You are on page 1of 13

HYDRAULIC FRACTURING IN VERTICAL WELLS: EFFECT

OF IN-SITU STRESS VARIATIONS USING COHESIVE ZONE


THREE-DIMENSIONAL FINITE ELEMENT

Marko A. L. Bendezua, Eulher C. Carvalhoa, Deane M. Roehla , Maria F. F. de


Oliveiraa and Luis C. de Sousa Jrb
a
Department of Civil Engineering, Computer Graphics Technology Group (Tecgraf)
Pontifical Catholic University of Rio de Janeiro (PUC-Rio). Rua Marquês de São Vicente,
225, Gávea, 22453-900, Rio de Janeiro, RJ, Brazil,
b
Exploration and Production Business Area, E&P, Petrobras

Keywords: hydraulic fracturing, cohesive zone model, in-situ stress.

Abstract. Hydraulic fracturing is a process mainly used in the Petroleum Industry for
the stimulation of wells. It involves the injection of a variety of fluids and other
materials into the well at rates that actually cause the cracking or fracturing of the
reservoir formation. This process is done under high enough pressures to exert tension
stresses on the rock that overcome its tensile failure stress. The estimation or
determination of the fracture geometry of is one of the most difficult technical
challenges in the treatment of hydraulic fracturing. The aim of this work is to study the
effects of the in-situ stress variations on the hydraulic fracture development using a
three-dimensional finite element model. The non-linear effects associated to the fluid-
rock coupling are considered. A vertical fracture is modeled with the Cohesive Zone
Model (CZM) available in ABAQUS software, making use of the traction-separation
law and linear damage law for initiation and evolution of material degradation.
Solutions for 2D and 3D models are compared and checked against an analytical
solution.
1 INTRODUCTION
Hydraulic fracturing is a process mainly used in the Petroleum Industry for the
stimulation of wells. It involves the injection of a variety of fluids and other materials
into the well at rates that cause the cracking or fracturing of the reservoir formation.
Modeling of fracture propagation by hydraulic induced fractures is of great interest in
order to define the required amount of injection fluid, the injection pressure, the
required proppant volume and to predict the effectiveness and feasibility of the
treatment. This process is done under pressures high enough to induce tension stresses
on the rock that reach its tensile failure stress. Hydraulic fracturing is intrinsically a
three dimensional non linear coupled problem, where fluid flow and diffusion into
rock formation, fracture propagation, and inelastic rock deformation are mechanisms
to be described by the model.
Most hydraulic fracturing simulators are based on analytical methods or numerical
methods relying on linear elastic fracture mechanics (LEFM). For hard rocks, for
which brittle fracture prevails, these models give results which are in good agreement
with in field observation. For ductile rocks or under high stress conditions of
confinement, however, these simulators provide conservative results (Yao et al., 2010
and Zhang et al., 2010).
One reason for this is the fact that for the ductile behavior of rock, the fracture
process is associated with the development of a region around the crack tip that
presents plastic deformation prior to fracture propagation. This region, called the
fracture process zone, has two non-linear deformation regions: a localized region
characterized by softening behavior of the material, and a region where the perfect
plasticity or even plastic hardening occurs (Papanastasiou, 1999). For brittle materials,
the plastic process zone is small, so LEFM is applicable. For ductile rocks, however,
although the softening zone is still small, the plastic process zone is not negligible and
calls for the application of elastic-plastic fracture mechanics.
In order to take into account the inelastic deformations that arise in the fracture
process zone, the cohesive zone model (CZM), originally proposed by Barrenblatt,
1962 can be employed. With this method, the deformations at the crack tip prior to
fracture propagation are accounted for and energy dissipation occurs in a finite region.
The size of this region is a function of the rock material linear damage criteria
(Young's Modulus and fracture toughness), the viscosity of the fluid and the in-situ
stresses. Moreover, the singularity 1/r at the crack tip, inherent to LEFM, is not
present. As cited in (Chen et al., 2009) the cohesive zone method not only provides
more realistic results, but also simplifies the issues of finite element modeling, once
pre-determination of the crack tip is not required. The fracture initiation and
propagation are the natural solution of CZM.
In this paper, a full fluid-solid coupled two- and three-dimensional finite element
model is proposed for the analysis of hydraulic fracture. The effect of the in-situ stress
contrast is studied with the models. The results of both models are compared with the
analytical solution given in (Geertsma and Klerk, 1969) and (Carvalho, et al., 2010).

