ = 2.0 and ξ = 5/9. It indicates that the combined effect of fracture toughness
contrast and the proportion of the pay zone is very significant for the fracture profile.
It should be noted that the fracture profiles under ξ = 3/9, 5/9 in Figure 18a–c are very
similar to the fracture shape of the classical Perkins–Kern–Nordgren (PKN) model [39,40].
This further proves the importance of the effects of the fracture toughness contrast and the
proportion of pay zone on the fracture profile.
In this model, the radii of radial zones θ = 0 and θ = π/2 are the most representative
lengths to describe the fracture geometry because of symmetry, so here we take the radii of
radial zone θ = 0 and θ = π/2 as the “half-length” (Rl ) and the “half-height” (Rh ) of the
fracture according to the traditional practice in most models of hydraulic fracturing. The
ratio of Rl and Rh is called the fracture shape coefficient η = Rl /Rh .
Figures 19–21 show the evolution of Rl , Rh and η with time t for the experiments of
Groups 1, 2 and 3 corresponding to ξ = 1/9, 3/9 and 5/9, respectively. It is known from
Figure 19 (ξ = 1/9) that when ζ < 1, Rl is always greater than Rh , but with the increase of
the fracture toughness contrast from 0.5 to 1, the difference between Rl and Rh decreases
and approaches zero. When ζ > 1, Rh is in turn greater than Rl and the difference between
Rl and Rh appears again, even rapidly increases with fracture toughness contrast changing
from 1 to 2. The fracture shape coefficient η initially changes rapidly and then approaches
one certain limit value for all fracture toughness contrasts. The difference for ζ < 1 and
ζ > 1 is that when ζ < 1, η is an increasing function of time from unit to the certain
limit value, while ζ > 1, η is a decreasing function of time from unity to the limit value.
According to the data in Figure 19, it can be estimated that the limit values of the fracture
shape coefficient η under ξ = 1/9 are about 2.55, 1.58, 1.27, 1.11, 1, 0.84, 0.71, 0.61 and 0.53
for the corresponding fracture toughness contrasts ζ = 0.5, 0.625, 0.75, 0.875, 1.0, 1.25, 1.5,
1.75 and 2.0, respectively.
The results of Figures 20 and 21 under ξ = 3/9, 5/9 show the same evolution law as
Figure 19 under ξ = 1/9. The main difference is that with the increase of the proportion
of pay zone, the velocities of fracture “half-length” and “half-height” change a lot, which
greatly affects the evolution of the fracture shape coefficient. Table 2 gives the estimated
limit value of the fracture shape coefficient with respect to the fracture toughness contrast
and the proportion of the pay zone. From this table, we know that when ζ > 1, the limit
values of η are inversely proportional to parameters ξ and ζ; when ζ < 1, the values of η
are proportional only to the parameter ζ. Under the same fracture toughness contrast, the
limit values of η reach the maximum with ξ = 3/9, not with ξ = 5/9. Comparing fracture
profiles in Figures 15–18 with the limit values of fracture shape coefficient, we find that
the formation with the moderate value of the proportion of the pay zone (ξ = 3/9) and
the small fracture toughness contrast (ζ = 0.5) is the best for hydraulic fracturing, and the
fracture profile is the most ideal.
Table 2. Estimated limit values of the fracture shape coefficient η with respect to the fracture
toughness contrast ζ and the proportion of pay zone ξ.
ζ
0.5 0.625 0.75 0.875 1.0 1.25 1.5 1.75 2.0
ξ
1/9 2.55 1.58 1.27 1.11 1.0 0.84 0.71 0.61 0.53
3/9 4.0 2.1 1.57 1.2 1.0 0.77 0.59 0.44 0.35
5/9 2.8 1.85 1.43 1.17 1.0 0.76 0.57 0.38 0.2
Energies2022,
Energies 2022,15,
15,1967
x FOR PEER REVIEW 22 of
22 of 35
36
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
Figure 19.
Figure 19. Relationships
Relationships of
of the
the fracture
fracture“half-length”
“half-length”(R
(𝑅 ),),“half-height”
“half-height”(R
(𝑅 )) and
and the
the fracture
fracture shape
shape
l h
coefficient η𝜂with
coefficient withrespect
respecttototime
timefor
forGroup
Group1 1experiments
experimentswith 𝜉=
withξ = 1⁄9.
1/9.
The shape
The resultscoefficient
of Figuresη 20
canand 21 under 𝜉 as
be formulated =3 ⁄9 , 5⁄9 show the same evolution law
follows
as Figure 19 under 𝜉 = 1⁄9. The main difference is that with the increase of the proportion
of pay zone, the velocities of fracture Rl vt vl “half-height” change a lot, which
η =“half-length”
= l = and ,
greatly affects the evolution of the fractureRh shapevh t coefficient.
vh Table 2 gives the estimated
limit value of the fracture shape coefficient with respect to the fracture toughness contrast
where vl and vh are the average propagation velocities of horizontal and vertical directions
and the proportion of the pay zone. From this table, we know that when 𝜁 > 1, the limit
at the moment t. Another important feature of the curves of fracture shape coefficient η in
values of 𝜂 are inversely proportional to parameters 𝜉 and 𝜁; when 𝜁 < 1, the values of
Figures 19–21 is that: whether ζ < 1 or ζ > 1, the shape coefficient η always increases or
𝜂 are proportional only to the parameter 𝜁. Under the same fracture toughness contrast,
decreases rapidly only at the initial period of fracture propagation, after which it changes
the limit values of 𝜂 reach the maximum with 𝜉 = 3⁄9, not with 𝜉 = 5⁄9. Comparing
(increases or decreases) very slightly and just fluctuates in a small range around the limit
fractureItprofiles
values. in Figures
means that 15–18
the ratio with the
between the horizontal
limit values andof the
fracture shape
vertical coefficient,
velocities we
changes
find that
rapidly at the
the formation
beginning,with the moderate
and after value
approaching theoflimit
the proportion
value of η inofTable
the pay zone
2, the =
(𝜉 of
ratio
3 ⁄9 ) and the small fracture toughness contrast (𝜁=0.5) is the best for hydraulic
velocities is basically unchanged. In a word, after the initial uneven propagation with a fracturing,
and the changed
rapidly fracture profile
velocityisrate,
the most ideal. propagation will enter into a stable stage with
the fracture
an approximately constant velocity ratio between the horizontal and vertical directions,
Table 2. Estimated limit values of the fracture shape coefficient 𝜂 with respect to the fracture tough-
and in this stage, the fracture profile has the similarity with respect to time.
ness contrast 𝜁 and the proportion of pay zone 𝜉.
Figures 22–24 give the evolution of fluid pressure (subgraphs (a)) and the fracture
opening
ζ (subgraphs (b)) at the wellbore corresponding to experiments of Group 1, 2 and 3,
in which
ξ ξ0.5
= 1/9, 0.625 0.75respectively.
3/9 and 5/9, 0.875 In 1.0 1.25of Figures
each figure 1.5 22–241.75
subgraphs 2.0(a)
and
1/9(b) have
2.55the same
1.58legend,
1.27so it was drawn only
1.11 1.0 in subgraph
0.84 (a).
0.71 0.61 0.53
3/9 4.0 2.1 1.57 1.2 1.0 0.77 0.59 0.44 0.35
5/9 2.8 1.85 1.43 1.17 1.0 0.76 0.57 0.38 0.2
The shape coefficient 𝜂 can be formulated as follows

