Journal of Petroleum Science and Engineering 168 (2018) 400–408

Contents lists available at ScienceDirect

Journal of Petroleum Science and Engineering

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

Experimental simulation of the hydraulic fracture propagation in an T

anthracite coal reservoir in the southern Qinshui basin, China
Jun Liua,b,d, Yanbin Yaoa,b,∗, Dameng Liua, Lulu Xuc, Derek Elsworthd, Saipeng Huanga,
Wanjing Luoa
a
School of Energy Resource, China University of Geosciences (Beijing), Beijing 100083, China
b
Coal Reservoir Laboratory of National Engineering Research Center of CBM Development & Utilization, China University of Geosciences, Beijing 100083, China
c
Hubei Geological Survey, Wuhan, Hubei 430034, China
d
Department of Energy and Mineral Engineering, G3 Center and EME Energy Institute, Pennsylvania State University, University Park, PA 16802, USA

A R T I C LE I N FO A B S T R A C T

Keywords: As coalbed methane (CBM) reservoirs have extremely low permeability, hydraulic fracturing is a common sti-
Coalbed methane mulation process for enhancing the CBM production. An in-depth understanding about the propagation me-
Anthracite coal chanism of hydraulic fractures in coal is important for designing a hydraulic fracturing process and thus for
Hydraulic fracture propagation improving the CBM production. This study performed a set of true tri-axial fracturing experiments on six block
Reservoir stimulation
samples (300 mm × 300 mm × 300 mm, including raw coal and artificial roof/floor) with consideration of in
Qinshui basin
situ conditions, aiming at simulating the propagation of hydraulic fractures in the CBM reservoir in the Changzhi
field, southern Qinshui basin. Four groups of experiments were organized to evaluate the influences from the
pre-existing natural fracture, in situ stresses and injection flow rates on the hydraulic fracture propagation.
Meanwhile, five series of numerical simulations were constructed to model the relationship between in situ
horizontal stresses and hydraulic fracture propagation. The results show that the hydraulic fracture propagates
only along the pre-existing natural fracture direction under a small approaching angle, while it propagates along
both the directions of the pre-existing natural fracture and the maximum horizontal principal stress
(σ < SUB > H < /SUB >) under a large approaching angle. Whether pre-existing natural fractures exist or not
can result in a distinct influence on hydraulic fracture propagation. Hydraulic fractures straightly propagate
along the σ < SUB > H < /SUB > direction under a high value of horizontal stresses difference coefficient (Kh),
while they tend to deviate from the σ < SUB > H < /SUB > direction under a low Kh value. The influence of Kh
is greater than that of the horizontal stresses difference (Δσ) in determining fracture propagation to extend along
the σ < SUB > H < /SUB > direction in the coal seam. Large approaching angle, high in situ stresses and a high
injection flow rate are three major factors to cause the roof/floor broken by hydraulic fluids.

1. Introduction
Hydraulic fracturing is usually defined as a process by which frac-
tures are initiated and propagate resulted from hydraulic loading ap-
plied by fluid inside the fractures, which is a primary means in en-
hancing gas production from low-permeable rocks, such as tight
sandstone, shale and coal (Ferrill et al., 2014; Yuan et al., 2015). Hy-
draulic fracturing can significantly increase the number of inter-
connected cracks and apertures and result in the improvement of gas
permeability even though that some gas/oil is extruded by high-pres-
sure fluid (Abass et al., 1992; Gan and Elsworth, 2016; Yuan et al.,
2017). The hydraulic fracturing technology has hitherto been ex-
tensively employed in the petroleum industry to stimulate oil or gas
productivity (Yuan et al., 2015). For example, almost all shale gas wells
and more than 90% of coalbed methane (CBM) wells are improved by
hydraulic fracturing in United States (Zhang and Bian, 2015).
Although gas extractions from coal and shale are both highly de-
pendent on hydraulic fracturing, production requirement for the pro-
pagation of hydraulic fractures in coal seam is different from that in
shale because coal and shale have large differences in petrological and
mechanical characteristics (Penny et al., 1991; Liu et al., 2017; Wu
et al., 2018). Hydraulic fracturing is expected to create extensive and
complex fracture pathways in shale, which aims at minimizing the
spacing between fractures and maximizing the total stimulated re-
servoir volume (Sobhaniaragh et al., 2016; Yuan et al., 2016). In con-
trast, in coal seam, hydraulic fracturing is supposed to stimulate long

∗
Corresponding author. School of Energy Resource, China University of Geosciences (Beijing), Beijing 100083, China.
E-mail address: yyb@cugb.edu.cn (Y. Yao).

