Accepted Manuscript

An advanced continuum damage mechanics model for predicting the

crack progress process based on the consideration of the influence of
crack direction under quasi-static load

Kumchol Yun, Zhenqing Wang, Sakaya Ronald, Yongchol Pak

PII: S0020-7403(17)30288-6
DOI: 10.1016/j.ijmecsci.2017.05.021
Reference: MS 3696

To appear in: International Journal of Mechanical Sciences

Received date: 1 February 2017

Revised date: 5 May 2017
Accepted date: 7 May 2017

Please cite this article as: Kumchol Yun, Zhenqing Wang, Sakaya Ronald, Yongchol Pak, An advanced
continuum damage mechanics model for predicting the crack progress process based on the consid-
eration of the influence of crack direction under quasi-static load, International Journal of Mechanical
Sciences (2017), doi: 10.1016/j.ijmecsci.2017.05.021

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service
to our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and
all legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

T
IP
CR
US
AN
M
ED
PT
CE
AC

1
 ACCEPTED MANUSCRIPT

An advanced continuum damage mechanics model for predicting

the crack progress process based on the consideration of the
influence of crack direction under quasi-static load
Kumchol Yuna,b , Zhenqing Wanga,∗, Sakaya Ronalda,c , Yongchol Pakb
a
College of Aerospace and Civil Engineering, Harbin Engineering University, Harbin, 150001, P.R.China
b
Faculty of Mechanics, Kim Il Sung University, Pyongyang 950003, D.P.R. Korea

T
c
College of Civil and Building Engineering, Kyambogo University, Kampala, Uganda

IP
CR
Abstract
In reality the wrong crack path can be generally obtained in the case of arbitrary crack
propagation by traditional continuum damage mechanics (CDM). In this paper a novel ad-

US
vanced continuum damage mechanics (ACDM) method is proposed, which can predict the
crack propagation and fracture behavior correctly for the structures. The material prop-
erty degradation method, which is usually used when simulating the structures within the
AN
framework of CDM, is advanced based on considering the influence of crack direction. The
maximum tensile stress criterion is used to predict the damage initiation and crack prop-
agation direction and the advanced CDM used to predict the damage evolution process in
meso-level under the quasi-static load. It can directly evaluate the propagation process of
M

the discrete crack and the fracture strength for structures using the continuum model as
well as not using discontinuum model. The algorithm for the application of our advanced
CDM theory in the numerical simulation based on finite element method (FEM) is presented.
ED

ACDM model is not only a simple and useful model which can easily be used in FEM frame-
work but also a phenomenological model based on the concept of crack propagation. The
simulation results by our ACDM are compared with the experiment results and the ones and
from cohesive zone method and extended finite element method for good agreements to be
PT

achieved.
Keywords: continuum damage mechanics(CDM); discrete crack; damage initiation; crack
CE

direction parameter

1. Introduction
AC

It is very important to predict the crack propagation and failure behavior in structures
correctly under quasi-static load. The numerical simulation using finite element method(FEM)
has become a powerful tool for predicting the macrocrack initiation and propagation and
rendering the failure mechanisms, because it is difficult to get theoretical results under the
complex loading and boundary conditions.

∗
The corresponding author: Zhenqing Wang, Tel: +86 0451 82589364
Email address: wangzhenqing@hrbeu.edu.cn (Zhenqing Wang)

Preprint submitted to International Journal of Mechanical Sciences

 ACCEPTED MANUSCRIPT

Various numerical methods which can simulate the crack problem have been proposed
and demonstrated. Virtual crack closure technique (VCCT) based on linear elastic fracture
mechanics (LEFM) has been widely used to simulate the crack propagation problems [1-3].
It can represent the crack propagation process effectively and does not require a very good
mesh, but it has some disadvantages such as re-meshing and a presence of initial crack[4,16].
Cohesive zone model (CZM) based on the nonlinear traction-separation law is also an widely
used finite element (FE) approach to simulate the crack propagation process in the past three
decades [5-7]. CZM was first proposed by Dugdale [8] and Barenblatt [9] and investigated
by many researchers. It does not require the presence of initial crack and re-meshing for

T
crack propagation, and can predict both initiation and propagation of crack to solve the

IP
problem which has the pre-known crack propagation path such as interface crack problem
for laminated composites [6,10].