around the limit values. It means that the ratio between the horizontal and the vertical
velocities changes rapidly at the beginning, and after approaching the limit value of 𝜂 in
Table 2, the ratio of velocities is basically unchanged. In a word, after the initial uneven
propagation with a rapidly changed velocity rate, the fracture propagation will enter into
a stable stage with an approximately constant velocity ratio between the horizontal and
Energies 2022, 15, 1967 23 ofto
vertical directions, and in this stage, the fracture profile has the similarity with respect 35
time.
(a): =3/9, =0.5 (b): =3/9, =0.625 (c): =3/9, =0.75

4 4 3 2.5 2.5 1.5
3 1.4
2
3 2 2
Rl
R (m)
R (m)
1.3
R (m)
2 Rh 1.5
=Rl/ Rh Rl Rl 1.2
2 1 1.5
1 Rh 1 Rh
1.1
=Rl/ Rh =Rl/ Rh
0 1 0 1 0.5 1
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
t (s) t (s) t (s)
(d): =3/9, =0.875 (e): =3/9, =1.0 (f): =3/9, =1.25

1.25 2 2 1
2
Rl Rl
1.2
Rh Rh
1.5 1.5 1.5 0.9
1.15 =Rl/ Rh =Rl/ Rh
R (m)
R (m)
R (m)
1
Rl 1.1
1 1 1 0.8
Rh
1.05
=Rl/ Rh
0.5 1 0.5 0.5 0.7
0 5 10 15 20 0 5 10 15 0 5 10 15 20
t (s) t (s) t (s)
(g): =3/9, =1.5 (h): =3/9, =1.75 (i): =3/9, =2.0

2 1 2.5 1 2.5 1
Rl Rl Rl
Rh 0.9 Rh
2 2 Rh 0.8
1.5 0.8
=Rl/ Rh 0.8 =Rl/ Rh =Rl/ Rh
R (m)
R (m)
1.5 R (m) 1.5 0.6

0.7
1 0.6
1 1 0.4
0.6
0.5 0.5 0.5 0.4 0.5 0.2

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
t (s) t (s) t (s)
Figure
Figure 20.
20. Relationships
Relationships of
of the
the fracture
fracture “half-length” (𝑅 ),
“half-length” (R “half-height” (𝑅 ) and the fracture shape
l ), “half-height” (R h ) and the fracture shape
coefficient 𝜂 with respect to time for Group 2 experiments with 𝜉 = 3⁄9.
coefficient η with respect to time for Group 2 experiments with ξ = 3/9.
As we can see from Figures 22a, 23a and 24a, after entering into propagation stage
pressure at the wellbore decreases with time because of fracture propagation. Furthermore,
for all values ξ = 1/9, 3/9 and 5/9, the pressure at the wellbore p0 increases with the
increase of fracture toughness contrast. The fracture opening at the wellbore w0 has the
same evolution characteristic as the pressure at the wellbore, but there are two exceptions:
ζ = 1.75 and ζ = 2.0, in which w0 are smaller than that of experiments ζ = 1.5. These
exceptions are evident in Figures 23 and 24, but not in Figure 22. The reason is that with the
increase of the proportion of the pay zone (ξ = 3/9 and 5/9), the ability to hinder fracture
propagation is significantly enhanced when ζ = 1.75 and ζ = 2.0 (see Figures 16 and 17:
the pay zone with the larger fracture toughness propagated very slowly), even can lead
to no propagation of the pay zone (see the fracture profile of experiment 27 with ξ = 5/9
and ζ = 2.0). Therefore, when the fracture toughness of the pay zone is large enough as
ζ = 1.75 and ζ = 2.0, the difference between radii of the pay zone and the barrier zones is
great, which leads to non-monotonically decrease of fracture opening with respect to the
radial coordinate along some radial zones, even near the wellbore the fracture opening is a
monotonically increasing function of the radial coordinate (see Figure 13c–f). This indicates
that for some special cases (ξ = 3/9 and 5/9, ζ = 1.75 and 2.0) the fracture opening at the
wellbore is not the maximum, which is significantly different from the 2D and 3D models
for uniform fracture propagation.

Energies 2022, 15, 1967 24 of 35
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
R (m)
Figure 21. Relationships of the fracture “half-length” (𝑅 ), “half-height” (𝑅 ) and the fracture shape
coefficient 𝜂 with respect to time for Group 3 experiments with 𝜉 = 5⁄9.
opening (subgraphs (b)) at the wellbore corresponding to experiments of Group 1, 2 and
Figure 21.
21.Relationships
Figure
3, in which 𝜉 = 1⁄9 , of
Relationships 3⁄the
of the fracture
9 and 5⁄9“half-length”
fracture “half-length” (𝑅
(Rl),),In
, respectively. “half-height”
“half-height”
each figure (𝑅 ) )and
(Rhof andthe
thefracture
Figures 22–24shape
fracture shape
sub-
coefficient
coefficient 𝜂 with
with respect
respect to
to time
time for
for Group
Group 33 experiments
experiments with
with 𝜉== 5⁄9
5/9. .
graphs (a) and (b) have the same legend, so it was drawn only in subgraph (a).
η ξ
opening (subgraphs (b)) at the wellbore corresponding to experiments of Group 1, 2 and
3, in which 𝜉 = 1⁄9 , 3⁄9 and 5⁄9 , respectively. In each figure of Figures 22–24 sub-
graphs (a) and (b) have the same legend, so it was drawn only in subgraph (a).
p0 (Pa)
w 0 (m)
p0 (Pa)
w 0 (m)
Figure 22. Evolution