https://doi.org/10.1016/j.petrol.2018.05.035
Received 30 November 2017; Received in revised form 5 May 2018; Accepted 11 May 2018
Available online 16 May 2018
0920-4105/ © 2018 Elsevier B.V. All rights reserved.
J. Liu et al.
Fig. 1. Propagation mechanism of a hydraulic fracture in layered formations (Zhao and Chen, 2010).
and stable cracks that can be connected with the pre-existing natural
fractures to improve drainage efficiency and gas extraction (Wu et al.,
2018). Due to the special demand for hydraulic fracturing in coal seam,
better understanding the propagation mechanism of hydraulic fractures
in coal seam is of great significance for designing an optimal hydraulic
fracturing process and thus for enhancing CBM production.
Existed achievements suggested the propagation of hydraulic frac-
ture is influenced by various factors, among which the in situ stresses
and the pre-existing natural fractures are two dominant factors
(Renshaw, 1994; Yang et al., 2011; Zou et al., 2016). According an
experimental study reported by Tan et al. (2017), larger horizontal
stresses difference (Δσ) exerts more control on induced fracture mor-
phology and makes the hydraulic fractures expand vertically in coal.
Other experimental research (Beugelsdijk et al., 2000; Wu et al., 2018)
also suggested that higher Δσ tend to straighten the hydraulic fracture
in coal and make the fracture surfaces more regular. The pre-existing
natural fracture also acts important influence on the propagation of
hydraulic fracture in coal. Based on a numerical study, Wang et al.
(2017) stated that natural fracture contributes greatly to form a com-
plex fracture network and increase coal porosity in the fractured region,
which indirectly improving coal permeability. The third influence
factor for the propagation of hydraulic fracture is the mechanical
strength properties of coal seam and roof/floor. Wang et al. (2014) ever
did a numerical simulation study and found that both the elasticity
modulus and Poisson's ratio have a positive relationship with the scale
of induced hydro-fracture in coal. Different from the view of Wang et al.
(2014), Jiang et al. (2016) pointed out higher elasticity modulus im-
pedes the development of induced hydro-fracture in coal. According to
Jiang et al. (2016), the length, width and height of main hydraulic
fracture respectively decreases about 21.3%, 13.8% and 14.2% when
the elasticity modulus increases from 1.43 GPa–2.43 GPa. Previous re-
searches also suggested that other mechanical parameters such as ten-
sile strength and uniaxial compressive strength of coal have negligible
influence on the development of hydraulic fractures (Wang et al.,
2014). The mechanical strength of the roof/floor has certain influence
on hydro-fracturing because the height of hydraulic fractures will be
significantly controlled by the roof/floor and the fractures can only
propagate in coal seam if the strength of roof/floor is stronger than that
of coal (Wu et al., 2018). In addition, the propagation of hydraulic
fracture is also related to the injected fracturing fluid: the injected fluid
with high-viscosity commonly forms short and wide hydro-fractures in
coal (Shimizu et al., 2011; Fan and Zhang, 2014); and different injec-
tion flow rate, time and pressure can induce different propagation rule
of induced hydro-fractures in coal (Wang et al., 2014).
401
Journal of Petroleum Science and Engineering 168 (2018) 400–408
Although there are extensive investigations on the influences of the
hydraulic fractures propagation in coal, most of these researches are
based on mathematic modeling or simulation (e.g., Wang et al., 2014;
Wang et al., 2017), other experimental researches only discussed some
limited influence factors such as injection fluid (Fan and Zhang, 2014),
natural fractures (Wang et al., 2017), or in situ stress (Tang et al.,
2011). Moreover, there are also some other deficiencies for the existed
simulation experiments on coals, for example ignoring the influences of
the roof/floor (Jiang et al., 2016), or using artificial specimens (Lin and
Du, 2011). As for as the anthracite coals of southern Qinshui basin in
China, there is few report relating to direct predicting the hydraulic
fracture propagation. In this study, a set of true tri-axial hydraulic
fracturing simulation experiments are conducted on large natural an-
thracite coal samples from southern Qinshui basin. Different from ex-
isting researches, this study considered much more comprehensive in-
fluences on hydraulic fractures propagation, including the depth, tri-
axial stresses, pre-existing natural fracture, roof-floor, and injection
flow rate. Thus, this study is helpful for forecasting the hydraulic
fractures propagation in study coal seam, which would provide theo-
retical basis and technical support for the efficient CBM extraction in
the southern Qinshui basin.
2. Mechanism of fracture propagation
Zhao and Chen (2010) established a model to explain the hydraulic
fracture propagation in coal-bearing strata (Fig. 1), which is employed
in this study. For the fracture propagation in coal and its roof and floor,
in this model, the critical hydraulic pressures can be obtained by
K1c π
P1 = + σh
2 h (1)
P2 = σv cos2 θ + σh sin2 θ + σt (2)
K2c π
P3 = + σh
2 h (3)
where P1 is the critical pressure (MPa) above which a hydraulic fracture
stops extending at the interface, P2 is the critical pressure (MPa) above
which a hydraulic fracture extends along the interface I or II, P3 is the
critical pressure (MPa) above which a hydraulic fracture penetrates
through the roof or floor, K1c is the fracture toughness of coal (MPa·m1/
2
), K2c is the fracture toughness of the roof or floor (MPa·m1/2), h is the
fracture half-height (m), σt is the tensile strength of interface I or II
(MPa), and θ is the stratigraphic dip of the interface (°).
The values of P1, P2 and P3 are used to predict the patterns of
J. Liu et al.
Fig. 2. Illustration of sample preparation by combining coal and artificial roof/floor (a), and picture of an actual specimen (Sample 4#) after fracturing experiment
(b).
fracture propagation when hydraulic fractures reach formation inter-
faces, such as stopping the extension, extending along the formation
interface, and directly penetrating through the roof and floor of the coal
seam.
3. Experiments
3.1. Samples preparation
In this study, hydraulic fracturing experiments are performed in a
true tri-axial pressure system descried by Zhou et al. (2008). The nat-
ural anthracite coal samples were collected from the No. 3 coal seam of
the Shanxi formation in the Changzhi CBM field, southern Qinshui
basin. Six cubic samples (each with a size of
300 mm × 300 mm × 300 mm) were specially made, in which each
sample was composed of a 75-mm-thick layer of artificial roof, a 150-
mm-thick layer of coal and a 75-mm-thick layer of artificial floor
(Fig. 2). Among the six samples, the coal of 1#, 2# and 3# was ob-
tained from a burial depth of 930 m, while that of 4#, 5# and 6# was
collected from a burial depth of 650 m.
To model the in situ reservoir conditions in the Changzhi CBM field,
both the artificial roof and floor parallelly contact the coal seam be-
cause the Shanxi formation has a conformable contact with upper and
bottom sedimentary rocks (Cai et al., 2011). In this study, we define the
in situ roof and floor of 1#, 2# and 3# samples as ‘sandstone’ and those
of 4#, 5# and 6# samples as ‘shale’. According to test results of the in-