CR
Recently extended finite element method (XFEM), which firstly developed by Belytschko
et al. [12] in 1999 based on the combination of partition of unity approach (PUA) and
FEM, is the most widely used approach to simulate the crack propagation problems [4,13-
14]. It can capture the crack within finite element by using the enriched shape function

US
and consider the crack defined independently of mesh. XFEM has advantages that can
capture the displacement field of discontinuity reflecting the material response in front of
the crack tip and can use the standard finite elements without re-meshing [15]. XFEM using
AN
discontinuum model has been recognized as a very excellent method to predict the arbitrary
crack propagation, but it is necessary to be known as the enrichment function. XFEM has
disadvantage that it requires the different enrichment functions and failure criteria with the
different cracks and materials, so it has been generally used for the static and quasi-static
M

problems in homogeneous materials, but the enrichment functions are not readily available
except for some very special cases in heterogeneous ones [16].
Continuum damage mechanics (CDM) has been extensively used in fracture analysis
ED

problems combined with multi failure modes of materials such as composite laminates applied
to quasi-static load or impact loads [18,29,35]. Unlike fracture mechanics that can only
capture individual crack initiation and propagation, the CDM approach has captured the
PT

overall system response with the different damage mechanisms [21]. CDM theory was first
applied by Kachanov [22] and Rabotnov [23] to study creep rupture of metal materials and
it can predict the multi damage fracture process by introducing the damage variables. In
general, damage mechanics deals with the average quantity called representative volume
CE

element(RVE) and the damage variable has been defined by using this averaging method.
In damaged material, effective stress concept and effective strain concept have been re-
spectively applied to represent stress and strain field reflecting the damaged state and among
AC

two concepts effective stress concept has been more extensively used in engineering appli-
cations [24]. Based on the effective stress method of Kachanov, Lemaitre[26-27] suggested
equivalent stress hypothesis and material property degradation model which can directly
obtain the strain field from effective stress field. Material property degradation matrix
was determined based on the damage variable and several method was proposed to obtain
the damage variable [19,28,33,37]. The method, by which the gradually increasing damage
variable based on the energy amount dissipated for various damage modes of composite
materials is got, was introduced by [31]. In the isotropic materials, the isotropic damage
model have been still used in many engineering applications, because it is more simple than

3
 ACCEPTED MANUSCRIPT

anisotropic damage. CDM approach in conjunction with FEA was used to estimate fretting
fatigue crack initiation lifetime of double bolted lap joint with isotropic damage model by
Ferjaoui and Wahab [33]. The isotropic damage model of CDM is used to describe the post-
elastic behavior of low-basis-weight paper by Hägglund and Isaksson [34] in order to enable
characterization, optimization and quality control. Anisotropic damage model was widely
studied to consider the degradation of the material property at the different direction in the
isotropic material. Murakami and Kamiya [35] developed the anisotropic damage model on
the basis of irreversible thermodynamic theory to describe the elastic damage material and
Ismail and Zaı̈ri [36] had used finite element analyses explicitly taken into account by the

T
anisotropic CDM model which is enhanced by coupling with a vanishing element technique.

IP
An anisotropic damage model with bi-linear damage evolution law for CDM was proposed
by Sun [37] to model the response of glass subjected to the static load. The stress-based

CR
anisotropic damage tensor was studied by employing normal principal stresses and shear
stresses, but it emphasized that its CDM model did not address issues such as discrete
crack growth, crack propagation, or crack kinking. A study for advancing CDM theory has
been widely researched, but majority of studies cannot predict the discrete crack progress

US
direction correctly, for these models have no direct parameters reflecting the crack direction.
They said that it is difficult by continuum model to simulate crack propagation process
which is the discontinuum process phenomenologically. Little studies have been performed
AN
to simulate the crack propagation process using CDM recently. Özden [38] implemented an
isotropic damage model based on a CDM approach in Abaqus/Explicit and simulated the
crack propagation in the material with the WC and the Co phases by using the separate
damage laws based on brittle failure and fatigue respectively. Lu summarized three types
M

of formulation for the embedded finite element method (EFEM) that incorporated strain
or displacement discontinuities into standard finite element and proposed a multiscale finite
element model coupled with technique of embedded elements [44]. Han and Lubineau devel-
ED