Figure 22. Evolution of
of the
the fluid pressurep𝑝 and
fluidpressure andthe
the fracture
fracture w0 𝑤at the
opening
opening at the wellbore
wellbore under
under the
0
the proportion of pay zone 𝜉 = 1⁄9.
proportion of pay zone ξ = 1/9.
Figure 22. Evolution of the fluid pressure 𝑝 and the fracture opening 𝑤 at the wellbore under
Energies2022,
Energies 2022,15,
15,1967
x FOR PEER REVIEW 25 of 36
25 of 35
ww(m)(m)
(Pa)
p p(Pa)
0 0
0 0
Figure
Figure 23.
23. Evolution
Evolutionofofthe
thefluid pressure𝑝p0and
fluidpressure andthe
thefracture opening𝑤w atatthe
fractureopening thewellbore
wellboreunder
under the
Figure 23. Evolution
the proportion of pay of the 𝜉fluid
zone = 3⁄pressure
9. 𝑝 and the fracture opening 𝑤 0at the wellbore under
proportion of pay zone = 3/9.
ξ
(Pa)
ww(m)(m)
p0p(Pa)
0 0
0
Figure 24. Evolution of the fluid pressure 𝑝 and the fracture opening 𝑤 at the wellbore under
Figure
the 24.
24. Evolution
proportion of pay of
ofthe
zonethe𝜉fluid
= 5⁄pressure
9pressure
. 𝑝p and the fracture opening 𝑤 at the wellbore under
Figure Evolution fluid 0 and the fracture opening w0 at the wellbore under the
proportion of pay zone ξ = 5/9.
Energies 2022, 15, 1967 26 of 35
5. Hydraulic Fracture Cleaning Process

The article considers the process of cleaning a fracture from hydraulic fracturing
fluid. This process is performed by oil under the effect of inverse pressure drop, and is
regarded within the frame of a planar 2-D approximation in the plane orthogonal to a
vertical fracture [41]. The following geometry of the position of the wells is considered:
four injection wells in the corners of the computational domain and one production well in
the center. The central well is connected with a symmetrical fracture filled in with proppant.
The reservoir area around the crack is filled with fracturing fluid. Using numerical modeling
on the basis of the constructed mathematical model, a relationship is established between
the quality of hydraulic fracture cleaning and the geometrical parameters of the fracture
and the region filled with the hydraulic fracturing fluid.
5.1. Mathematical Model for Seepage Flows of Multi-Phase Fluids

The problem was solved under the following assumptions:
1. All liquids are incompressible;
2. The viscosity of the displaced fluid is greater than the viscosity of the displacing fluid
in the reservoir;
3. Capillary effects at the interface are not taken into account;
4. Permeability and porosity are location dependent;
5. Thermal effects and gravity are not taken into account;
6. The outer boundary is impermeable;
7. Seepage is modeled by Darcy’s law, taking into account the relative permeabilities of
the phases
8. Hydraulic fracture—area of increased permeability and porosity
9. Permeability has a random “ripple” that contributes to the onset of displacement instability.
The following system of equations was solved.
For each fluid phase, the law of mass conservation is written as:
∂ϕsk $k ∂
+ $k uk,j = 0. (21)
∂t ∂x j
Here, ϕ is porosity, s is saturation, $ is density (intrinsic), uk,j is j-th component of the

seepage velocity of the k-th phase.
The average volumetric seepage velocity of the fluid is defined as:
uj = ∑ ukj . (22)
k
Summation of Equations (21), taking into account the liquid’s incompressibility and
constant in time porosity, leads to:
∂u j
= 0. (23)
∂x j
Darcy’s law for each phase is as follows:
KKkR ∂pk
uk,j = − , (24)
µk ∂x j
where µk is the dynamic viscosity of the phase, K is absolute permeability of the medium,
KkR is the relative permeability of k-th phase, p is the pressure in the pores.
Substitution of Equations (24) into (22) leads to:
∂p
u j = −K ∑ mk . (25)
k
∂x j
Energies 2022, 15, 1967 27 of 35
kr
where mk = µk —relative mobility.
k
Substituting expression (24) into equation (23) gives an equation for the pressure:
!
∂ ∂p
K ∑ mk =0 (26)
∂x j k
∂x j
Equation (21), taking into account the introduced definitions (22,25,26), leads to an
equation for saturation dynamics:
∂sk ∂
+ ϕ −1

f k u j = 0, (27)
∂t ∂x j
where f k = m m+km2 is the proportion of relative phase mobility.

1
The relative permeabilities KkR are calculated using the Brooks–Corey model [42]:
0
(
n
k0k Sk k , sk ≥ sres sk − sres
KkR = k , S k = k
. (28)
0 , sk < sres 1 − s1res − s2res
k
Here, k0k > 0 and n0k > 0 are the model parameters, and the effective saturation Sk is
determined by the residual saturations of 0 ≤ sres res res
k ≤ 1 (s1 + s2 < 1).
A hydraulic fracture is modeled as an area of increased permeability and porosity.
Thus, absolute permeability and porosity are expressed as:

K f , ( x, y) ∈ F
K ( x, y) = (29)
K0 , ( x, y) ∈/F

ϕ f , ( x, y) ∈ F
ϕ( x, y) = (30)
ϕ0 , ( x, y) ∈/ F,
where F is the set of points lying inside the fracture.
It is also worth noting that a weak random “ripple” is superimposed on the perme-
ability, so that finally the absolute permeability is defined as:
K̂ (x, y) = K ( x, y)· exp(δξ ), (31)
where ξ is a random variable uniformly distributed in the interval [−1; +1], δ is a rather
small quantity. This random “ripple” contributes to the onset of displacement instability.
Figure 25 shows the geometry of the model problem. The reservoir is initially saturated
with pore fluid (oil). There are four wells in the corners with constant pressure in it. In the
center of the area there is a production well and a hydraulic fracture filled with proppant
and hydraulic fracturing fluid. It is assumed that part of the fracturing fluid has leaked
into the surrounding rock. It is also assumed here that the region impregnated with the
hydraulic fracturing fluid is elliptical (region A). The problem under consideration is
cleaning the region A from the remains of hydraulic fracturing fluid.
Due to the pressure drop, pore fluid is filtered in an isotropic porous medium to the
producing well and to the outside. The viscosity of the fracturing fluid is greater than the
viscosity of the oil. Therefore, the process is unstable. The outer walls are impermeable.
The boundary conditions:
Σin : s = 0, P = Pin
Σout : P = Pout < Pin (32)
Σw : un = 0 →
∂p
∂n Г =0
Energies 2022, 15, 1967 28 of 35