situ roof/floor samples at each sampling location, the compressive
strength of sandstones varies from 29.8 MPa to 35.2 MPa with an
average of 33.0 MPa, while that of shales averages 21.5 MPa with the
range of 18.6–24.3 MPa. Since it is hard to bond the sandstone/shale
with coal, we used artificial rocks as roof or floor of coal during the
experiments. Made by mixing cement (high clay content) with quartz
sand and water, the artificial rocks with higher and lower mechanical
strength are employed to simulate the in situ sandstone and shale, re-
spectively. The artificial roof/floor was made as far as possible to match
the mechanical strength of in-situ roof/floor. To choose ideal artificial
roof/floor, we did uniaxial compressive strength tests for a series of
artificial cores with different sand/cement mass ratios. According to the
results shown in Fig. 3, we chose sand/cement mass ratios of 1:3 and
1:7 to make artificial rocks and to simulate the real sandstone and shale,
respectively.
3.2. In situ stresses measurement
The depth-dependent vertical stress (σv) is obtained from the density
logging, while the maximum (σ < SUB > H < /SUB >) and minimum
402
Journal of Petroleum Science and Engineering 168 (2018) 400–408
(σh) horizontal stresses are measured on the basis of the Kaiser Effect.
By the Kaiser Effect theory, when the exerted stress reaches the max-
imum value of the coal seam loaded in the past, acoustic emission
would noticeably increase (Lavrov, 2003; Lehtonen et al., 2012; Hsieh
et al., 2014). Prior to the Kaiser Effect test, three small coal cores
(diameter of 25 mm and length of 50 mm) were horizontally drilled
from a major coal sample along the radial directions at increments of
45° in the horizontal plane. Then, for these coal cores, normal stresses
were measured to calculate σ < SUB > H < /SUB > and σh.
The Kaiser Effect tests were performed for samples 3# and 4#,
whose tri-axial stresses (σv/σ < SUB > H < /SUB > /σh) are 18.2/
20.4/16.4 MPa and 12.7/15.8/11.8 MPa, respectively. These results
were used in the design of the fracturing experiments.
3.3. Fracturing experiment
During the fracturing experiments, guar fracturing fluid with a
viscosity of 6 mp·s was employed. Meanwhile, a yellow fluorescent dye
was added into the fracturing fluid for better tracing of the hydraulic
fractures. According to the principle of similarity criterion (Liu et al.,
2000), the laboratory injection flow rate was 0.33 mL s−1, which is
used to simulate the injection flow rate of 6.0 m3 min−1 during a field
fracturing operation.
In this study, six samples were organized into four groups to eval-
uate the influences on hydraulic fractures propagation from four dif-
ferent factors (Fig. 4). Among which, Group I (1# and 2#) is designed
to model the influence from the approaching angle of pre-existing
natural fracture, Group II (4# and 6#) simulates the influence from the
existence of pre-existing natural fracture, Group III (3# and 4#) sym-
bolizes the influence from different tri-axial stresses, and Group IV (4#
and 5#) characterizes the influence from different injection flow rates.
After the hydraulic fracturing experiments, the fractured coal blocks
were separated from the roof and floor to observe the propagation of
fractures within the samples.
3.4. Stress simulation analysis
The effect of horizontal stress on hydraulic fracture propagation was
simulated by using the finite element analysis software ANSYS 13.0.
The simulation was based on the assumptions that 1) the simulated
rocks are treated as isotropic linear elastic materials and the strata dip
is close to horizontal, and 2) the maximum tensile-stress criterion was
taken as the failure criterion with no consideration of the interaction
among the fluid field, geothermal field and stresses field. To simulate
the in situ reservoir conditions, the input parameters were set as fol-
lows: Young's modulus of 3800 MPa, Poisson's ratio of 0.37, tensile
J. Liu et al.
Fig. 3. Compressive strength test for artificial cores with different sand/cement mass ratios. The dash line represents the average compressive strength of in-situ roof/
floor at each sampling location.
strength of 0.55 MPa, coal density of 1.40 g/cm3, fracturing fluid visc-
osity of 6 mp·s, and injection flow rate of 0.33 mL/s. In total, five series
of models were constructed. Because the experimental Group III (3#
and 4#) was set to evaluate the influence from different in-situ stresses,
in this stress simulation analysis, the input mechanical properties were
the average values of experimental tests for coal samples 3# and 4#,
while the injection parameters were the same with the practical op-
eration in Group III.
4. Experiment results and discussions
The results of the fracturing experiments are exhibited in Table 1,
and the fracture propagations in the six samples are shown in Fig. 5. It
can be found that vertical fractures are common in the selected samples
(1# through 5#), whereas horizontal fractures only develop in limited
samples (e.g., 3# and 6#). The extension of vertical fractures occurs
along the natural fracture direction in 1#, while it occurs along the
σ < SUB > H < /SUB > direction in 4# and 5#. The vertical fractures
in 2# extend along two directions, i.e., the directions of the natural
fracture and σ < SUB > H < /SUB > . The horizontal fracture in 6#
extends along the direction of the natural fracture. The fractures in 3#
are complex, with vertical fractures extending along the directions of
the natural fracture and σ < SUB > H < /SUB > and horizontal
403
Journal of Petroleum Science and Engineering 168 (2018) 400–408
fractures occurring along bedding plane in the coal seam.
The occurrence of yellow fluorescent dye in the roof or floor in-
dicates the roof or floor is cut through by hydraulic fractures, and vice
versa. The experiment results demonstrate that both the roofs and floors
are intact for 1#, 4# and 6#, the roof are intact but the floors are cut
through for 2# and 3#, while both the roof and floor are cut through
for 5# (Table 1).
In general, multiple types of fractures are induced in coal reservoirs
during the hydraulic fracturing experiments. It seems that the propa-
gation of hydraulic fractures in coal reservoirs is much more complex
than those in sandstone, carbonate, and shale reservoirs. Vertical frac-
tures commonly develop within the coals, and limited horizontal frac-
tures usually occur at the upper section close to interface between the
roof and the coal. The roof or floor can be cut through only when there
is a large approaching angle, high tri-axial stresses, or a high injection
flow rate. The details about the experimental results are given below.
4.1. Impacts of pre-existing natural fracture on fracture propagation
4.1.1. The approaching angle of pre-existing natural fracture
Group I (1# and 2#) provides an example of the influence from the
approaching angle on fracture propagation. As shown in Fig. 5, a hy-
draulic fracture propagates only along the natural fracture direction
Fig. 4. Hydraulic fracturing experiments simulating
different in situ reservoir situations (θ1, approaching
angle; NF, pre-existing nature fracture; Q, injection
flow rate; S, in situ stresses). Note the artificial roof/
floor of samples 1#/2#/3# and samples 4#/5#/6#
are made with sand/cement mass ratios of 1:3 and 1:7
to simulate the sandstone and shale, respectively.
J. Liu et al. Journal of Petroleum Science and Engineering 168 (2018) 400–408

