Development and Validation of Fully-Coupled

Hydraulic Fracturing Simulation Capabilities

Matias G. Zielonka, Kevin H. Searles, Jing Ning and Scott R. Buechler

ExxonMobil Upstream Research Company

3120 Buffalo Speedway, Houston, TX 77098

Abstract: The problem of the propagation of a hydraulically driven fracture in a fully saturated,
permeable, and porous medium is investigated. Fluid driven fracture propagation in porous media
is a coupled problem with four unknown fields: the flow of the fracturing fluid within the fracture,
the flow of the pore fluid within the pores, the porous medium deformation, and the fracture
configuration. The corresponding governing equations are the mass balance of the fracturing
fluid, mass balance of the pore fluid, equilibrium of the porous medium, and fracture initiation
and propagation criteria. In this work, the recently co-developed Abaqus fully-coupled hydraulic
fracturing modeling capabilities are evaluated by assessing their consistency, convergence, and
accuracy qualities. The Abaqus “coupled pressure/deformation cohesive elements” and “coupled
pressure/deformation extended finite elements (XFEM)” are used to model the propagation of the
fracture and the flow of the fracturing fluid, while the porous medium deformation and pore-fluid
flow are modeled with coupled “pore-pressure/deformation” continuum finite elements. The
propagation of a vertical planar fluid-driven fracture with constant height and vertically uniform
width within a prismatic-shaped reservoir (KGD model), and the propagation of a horizontal,
circle-shaped, planar, fluid-driven fracture within a cylindrical reservoir (“Penny-Shaped”
model) are simulated in both two and three dimensions. The Abaqus numerical solution obtained
with each modeling technique (cohesive and XFEM) is compared with asymptotic analytical
solutions for both the KGD and Penny-shaped models in the toughness/storage dominated and
viscosity/storage dominated propagation regimes. Both methods are found to accurately
reproduce the analytical solutions, and converge monotonically as the mesh is refined. This
validation of the newly developed hydraulic fracturing capabilities within Abaqus provides
confidence in its ability and readiness to simulate fluid driven fracturing applications for the oil
and gas industry including injection, stimulation, and drilling operations.

Keywords: geomechanics, soil mechanics, fracture mechanics, hydraulic fracturing, fluid-driven

fracturing, geostatic, soils, pore pressure, cohesive elements, extended finite elements, XFEM,
reservoir, drilling, injection.

1. Introduction
Hydraulic fracturing is a fundamental problem in Petroleum Engineering and plays a critical role
in many applications within the oil and natural gas industry. The process can be generally defined
as the intentional (or unintentional) initiation and propagation of a fracture due to the
2014 SIMULIA Community Conference 1
www.3ds.com/simulia
pressurization of fluid that flows within the fracture. Examples of applications include (a) the
stimulation of rock formations with poor or damaged permeability to increase conductivity
between the reservoir and the producing wells, (b) improvement of produced water re-injection
(PWRI) where water is injected to replace produced fluids and maintain reservoir pressure or
provide enhanced oil recovery, (c) cuttings reinjection (CRI) where a slurry of drill cuttings is
injected into a formation to mitigate the cost and risk of surface disposal, (d) in-situ stress
measurement by balancing the fracturing fluid pressure in a hydraulically opened fracture with the
geostatic stresses, and (e) wellbore integrity analysis of drilling operations to avoid propagating
near-wellbore fractures that could result in drilling fluid losses to the formation and an inability to
effectively clean the wellbore.
Knowledge of the fracture dimensions (length/width/height), fracture geometry, and wellbore
pressure is crucial for both the design and integrity of hydraulic fracturing field operations. For
stimulation, PWRI and CRI, one of the fundamental questions is whether or not fracture
containment is achieved. This means that the injection fluid and fracture are confined to a target
interval or “pay” zone for PWRI and stimulation, or a dedicated disposal domain for CRI. Other
important considerations include predictions of the injection rate, pressure, or injected volume
required to initiate fractures, inject under matrix conditions, or minimize the potential for inducing
fractures while drilling.
Currently, there are no reliable techniques to measure fracture geometries during or after the
hydraulic fracturing process. Furthermore, direct solutions of the underlying differential equations
representing the different physical processes occurring during fracturing are difficult to construct,
even in their most simplified forms. Therefore, the development of a numerical simulator with
accurate predictive capability is of paramount importance.
The computational modelling of hydraulic fracturing of porous media is a challenging endeavor.
The difficulty originates primarily from the strong non-linear coupling between the governing
equations, as the process involves at least the interaction between four different phenomena: (i) the
flow of the fracturing fluid within the fracture, (ii) the flow of the pore fluid and seepage of
fracturing fluid within the pores, (iii) the deformation of a porous medium induced by both the
hydraulic pressurization of the fracture and the compression/expansion and transport of pore fluid
within the pores, and (iv) the fracture propagation which is an inherently an irreversible and
singular process. Additionally, fracture propagation typically occurs in heterogeneous formations
consisting of multiple layers of different rock types, subjected to in-situ confining stresses with
non-uniform magnitudes and orientations. Furthermore, fracturing fluids typically exhibit
nonlinear rheologies and the leakoff of these fluids from the fracture into the surrounding rock is
often history dependent.
There are a number of commercial hydraulic fracture simulators used in the oil and natural gas
industry for rapid design, analysis and prediction of fracture size, treating pressures, and flows
(Clearly 1980, Meyer 1989, Warpinski 1994). These simulators rely in strong simplifying
assumptions to render the problem solvable in realistic times:
 Fractures are assumed to be planar and symmetric with respect to the wellbore 
 Fracture geometries are represented with few geometric parameters 