oped a morphing method coupled local-continuum damage mechanics with peridynamics to

objectively simulate the damage nucleation, crack formation and propagation for material
failure [40-43]. Cervera [29] proposed a tracking algorithm that can simulate starting and
PT

propagating of localized tensile cracks along one or two orientations with an isotropic dam-
age constitutive model based on the CDM on quasi-brittle materials. This crack-tracking
method inserted in the continuum model need many parameters defined by the user such as
exclusion radius(rexcl ), crack tip neighbourhood radius(rneigh ) and maximum curvature angle
CE

which can influence the result of crack path prediction, and the crack can only increase from
one side of an element to another side of the same element at each time step. In a word,
there is no effective CDM model for capturing the crack progression process correctly and
AC

also reflecting the different damage mechanisms together.

In this paper, ACDM method which can correctly predict the crack propagation and
fracture behavior for the structures, is proposed to prove the accuracy of CDM model.
Damage initiation and damage evolution process were predicted by MS criterion and the
material property degradation method under the quasi-static load based on considering the
influence of crack direction respectively. The propagation process of discrete crack and
fracture strength for structures can be directly evaluated using the continuum model. The
algorithm for the application of our advanced CDM theory in the numerical simulation based
on finite element method (FEM) is presented and can be easily used in FEM framework.

4
 ACCEPTED MANUSCRIPT

The simulation results using our ACDM are compared with the experiment results and ones
using cohesive zone method and XFEM, and good agreements are achieved.

2. Advanced continuum damage mechanics (ACDM) model

In this section, the advanced continuum damage mechanics model to consider the prop-
agation path of discrete crack when constructing the material property degradation matrix
as in the damage propagation was investigated. The general form of model was introduced
in 3D damage problem and then the detail of applied techniques in 2-D damage problem

T
was presented. The isotropic material was only considered and the detailed expressions were

IP
all presented in this research in order to help the comprehensive understanding and give the
basis for the more extensive applications in future.

CR
2.1. Space damage problem
The effective stress concept has still been the basic key part in CDM theories and has
succeeded in most advanced damage mechanics model recently. In the case of isotropic

US
damage, the effective stress is expressed as follows.

σ̃ = (1 − D)Cε (1)
AN
where σ̃ and ε are the effective Cauchy stress and strain tensors respectively, D is the
damage variable which is concerned with the ratio between the effective stress and Cauchy
stress and constant in isotropic damage, and C is the stiffness matrix. For the anisotropic
M

damage of anisotropic materials such as composites, D is not scalar value and converted to
the matrix and 1 is converted to the unit matrix I. The plane damage model of orthotropic
material which is mostly employed in FEM simulation was presented in [31], where the fiber
ED

tension damage, fiber compression damage, matrix tension and matrix compression were
defined separately and the shear damage is dependent on the others. We have introduced
the anisotropic damage concept for the discrete crack problem in isotropic materials and then
made the new damage mechanics model, which can predict the discrete crack propagation
PT

direction and path in continuum model.

The stress-based anisotropic damage model for CDM was proposed by Sun [37] in isotropic
material by using normal principal stresses and shear stresses, but it cannot address problems
CE

such as discrete crack growth, crack propagation. In order to amend these drawbacks the
crack direction parameter is also introduced as a main damage factor with damage variable
in CDM frameworks.
AC

Two hypotheses were built up to get the elasticity matrix for the element with crack as
follows. First, the element size is small enough, so in an element, the crack direction is not
changed and the effect of the crack situation can be neglected. Second, the crack begins to
open and then propagate in the fixed direction in an element, if the energy dissipated per
unit area within an element reaches the critical energy dissipation rate of material.
The crack has the direction and location in an element, so the effect of the crack location
can influence the damage evolution process and crack propagation path. The first hypothesis
was employed to neglect this effect in predicting the crack path with raising accuracy when
having small element size. The second hypothesis was used to obtain the damage variables

5
 ACCEPTED MANUSCRIPT

for damaged elements using fracture mechanics techniques. In order to predict the mixed
mode failure, Erdogan and Sih[25] proposed a maximum tension stress(MS) criteria in which
the crack extension occurs in the direction at which, the first principal stress (FPS) σ1 occurs
near the crack tip and fracture takes place when MS is equal to the stress that leads to Mode
I fracture. In this paper, MS criteria is used to predict the damage initiation and the crack
direction within elements. When FPS σ1 reaches its own critical value, damage is initiated
and at this time the crack plane is perpendicular to the FPS direction.
Damage variables are obtained based on the criteria of energy dissipation rate in linear
elastic fracture mechanics (LEFM) and equivalent stress σe and equivalent displacement δe

T
are defined as follows.