2 FINITE ELEMENT HYDRAULIC FRACTURE MODEL


The criterion for fracture propagation is usually given either by conventional
energy approach which states that a fracture propagates when the energy release rate
reaches a critical value related to the critical stress intensity factor or by the stress
intensity approach which states that a fracture propagates when the stress intensity
factor at the crack tip exceeds the rock toughness. The most robust criterion for non-
linear mechanics is described by the cohesive zone constitutive model.
The constitutive behavior of the cohesive zone is defined by the traction-separation
relation derived from laboratory tests. This traction-separation law is presented in
Figure 1. Simple cohesive zone models can be described by two independent
parameters which are usually, for mode-I plane strain, the normal work of separation
or the fracture energy GIC and either the tensile strength σd or the complete separation
length δC. An additional parameter in these models is the slope of the initial loading
which may define a range from rigid-softening to elastic-softening response under
tensile stress-state.
σ
Kn
G IC
σd

δd δc δ

σo: horizontal in-situ stress


σo

Figure 1: Bi-linear cohesive element traction-separation () law

The crack opening criterion is governed by the fracture energy of rock, which in a
CZM is equal to the area under the rock damage curve. For elastic material behavior,
the critical stress intensity factor KIC is related to GIC through the following relation:
K IC 2 1  2   c   d   d
GIC           (1) 
E 2
where E is the young modulus and ν is the Poisson ratio. The critical stress
intensity factor can be calculated from laboratory tests. For given values of the
fracture toughness, and rock tensile strength, d the stress-displacement law is
uniquely defined.
The initial cohesive stiffness (Kn) is given by:
d
Kn              (2)
d
where d is the critical opening displacement to activate material damage.

The opening displacement, c for which the fracture strength falls to zero is
calculated with the expression:
2G
 c  IC   d           (3) 
d
Substituting Equation 2 into Equation 1, we arrive at an equation that relates both
criteria, thus:
d2 
Kn              (4) 
2 GIC
where  is a coefficient representing the behavior of the rock. Small values of 
result in cohesive elements with low initial stiffness (ductile rocks) and large values of
 result in cohesive elements with high stiffness (brittle rocks).

Damage initiation refers to the beginning of degradation of the response of a


material point. The process of degradation begins when:
  
  max  n0 , s0   1           (5)
 n  s 
where nands represent the normal and the shear tractions. n0 and s0 represent
the peak values of the nominal stress when the deformation is either purely normal to
the interface or purely in the shear direction.
The damage evolution law describes the rate at which the material stiffness is
degraded once the corresponding initiation criterion is reached. In the post-peak
softening regime the cohesive constitutive relation is given by:
1  D   n , n 0
n   .        (6) 
 n , n 0
 s  1  D   s .             (7)
where  n   and  s are the normal and shear stresses corresponding to the openings
 n  and  s , respectively, assuming that the material had not suffered previous damage.
For linear softening, the damage variable is calculated by the following equation
(Turon et al., 2006):
    
D c m d           (8)
 m  c   d 
where  m is the maximum effective relative displacement in the cohesive element
during the treatment history.