2 2014 SIMULIA Community Conference

www.3ds.com/simulia
  The formation is assumed to be unbounded and modeled using linear elasticity theory
resulting in an integral equation relating fracture opening and fluid pressure 
 The fracture propagation is modeled within the framework of linear elastic fracture
mechanics without any consideration of pore fluid pressure effects 
 Leakage of fracturing fluid from the fracture into the rock is modeled as one dimensional
and decoupled from the porous medium deformation. 
Although these simulators are useful in predicting broad trends and upper/lower bounds in
operational parameters, their reliability and accuracy are restricted to unrealistic scenarios intrinsic
simplistic assumptions apply, i.e., situations where some of the coupling between the many
different processes involved can be neglected, and with strong symmetry in confinement stresses
and geology.
The accurate modelling of the hydraulic fracturing process under realistic geologies, wellbore
configurations, confining stress states, and operational conditions calls for a more advanced,
multi-physics numerical simulator that incorporates the complex coupling between the injected
fluid, the pore fluid, the rock deformation, and the fracture configuration, thus overcoming the
limitations of currently available commercial simulation tools.
To this end, fully-coupled hydraulic fracturing simulation capabilities that leverage (i) the existing
Abaqus non-linear soil consolidation analysis solver, (ii) Abaqus cohesive elements for modelling
interface decohesion, and (iii) Abaqus extended finite element method (XFEM) for modelling
propagating discontinuities, are being co-developed between ExxonMobil Upstream Research
Company and Dassault Systemes Simulia Corporation.
Specifically, two new element classes have been integrated into the existing Abaqus/Standard
coupled pore fluid diffusion and solid stress porous media analysis solver:
i. A coupled pressure/deformation cohesive element that models the progressive damage of
normal mechanical strength and normal hydraulic conductivity as well as the flow of
fracturing fluid within the opening fracture.
ii. An enriched version of the continuum coupled pore fluid diffusion/stress elements
capable of activating arbitrarily oriented discontinuities in both displacements and pore
pressures while simultaneously modelling the fracturing fluid flow along the fracture.
This work describes and validates these two new formulations for hydraulic fracturing modeling
by assessing consistency, accuracy and convergence qualities. The propagation of a fluid-driven
vertical planar fracture of uniform width and constant height within a prismatic-shaped rock
formation (Khristianovich-Geertsma-de Klerk, or KGD model) and the propagation of a
horizontal, circle-shaped, planar, fluid-driven fracture within a cylindrical reservoir (radial or
“Penny-Shaped” model) are simulated for both two and three dimensions (Clearly 1980, Geertsma
1969, Yew 1997). The numerical solution obtained with each new modeling technique (cohesive
and XFEM) are then compared with available asymptotic analytical solutions for both the KGD
and Penny-shaped models in the toughness/storage dominated and viscosity/storage dominated
propagation regimes. Finally, the consistency, accuracy and convergence attributes are assessed
for both methods.

2014 SIMULIA Community Conference 3

www.3ds.com/simulia
Section 2 describes the governing equations for each of the coupled processes as well as the
constitutive and kinetic relations assumed for the porous medium, pore fluid and fracturing fluid,
including:
i. Equilibrium equation for the porous medium
ii. Constitutive equation for the porous medium (Biot’s theory of poroelasticity)
iii. Continuity equation for the pore fluid
iv. Continuity equation for the fracturing fluid
v. Momentum equation for the pore fluid (Darcy’s Law)
vi. Momentum equation for the fracturing (Lubrication Equation)
Section 3 details the procedures employed by both formulations (cohesive and extended finite
element methods) and the fracture initiation and propagation criteria. Section 4 defines the test
models (KGD plane-strain and the Penny-Shaped models) and the model set-up and assumptions
used within Abaqus, while Section 5 presents numerical results and assesses accuracy and
convergence by comparing the main solution variables obtained with meshes of different
resolutions with available asymptotic analytical solutions. Finally, some concluding remarks are
summarized in Section 6.

2. Governing Equations
As stated in the introduction, hydraulic fracturing involves the interaction between four different
phenomena:
i. Porous medium deformation
ii. Pore fluid flow
iii. Fracturing fluid flow
iv. Fracture propagation
The equations and constitutive relation governing these coupled processes, i.e., Biot’s theory of
poroelasticity for porous media, Darcy’s Law for pore fluid flow, Reynold’s lubrication theory for
fracturing fluid flow and the cohesive zone model to characterize fracturing (Abaqus 2013,
Charlez 1997)) are summarized in what follows.
2.1 Porous Media Deformation
Porous media can be modelled as an isotropic, poroelastic material undergoing quasistatic
deformation. The equilibrium equation enforced by Abaqus, when body forces are neglected is,
, 0 (1)
while the poroelastic constitutive relation, assuming small strains, is given by,
2 (2)

2
1
4 2014 SIMULIA Community Conference
www.3ds.com/simulia
 3
1 2
in which is Biot’s coefficient, and are the dry elastic shear and bulk moduli, is the dry
Young’s modulus, and is the dry Poisson’s ratio. Abaqus is formulated in terms of Terzaghi
effective stresses ′, defined for fully saturated media as (Abaqus 2013, Charlez 1997)

In terms of the latter, the constitutive relation takes the form

2 1
Defining effective strains as
1
3
the constitutive relation simplifies to
2
This identity is identical to the constitutive relation for linear elastic materials, but expressed in
terms of Terzaghi effective stresses ′ and effective strains ′. Abaqus internally translates total
stresses and strains into Terzaghi effective stresses and strains to leverage this equivalence
(Abaqus 2013).
2.2 Pore Fluid Flow
The continuity equation for the pore fluid is, assuming small volumetric strains, given by
1
, 0

where is the pore fluid seepage velocity, and and are Biot’s modulus and Biot’s
coefficient, respectively. These two poroelastic constants are defined by the identities
1

1 1

where is the pore fluid bulk modulus, is the porous medium solid grain bulk modulus, and
is the initial porosity. In Abaqus, the two compressibilities and are specified using the
*POROUS BULK MODULI keyword. Pore fluid is assumed to flow through an interconnected pore
network according to Darcy’s law

, ,

in which is the permeability, is the pore fluid viscosity, is the hydraulic conductivity and is
the pore fluid specific weight. Combining with the continuity equation, the pore fluid diffusion
equation is obtained

2014 SIMULIA Community Conference 5

www.3ds.com/simulia
 1
, (3)

Within Abaqus, the hydraulic conductivity and specific weight are specified through the
*PERMEABILITY keyword.