Figure 25 shows the geometry of the model problem. The reservoir is initially satu-
rated with pore fluid (oil). There are four wells in the corners with constant pressure in it.
In the center of the area there is a production well and a hydraulic fracture filled with
Theand
proppant initial conditions:
hydraulic fracturing fluid.
It is assumed that part of the fracturing fluid has
leaked into the surrounding rock. It is also 1, ( x, y) ∈
s =assumed A that the region impregnated with
here (33)
s = 0, ( x, y ) ∈
/ A
the hydraulic fracturing fluid is elliptical (region A). The problem under consideration is
cleaning the
The region
model A fromofthe
consists remains(25)–(28),
Equations of hydraulic withfracturing fluid. (29)–(33).
the conditions
Figure 25. Calculation area for modeling the fracture cleaning process.
Figure 25. Calculation area for modeling the fracture cleaning process.
5.2. Effect of Fracture Length on Displacement Dynamics
Due to the pressure drop, pore fluid is filtered in an isotropic porous medium to the
The purpose of the first numerical experiment is to investigate the effect of fracture
producing
length (L fwell and to the outside. The viscosity of the fracturing fluid is greater than the
racture ) on the cleaning process dynamics. Two calculations were carried out with
viscosity of the oil. Therefore,
different fracture lengths. the
The process isparameters
calculation unstable. The
areouter
givenwalls are 3.
in Table impermeable.
The boundary conditions:
Table 3. Calculation parameters for the𝛴first
: numerical
𝑠 = 0, 𝑃 experiment
=𝑃 with different fracture lengths.
Number of cells 𝛴 : 𝑃=𝑃 <𝑃 Nx = 1001, Ny = 1001

(32)
The dimensions of the field 𝜕𝑝 Lx = 100 m, Ly = 100 m
𝛴 : 𝑢 =0 → =0
Absolute permeability of the formation 𝜕𝑛 Г K0 = 0.5 Darcy
The initial Formation porosity

conditions: φ0 = 0.2
Oil viscosity µ1 = 0.01 Pa ∗ c
𝑠 = 1, (𝑥, 𝑦) ∈ 𝐴
(33)
Fracturing fluid viscosity 𝑠 = 0, (𝑥, 𝑦) ∉ 𝐴 µ2 = 0.1 Pa ∗ c
Pressure
The model in injection
consists wells (25)–(28), with the conditions
of Equations Pin = (29)–(33).
500 atm
Pressure in the production well Pout = 5 atm
5.2. Effect of Fracture Length
Fracture on Displacement Dynamics Lfracture = 20 m/Lfracture = 30 m
length
The purpose
Fracture of the first
absolute numerical experiment is to investigate
permeability Kf = 50the effect of fracture
Darcy
length (𝐿 ) on the cleaning process dynamics. Two calculations were carried out
Fracture porosity φf = 0.4
with different fracture lengths. The calculation parameters are given in Table 3.
Figure 26 shows the dependence of mean saturation of hydraulic fracturing fluid in

elliptic zone on time.
Energies 2022, 15, 1967 29 of 35
The dependence in Figure 26 initially behaves linearly, after that it continues to de-
crease, but never reaches a constant: the asymptote in this case is s = 0. Therefore, any
sufficiently
Table small orparameters
3. Calculation satisfactory forsaturation value canexperiment
the first numerical be taken as a criterion
with differentwhen it is
fracture lengths.
considered that the fracture is already cleaned. If we assume that the fracture is already
cleaned when the Number
saturationof ofcells
the hydraulic fracturing fluid reached Nx =0.2,1001,
thenNity took
= 1001about
5 days 8 h to clean the fracture 20
The dimensions of the field m long, and it took 9 days 13 hLx = 100 m, Ly = 100 30
to clean the fracture mm
long.Absolute
It turns out that the ratio
permeability ofisthe
1.79. Therefore, the length of theKfracture
formation significantly
0 = 0.5 Darcy
affects the duration of the cleaning process.
Formation porosity ϕ0 = 0.2
Figures 27–29 show the distribution patterns of the oil seepage velocity for various
Oil viscosity
moments of time. The instability of displacement is observed in all figures. µ 1 = 0.01 Pa ∗ c
In FigureFracturing fluidflows
27, the largest viscosity
are observed in the fracture, asµwell 2 = 0.1
as Pa
near∗ cthe pro-
duction well. It can be seen that for
Pressure in injection wells a larger fracture, the fluid flows faster in the
Pin = 500 atm area near
the well.
In Figure 28, for the case when the length is 20 m, the flow occurs along the entire
fracture. In the case Fracture length fracture, a breakthroughLoccurred
of a longer fracture = 20 m/Lfracture = 30 m
in the area of the
Fracture absolute permeability
production well. Thus, at this time moment, not the entire hydraulic Kffracture
= 50 Darcyis involved
in the seepage process.
FractureInporosity
Figure 29 it can be noted that seepage occurs in almost the
ϕf = 0.4
entire area for both fracture lengths. Oil is accumulated strongly in the rock near the
well,Figure
continues to flow the
26 shows intodependence
it. Besides, there are areas
of mean that have
saturation of not changed
hydraulic much: thefluid in
fracturing
accumulation of oil and the displacement of hydraulic fracturing fluid in them occurs
elliptic zone on time.
very slowly.
26. The
Figure 26.
Figure Thedependence
dependenceofof
mean saturation
mean of hydraulic
saturation fracturing
of hydraulic fluid in
fracturing elliptic
fluid zone onzone
in elliptic time.on time.
The dependence in Figure 26 initially behaves linearly, after that it continues to de-
crease, but never reaches a constant: the asymptote in this case is s = 0. Therefore, any
sufficiently small or satisfactory saturation value can be taken as a criterion when it is
considered that the fracture is already cleaned. If we assume that the fracture is already
cleaned when the saturation of the hydraulic fracturing fluid reached 0.2, then it took
about 5 days 8 h to clean the fracture 20 m long, and it took 9 days 13 h to clean the fracture
moments of time. The instability of displacement is observed in all figures.
Energies 2022, 15, 1967 moments of time.
Figures 27–29The instability
show of displacement
the distribution is observed
patterns of the in
oilallseepage 30 for
figures. velocity of 35various
moments of time. The instability of displacement is observed in all figures.
Figure 27. Oil seepage velocity approximately 5 h after the start of cleaning.
In Figure 27, the largest flows are observed in the fracture, as well as near the pro-
duction well. It27,
In Figure canthebelargest
seen that
flowsforare
a larger fracture,
observed in thethe fluid flows
fracture, faster
as well in the
as near thearea
pro-near
Figure
the 27.
well. Oil seepage velocity approximately 5 h after the start of cleaning.
duction
Figure 27.well. It can bevelocity
Oil seepage seen that for a larger fracture,
approximately 5 h afterthe
thefluid
startflows faster in the area near
of cleaning.
the well.
In Figure 27, the largest flows are observed in the fracture, as well as near the pro-
duction well. It can be seen that for a larger fracture, the fluid flows faster in the area near
the well.
Figure 28.Oil
Figure28.
Figure 28. Oilseepage
Oil seepagevelocity
seepage velocityapproximately
velocity 66hh6after
approximately
approximately the
h after
after start
thethe of
ofcleaning.
start
start of cleaning.
cleaning.
Figure 29.Oil
Figure29. Oilseepage
seepagevelocity
velocityin
inapproximately
approximately15
15hhafter
afterthe
thestart
startofofcleaning.
cleaning.
Figure 29. Oil seepage velocity in approximately 15 h after the start of cleaning.
Figure 29. Oil seepage velocity in approximately 15 h after the start of cleaning.
In Figure 28, for the case when the length is 20 m, the flow occurs along the entire
fracture. In the case of a longer fracture, a breakthrough occurred in the area of the pro-
duction well. Thus, at this time moment, not the entire hydraulic fracture is involved in
the seepage process. In Figure 29 it can be noted that seepage occurs in almost the entire
Energies 2022, 15, 1967 area for both fracture lengths. Oil is accumulated strongly in the rock near the well,31con-
of 35
tinues to flow into it. Besides, there are areas that have not changed much: the accumula-
tion of oil and the displacement of hydraulic fracturing fluid in them occurs very slowly.
5.3. Effect of Fracture Fluid Viscosity on Displacement Dynamics
5.3. Effect of Fracture Fluid Viscosity on Displacement Dynamics
The purpose of the second numerical experiment is to investigate the effect of fracture
fluidThe purpose
viscosity ofon
(µ2) thethe
second numerical
cleaning processexperiment is to calculations
dynamics. Two investigate the
wereeffect of frac-
carried out
ture
withfluid viscosity
different (µ2) on
viscosities of the cleaning
fracturing process
fluid. dynamics. Two
The calculation calculations
parameters were
are given incarried
Table 4.
out with different viscosities of fracturing fluid. The calculation parameters are given in
Table
Table 4.
4. Calculation parameters for the second numerical experiment with different fracture fluid
viscosities.
Table 4. Calculation parameters for the second numerical experiment with different fracture fluid
viscosities. Number of cells Nx = 1001, Ny = 1001
The dimensions
Number ofofcells
the field L
Nxx==100 m, N
1001, Lyy == 100
1001m
The dimensions
Absolute permeability of
of the formation
field Lx = 100
K0 =m, 0.5LDarcy
y = 100 m
Absolute permeability of the formation