Table 1
Fracture propagation in the coal, roof and floor.
Samples σv/σ < SUB > H < /SUB > /σh (MPa) Q (mL·s−1) θ1 (°) Location of hydraulic fracture propagation

Coal Roof/Floor

1# 18.2/20.4/16.4 0.33 30° VF along NF None/None

2# 18.2/20.4/16.4 0.33 70° VF along NF and σ < SUB > H < /SUB > None/VF
3# 18.2/20.4/16.4 0.33 / VF along NF and σ < SUB > H < /SUB > HF along bedding plane None/VF
4# 12.7/15.8/11.8 0.33 / VF along σ < SUB > H < /SUB > None/None
5# 12.7/15.8/11.8 0.66 / VF along σ < SUB > H < /SUB > VF/VF
6# 12.7/15.8/11.8 0.33 / HF along NF None/None

Note: Q is injection flow rate; θ1 is approaching angle; VF is vertical fracture; HF is horizontal fracture; NF is pre-existing natural fracture.

Fig. 5. Fracture propagation in six samples. The red lines schematically indicate the main hydrologic fracturing directions. NF is pre-existing natural fracture. Note,
in figure (e), the fracture in the roof/floor along the σh direction are artificial during separation of coal and roof/floor after fracturing. (For interpretation of the
references to color in this figure legend, the reader is referred to the Web version of this article.)