2.3 Fracturing Fluid Flow

Longitudinal fluid flow within the fracture is governed by Reynold’s lubrication theory defined by
the continuity equation

0
and the momentum equation for incompressible flow and Newtonian fluids through narrow
parallel plates (i.e., Poiseuille flow)

12
where is the fracture gap (Figure 1), ∙ is the fracturing fluid flow (per unit width)
across the fracture, and are the normal flow velocities of fracturing fluid leaking through the
top and bottom faces of the fracture into the porous medium, is the fracturing fluid viscosity,
and is the fracturing fluid pressure along the fracture surface parameterized with the curvilinear
coordinate, .

Figure 1: Fracture aperture, width and fracturing fluid flow

Abaqus computes the normal fracturing fluid velocities as

(4
where and are the pore fluid pressures on the top and bottom surface of the fracture and
and are the so-called “leakoff coefficients”. This simple leakoff model simulates a layer of
filtrate that might accumulate and reduce the effective normal permeability of the fracture
surfaces.
Inserting the Poiseuille flow equation and the simplified leakoff model into the continuity equation
for the fracturing fluid yields the final form:

6 2014 SIMULIA Community Conference

www.3ds.com/simulia
 (5)
12
Abaqus specifies the fracturing flow viscosity and leakoff coefficients with the *GAPFLOW and the
*LEAKOFF keywords, respectively.

2.4 Fracture Initiation and Propagation

Fracturing can be conceptualized as the transition between two limiting states: the undamaged
state with continuous displacements and non-zero tractions in all directions and the fully damaged
state characterized by the presence of a displacement discontinuity along a material interface with
zero tractions in the direction normal to the interface. In Abaqus, this transition process is modeled
as a progressive degradation of cohesive strength along a zero-thickness interface whose
orientation and extent is either predefined (cohesive element method) or calculated during the
simulation (extended finite element method). The gradual loss of strength in the interface with
increasing separation is defined with an interface traction/interface separation relation or cohesive
law (Abaqus 2013, Ortiz 1999).
For the purpose of this study, a traction-separation cohesive law with linear softening (Figure 2) is
assumed, defined by the cohesive energy (area under the softening part of the traction
separation curve) and the cohesive strength . For the cohesive element procedure, it is also
required to define the traction-separation behavior prior to damage initiation, which is assumed to
be linear with initial stiffness . The cohesive traction of the interface thus evolves from a
maximum tensile strength at damage initiation, down to zero when the interface is fully
damaged and free to open beyond the total separation . If the interface is unloaded prior to
complete damage, the traction is assumed to ramp down linearly with a damaged stiffness . The
interface effective tractions are therefore given by
0

COHESIVE XFEM
Traction

Separation

Figure 2: Cohesive law for Cohesive and XFEM procedures

Upon damage initiation, the fracture is pressurized by instantaneously applying the fracturing fluid
pressure, , calculated from the fracturing fluid equations in Equation 5. The total tractions
resisted by (and acting on) the interface elements are therefore given by

2014 SIMULIA Community Conference 7

www.3ds.com/simulia
 0
As stated in Section 2.1, the Abaqus porous media analysis solver is formulated in terms of
Terzaghi effective stresses. Therefore, the cohesive strength defining the onset of interface
decohesion must be understood in terms of an effective strength (and not total strength).

3. Cohesive Element and Extended Finite Element Methods

The evolution of a fracture is modelled in Abaqus through zero-thickness interface elements with
separation resisted by gradually decreasing tensile tractions. For the cohesive element procedure,
these interface elements are defined a priori and placed between continuum element faces,
whereas in the extended finite element method (XFEM), they are inserted and oriented
automatically during the course of the simulation within existing continuum elements.
3.1 Cohesive Element Method
The coupled pressure/deformation cohesive elements implemented in Abaqus (COHPE4P, COHAX4P,
COH3D8P) are standard linear isoparametric elements with displacement and pore pressure degrees
of freedom associated with their corner nodes, as depicted in Figure 3 (nodes 1,2,3,4). These
elements must be inserted a priori between the faces of adjacent pressure diffusion/stress elements
(CPE4P, CAX4P, C3D8P) in order to model the yet to open fracture. To accommodate the coupling of
the fracturing fluid flow equations, the elements are equipped with additional pressure degrees of
freedom (attached to the center of the element edges perpendicular to the fracture) to interpolate
the fracturing fluid pressure after damage initiation (nodes 5 and 6, Figure 3).
3
3 4 4
5 6 5 6

1 2
1 2
Undeformed Deformed
Configuration Configuration

Figure 3: Coupled pressure/deformation cohesive elements for hydraulic fracturing

The cohesive elements can have arbitrary undeformed geometric thickness as the instantaneous
gap coupled in the fracturing fluid flow equation (Equation 5) is defined in Abaqus as the
difference between the deformed and underfed thickness, i.e., . Prior to damage, the top
and bottom faces of the unopened fracture are subjected to the pore fluid pressure acting towards
increasing separation and the cohesive effective tractions resisting separation,

where is the stiffness of the cohesive element prior to failure (Figure 2). After damage
initiation, the pore fluid is displaced by the fracturing fluid pressurizing the interface. The total
tractions acting on the top and bottom faces of the opening fracture are then substituted by