Formation porosity K0 =φ0.50 = Darcy
0.2
Formation porosity
Oil viscosity ϕ 0 = 0.2
µ1 = 0.01 Pa ∗ s
Oil viscosity µ1 = 0.01 Pa ∗ s
Fracturing fluid viscosity µ2 = 0.1 Pa ∗ s/µ2 = 0.08 Pa ∗ s
Fracturing fluid viscosity µ2 = 0.1 Pa ∗ s/µ2 = 0.08 Pa ∗ s
Pressure in injection wells Pin = 500 atm
Pressure in injection wells Pin = 500 atm
Fracture length
Fracture length L
Lfracture = 20 m
fracture = 20 m
Fractureabsolute
Fracture absolute permeability
permeability KKff==50 50 Darcy
Darcy
Fracture
Fracture porosity
porosity ϕ
φff == 0.4
0.4
In
In Figure 30 we
Figure 30 we can
cansee
seehow
howthethesaturation
saturationofofthe
the displaced
displaced fluid
fluid in in
thethe elliptical
elliptical re-
region
gion changes
changes with with
time. time. It is noticeable
It is noticeable that
that on on graph
each each graph three can
three parts parts
be can be distin-
distinguished.
guished. The first part, similar to the linear part, ends when there is a breakthrough
The first part, similar to the linear part, ends when there is a breakthrough of oil to of the
oil
to the production well. The second part is the curved up section, and
production well. The second part is the curved up section, and the third is the curvedthe third is the
curved down section.
down section.
Figure
Figure 30.
30. The
Thedependence
dependenceof
ofmean
meansaturation
saturationof
ofhydraulic
hydraulicfracturing
fracturingfluid
fluidin
inelliptic
ellipticzone
zoneon
ontime.
time.
The section curved up is the time interval when the displacement velocity increases.
This phenomenon can be explained by the fact that during this time interval, the liquid is
displaced from those areas that were not previously involved. These areas are shown in
Figure 31. Then, when these hard-to-reach areas are partially cleared, displacement again
slows down (third area). For the considered cases, the difference between the two graphs
(Figure 30) is not as significant as for the case of different fracture lengths. Using the same
criterion (s = 0.2), it took about 5 days 8 h to clean the fracture for µ2 = 0.1 Pa s, and it took
The
The section
section curved
curved upupis is
the time
the time interval
interval when
when the the displacement
displacement velocity
velocity increases.
increases.
This
Thisphenomenon
phenomenon cancanbebeexplained
explained byby thethefact
factthat during
that during this time
this time interval,
interval, the
the liquid
liquid is is
displaced from those areas that were not previously involved.
displaced from those areas that were not previously involved. These areas are shown in These areas are shown in
Figure
Figure 31.31.Then,
Then,when
when these
thesehard-to-reach
hard-to-reach areas
areas are partially
are partially cleared,
cleared, displacement
displacement again
again
Energies 2022, 15, 1967 slows down (third area). For the considered cases, the difference
slows down (third area). For the considered cases, the difference between the two 32 between the two graphs
graphs
of 35
(Figure
(Figure 30) is is
30) not asas
not significant
significant asas
forfor
thethecase
caseofofdifferent
different fracture
fracture lengths.
lengths. Using
Using the thesame
same
criterion
criterion (s (s
= 0.2),
= 0.2),it it
took
took about
about 5 days
5 days 8h 8h totoclean
cleanthethe fracture
fracture forfor𝜇 𝜇= = 0.1 𝑃𝑎𝑃𝑎
0.1 ∙ s,∙ s,
and
and it it
took 4 days 11 h to clean the fracture for 𝜇 =
𝜇 = 0.08 𝑃𝑎 ∙ s. It turns out that the ratio is 0.84,
4took
days411 daysh to11clean
h to the
clean the fracture
fracture for µ2 for= 0.08 Pa0.08
s. It𝑃𝑎
turns∙ s. It
outturns
thatout
thethat
ratiothe ratio which
is 0.84, is 0.84,
which
which is isgreater
greater than
than a aviscosity
viscosity ratio
ratio ofof0.8. Therefore,
0.8. Therefore, fracturing
fracturing fluid
fluidviscosity
viscosity has
hasa a
is greater than a viscosity ratio of 0.8. Therefore, fracturing fluid viscosity has a slightly
slightly
slightlylesser
lessereffect
effecton the duration
on the duration of the cleaning
of the process, process,
cleaningthan process, than fracture
thanlength. length.
fracture length.
lesser effect on the duration of the cleaning fracture
Figure
Figure31.
Figure31.The
31.Thedistribution
The ofofof
distribution
distribution saturation atatat
saturation
saturation different moments
different
different of
moments
moments time.
ofof time.
time.
Figure
Figure
Figure 32 shows
3232shows
showsthe presence
presence
the presence ofof
of aabreakthrough
breakthrough
a breakthrough ininthe
the
in case
case
the ofof
case the
ofthehigh
thehigh viscosity
viscosity
high and
viscosity and
the
and
the absence
absence of a of a breakthrough
breakthrough in in
the the
case case
of theof the
low low viscosity.
viscosity. It It
turns turns
out
the absence of a breakthrough in the case of the low viscosity. It turns out that a break- out
that a that a break-
breakthrough
through
for higher
through forforhigher
viscosity
higherviscosity
occurs
viscosity occurs
earlier earlier
than
occurs than
for lower
earlier thanfor lower
viscosity.
for lower viscosity.
When theWhen
viscosity. the
fracturing
When thefracturing
fluid has a
fracturing
high
fluid viscosity,
has a high a breakthrough
viscosity, a through
breakthrough areas near
through the
areas production
near the well
production
fluid has a high viscosity, a breakthrough through areas near the production well is pos- is possible.
well is Thus,
pos-
the
sible.hydraulic
Thus,
sible. Thus, the fracture
the effectively
hydraulic
hydraulic fracture
fracturedoes not work,
effectively
effectively does since
doesnotnot the
work, fracture
work, since
since isthe
the not connected
fracture
fracture is is
not
notto the
con-
con-
seepage
nected
nected process.
toto
thetheseepage
seepage process.
process.
Figure
Figure
Figure32.
32.
32.Oil
Oil
Oilseepage
seepage velocity
velocity
seepage approximately
approximately
velocity 55h5h38
approximately h3838
min
min after
after
min the
the
after start
start
the ofof
of
start cleaning.
cleaning.
cleaning.
6. Discussion
A planar-3D non-uniform propagation model with consideration of rock’s inhomoge-
neous fracture toughness has been developed for studying the effect of inhomogeneous
fracture toughness on the fracture propagation. This model allows fluid flow along the
radial and angular directions and considers the uneven propagation at different fracture tip
locations. The inhomogeneous fracture toughness of formation is described by two param-
eters (the fracture toughness contrast ζ and the proportion of the pay zone ξ). According
to these two controlling parameters, three groups of numerical experiments are designed
Energies 2022, 15, 1967 33 of 35
and classified by the proportion of pay zone, in each group, there are nine subgroups of ex-
periments corresponding to nine different levels of fracture toughness contrast. Computer