The fluid constitutive response comprises: tangential flow within the gap and
normal flow across the gap (Figure 2). In order to avoid this complex fluid behavior, a
simple appropriate model for fluid flow in a fracture is embodied in lubrication theory.
It assumes laminar flow (uniformly viscous Newtonian flow), the fluid is
incompressible and it accounts for the time dependent rate of crack opening. The
continuity equation which imposes the conservation of mass in one dimensional flow
is (Boone and Ingraffea 1990):
dqt dw
  ql  0           (9)
dx dt
where qt is the local flow rate along the fracture in direction x, ql is the local fluid
loss in rock formation and w is the crack opening. Equation (9), which accounts for the
fluid leak-off from the fracture surface into the rock formation, can be used to
determine the local flow rate q.
ql

Figure 2: The flow patterns of the pore fluid in cohesive elements

The second equation is derived from the conservation of momentum balance. For a
flow between parallel plates the lubrication equation, which relates the pressure
gradient to the fracture width with smoothly varying aperture is laminar for a
Newtonian fluid of viscosity μ, yields (Boone and Ingraffea 1990)
w3
qt   p         (10)
12 
where p denotes the fluid pressure. Equation (9) determines the pressure profile
along the fracture from the local width and local flow rate. According to Eq. (9), the
pressure gradient and hence the solution, is very sensitive to fracture width. The effect
of leak-off from the fracture surface into the rock formation is left to a further study.

3 SIMULATION MODEL
A fracturing process of a vertical well is simulated with the software ABAQUS.
The true vertical depth of the well is 461.0 m, the pay zone is located at the depth of
464.0–467.0 m. The dimension of the model is 12.2 m, 12.2 m and 9 m in the X, Y, Z
directions, respectively. The lithology of the analyzed reservoir is composed of three
layers: two barriers interspaced with a pay zone. The vertical fracture is in the center
of the model, with an initial thickness of 2·10-5 m. Figure 4 shows a sketch of the
model.
Horizontal  z
in‐situ Overburden pressure
stress
2∙10‐5 m ‐461 m
x
vertical fracture
3.0 m

b Barrier
3.0 m

r Pay zone Fluid flow


3.0 m

b Barrier

6.1 m 6.1 m

Figure 3: Schematic representation of the formation


The mesh is composed of three dimensional eight-node hexahedral elements for the
rock formation and twelve-node cohesive element having both displacement and pore
pressure as degrees of freedom. Altogether the finite element mesh has 15180
elements and 17825 nodes (Figure 4). Fluid flow and deformation of porous medium
are fully coupled. The fracture is composed with cohesive elements to describe the
process of fracture initiation and propagation during treatment history.

Cohesive
elements

9m 3m

.2 m
12.2 12
m

Figure 4: Mesh of the model 3D

In-situ stresses of pay zone in the Z, Y, X directions, are -6.5 MPa, -4.5 MPa and -
4.5 MPa respectively. In-situ stress of the barriers have the same values in the Z, Y, X
directions, are -6.5 MPa, -4.5 MPa and -4.5 MPa respectively for scenario one.

The formation mechanical properties as well as those of the cohesive element in


were taken from (Chen et al., 2009).

Table 1: Formation geologic parameters

E  Kw w
Layer
(GPa) (-) (mD) (kN/m3)
Pay zone 30.00 0.20 100 9.81
Barrier 30.00 0.20 100 9.81
Kw: Permeability; γw: Fluid specific gravity

Table 2: Material properties of cohesive element in the pay zone and the barrier

E  d KIC
Layer
(GPa) (-) (MPa) (MPa.m1/2)
Pay zone 30.00 0.20 2.00 1.00
Barrier 30.00 0.20 2.00 1.00
The analysis was carried out in two load stages. Initialization of the in-situ stresses
with the geostatic load, followed by fluid flow, prescribed along the fracture. In the
hydraulic fracturing stage, flow is prescribed along the whole pay zone. Fracture fluid
is injected at the center of the fracture with a constant rate of 5·10-4 m3/s and lasts 10 s.