8 2014 SIMULIA Community Conference

www.3ds.com/simulia
where is the damaging stiffness (Figure 2). A coupled cohesive element method for hydraulic
fracturing similar to the formulation just outlined is described by Boone, 1990 and Carrier, 2012.
3.2 Extended Finite element method (XFEM)
The Extended Finite Element Method (XFEM) is implemented within Abaqus using the so called
“phantom node” approach (Abaqus 2013, Remmers 2008, Song 2006, Sukumar 2003). In this
implementation, each enriched pressure diffusion/stress element (CPE4P, CAX4P, C3D8P) is
internally duplicated with the addition of corner phantom nodes, as depicted in Figure 4, in which
original nodes are represented with full circles and corner phantom nodes with hollow circles.
Prior to damage initiation only one copy of the element is active. Upon damage initiation the
displacement and pore pressure degrees of freedom associated with the corner phantom nodes are
activated and both copies of the element are allowed to deform independently, pore pressures are
allowed to diffuse independently, and the created interface behavior is enforced with a traction-
separation cohesive law.

4

4 3 4, 4 3, 3
4 3

1 1, 1 2
2 2, 2 1
Undeformed 1
Undeformed Deformed 2
configuration configuration configuration
before damage after damage after damage

Figure 4: Implementation of the XFEM with “corner” and “edge” phantom nodes
In order to enable the solution of the fracturing fluid flow equations, the enriched elements also
incorporate new “edge-phantom nodes” (depicted as red triangles in Figure 4) that interpolate the
fracturing fluid pressure within the fracture. The pore fluid pressure and at the top and
bottom faces of the fracture are interpolated from the pore pressure degrees of freedom at the
corner real nodes and phantom nodes. The difference with the fracturing fluid pressure
(interpolated at the edge-phantom node) is the driving force that controls the leakage of fracturing
fluid into the porous medium (Equation 4).
The fracture is extended to a new element ahead of the fracture tip when the maximum effective
principal stress at this element (interpolated to the tip) in a given iteration is equal to the cohesive
strength . The orientation of the fracture segment to be extended into the tip element is set to the
2014 SIMULIA Community Conference 9
www.3ds.com/simulia
direction perpendicular to the minimum principal stress of the current iteration. This fracture
initiation/orientation criterion is defined in Abaqus through the keyword
*DAGAMAGE INITATION, CRITERION=MAXPS, POSITION=CRACKTIP
As in the cohesive element formulation, the fracturing fluid pressure is applied to the top and
bottom faces of the fracture and superposed to the cohesive tractions.

4. Benchmark Models
In this section, the two formulations previously outlined (coupled pressure/displacement cohesive
and extended finite elements) are applied to model the propagation of a hydraulically driven
fracture in two different configurations:
i. Horizontal, circle-shaped, planar, fracture within a cylindrical domain, (radial or “Penny-
Shaped” model (Clearly 1980, Charlez 1997, Yew 1997))
ii. Vertical, rectangle-shaped, planar fracture within a prismatic-shaped domain
(Khristianovich-Geertsma-de Klerk, or KGD model (Charlez 1997, Geertsma 1969, Yew
1997)).
These models serve as benchmark examples to assess the consistency, convergence and accuracy
of the numerical solution obtained with Abaqus.
4.1 Fracture Propagation Regimes
Despite the simplicity of the fracture geometry and strong symmetry in the chosen benchmark
problems, there are no available closed-form analytical solutions for these problems when all
coupled processes are considered in the analysis, i.e., when the formation is assumed to be porous
and permeable with pore fluid flow and fracturing fluid is leaking into the pore space displacing
the pore fluid. However, using the more restrictive theoretical framework resulting from assuming
(i) an infinite domain, (ii) material fully impermeable, (iii), material linear elastic, (iv) linear-
elastic fracture mechanics, and (v) Carter’s leakoff model (Howard 1957, Charlez 1997),
approximate analytical solutions exist in the form of regular asymptotic expansions (Bunger 2005,
Detournay 2006, Garagash 2006, Hu 2010, Garagash 2011, Peirce 2008, Savitski 2002). The
governing equations then simplify to (i) the equilibrium equation for the linear elastic material,
that for an infinite domain can be represented as an singular integral equation relating fracture
opening and fluid pressure, (ii) the local and global mass balance equations for the fracturing fluid,
and (iii) the fracture propagation criterion, also expressible as a singular integral equation relating
fracturing pressure and fracture toughness. A non-dimensional analysis of this reduced system of
equations uncovers the presence of two pairs of competing physical processes. The first pair
consists of competing dissipative mechanisms: (a) energy dissipated due to fluid viscosity and (b)
energy dissipated due to fracture propagation; the second pair consists of competing components
of fluid balance: (a) fluid storage within the fracture and (b) fluid leakage from the fracture into
the surrounding material. Depending on which of the two dissipative mechanisms and which of
the two storage mechanisms dominate, four primary limiting regimes of propagation emerge:
 Viscosity dominated and storage dominated propagation regime ( ).
 Toughness dominated and storage dominated propagation regime ( ).
10 2014 SIMULIA Community Conference
www.3ds.com/simulia
  Viscosity dominated and leak-off dominated ( ).
 Toughness dominated and leak-off dominated regime ( ).

These four fracture propagation regimes can be conceptualized in a rectangular parametric space
where each limiting regime corresponds to each of the vertices of the rectangle with one
dissipation mechanism dominating and the other being neglected, and one component of fluid
global balance dominating with the other also neglected (Figure 5).