404
J. Liu et al.
under the small approaching angle (30°) 1# (Fig. 5a). By contrast,
hydraulic fractures propagate along the directions of σ < SUB > H < /
SUB > and the natural fracture and even cut through the floor under
the high approaching angle (70°) of 2# (Fig. 5b). Zhou et al. (2008)
ever did some experimental simulations on artificial blocks which were
made by cement and fine sand with a mass ratio of 1: 1. The results of
this study agree well with the Zhou et al. (2008), in which the hydraulic
fractures crossed the pre-fractures only at high approaching angles of
60° or greater. It seems that larger approaching angle (> 60°) tends to
induce more complex fracture system.
Because the tensile strength is weak on the interface between the
coal seam and its roof and floor, the tension failure criterion can be
used to explain the mechanism of the experimental results of 1# and 2#
(Zhao and Chen, 2010). According to Eq. (2), the critical hydraulic
pressure of the natural fracture plane (Pn) is given by
Pn = σH cos2 θ1 + σh sin2 θ1 + σt1 (4)
where θ1 is the approaching angle (°) and σt1 is the tensile strength of
the natural fracture (MPa). After a trigonometric function transforma-
tion, Eq. (4) can be written as
σH + σh σ − σh
Pn = − H cos 2θ1 + σt1
2 2 (5)
According to Eq. (5), Pn increases with the approaching angle ran-
ging from 0° to 90°. Sample 2# has a lager approaching angle and thus a
higher Pn than sample 1#. In contrast, the lower Pn of sample 1# made
the hydraulic fracture propagate easily along the natural fracture in 1#,
and the fracturing fluid leak-off behavior in the natural fracture re-
sulted in the rapid decrease in the fluid pressure (Pw) (point N in
Fig. 6a). The fluid leak-off pulls down the Pw, so that the hydraulic
fracture can hardly propagate along the σ < SUB > H < /SUB >
direction and cut through the roof or floor in sample 1# (Fig. 6a).
Regarding to sample 2#, due to the larger Pn, the hydraulic fracture can
hardly propagate along the natural fracture and thus there is few fluid
leak-off in natural fracture, so the Pw just decreases a little at the point
N in Fig. 6b when the fracturing fluid reaches to the natural fracture.
The continuous accumulation of Pw results in the broken of floor by
hydraulic fluids at Point M in Fig. 6b.
4.1.2. The existence of pre-existing natural fracture
Samples 4# and 6# in Group II can be used to investigate the in-
fluences from the existence of natural fractures. There is no pre-existing
natural fracture in 4#, while a major pre-existing natural fracture exists
in 6# and is parallel to the bedding plane. For 4#, a vertical fracture is
created along the σH direction (Fig. 5d), which is consistent with the
numerical simulations based on conceptual coal models with 2D size of
Fig. 6. Fracturing curve of the six samples.
405
Journal of Petroleum Science and Engineering 168 (2018) 400–408
10 m × 10 m (Wang et al., 2017). The experimental results also indicate
that the fluid leak-off occurs in the interface of coal and roof/floor
(Fig. 5d), resulting in that the Pw has a transitory breakdown at point N
in Fig. 6d, then a slight increase, followed by a smooth phase with rare
variations (Fig. 6d). By contrast, in sample 6#, a horizontal fracture is
created along the major pre-existing natural fracture parallel to the
bedding plane (Fig. 5f), which differs from the numerical simulation
study by Wang et al. (2017) where both vertical and horizontal cracks
are induced. During the fracturing process, there is a rapid decrease of
Pw caused by fluid leak-off when the fracturing fluid reaches the pre-
existing natural fracture of sample 6# (Point N in Fig. 6f). Then the Pw
increases slowly and finally tends to remain stable when the fracturing
fluid reaches the boundary of the sample (Fig. 6f). The relative low Pw
value after Point N in Fig. 6f makes it difficult for a hydraulic fracture to
develop along another orientation, resulting in that the hydraulic
fractures propagate only along the pre-existing natural fracture in 6#.
The comparison indicates whether the pre-existing natural fracture
exists or not can result in a distinct influence on the hydraulic fracture
propagation.
4.2. Tri-axial stresses' impact on fracture propagation
Group III (3# and 4#) was used to investigate the influences from
tri-axial stresses on hydraulic fractures propagation. As shown in Fig. 5,
the single hydraulic fracture is formed along the σH direction and the
roof and floor remain intact in 4# (Fig. 5d), while many fractures are
formed along multi-directions in 3# (Fig. 5c). The major vertical frac-
ture in 3# propagates along the σH direction at first, then deviates from
the σH direction, and finally breaks through the floor (Fig. 5c).
4.2.1. Explanations for formation of multi-fractures
The development of multi-fractures in coal seam is related to the
breakdown pressure (Pf), in which the larger Pf value usually means the
hydraulic fracture is harder to be created. Pf is given by the following
equation (Haimson and Cornet, 2003):
Pf = 3σh − σH + σt 2 − αPp (6)
where σt2 is the tensile strength of coal (MPa), α is the biot coefficient
(dimensionless), and Pp is the pore pressure of coal (MPa).
According to Eq. (6), for its lager σh value, 3# has a greater Pf value
than 4#, resulted in the hydraulic fracture is harder formed in 3# when
compared to the situation of 4#. Besides, the fluid leak-off occurs in the
interface between coal and roof/floor of 4# (Fig. 5d) but not in that of
3# (Fig. 5c). Thus the sample 3# has a greater Pw accumulation than
sample 4# during the fracturing process (Fig. 6c and d). In this situa-
tion, during the fracturing process of sample 3#, multi-fractures can be
J. Liu et al.
formed when the accumulated Pw (Fig. 6c) is larger than the critical
hydraulic pressures in the micro-fractures and coal bedding plane
(Fig. 5c).
In this study, hydraulic fracturing experiments indicate that multi-
fractures are created simultaneously along multiple directions for 3#,
including σH, the coal bedding plane, and the direction of the micro-
fractures. This is discrepant from the previous view that the layered
formation can only fracture at the location with the minimum critical
pressure and the hydraulic fractures can hardly propagate along mul-
tiple directions at a same time (Zhao and Chen, 2010). Hence we as-
sume the fracture propagation patterns are determined not only by the
values P1, P2 and P3 but also by the relationships between Pw and P1,
P2 and P3. The specific mechanism needs more investigation as the Pw
is an important influential parameter for determining fracture propa-
gation.
4.2.2. Explanations for breaking through the floor
Even though the roof and floor of 3# have greater compressive
strength than those of 4#, the floor of 3# is broken through by frac-
turing fluid, while the roof and floor of 4# remain intact (Table 1;
Fig. 5). This reveals that tri-axial stresses have a great effect on the
behavior of a hydraulic fracture breaking through the roof and floor.