Three sets of analysis were carried out. In the first set the evolution of the fracture
width and length are compared with an analytical solution KGD (Geertsma and Klerk,
1969). In the second set the effect of different in-situ stresses between layers is
studied. Finally the model 3D is compared with the model 2D also using CZM.
The results of the simulations and the analytical solutions are plotted in Figure 5
and Figure 6. It is clear that the evolution of the fracture width and the length of
fracture obtained from the simulation are in agreement with the analytical solutions.
These comparisons are important by the judgment of success or failure of a hydraulic
fracturing treatment, but this is not the sole alternative. The analytical solution
presents a solution for the propagation of a hydraulic fracture by assuming the width
of the crack at any distance from the well is independent of vertical position and can
be derived form linear elastic stress-stain relations.

0.90

0.80

0.70
Fracture width [mm]

0.60

0.50

0.40

0.30

0.20 Numerical (CZM)


0.10 Analytical (KGD)

0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]
Figure 5: Comparison with analytical solution
3.0

2.5

Half length of fracture [m]


2.0

1.5

1.0

Numerical (CZM)
0.5
Analytical (KGD)

0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0

Time [s]
Figure 6: Comparison with analytical solution

The effect of in-situ stresses contrast on fracture growth. The in-situ least principal
stress in the barrier in taken as 4.5 MPa and 6.2 MPa, respectively in two cases and all
other parameters keep the same as the standard model.
Figure 7, Figure 8 and Figure 9 are shown the evolution of the fracture width, the
half length of fracture and the fracture height for two cases different in-situ stresses. It
is observed from the above results that both fracture half length and fracture width
increase, while the fracture height decreases, when the in-situ least principal stress
contrast between the barrier and the pay zone grows.

0.90

0.80

0.70
Fracture width [mm]

0.60

0.50

0.40

0.30
minimum principal stress 4.5 MPa
0.20
minimum principal stress 6.2 MPa
0.10

0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]

Figure 7: Evolution of the fracture width of the two different least principal stress cases
6.0

Half length of fracture [m] [m]


minimum principal stress 4.5 MPa
5.0
minimum principal stress 6.2 MPa

4.0

3.0

2.0

1.0

0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0

Time [s]

Figure 8: Evolution of the half length of fracture of the two different least principal stress cases

8.0

7.0

6.0
Fracture height [m]

5.0

4.0

3.0

2.0
minimum principal stress 4.5 MPa
1.0
minimum principal stress 6.2 MPa
0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time [s]

Figure 9: Evolution of the fracture height of the two different least principal stress cases

Comparison with a model 2D (Carvalho, et al., 2010), Figure 10 shows the


evolution of the poro pressure, the fracture width and fracture height for the zone
cohesive model in 2D and 3D. The 2D case corresponds to the case where the fracture
is infinite along Y direction. The parameters are the same for both models. The energy
consumed in rock fracturing is higher in the 3D model. The geometry and evolution of
the pore pressure are not in good agreement.
Figure 11 and Figure 12 show the fracture prediction in different time stages using
the CZM model. The blue elements were not fractured and the reds are fractured. For
time 0.17s, the breakdown pressure (pb) is 7202 kPa and the process of fracture
initiation starts. The second peak of pressure (pw) is 9160 kPa (t = 0.79s). It is
observed that the first elements breakdown completely and fracture propagation
follows. For the large stress contrast case, the propagation pressure (pprog) is 8159 kPa
(t=6.8s) and it starts the fracture initiation and propagation of the barrier. Fracture
shapes are shown at the end of the treatment.
10000 10000

Pore pressure [kPa]


Pore pressure [kPa]

7500 7500

5000 5000

2500 2500
2D 2D
3D 3D
0 0
0.0 2.0 4.0 6.0 8.0 10.0 0.0 2.0 4.0 6.0 8.0 10.0

Time [s] Time [s]


a) cap – res = 0 kPa b) cap – res = 1700 kPa
0.001 0.001

0.0008 0.0008
Fracture width [m]

Fracture width [m]


0.0006 0.0006

0.0004 0.0004

0.0002 2D 0.0002
2D
3D
3D
0 0
0.0 2.0 4.0 6.0 8.0 10.0 0.0 2.0 4.0 6.0 8.0 10.0

Time [s] Time [s]


c) cap – res = 0 kPa d) cap – res = 1700 kPa
9.0 9.0
8.0 8.0
[m]