codes for these numerical experiments were developed and implemented using the Matlab
software package. The code was verified by comparing with exact solutions of a 2-D plane
fracture problem.
Numerical results of experiments show that the fracture propagation properties in-
cluding the fracture geometry, fluid pressure distribution and the fracture opening are
significantly dependent on these two controlling parameters, especially on the fracture
toughness contrast, which mainly controls the overall direction of fracture propagation:
when ζ < 1, i.e., the fracture toughness of pay zone is less than that of barrier zones, the frac-
ture mainly propagates along the pay zone (along the horizontal direction, Rl > Rh ); when
ζ > 1, i.e., the fracture toughness of pay zone is larger than that of barrier zones, the fracture
mainly propagates along the barrier zones (along the vertical direction, Rl < Rh ). Besides,
the controlling parameter ξ mainly enhances the capability of the fracture toughness con-
trast determining the overall fracture direction: when the value of ξ is small (ξ = 1/9), no
matter what the fracture toughness contrast is, it can only affect the partial fracture profile,
the overall fracture profile is still a circle with or without a pair of narrow wedges. When ξ
increases to 3/9, or 5/9, the combined effect of parameters ξ and ζ on the fracture profile
is very significant: the fracture profile is transformed into an approximate ellipse or an
approximate rectangle under some values of ζ (see Figures 16 and 17). The comparison of
the results of the fracture “half-length” (Rl ) and the fracture “half-height” (Rh ) indicates
that the velocity difference of uneven fracture propagation between the horizontal and
vertical directions is mainly accumulated in the early stage of fracture propagation, after
which the fracture non-uniformly propagates with an approximately constant velocity ratio
between the horizontal and vertical directions. The analysis of the fluid pressure and the
fracture opening at the wellbore shows that the greater the fracture toughness contrast, the
larger the fluid pressure at the wellbore is obtained; the fracture opening has the similar
evolution characteristic with two exceptions ζ = 1.75 and ζ = 2.0, because under the two
values of fracture toughness contrast, the pay zone almost does not propagate, which leads
to partially monotonous increase of fracture opening near the wellbore along the radius
(see Figure 13), which further leads to some decrease of fracture opening at the wellbore. It
should be noted that the presented results show the dependence of the hydraulic fracturing
process only on the inhomogeneity of the fracture toughness of the medium, while in real
conditions such inhomogeneity will also be associated with other inhomogeneities, such as
stiffness or permeability and porosity. On the other hand, the obtained results show that
the inhomogeneity of fracture toughness should not be ignored when taking into account
real heterogeneous environment without proper justification.
In the hydraulic fracture cleaning process, when the fracturing fluid is replaced by oil
from the deposit, instability of the displacement front takes place in the case of liquid with
higher viscosity (fracturing fluid) being displaced by a liquid of lower viscosity (oil). In a
small fracture, the cleaning process is faster, and the seepage process looks uniform. When
the fracturing fluid has a higher viscosity than the displacing agent does, it is possible that
the breakthrough will happen near the well, and not through the fracture. That means that
the hydraulic fracture does not work properly as an oil collector; the fracture remains filled
in by fracturing fluid not being involved in the oil seepage process. The breakthrough of
the fluid to the production well is the faster the greater is the viscosity of the fracturing
fluid. For the investigated hydraulic fracturing fluids with viscosities µ = 100 Mpa∗s and µ
= 80 Mpa∗s, the effect of fracture length on the average saturation of hydraulic fracturing
fluid is more noticeable than the effect of fracturing fluid viscosity. Due to the simplified
problem statement, quantitative results should be treated with caution, but qualitative
results should be kept in mind when evaluating the effectiveness of hydraulic fracturing.
Energies 2022, 15, 1967 34 of 35
Author Contributions: Conceptualization, N.S., D.P. and E.S.; methodology, D.P., K.L., A.S. and
E.S.; software, K.L. and E.S.; validation, K.L. and E.S.; formal analysis, A.K.; investigation, K.L.,
E.S., D.P. and A.S.; resources, N.S.; data curation, A.S.; writing—original draft preparation, K.L.;
writing—review and editing, D.P.; visualization, K.L. and E.S.; supervision, N.S. and C.Q.; project
administration, N.S.; funding acquisition, N.S. and C.Q. All authors have read and agreed to the
published version of the manuscript.
Funding: This work was carried out with support of the subsidy from Ministry of Science and
Education of the Russian Federation given to the Federal Science Center Scientific Research Institute
for System Analysis of the Russian Academy of Sciences to implement the state assignment on
the topic No. 0580-2021-0021 “Development of algorithms and codes for multiscale processes and
combustion simulations”.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Acknowledgments: The support of China Scholarship Council given to K.L. is gratefully acknowledged.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Independent Statistics & Analysis. Technically Recoverable Shale Oil and Shale Gas Resources: An Assessment of 137 Shale Formation in
41 Countries Outside the United States; Independent Statistics & Analysis, U.S. Department of Energy: Washington, DC, USA, 2013.
2. U.S. Energy Information Administration. Annual Energy Outlook 2020 with Projections to 2050; U.S. Energy Information Adminis-
tration: Washington, DC, USA, 2020.
3. Jia, C.; Zheng, M.; Zhang, Y. Unconventional hydrocarbon resources in China and the prospect of exploration and development.
Pet. Explor. Dev. 2012, 39, 139–146. [CrossRef]
4. Gerke, K.M.; Vasilyev, R.V.; Korost, D.V.; Karsanina, M.V.; Balushkina, N.S.; Khamidullin, R.; Kalmykov, G.A.; Mallants, D.
Determining physical properties of unconventional reservoir rocks: From laboratory methods to pore-scale modelling. In
Proceedings of the SPE Unconventional Resources Conference and Exhibition-Asia Pacific, Brisbane, Australia, 11–13 November
2013; p. SPE 167058. [CrossRef]
5. Economides, M.J.; Nolte, K.G. Reservoir Stimulation; Schlumberger: Houston, TX, USA; John Wiley & Sons: Hoboken, NJ, USA, 2000.
6. Smith, M.B.; Bale, A.B.; Britt, L.K.; Klein, H.H.; Siebrits, E.; Dang, X. Layered modulus effects on fracture propagation, proppant