∞
0

0
∞

∞ 0
∞

Figure 5: Parametric diagram representing the four limiting propagation regimes of

hydraulically driven fractures
This work analyzes each benchmark problem in both the toughness/storage-dominated (near- )
and the viscosity/storage-dominated (near- ) propagation regimes. The near- and the near-
asymptotic solutions (small time solutions in the toughness and viscosity regimes) are used to
compare to Abaqus numerical solutions for each formulation (cohesive element method and
XFEM) with material parameters, loads, and boundary conditions that reproduce each of these
propagation regimes.
In order to render the Abaqus solution comparable with the asymptotic solutions, the model
dimensions and material properties are selected such that the more restrictive conditions for which
these solutions apply are adequately recreated. Specifically, the dimensions of the domain of
analysis are much bigger than the fracture aperture and length, the permeability is defined to
minimize the influence of poroelastic effects ahead of the fracture tip, and cohesive properties are
selected to ensure a small cohesive zone relative to the size of the fracture.
4.2 Radial (Penny-Shaped) Model
The first benchmark problem consists of an axisymmetric, penny-shaped, hydraulically-driven
fracture propagating in a cylindrically shaped poroelastic formation as illustrated in Figure 6.

2014 SIMULIA Community Conference 11

www.3ds.com/simulia
 0

,
,

Figure 6: Cylindrical domain with a horizontal, circular-shaped, hydraulically driven

fracture
The domain of analysis is characterized by the inner radius , outer radius , and height . The
porous medium is characterized by Young’s modulus , Poisson ratio , fracture toughness ,
porosity , Biot’s coefficient , Biot’s modulus , and hydraulic conductivity . An
incompressible Newtonian fluid with viscosity is injected at a constant rate at the center of
the fracture from a vertical wellbore. The fracture aperture , , the net pressure ,
(defined as the difference between the fracturing fluid pressure , and the confining stress
), and the fracture radius are the sought quantities.
4.3 Plane Strain (KGD) Model
The second benchmark problem considers a hydraulically-driven vertical fracture propagating in a
poroelastic prismatic-shaped formation of length L, depth R and height H as illustrated in Figure
7.
12 2014 SIMULIA Community Conference
www.3ds.com/simulia
 ,

Figure 7: Prismatic domain with a planar vertical, hydraulically driven fracture

Again, the porous medium is characterized by Young’s modulus , Poisson ratio and fracture
toughness , porosity , Biot’s coefficient , Biot’s modulus , and hydraulic conductivity
with an incompressible Newtonian fluid with viscosity injected along a vertical wellbore at a
constant rate per unit of vertical length . The unknowns of the problem are the fracture aperture
, , the net pressure , , , and the fracture radius .
4.4 Model Setup
4.4.1 Geometry and Mesh for the Radial (Penny-Shaped) Fracture Model
For each propagation regime (toughness/storage and viscosity/storage dominated) and modeling
procedure (cohesive element method and XFEM) two models are constructed: (a) 2D
axisymmetric model (Figure 8) and (b) 3D model (Figure 9). Both models are expected to render
similar solutions. Due to the axial symmetry of the problem only one quarter of the 3D cylindrical
domain is considered.

2014 SIMULIA Community Conference 13

www.3ds.com/simulia
 Coupled pressure/deformation Coupled pressure/deformation
cohesive elements (COHAX4P) continuum elements (CAX4P) enriched

Figure 8: Axisymmetric and plane strain model meshes, cohesive element and
XFEM procedures

Coupled pressure/deformation Coupled pressure/deformation

cohesive elements (COH3D8P) continuum elements (C3D8P) enriched

Figure 9: 3D model meshes, cohesive element and XFEM procedures

The formation is discretized with linear coupled pore fluid diffusion/stress elements (CAX4P for the
axisymmetric model, and C3D8P for the 3D model). For the cohesive element method, coupled
pressure/deformation cohesive elements (COHAX4P for the axisymmetric model, and COH3D8P for
the 3D model) are inserted along the horizontal mid-plane of the domain in which the fracture is
expected to propagate. Continuum elements and cohesive elements are connected with shared
nodes.

14 2014 SIMULIA Community Conference

www.3ds.com/simulia
For the XFEM model, an equivalent mesh is used where the layer of cohesive elements is replaced
by a set of enriched coupled pore fluid diffusion/stress elements (CAX4P and C3D8P), as depicted in
Figures 8 and 9. To declare a set of continuum elements as enriched in which a fluid-driven
fracture may propagate the following keyword is required:
*ENRICHMENT NAME=enrichmentname, TYPE=PROPAGATION CRACK, ELSET=elsetname
The dimensions of the analysis domain are chosen large enough to minimize boundary effects
(Table 1).
Table 1: Penny-Shaped fracture model dimensions
Dimension Value
Inner radius 0.01
Outer radius 45
Height 30
Angle 90°

In order to analyze the convergence of the two modeling procedures as the mesh is refined, three
meshes of decreasing element size are recreated for the 2D axisymmetric and 3D continuum cases
(Figures 10 and 11). To minimize model size, only the elements in the vicinity of the fracture
plane are refined. These elements are generated by subdividing elements in the parent (coarse)
mesh into four (axisymmetric model) or eight (3D model) child elements. The compatibility of
displacements and pore fluid pressures on the nodes lying at the intersection between coarse and
fine regions of the resultant meshes are enforced with multi point constraints (*MPC, LINEAR, and
*MPC, BILINEAR). This refinement technique is chosen to minimize element distortion. Table 3
lists the number of elements and nodes for the respective meshes.