The coal seam commonly has high stress sensitivity (Jose, 1997;
Meng et al., 2011; Li et al., 2013). Due to the case that 3# has higher in
situ stresses than 4#, 3# readily has the low leak-off of a fracturing
fluid. The lower fluid leak-off makes the continuous Pw accumulation in
3#, which explains why the floor is fractured (point M in Fig. 6c).
Another reason for the floor of 3# is cut through is that 3# has a
high vertical stress. Because the experiment is modeled under the
condition that the coal seam is in parallel contact with the roof and
floor, P2 in Eq. (2) can be expressed as
P2 = σv + σt (7)
In Eq. (7), σt is so low that it can be neglected. Thus, P2 is mainly
related to σv. Because 3# has a larger σv value than 4#, the P2 value of
3# is higher than that of 4#. Thus it is hard for the hydraulic fracture in
3# to extend along the interface between the coal and the floor, re-
sulting in the accumulation of Pw. This also can cause hydraulic frac-
tures break through the floor in 3#.
4.2.3. Effects of Kh and Δσ
The horizontal stress difference coefficient (Kh) and the horizontal
stress difference (Δσ) are defined by
σ − σh Δσ
Kh = H = 4# are intact under the low injection flow rate (0.33 mL/s) (Fig. 5d),
σh σh (8)
Kh and Δσ are two key parameters influencing hydraulic fracture
propagation along the σ < SUB > H < /SUB > direction. It is still dif-
ficult to compare the importance of the Kh and Δσ. Pan et al. (2015)
insists that the Δσ is important because larger Δσ value usually increases
the distance of induced fracture in the σ < SUB > H < /SUB >
direction, while some achievements by Zhou et al. (2010) and Guo et al.
(2014) suggested that Kh value controls the propagation of hydraulic
fracture, because induced fracture would propagate along the σ <
SUB > H < /SUB > direction at greater Kh. According to the mea-
surements on shale samples, there is an obvious trend to form a main
fracture when Kh is greater than 0.25 (Guo et al., 2014). In this study,
for 3# and 4#, both samples have a Δσ of 4.0 MPa, while the values of
Kh are 0.24 and 0.34, respectively. The experimental results show that
some hydraulic fractures deviate from the σ < SUB > H < /SUB >
direction in 3# (with lower Kh), whereas almost all induced fractures
propagate straight along the σ < SUB > H < /SUB > direction in 4#
(with higher Kh). It seems that under the constant Δσ, a higher Kh value
is more favorable for a hydraulic fracture extending along the σ <
SUB > H < /SUB > direction.
To confirm the combined effects of Δσ and Kh on fracture
406
Journal of Petroleum Science and Engineering 168 (2018) 400–408
propagation, in this study, a series of numerical simulations on the
conditions of different horizontal stresses were constructed using the
ANSYS Finite Element Software. The boundary conditions for the si-
mulation are defined in Fig. 7a. All loadings were applied on the model
surfaces and displacement in each element was constrained along the
directions of x-axis and y-axis. Since the simulated blocks were con-
sidered to be linearly elastic, the finite element analysis was valid be-
fore initiating any fractures (Huang et al., 2017). In this study, the in
situ stress preconditions for five simulations are given in Table 2, and
the simulated results are shown in Fig. 7b and f. Five series of simu-
lations are organized into two groups: series A, B and C are for ana-
lyzing constant Δσ but varying Kh, whereas series C, D and E are for
analyzing varying Δσ but constant Kh (Table 2). To obtain a reliable
simulation results, we defined the σ < SUB > H < /SUB > of 4–20 MPa
and σh of 2–10 MPa, which values are within the accrual values of
σ < SUB > H < /SUB > (4.50–41.86 MPa) and σh (2.07–26.40 MPa) in
the study area (Meng et al., 2010; Yan et al., 2013).
In the stress nephograms (Fig. 7), variable colors indicate the dif-
ferent value of compressive stress in each simulated situation, in which
darker color means greater compressive stress and vice versa. The po-
tential fractures tend to be induced initially nearby the wellbore and
then propagate along the direction with high stress concentration. From
the simulated series A to B to C, under the conditions that Δσ value
remains constant at 2 MPa, the stress concentration degree at the crack
tip increases steadily along the σ < SUB > H < /SUB > direction with
increasing values of Kh (Fig. 7b and d). This means that it is easy for a
hydraulic fracture to propagate straight along the σ < SUB > H < /
SUB > direction in series C, but it is prone to deviate from the σ <
SUB > H < /SUB > direction in series A. The simulated results agree
with the experimental results of 3# and 4# (Table 1). From series C to
D to E, by contrast, Kh remains constant and Δσ increases from 2 to
10 MPa (Table 2), in which the change in the stress concentration de-
gree at the crack tip along the σ < SUB > H < /SUB > direction is too
small to be distinguished (Fig. 7d and f), indicating that a hydraulic
fracture can selectively propagate along the σ < SUB > H < /SUB >
direction in the three horizontal stresses conditions. In general, the si-
mulations results confirm that Kh is much more important than Δσ in
determining hydraulic fracture propagation along the σ < SUB > H < /
SUB > direction.
4.3. Injection flow rate impacts on fracture propagation
Group IV, which includes 4# and 5#, was set to study the influence
of the injection flow rate on fracture propagation. The roof and floor of
while both of them for 5# are broken (Fig. 5e) during hydraulic frac-
turing process with a high injection flow rate (0.66 mL/s) (Table 1). The
results suggest the Pw of 4# is lower than P3, while that of 5# is larger
than P3. This demonstrates that a high injection flow rate may result in
breaking through the roof and floor by hydraulic fracturing for the coal
reservoir with constant P3. Compared with the situation of sample 4#,
the Pw is accumulated more rapid because of the larger injection flow
rate (Fig. 6d and e). Regarding to sample 5#, the Pw is accumulated
continuously after a transitory decline at Point N in Fig. 6e because of
the breakdown of coal seam, which finally results in the breakdowns of
the roof and floor (Point M in Fig. 6e).
In this experiment, the comparison between 4# and 5# also in-
dicates that high injection flow rate (0.66 mL/s) makes similar types of
induced fractures with the low one (0.33 mL/s) (Fig), i.e., vertical
fractures extending along the σ < SUB > H < /SUB > direction
(Fig. 5d and e). Hence, during the field operation of hydraulic frac-
turing in study area, the injection flow rate should be limited in a
certain extent to avoid fracturing the roof/floor which may lead to high
water production. High flow rate can stimulate extending single branch
fracture for coals seam, which may be different from the situation in
shale (Wang et al., 2016) or synthetic mortar samples (Fallahzadeh
J. Liu et al. Journal of Petroleum Science and Engineering 168 (2018) 400–408