IP
For tension: q
2 2
δe = Le < ε11 >2 +γ12 + γ13 , σe =< σ1 > (2)

CR
For compression:
q
2 2
δe = Le < −ε33 >2 +γ31 + γ32 , σe =< −σ3 > (3)

US
where σ1 and ε11 are the first principal stress and the tension strain in the local element
coordinates respectively, σ3 and ε33 are the third principal stress and the compression strain
in the local element coordinates respectively, γ12 , γ13 and γ32 are the shear strains in the
AN
local element coordinates, and < · > is the McCauley operator. Le is √ characteristic length of
element, which is √calculated from the element area A as Le = 1.12 A for a square element
and as Le = 1.52 A for a triangular element respectively [39]. The constitutive relation
is converted from stress-strain to stress-displacement relation with the characteristic length
M

and it can decrease mesh dependency on the result for finite element analysis[32]. Energy
dissipated per unit area Gc , which is specified individually for two damage modes(tension and
compression), is expressed as a function of the equivalent stress and equivalent displacement.
ED

Z δef
Gc = σe dδe (4)
0
PT

where δef is the equivalent displacement when the material reaches its ultimate strength.
The individual damage variables are expressed as follows.

CE

 0f
 δe < δe∗
∗
δe (δe −δe )
D= δe (δef −δe∗ )
δe∗ ≤ δe < δef (5)

 f
1 δe ≥ δe
AC

where δe∗ is the equivalent displacement when damage initiates. The equivalent displacement
for each damage mode is calculated only once at the damage initiation when the effective
stress is the same as the maximum admissible stress, and this value is used for damage
evolution process. The crack surface is set as the surface at where the FPS becomes maximum
from MS criteria and the principal stress coordinates are set as the element local coordinates.
We set the x0 -axis as perpendicular to the crack surface and y 0 -axis as the crack direction in
local coordinates.
At the same time the material properties are degraded in damaged elements. This degra-
dation is not isotropic and concerned with the crack direction in elements. Ex0 was set as

6
 ACCEPTED MANUSCRIPT

E(1 − D) in the FPS direction and Ey0 and Ez0 were kept to the initial state as E, so the
damaged elements are turned from isotropic to orthotropic. In local coordinates, the stiffness
matrix C 0d of a damaged element is turned as follows.
 
C11 (1 − D) C12 (1 − D) C13 (1 − D) 0 0 0
 C12 (1 − D) C22 C23 0 0 0 
 
 C (1 − D) C C 0 0 0 
Cd = 
0

13 23 33  (6)

 0 0 0 G23 0 0 
 0 0 0 0 G31 (1 − Ds ) 0 

T
0 0 0 0 0 G12 (1 − Ds )

IP
where Cij , i, j = 1, 2, 3 and Gij , i 6= j are the elasticity stiffness coefficients and the shear
stiffness coefficients respectively, and D and Ds are the tension damage variable and shear

CR
damage variable respectively. In compression the constitutive model is set as isotropic dam-
age mode such as C 0d = (1 − D)C.
Rewriting Eq. (6) in isotropic material

Cd = 
0

 C12 (1 − D)

 C 12 (1 − D)
C11
C 12
US
C11 (1 − D) C12 (1 − D) C12 (1 − D) 0
C12
C 11
0
0
0
0
0
0
0
0





AN
  (7)
 0 0 0 G 0 0 
 0 0 0 0 G(1 − Ds ) 0 
0 0 0 0 0 G(1 − Ds )
M

The elasticity stiffness coefficients and the shear stiffness coefficients are as follows: C11 =
E(1 − ν)/((1 − 2ν)(1 + ν)), C12 = Eν/((1 − 2ν)(1 + ν)), G = 0.5E/(1 + ν), where E, ν are
the elasticity modulus and Poissons ratio respectively, G is the shear modulus. The elastic
ED

property at a direction perpendicular to crack surface are decreased with the increment
of damage variable, while the elastic properties at the other directions are conserved and
not concerned with the change of damage variable at a element. When damage variable D
PT

reaches to 1 at an element, the element is fully fractured and the killing element technique
introduced by Ismail [36] is applied. It means that the crack is formed by killing element in
continuum medium model and the artificially crack control is not performed in this work.
CE

The stiffness matrix equation (7) at a damaged element expressed in local coordinates which
is placed on the crack surface can be re-expressed in global Cartesian coordinates.