2014 SIMULIA Community Conference 15

www.3ds.com/simulia
 COH XFEM
E

Coarse

Intermediate

Fine

Figure 10: Meshes of different resolution used for the 2D penny-shaped and planar
fracture models (axisymmetric and plane strain)

16 2014 SIMULIA Community Conference

www.3ds.com/simulia
 COH XFEM
E

Coarse

Intermediate

Fine

Figure 11: Meshes of different resolution used for the Penny-shaped fracture
model
4.4.2 Geometry and Mesh for the Planar (KGK) Fracture Model
For the second benchmark problem (planar fracture geometry) two pairs of models are also
constructed: (a) 2D plane-strain model with nodal coordinates and element connectivity identical
to the axisymmetric model (Figures 8 and 10) and (b) 3D model as depicted in Figure 12. As in
the case of the penny-shaped mesh, the formation is discretized with linear coupled pore fluid
diffusion/stress elements (CPE4P for the plane strain model and C3D8P for the 3D model), while the
fracture plane is modeled with either plane strain (COH2D4P) or 3D (COH3D8P) coupled
pressure/deformation cohesive elements (i.e. cohesive element method), or the enriched version of
the coupled pore fluid diffusion/stress continuum elements (i.e. XFEM).
Table 2 summarizes the planar fracture model dimensions chosen to minimize boundary effects.
Table 2: Planar fracture model dimensions
Dimension Value
Width 45
Height 45
Length 60

2014 SIMULIA Community Conference 17

www.3ds.com/simulia
Figure 13 details the sequence of meshes of different resolution constructed to analyze
convergence for this second benchmark model.

Coupled pressure/deformation Coupled pressure/deformation

cohesive elements (COH3D8P) continuum elements (C3D8P) enriched

Figure 12: 3D mesh for the KGD model, cohesive procedures

Figure 13: Meshes of different resolution used for the KGD fracture model

18 2014 SIMULIA Community Conference

www.3ds.com/simulia
 Table 3: Number of elements and nodes for each mesh
Model Mesh Elements Nodes
Coarse 2745 2898
2D Intermediate 3330 3577
Fine 4500 4931
Coarse 54900 60858
3D Intermediate 82800 95137
FIne 194400 230291

4.4.3 Material Properties

The poroelastic material parameters used for all simulations are summarized in Table 4.
Table 4: Poroelastic material parameters
Parameter Value
Young modulus 17
Poisson’s ratio 0.2
Biot coefficient 0.75
Porosity 0.2
Pore fluid specific weight 0.0098
Hydraulic conductivity 9.8 10
Biot modulus 68.7

To define the cohesive behavior of the material, a traction-separation law with linear softening is
used (*DAMAGE EVOLUTION, TYPE=ENERY, SOFTENING=LINEAR) defined by a cohesive energy
120 ∙ . Fracture toughness can be estimated as 1.46 √ . The
quadratic nominal stress fracture initiation criterion (*DAMAGE INITIATION, CRITERION=QUADS)
with a cohesive strength of 1.25 is used in all simulations. For the cohesive element
method, the elastic stiffness prior to failure (*ELASTIC, TYPE=TRACTION) is defined as 100 times
the Young’s modulus of the material.
The fracturing fluid is modeled as a Newtonian fluid (*GAPFLOW, TYPE=NEWTONIAN) with viscosity
equal to 100 for the viscosity/storage dominated propagation regime and
0.1 for the toughness/storage dominated regime.
4.4.4 Initial Conditions, Boundary Conditions and Loads
The following initial conditions are specified for all solid sections with poroelastic constitutive
behavior (Table 5):
Table 5: Initial conditions
Pore fluid Initial saturation 1.0
Pore fluid initial pressure 0.0
Rock initial void ratio 0.25
Initial principal in-situ stresses , , 0.0, 0.0, 0.0
The media is assumed to be initially fully saturated and with no confinement stresses. Abaqus
requires an explicit specification of the set of cohesive elements or enriched elements where the
fracturing fluid flow equations will be initially solved. These initial conditions are specified
2014 SIMULIA Community Conference 19
www.3ds.com/simulia
through the keyword *INITIAL CONDITIONS, TYPE=INITIALGAP, and *INITIAL CONDITIONS,
TYPE=ENRICHMENT, respectively. Only the set of elements containing the nodes where the fluid is
injected are needed to meet this requirement (Figure 14).

Cohesive Element Model XFEM model

Figure 14: Initial set of cohesive elements and initial enrichment required to define
the initial solution domain for the fracturing fluid flow equations
The displacements in the normal direction to all boundary surfaces and symmetry planes are
constrained. Additionally, pore fluid pressures are fixed to the uniform value 0 on all model
boundaries. For the XFEM model, the corner phantom nodes on the symmetry surfaces and
boundary surfaces are constrained to move within these surfaces (Figure 15). This special type of
boundary condition is enforced with the keyboard *BOUNDARY, PHANTOM=NODE.

Corner phantom Corner phantom

nodes constrained nodes unconstrained

Figure 15: Displacement boundary conditions on corner phantom nodes

In all cases, fracturing fluid is injected at a constant rate 0.001 and injection is simulated
for 40 . As explained in the previous section, the fracturing fluid pressure degrees of freedom are
associated with the mid-edge nodes of the cohesive elements and the edge phantom nodes of the
enriched elements. Therefore, concentrated fracturing fluid flow must be applied directly to these
mid-edge (*CFLOW) and phantom edge nodes (*CFLOW, PHANTOM=EDGE), as depicted in Figure 16.