placement, and fracture modelling. In Proceedings of the SPE Annual Technical Conference and Exihibition, New Orleans, LA,
USA, 30 September–3 October 2001; Society of Petroleum Engineers: Houston, TX, USA, 2001; p. SPE 71654.
7. Smirnov, N.N.; Tagirova, V.P. Analysis of power-law self-similar solutions to the problem of hydraulic fracture crack formation.
Mosc. Univ. Mech. Bull. 2007, 62, 48–54, reprinted from Cмирнoв, Н.Н.; Тaгирoвa, B.Р. Aнaлиз степенных aвтoмoдельных
решений зaдaчи o фoрмирoвaнии трещины гидрoрaзрывa. Bестн. МГУ. Мaтемaтикa Мехaникa 2007, 1, 48–54. [CrossRef]
8. Smirnov, N.N.; Tagirova, V.P. Problem of propagation of a gas fracture in a porous medium. Fluid Dyn. 2008, 43, 402–417,
reprinted from Cмирнoв, Н.Н.; Тaгирoвa, B.Р. Зaдaчa o рaспрoстрaнении трещины гaзoвoгo рaзрывa в пoристoй среде. Изв.
РAН. МЖГ 2008, 3, 77–93. [CrossRef]
9. Li, K.; Smirnov, N.N.; Qi, C.; Kiselev, A.B.; Pestov, D.A. The numerical asymptotic solution to initial condition problem of
preexisting plane-strain hydraulic fracture with fluid lag. Eng. Fract. Mech. 2020, 239, 107296. [CrossRef]
10. Li, Q.; Xing, H.; Liu, J.; Liu, X. A review on hydraulic fracturing of unconventional reservoir. Petroleum 2015, 1, 8–15. [CrossRef]
11. Slatt, R.M. Important geological properties of unconventional resource shales. Cent. Eur. J. Geosci. 2011, 3, 435–448. [CrossRef]
12. Warpinski, N.R.; Schmidt, R.A.; Northrop, D.A. In-situ stresses: The predominant influence on hydraulic fracture containment. J.
Pet. Technol. 1982, 34, 653–664. [CrossRef]
13. Warpinski, N.R.; Clark, J.A.; Schmidt, R.A.; Huddle, C.W. Laboratory investigation on the effect on in-situ stresses on hydraulic
fracture containment. Soc. Pet. Eng. J. 1982, 22, 333–340. [CrossRef]
14. Fung, R.L.; Vijayakumar, S.; Cormack, D.E. Calculation of vertical fracture containment in layered formations. SPE Form. Eval.
1987, 2, 518–522. [CrossRef]
15. Van Eekelen, H.A.M. Hydraulic fracture geometry: Fracture containment in layered formations. Soc. Pet. Eng. J. 1982, 22, 341–349.
[CrossRef]
16. Zhang, X.; Wu, B.; Jeffrey, R.G.; Connell, L.D.; Zhang, G. A pseudo-3D model for hydraulic fracture growth in a layered rock. Int.
J. Solids Struct. 2017, 115–116, 208–223. [CrossRef]
17. Zhang, X.; Wu, B.; Connell, L.D.; Han, Y.; Jeffrey, R.G. A model for hydraulic fracture growth across multiple elastic layers. J. Pet.
Sci. Eng. 2018, 167, 918–928. [CrossRef]
18. Hanson, M.E.; Shaffer, R.J. Some results from continuum mechanics analyses of the hydraulic fracturing process. Soc. Pet. Eng. J.
1980, 20, 86–94. [CrossRef]
Energies 2022, 15, 1967 35 of 35
19. Baykin, A.N.; Golovin, S.V. Modelling of hydraulic fracture propagation in inhomogeneous poroelastic medium. J. Phys. Conf. Ser.
2016, 722, 012003. [CrossRef]
20. Golovin, S.V.; Baykin, A.N. Influence of pore pressure on the development of a hydraulic fracture in poroelastic medium. Int. J.
Rock Mech. Min. Sci. 2018, 108, 198–208. [CrossRef]
21. Gao, Q.; Han, S.; Cheng, Y.; Yan, C.; Sun, Y.; Han, Z. Effects of non-uniform pore pressure field on hydraulic fracture propagation
behaviors. Eng. Fract. Mech. 2019, 221, 106682. [CrossRef]
22. Budennyy, S. An enhanced pseudo-3D model of hydraulic fracture in multi-layered formation. In Proceedings of the SPE Annual
Technical Conference and Exhibition, San Antonio, TX, USA, 9–11 October 2017.
23. Adachi, J.; Siebrits, E.; Peirce, A.; Desroches, J. Computer simulation of hydraulic fractures. Int. J. Rock Mech. Min. Sci. 2007, 44,
739–757. [CrossRef]
24. Xiao, H.T.; Yue, Z.Q. A three-dimensional displacement discontinuity method for crack problems in layered rocks. Int. J. Rock
Mech. Min. Sci. 2011, 48, 412–420. [CrossRef]
25. Siebrits, E.; Peirce, A.P. An efficient multi-layer planar 3D fracture growth algorithm using a fixed mesh approach. Int. J. Numer.
Methods Eng. 2002, 53, 691–707. [CrossRef]
26. Warpinski, N. Fracture Growth in Layered and Discontinuous Media; United States Environmental Protection Agency: Washington,
DC, USA, 2011. Available online: https://www.epa.gov/hfstudy/fracture-growth-layered-and-discontinuous-media (accessed
on 1 March 2022).
27. Chong, K.P.; Kuruppu, M.D.; Kuszmaul, J.S. Fracture toughness determination of layered materials. Eng. Fract. Mech. 1987, 28,
43–54. [CrossRef]
28. Chandler, M.R.; Meredith, P.G.; Brantut, N.; Crawford, B.R. Fracture toughness anisotropy in shale. J. Geophys. Res. Solid Earth
2016, 121, 1706–1729. [CrossRef]
29. Su, X.; Chen, P.; Ma, T. Evaluation of shale fracture toughness based on micrometer indentation test. Petroleum 2019, 5, 52–57.
[CrossRef]
30. Kramarov, V.; Parrikar, P.N.; Mokhtari, M. Evaluation of fracture toughness of sandstone and shale using digital image correlation.
Rock Mech. Rock Eng. 2020, 53, 4231–4250. [CrossRef]
31. Thiercelin, M. Fracture toughness and hydraulic fracturing. Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 1989, 26, 177–183.
[CrossRef]
32. Thiercelin, M.; Jeffrey, R.G.; Naceur, K.B. Influence of fracture toughness on the geometry of hydraulic fractures. SPE Prod. Eng.
1989, 4, 435–442. [CrossRef]
33. Peirce, A.; Detournay, E. An implicit level set method for modelling hydraulically driven fractures. Comput. Methods Appl. Mech.
Eng. 2008, 197, 2858–2885. [CrossRef]
34. Salehi, S. Full fluid-solid cohesive finite-element model to simulate near wellbore fractures. J. Energy Resour. Technol. 2015, 137,
012903. [CrossRef]
35. Kiselev, A.B.; Zacharov, P.P. On numerical modeling of dynamics of irreversible deforming and fracture of oil bearing layer. Mat.
Modelirovanie 2013, 25, 62–74.
36. Sesetty, V.; Ghassemi, A. A numerical study of sequential and simultaneous hydraulic fracturing in single and multi-lateral
horizontal wells. J. Pet. Sci. Eng. 2015, 132, 65–76. [CrossRef]
37. Bui, H.D. An integral equations method for solving the problem of a plane crack arbitrary shape. J. Mech. Phys. Solids 1977, 25,
29–39. [CrossRef]
38. Savitski, A.A.; Detournay, E. Propagation of a penny-shaped fluid-driven fracture in an impermeable rock: Asymptotic solutions.
Int. J. Solids Struct. 2002, 39, 6311–6337. [CrossRef]
39. Perkins, T.K.; Kern, L.R. Widths of hydraulic fractures. J. Pet. Technol. 1961, 13, 937–949. [CrossRef]
40. Nordgren, R.P. Propagation of a vertical hydraulic fracture. Soc. Pet. Eng. J. 1972, 12, 306–314. [CrossRef]
41. Kolenkina (Skryleva), E.I.; Smirnov, N.N.; Nikitin, V.F.; Fakhretdinova, R.R.; Makeeva, M.N. Modeling of Liquid Displacement
from the Porous Medium Taking into Account the Presence of Hydraulic Fracture. In Advanced Problems in Mechanics; Institute for
Problems in Mechanical Engineering RAS: St. Petersburg, Russia, 2022; ISSN 2312-9921.
42. Brooks, R.H.; Corey, A.T. Hydraulic Properties of Porous Media; Hydrology Paper No. 3; Colorado State University: Fort Collins,
CO, USA, 1964.