[m]

7.0 7.0
Fracture height

Fracture height

6.0 6.0
5.0 5.0
4.0 4.0
3.0 3.0
2.0 2.0
2D 2D
1.0 1.0
3D 3D
0.0 0.0
0.0 2.0 4.0 6.0 8.0 10.0 0.0 2.0 4.0 6.0 8.0 10.0

Time [s] Time [s]


f) cap – res = 0 kPa g) cap – res = 1700 kPa
Figure 10: Comparison the zone cohesive model 2D and 3D
a) time = 0.17 s, pw,b = 7202 kPa

b) time = 0.79 s, pw = 9160 kPa

c) time = 10.0 s, pprog = 7098 kPa


Figure 11: Cohesive fracture propagation with c = 0 KPa
a) time = 6.8 s, pw = 8159 kPa

c) time = 10.0 s, pprog = 7806 kPa


Figure 12: Cohesive fracture propagation with c = 1700 KPa

4 CONCLUSIONS
The full-fluid-solid coupling finite element model for hydraulic fracturing was
established with a good agreement between simulation results and analytical solution..
The effect of the in-situ stress contrast between barriers and pay zone is a factor
controlling fracture height. Large in-situ stresses have a confining effect on the
fracture, this will hinder fracture height growth and increase fracture width, fracture
length and bottomhole pressure. Other effects of parameters on fracture propagation
can be studied based on the above model.
A finer mesh results in pore pressure curves with fewer peaks. Moreover, the
cohesive finite element model has advantages in dealing with 3D and T-shaped
hydraulic fractures in multilayer non-homogeneous formation and poroelastic media.

Acknowledgments
This research was carried out in the Computer Graphics Technology Group
Tecgraf/PUC-Rio and was funded by Petrobras and CNPq.
REFERENCES
ABAQUS user’s manual, version 6.11, SIMULA (2011)
Boone J.T., Ingraffea A.R.. A numerical procedure for simulation of hydraulically
driven fracture propagation in poroelastic media. Int J Numer Anal Methods
Geomech 14:27–47. 1990
Carvalho E.C., Bendezu M. L., Oliveira M. F., Roehl D. M., Sousa Jr L.C.. Finite
Element Modeling of Hydraulic Fracturing in Vertical Wells. Mecânica
Computacional. Volume XXIX. Number 88. Application of Computational Methods
in Petroleum Engineering (B). Abstracts, p. 8571-8578, 2010.
Geertsma, J and Klerk F. A rapid method of predicting width and extent of hydraulic
induced fractures. Journal of Petroleum Technology, 21: 1571-1581, 1969.
Papanastasiou, P., The effectice fracture toughness in hydraulic fracturing.
International Journal of Fracture, 96, 127-147, 1999.
Barrenblatt, G. I., The mathematical theory of equilibrium of cracks in brittle fracture.
Advances in Applied Mechanics, 7, 55-129, 1962.
Turon, A., Camanho, P. P., Costa, J., Davila, C. G., A damage model for the
simulation of delamination in advanced composites under variable-model loading.
Mech. Mater. 38(11), 1072-1089, 2006.
Chen, Z., Bunger, A. P., Zhang, X., Jeffrey, R. G., Cohesive zone finite element-based
modeling of hydraulic fractures. Acta Mechanica Solida Sinia, Vol.22, No.5 ,443-
452, 2009.
Yao, Y., Gosavi, S. V., Searles, K. H., Ellison, T. K., Cohesive fracture mechanics
based analysis to model ductile rock fracture. Proceedings of the 44th US Rock
Mechanics Symposium, Salt Lake City, 2010.
Zhang, G. M., Liu, H., Zhang, J., Wu, H. a., Wang, W. W., Three dimensional finite
element simulation and parametric study for horizontal well hydraulic fracture.
Journal of Petroleum Science and Engineering, 72, 310-317, 2010.

You might also like