20 2014 SIMULIA Community Conference

www.3ds.com/simulia
 Edge mid node Edge phantom

Cohesive Element Model XFEM

Figure 16: Injection rate applied to cohesive edge mid-node and enriched edge-
phantom node

5. Results
The results obtained for each benchmark problem and modeling procedures (cohesive and XFEM)
will now be presented. The numerical solutions obtained from Abaqus for the temporal and spatial
distributions of fracturing fluid pressure and aperture are compared with the small-time asymptotic
analytical solutions (Bunger 2005, Detournay 2006, Garagash 2006, Hu 2010, Garagash 2011,
Peirce 2008, Savitski 2002) for both the viscosity-storage and toughness-storage propagating
regimes.
5.1 Radial (Penny-shaped) Fracture Model
Figures 17, 18, 19, and 20 display the variations of injection pressure , , fracture mouth
opening , in time (evaluated at the inner radius 0.01 ), fracturing fluid pressure
distribution , and fracture aperture , (evaluated at the final time 40 )
along the crack surface. These 2D axisymmetric results (cohesive element and extended finite
element procedures) are obtained for the toughness-storage dominated propagation regime ( -
vertex) using the three different mesh resolutions depicted in 10. Similarly, Figures 21, 22, 23, and
24 depict the evolution and distribution of the same variables obtained for the viscosity-storage
dominated propagation regime ( -vertex). In all of these figures, the results are compared with
the asymptotic analytical solutions. Good agreement between the Abaqus cohesive element
method solution, the XFEM solution, and the analytical solution is found for all meshes, both
methods, and -vertex and -vertex limits. The relative error between Abaqus simulations and
analytical solutions for all variables decreases with more mesh refinement. The pressure
distribution along the fracture is nearly uniform for the -vertex regime as viscous dissipation is
relatively negligible (Figure 19). By contrast, the flow along the fracture in the -vertex limit is
characterized by viscous pressure loss that increases towards the crack tip. This non-uniform
pressure distribution and gradients are captured accurately using the Abaqus procedures (Figure
23).

2014 SIMULIA Community Conference 21

www.3ds.com/simulia
 Figure 17: Time evolution of the injection pressure ( -vertex)

Figure 18: Time evolution of fracture aperture near injection point ( -vertex)

Figure 19: Pressure distribution along fracture at the final time ( -vertex)

22 2014 SIMULIA Community Conference

www.3ds.com/simulia
 Figure 20: Fracture aperture at the final time ( -vertex)

Figure 21: Time evolution of the injection pressure ( -vertex)

Figure 22: Time evolution of fracture aperture near injection point ( -vertex)

2014 SIMULIA Community Conference 23

www.3ds.com/simulia
 Figure 23: Pressure distribution along fracture at the final time ( -vertex)

Figure 24: Fracture aperture at the final time ( -vertex)

Contour plots of the maximum principal effective stresses, evaluated for the final configuration
and final time 40 , are shown in Figures 25, 26, 27, and 28. Since the stresses are
compressive (negative), the minimum (most negative) principal stresses must be interrogated from
the Abaqus output database in order to obtain the maximum (compressive) principal stresses.
Similarly, the contours plots of maximum principal stress obtained for the 3D fine mesh models
for both procedures and both propagation regimes are displayed in Figures 29 and 30. The
equivalence between the 2D axisymmetric and 3D Abaqus solutions is noteworthy.

24 2014 SIMULIA Community Conference

www.3ds.com/simulia
 Figure 25: Maximum effective principal stresses at the final time for the cohesive
element method ( -Vertex propagation regime)

Figure 26: Maximum effective principal stresses at the final time for the XFEM ( -
Vertex propagation regime)

Figure 27: Maximum effective principal stresses at the final time for the cohesive
element method ( -Vertex propagation regime)

2014 SIMULIA Community Conference 25

www.3ds.com/simulia
Figure 28: Maximum effective principal stresses at the final time for the XFEM ( -
Vertex propagation regime)

Figure 29: Maximum effective principal stresses for the 3D, fine mesh model, for
both the cohesive element method (left) and XFEM (right) and the -Vertex
propagation regime

26 2014 SIMULIA Community Conference

www.3ds.com/simulia
 Figure 30: Maximum effective principal stresses for the 3D, fine mesh model, for
both the cohesive element method (left) and XFEM (right) for the -Vertex
propagation regime
5.2 Planar (KGD) Fracture Model
Figures, 31 and 32 display the variations of the injection pressure , and fracture mouth
opening , in time (evaluated at the injection node 0) obtained for the planar (KGD)
fracture problem for each modeling technique (cohesive element method and XFEM) and each of
the meshes described in Figure 10. In this section only the results for the toughness-storage
dominated propagation regime ( -vertex limit) are presented for brevity. As in the penny-shaped
fracture model, good agreement is found between Abaqus and the analytical solution for both
modeling procedures and the relative error between solutions decrease monotonically as the mesh
is refined. Similar performance is observed for the viscosity-storage dominated propagating
regime ( -vertex limit).

Figure 31: Time evolution of the injection pressure ( -vertex)

2014 SIMULIA Community Conference 27

www.3ds.com/simulia
 Figure 32: Time evolution of fracture aperture near injection point ( -vertex)
Figures 33 and 34 show the final fracture configuration and contour plots of the maximum
compressive effective principal stresses for each modeling technique and each mesh. Additionally,
Figure 35 presents the corresponding contour plots for the 3D Abaqus models of the finer mesh.
Again, it is worth noting the excellent agreement achieved between the solutions of each method
and each element type (plane-strain and 3D).

Figure 33: Maximum effective principal stresses at the final time for the cohesive
element method ( -Vertex propagation regime)

Figure 34 Maximum effective principal stresses at the final time for the XFEM ( -
Vertex propagation regime)
28 2014 SIMULIA Community Conference
www.3ds.com/simulia
 Figure 35: Maximum effective principal stresses for the 3D, fine mesh model, for
both the cohesive element method (left) and XFEM (right) and the -Vertex
propagation regime