C d = T σ C 0d T −1 (8)
AC

where T σ and T ε are the transform matrix of stress and strain respectively.

2.2. Plane damage problem

In this section the plane stress problem is typically investigated, and because the plane
strain problem can be converted to plane stress problem with substituting E and ν to
E(1 + 2ν)/(1 + ν)2 and ν/(1 + ν) respectively in isotropic materials, if the effort of through-
thickness stress can be neglected. The undamaged elasticity matrix can be expressed as
follows.

7
 ACCEPTED MANUSCRIPT

 
1 ν 0
E 
C= ν 1 0  (9)
1 − ν2 1−ν
0 0 2
Stress tensor σ and strain tensor ε are given as:
 T  T
σ = σx , σy , τxy , ε = εx , εy , γxy
The element with crack is shown in Fig. 1, where the local coordinates placed on the

T
crack surface are angled with the global Cartesian coordinates properly and the origin of local
coordinates laid on the centre of the element. The axis x0 is set as the direction perpendicular

IP
to the crack surface and axis y 0 as crack direction.
Damaged elasticity matrix in local coordinates placed in crack surface, can be expressed

CR
from Eq. (8).
 
1−D ν(1 − D) 0
E  
C 0d = ν(1 − D) 1 0 (10)
1 − ν2
0
US 0 1−ν
2
(1 − Ds )
where shear damage variable Ds is not independent with damage variable D, so to simplify
the analysis, the linear relationship is assumed to Ds = βD [17].
AN
The parameter β will be called as shear relation parameter in this paper. The transform
matrix T σ and T ε can be simply expressed as the function of the angle α between the direc-
tion perpendicular to crack surface (x0 -axis) and transverse direction of global coordinates
M

(x-axis).
 2   2 
l m2 −2lm l m2 −lm
T σ =  m2 l 2 2lm  , T ε =  m2 l2 lm  (11)
ED

2 2
lm −lm l − m 2lm −2lm l − m22

The inverse matrices of transform matrices are calculated as follows:

PT

T −1 T −1 T
σ = Tε , Tε = Tσ (12)

where l and m are cosine and sine of the angle α respectively.

CE

l = cosα, m = sinα (13)

where the angle α between x-axis and x0 -axis is expressed as follows.

AC

σ 1 − σx
π
α = arctan , θ =α+ (14)
τxy 2
The angle θ characterize the crack direction, while the angle α characterize the direction
perpendicular to the crack surface. The damaged elasticity matrix in global coordinates
is expressed from Eq. (8). The damaged elasticity matrix C 0d in local coordinates can be
expressed as the sum of two elasticity matrices from Eq. (10).
0
C 0d = C d + ∆C 0d (15)

8
 ACCEPTED MANUSCRIPT

where  
1 ν 0
0 E(1 − D) 
Cd = ν 1 0  (16)
1 − ν2
0 0 1−ν 2
 
0 0 0
ED  
∆C 0d = 0 1 0 (17)
1 − ν2 (1−ν)(1−β)
0 0 2
Substituting Eq. (15) to Eq. (8)

T
0
0
C d = T σ C d + ∆C 0d T −1 −1 0 −1
ε = T σ C d T ε + T σ ∆C d T ε (18)

IP
From Eq. (9) and (16)
0
C d = (1 − D)C (19)

CR
0
Substituting Eq. (16) to Eq. (8) and considering that the elasticity matrix has the Cd
same form with the elasticity matrix in isotropic material, we can get the isotropic damaged
elasticity matrix in global coordinates C d .