Fig. 7. Simulation results of five series of horizontal stress conditions in Table 2.

Table 2
Numerical simulations of five horizontal stress conditions.
Series σ < SUB > H < / σh/MPa Δσ(σ < SUB > H < /SUB > -
SUB > /MPa σh)/MPa
A 10 8 2
B 6 4 2
C 4 2 2
D 10 5 5
E 20 10 10
et al., 2015), where higher injection flow rate means more complex
network of induced fractures.
4.4. Implication for predicting hydraulic fracture propagation
Among the involved factors on hydraulic fracture propagation in
this study, the pre-existing natural fracture and in situ stresses are the
reservoir properties of coal seam, while the injection flow rate belongs
to the operation parameter during hydraulic fracture process. Because
of the heterogeneity, the coals at different sections or locations contain
variable properties and thus have different propagation mechanism of
hydraulic fracture.
During the hydraulic fracturing in the anthracite coal reservoir of
Changzhi CBM field, a small approaching angle (< 60°) tends to induce
single branch fracture along pre-existing natural fracture, while a large
one (> 60°) probably stimulates complex fracture systems. In terms of
Kh, the lower value (< 0.25) usually results in a complex fracture
system, while the higher value (> 0.25) tends to form a single main
fracture alone σ < SUB > H < /SUB > direction. Nevertheless, the
fracture orientation tends to extend along the pre-existing natural
fracture instead of the σ < SUB > H < /SUB > direction if the coal
sections contain pre-fractures, even though the Kh value is larger than
0.25. For example, only horizontal fracture is induced alone pre-ex-
isting natural fracture in sample 6# but vertical fracture is formed alone
σ < SUB > H < /SUB > direction in sample 4# and 5# even though
these samples have the same Kh value of 0.34. Regarding to the injec-
tion flow rate, two different rates stimulate similar fracture type in this
study. Relatively, the lower injection flow rate of 0.33 mL/s in labora-
tory is preferable, i.e., 6.0 m3/min for practical fracturing operation,
which is close to the operation parameter (6.0–7.0 m3/min) during
hydraulic fracturing process in field.
5. Conclusions
In situ tri-axial hydraulic fracturing simulation experiments were
carried out for large cubic samples composed of the coal and its roof
and floor. The influences from the natural fracture, in situ stresses and
injection flow rate on the hydraulic fracture propagation were dis-
cussed. The following conclusions can be made:
Kh
(1) A hydraulic fracture propagates in the coal seam along a pre-ex-
isting natural fracture direction under a small approaching angle,
0.25 while it propagates in the coal seam along the directions of a pre-
0.5
1
existing natural fracture and σ < SUB > H < /SUB > under a large
1 approaching angle. The pre-existing natural fracture exists or not
1 can also result in a distinct difference on the hydraulic fracture
propagation.
(2) A hydraulic fracture propagates straight along the
σ < SUB > H < /SUB > direction under a high value of Kh, and
tend to deviate from the σ < SUB > H < /SUB > direction under a
low value of Kh. The influence of Kh is greater than that of Δσ in
determining the fracture propagation along the σ < SUB > H < /
SUB > direction in the coal reservoir.
(3) The Pw accumulation can be resulted from large approaching angle,
high in situ stresses and high injection flow rate, which may break
the roof or floor by hydraulic fracture. The hydraulic fracture
propagation is related to Pw and the critical hydraulic pressures at
different interfaces, such as the pre-existing natural fracture, coal
petrological layered interfaces, or the interfaces between the coal
and rocks.
(4) Compared with lower injection flow rate, higher injection flow rate
makes similar types of induced fracture but more rapid Pw accu-
mulation. Hence the injection flow rate should be limited in a
proper extent to get more induced fractures and protect the roof/
floor, during the field operation of hydraulic fracturing in the coal
seam of Changzhi field.
Acknowledgments
We acknowledge financial support from the National Natural
Science Foundation of China (No. 41472137); the key project of
PetroChina Co Ltd (2017E-1404). We also gratefully acknowledge the
financial support for 1-year visiting scholar fellowship from the China
Scholarship Council (No. 201706400020).
References
Abass, H.H., Hedayati, S., Meadows, D.L., 1992. Nonplanar fracture propagation from
horizontal wellbore: experimental study. In: Paper SPE 24834 Presented at the 67th
ANNUAL SP Meeting, Washington D.C, pp. 133–137.
Beugelsdijk, L.J.L., de Pater, C.J., Sato, K., 2000. Experimental hydraulic fracture pro-
pagation in a multi-fractured medium. In: SPE Asia Pacific Conference on Integrated
Modelling for Asset Management, Yokohama, Japan, . https://doi.org/10.2118/
59419-MS.
Cai, Y.D., Liu, D.M., Yao, Y.B., Li, J.Q., Qiu, Y.K., 2011. Geological controls on prediction
of coalbed methane of No.3 coal seam in Southern Qinshui Basin, North China. Int. J.
Coal Geol. 88, 101–112.