6. Conclusions
This work describes and analyzes the Abaqus hydraulic fracturing capabilities co-developed
between ExxonMobil Upstream Research Company and Dassault Systemes Simulia Corporation.
Two new element classes have been implemented and integrated into the Abaqus general purpose
porous media analysis solver:
i. a coupled pressure/deformation, cohesive element
ii. An enriched version of the continuum coupled pore fluid diffusion/stress elements
(XFEM) tied with dynamically inserted (and oriented) zero-thickness interface elements.
Both elements incorporate new fracturing fluid pressure degrees of freedom to model the
fracturing fluid flow within the opening crack. The resulting hydraulic fracturing modelling
procedure accounts for the coupling between the deformation of the porous medium, the flow of
pore fluid within the pore network, the flow of fracturing fluid within the opening fracture, the
leakage of fracturing fluid into the adjacent pores, and the fracture initiation and propagation. The
consistency, accuracy and convergence qualities of the two modelling techniques have been
assessed by modelling two benchmark problems in 2D and 3D:
i. propagation of a penny-shaped fracture in a cylindrically shaped formation
ii. propagation of a vertical planar fracture in a prismatic-shaped formation, both in the
toughness/storage dominated and viscosity/storage dominated propagation regimes
Excellent agreement between Abaqus and asymptotic analytical solutions for these two benchmark
problems has been found for the different modelling procedures, dimensionality (i.e. 2D and 3D),
and propagation regimes. The Abaqus solution converges monotonically as mesh spatial
resolution is increased. The validation exercise presented in this work lays the foundation for the
successful development of fully coupled simulation capabilities for fluid driven fracturing

2014 SIMULIA Community Conference 29

www.3ds.com/simulia
applications in the oil and natural gas industry including injection, stimulation, and drilling
operations.

7. References
1. Abaqus User’s Manual, Version 6.13, Dassault Systèmes Simulia Corp., Providence, RI.
2013.
2. Boone T.J., Ingraffea, A. R., “A Numerical Procedure For Simulation Of Hydraulically‐
Driven Fracture Propagation In Poroelastic Media,” International Journal for Numerical and
Analytical Methods in Geomechanics, vol. 14, pp. 27-47, 1990.
3. Bunger, A. P., Detournay, E., Garagash, D. I., “Toughness-Dominated Hydraulic Fracture
with Leak-Off,” International Journal of Fracture, vol. 134, pp.175-190, 2005.
4. Carrier B., Granet S., “Numerical Modeling Of Hydraulic Fracture Problem In Permeable
Medium Using Cohesive Zone Model,” Engineering Fracture Mechanics, vol. 79, pp. 312–
328, 2012.
5. Clearly M. P., “Comprehensive Design Formulae for Hydraulic Fracturing,” Society of
Petroleum Engineers SPE-9259-MS, 1980.
6. Charlez P.A., “Rock Mechanics, Volume 2, Petroleum Applications,” Editions Technip, 1997.
7. Detournay, E., Adachi, J., Garagash, D. I., Savitski, A., “Interpretation and Design of
Hydraulic Fracturing Treatments,” United States Patent No. US 7111681B2, 2006.
8. Garagash, D. I., “Plane-Strain Propagation Of A Fluid-Driven Fracture During Injection and
Shut-In: Asymptotics Of Large Toughness,” Engineering Fracture Mechanics, vol. 73, pp.
456-481, 2006.
9. Hu, J., Garagash, D. I., “Plane-Strain Propagation Of A Fluid-Driven Crack in a Permeable
Rock With Fracture Toughness,” J. Eng. Mech., vol. 136(9), pp. 1152–1166, 2010.
10. Garagash, D. I., Detournay, E., Adachi, J., “Multiscale Tip Asymptotics in Hydraulic Fracture
with Leak-Off,” J. Fluid Mech., vol. 669, pp. 260–297, 2011.
11. Geertsma, J., de Klerk, F., “A Rapid Method Of Predicting Width and Extent of Hydraulically
Induced Fractures,” Society of Petroleum Engineers SPE 2458, 1969.
12. Howard G. C., Fast. R., Carter R.D. “Optimum Fluid Characteristics for Fracture Extension,”
Drilling and Production Practices, pp. 261-270, 1957.
13. Meyer B.R. “Three-Dimensional Hydraulic Fracturing Simulation on Personal Computers:
Theory and Comparison Studies,” Society of Petroleum Engineers, SPE 19329, 1989.
14. Ortiz, M., Pandolfi, A., “Finite-Deformation Irreversible Cohesive Elements For Three-
Dimensional Crack-Propagation Analysis,” Int. J. Numer. Meth. Engng. Vol. 44, pp. 1267-
1282, 1999.
15. Peirce, A., Detournay, E., “An Implicit Level Set Method for Modeling Hydraulically Driven
Fractures.” Comput. Methods Appl. Mech. Engrg, vol. 197, pp. 2858–2885, 2008.
16. Remmers, J. J. C., de Borst R., Needleman A., “The Simulation of Dynamic Crack
Propagation using the Cohesive Segments Method,” Journal of the Mechanics and Physics of
Solids, vol. 56, pp. 70–92, 2008
30 2014 SIMULIA Community Conference
www.3ds.com/simulia
17. Savitski, A. A., Detournay, E., “Propagation Of A Penny-Shaped Fluid-Driven Fracture In An
Impermeable Rock: Asymptotic Solutions,” International Journal of Solids and Structures,
vol. 39, pp. 6311-6337, 2002.
18. Song J.H., Areias P.M.A, Belytschko T.A., “A Method for Dynamic Crack and Shear Band
Propagation with Phantom Nodes,” Int J. Numer. Methods Eng., vol. 67, pp 868–893, 2006
19. Sukumar, N., Prevost J. H., “Modeling Quasi-Static Crack Growth with the Extended Finite
Element Method Part I: Computer Implementation,” International Journal for Solids and
Structures, vol. 40, pp. 7513–7537, 2003
20. Warpinski N.R., Moschovidis Z.A., Parker C.D., Abou-Sayed I.S., “Comparison Study of
Hydraulic Fracturing Models,” Society of Petroleum Engineers, SPE-25890-PA, 1994.
21. Yew, C. H., “Mechanics of Hydraulic Fracturing,” Gulf Publishing Company, Houston, TX,
1997.

2014 SIMULIA Community Conference 31

www.3ds.com/simulia