Energies 15 01967

Uploaded by

Copyright:

Available Formats

Energies 15 01967

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Energies 15 01967

Uploaded by

Copyright:

Available Formats

energies

Received: 21 January 2022

Energies 2022, 15, 1967. https://doi.org/10.3390/en15061967 https://www.mdpi.com/journal/energies

Energies 2022, 15, 1967 3 of 35

a plane-3D model considering an inhomogeneous fracture toughness for hydraulic frac-

2.1. Governing Equations

2.1.1. Elastic Deformation

where p( x, y, t) is the net pressure in the fracture, E0 = E/ 1 − ν2 is the plane strain

2.1.2. Fluid Flow

2.1.3. Boundary Conditions

w(sc , t) = 0, q(sc , t) = 0, (4)

2.1.4. Initial Conditions

3.1. Fracture Discretization

Energies 2022, 15, 1967 7 of 35

x = r cos θ, y = r sin θ, xi = ri cos θ j , y j = ri sin θ j . (14)

3.3. Implicit Algorithm

Energies 2022, 15, 1967 9 of 35

4.2. Experiment Results

Energies 2022, 15, x FOR PEER REVIEW 13 of 36

Figure 11.of Distributions of fracture opening of Group 2 numerical experiments (𝜉 = 3⁄9) w

Figure 12. Distributions

4.3. Validation of Model and Algorithm

4.4. Analysis of Results

Energies 2022, 15, x FOR PEER REVIEW 19 of 36

Figure 16. Fracture

(a): =0.5 (b): =0.625 (c): =0.75

x (m) x (m) x (m)

(d): =0.875 (e): =1.0 (f): =1.25

(g): =1.5 (h): =1.75 (i): =2.0

The shape coefficient 𝜂 can be formulated as follows

(a): =3/9, =0.5 (b): =3/9, =0.625 (c): =3/9, =0.75

(d): =3/9, =0.875 (e): =3/9, =1.0 (f): =3/9, =1.25

(g): =3/9, =1.5 (h): =3/9, =1.75 (i): =3/9, =2.0

1.5 R (m) 1.5 0.6

0.5 0.5 0.5 0.4 0.5 0.2

Energies 2022, 15, x FOR PEER REVIEW 24 of 36

Figure 22. Evolution

5. Hydraulic Fracture Cleaning Process

5.1. Mathematical Model for Seepage Flows of Multi-Phase Fluids

Here, ϕ is porosity, s is saturation, $ is density (intrinsic), uk,j is j-th component of the

where f k = m m+km2 is the proportion of relative phase mobility.

K̂ (x, y) = K ( x, y)· exp(δξ ), (31)

Energies 2022, 15, 1967 28 of 35

Number of cells 𝛴 : 𝑃=𝑃 <𝑃 Nx = 1001, Ny = 1001

The initial Formation porosity

Figure 26 shows the dependence of mean saturation of hydraulic fracturing fluid in

Energies 2022, 15, x FOR PEER REVIEW 29 of 36

moments of time. The instability of displacement is observed in all figures.

Absolute permeability of the formation

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.