407
J. Liu et al.
Fallahzadeh, S.H., James Cornwell, A., Rasouli, V., Hossain, M., 2015. The impacts of
fracturing fluid viscosity and injection rate on the near wellbore hydraulic fracture
propagation in cased perforated wellbores. In: 49th U.S. Rock Mechanics/
Geomechanics Symposium, 28 June-1 July, San Francisco, California.
Fan, T.G., Zhang, G.Q., 2014. Influence of injection rate and fracturing fluid viscosity on
hydraulic fracture geometry in coal. J. China Univ. Petrol 38 (4), 117–123.
Ferrill, D.A., Morris, A.P., Hennings, P.H., Haddad, D.E., 2014. Faulting and fracturing in
shale and self-sourced reservoirs: Introduction. AAPG Bull. 98 (11), 2161–2164.
Gan, Q., Elsworth, D., 2016. A continuum model for coupled stress and fluid flow in
discrete fracture networks. Geomech. Geophys. Geo-Energy Geo-Resour 2 (1), 43–61.
Guo, T.K., Zhang, S.C., Qu, Z.Q., Zhou, T., Xiao, Y.S., Gao, J., 2014. Experimental study of
hydraulic fracturing for shale by stimulated reservoir volume. Fuel 128, 373–380.
Haimson, B.C., Cornet, F.H., 2003. ISRM Suggested Methods for rock stress estimation-
Part3: hydraulic fracturing (HF) and/or hydraulic testing of pre-existing fractures
(HTPF). Int. J. Rock Mech. Min. Sci. 40, 1011–1020.
Hsieh, A., Dight, P., Dyskin, A.V., 2014. Ghost Kaiser effect at low stress. Int. J. Rock
Mech. Min. Sci. 68, 15–21.
Huang, S.P., Liu, D.M., Yao, Y.B., Gan, Q., Cai, Y.D., Xu, L.L., 2017. Natural fractures
initiation and fracture type prediction in coal reservoir under different in-situ,
stresses during hydraulic fracturing. J. Nat. Gas Sci. Eng. 43, 69–80.
Jiang, T.T., Zhang, J.H., Wu, H., 2016. Experimental and numerical study on hydraulic
fracture propagation in coalbed methane reservoir. J. Nat. Gas Sci. Eng. 35, 455–467.
Jose, G., 1997. Numerical simulation of coupled fluid-flow/geo-mechanical behavior of
tight gas reservoirs with stress sensitive permeability. In: Latin American and
Caribbean Petroleum Engineering Conference, 30 August-3 September, Rio de
Janeiro, Brazil.
Lavrov, A., 2003. The Kaiser effect in rocks: principles and stress estimation techniques.
Int. J. Rock Mech. Min. Sci. 40, 151–171.
Lehtonen, A., Cosgrove, J.W., Hudson, J.A., Johansson, E., 2012. An examination of in
situ rock stress estimation using the Kaiser effect. Eng. Geol. 124, 24–37.
Li, S., Tang, D.Z., Pan, Z.J., Xu, H., Huang, W.Q., 2013. Characterization of the stress
sensitivity of pores for different rank coals by nuclear magnetic resonance. Fuel 111,
746–754.
Lin, H.X., Du, C.Z., 2011. Experimental research on the quasi three-axis hydraulic frac-
turing of coal. J. China Coal Soc. 36 (11), 1802–1805.
Liu, G.H., Pang, F., Chen, Z.X., 2000. The principle of similarity criterion in the simulation
experiment during hydraulic fracturing. J. China Univ. Petrol 24 (5), 46–48.
Liu, J., Yao, Y.B., Liu, D.M., Pan, Z.J., Cai, Y.D., 2017. Comparison of three key marine
shale reservoirs in the southeastern margin of the Sichuan basin, SW China. Minerals
7 (10), 179.
Meng, Z.P., Tian, Y.D., Li, G.F., 2010. Characteristics of in-situ stress field in southern
qinshui basin and its research significance. J. China Coal Soc. 35 (6), 975–981.
Meng, Z.P., Zhang, J.C., Wang, R., 2011. In-situ stress, pore pressure and stress-dependent
permeability in the Southern Qinshui Basin. Int. J. Rock Mech. Min. Sci. 48, 122–131.
Pan, L.H., Cheng, L.J., Zhang, S.C., Guo, T.K., Liu, K.Y., 2015. Mechanism of fracture
propagation via numerical stimulation of reservoir volume fracture in shale re-
servoirs. Rock Soil Mech. 36 (1), 205–211.
Penny, G.S., Conway, M.W., Klem, R.C., 1991. Laboratory tests to determine parameters
forhydraulic fracturing of coalbed methane. In: Rock Mountain Regional Meeting and
Low-permeability Reservoirs Symposium, Denver, Colorado, . https://doi.org/10.
2118/21812-MS.
Renshaw, C.E., 1994. Are large differential stresses required for straight fracture propa-
gation paths. J. Struct. Geol. 16 (6), 817–822.
408
Journal of Petroleum Science and Engineering 168 (2018) 400–408
Shimizu, H., Murata, S., Ishida, T., 2011. The distinct element analysis for hydraulic
fracturing in hard rock considering fluid viscosity and particle size distribution. Int. J.
Rock Mech. Min. Sci. 48, 712–727.
Sobhaniaragh, B., Mansur, W.J., Peters, F.C., 2016. Three-dimensional investigation of
multiple stage hydraulic fracturing in unconventional reservoir. J. Petrol. Sci. Eng.
146, 1063–1078.
Tan, P., Jin, Y., Han, K., Zheng, X.J., Hou, B., Gao, J., Chen, M., Zhang, Y.Y., 2017.
Vertical propagation behavior of hydraulic fractures in coal measure strata based on
true triaxial experiment. J. Petrol. Sci. Eng. 158, 398–407.
Tang, S.H., Zhu, B.C., Yan, Z.F., 2011. Effect of crustal stress on hydraulic fracture in
coalbed methane wells. J. China Coal Soc. 36 (1), 65–69.
Wang, T., Hu, W.R., Elsworth, D., Zhou, W., Zhou, W.B., Zhao, X.Y., Zhao, L.Z., 2017. The
effect of natural fractures on hydraulic fracturing propagation in coal seams. J.
Petrol. Sci. Eng. 150, 180–190.
Wang, T., Zhou, W.B., Chen, J.H., Xiao, X., Li, Y., Zhao, X.Y., 2014. Simulation of hy-
draulic fracturing using particle flow method and application in a coal mine. Int. J.
Coal Geol. 121, 1–13.
Wang, Y., Li, X., Zhang, Y.X., Wu, Y.S., Zheng, B., 2016. Gas shale hydraulic fracturing: a
numerical investigation of the fracturing network evolution in the Silurian Longmaxi
formation in the southeast of Sichuan Basin, China, using a coupled FSD approach.
Environ. Earth. Sci. 75 (14), 1–18.
Wu, C.F., Zhang, X.Y., Wang, M., Zhou, L.G., Jiang, W., 2018. Physical simulation study
on the hydraulic fracture propagation of coalbed methane well. J. Appl. Geophys.
150, 244–253.
Yan, Z.F., Ju, Y.W., Tang, S.H., Hou, Q.L., Zhu, B.C., Wang, G.C., 2013. Numerical si-
mulation study of fracturing process in coalbed methane reservoirs in southern qin-
shui basin. Chin. J. Geophys. 56 (5), 1734–1744.
Yang, T.H., Zhu, W.C., Yu, Q.L., Liu, H.L., 2011. The role of pore pressure during hy-
draulic fracturing and implications for groundwater outbursts in mining and tun-
nelling. Hydrogeol. J. 19 (5), 995–1008.
Yuan, B., Moghanloo, R.G., Shariff, E., 2016. Integrated investigation of dynamic drai-
nage volume and inflow performance relationship (transient IPR) to optimize mul-
tistage fractured horizontal wells in tight/shale formations. J. Energy Resour.
Technol. 138 (5) 052901.
Yuan, B., Wood, D.A., Yu, W., 2015. Stimulation and hydraulic fracturing technology in
natural gas reservoirs: theory and case studies (2012–2015). J. Nat. Gas Sci. Eng. 26,
1414–1421.
Yuan, B., Zheng, D., Moghanloo, R.G., Wang, K., 2017. A novel integrated workflow for
evaluation, optimization, and production predication in shale plays. Int. J. Coal Geol.
180, 18–28.
Zhang, J.C., Bian, X.B., 2015. Numerical simulation of hydraulic fracturing coalbed me-
thane reservoir with independent fracture grid. Fuel 143 (10), 543–546.
Zhao, H.F., Chen, M., 2010. Extending behavior of hydraulic fracture when reaching
formation interface. J. Petrol. Sci. Eng. 74, 26–30.
Zhou, J., Chen, M., Jin, Y., Zhang, G.Q., 2008. Analysis of fracture propagation behavior
and fracture geometry using a tri-axial fracturing system in naturally fractured re-
servoirs. Int. J. Rock Mech. Min. Sci. 45, 1143–1152.
Zhou, J., Jin, Y., Chen, M., 2010. Experimental investigation of hydraulic fracturing in
random naturally fractured blocks. Int. J. Rock Mech. Min. Sci. 47, 1193–1199.
Zou, Y.S., Zhang, S.C., Tong, Z., Xiang, Z., Guo, T.K., 2016. Experimental investigation
into hydraulic fracture network propagation in gas shales using CT scanning tech-
nology. Rock Mech. Rock Eng. 49 (1), 33–45.