Rewriting Eq. (18)

US
C d = T σ C d T −1
0
ε = Cd
0
(20)
AN
C d = C d + ∆C d (21)
where elasticity matrix ∆C d is expressed as:
∆C d = T σ ∆C 0d T −1
ε (22)
M

C d is same as the elasticity degradation matrix in isotropic damage, while ∆C d is the

matrix characterizing the crack direction with in anisotropic damage. Substituting Eqs.
ED

(11), (12) and (17) to Eq. (22) and then considering Eq. (12), the matrix ∆C d is obtained.
 
sin4 α + 2p sin2 α cos2 α (1 − 2p) sin2 α cos2 α (p − 1) sin3 α cos α − p sin α cos3 α
ED 
∆C d = cos4 α + 2p sin2 α cos2 α (p − 1) sin α cos3 α − p sin3 α cos α 
1 − ν2
symmetry 0.5p + (1 − 2p) sin2 α cos2 α
PT

(23)
where parameter p is set as p = (1−ν)(1−β). The components C ij d (i, j = 1, 2, 3) of elasticity
matrix C d can be obtained by substituting Eqs. (23), (20) and (16) to Eq. (21).
CE

 11 E
 4 2 2

 C d = 1−ν 2 1 + D(−1 + sin α + 2p sin α cos α)





 C 12 E 2 2
d = 1−ν 2  ν + D[−ν + (1 − 2p) sin α cos α] 

 22 E 2
 C d = 1−ν 2 1 + D(−1+ cos4 α + 2p sin α cos2α)
AC

C 13 −ED 2 2
d = 1−ν 2 sin α cos αp cos α + (1 − p) sin α (24)

 23 −ED 2 2

 C d = 1−ν 2 sin α cos α p sin α + (1 − p) cos α
 C 33 = E 0.5(1 − ν) + D[0.5β(ν − 1) + (1 − 2p) sin2 α cos2 α]



 d21 1−ν122
C d = C d , C 31 13 32
d = Cd , Cd = Cd
23

The calculation of equivalent stress, equivalent displacement and damage variable are
the same in the case of the previous section, which can be obtained from Eqs. (2),(3) and
(5). In the elements with compression damage the isotropic damage model is used, so the
damaged elasticity matrix in global coordinates C d is set as (1 − D)C.

9
 ACCEPTED MANUSCRIPT

3. Algorithm for finite element modeling and implementation

3.1. Algorithm
In previous sections the detailed characterization concepts and expressions for ACDM
were described to predict the growth and propagation of discrete crack with continuum
model. The algorithm for implementing it within FEM framework is presented in this section.
The algorithm for ACDM approach is shown as Fig. 2.
ND is the number of damaged elements. Total time is set as 1, because it does not
influence the result under quasi static loading. Time step array time(i) is set by user and

T
the simplest form is that time(i) = i/NT . NT is the total number of time steps.

IP
The fundamental difference between the traditional CDM and our ACDM is that in our
theory the angle of discrete crack direction α (or θ) is calculated and employed in all damaged
elements with considering the material property degradation and this parameter is reflected

CR
in the damaged elasticity property matrix. The angle of discrete crack direction is a scalar
parameter in 2D damage problem and a vector with three components in 3D problem. In
generally the sign change of axial stresses differences, for example the sign change of axial

US
stress difference σx − σy in 2D problem, cause the sudden change of FPS direction when the
shear stresses are small, so produce spurious changes of crack propagation direction and this
is obviously erroneous.
In this algorithm the maximum crack inclined angle criteria was used to correct spurious
AN
changes of crack propagation direction . This criteria was also used to correct wrong crack
propagation by Cervera [29] with selecting the damaged elements whose centroids lie inside
a neighbourhood of the crack tip element and varying the crack direction vector as crack
M

average direction vector in these elements when there is the sudden change of curvature in
the crack track, where crack tip neighbourhood radius rneigh defined by user was proposed.
In this work the following criteria is used.
ED

|α − α| ≤ αmax (25)

where α is the crack direction angle parameter for the current newly damaged element, α
PT

is the average value of crack direction angle parameters for elements at which the following
equation ∆D = Di − Di−1 > 0 is satisfied, and αmax is the maximum inclined angle defined
by user at the beginning of the calculations. While equation (25) is satisfied, the crack
CE

direction parameter is set as the crack direction angle parameter α for the current element.
If equation (25) is not satisfied, the crack direction is deflecting sharply and a correction
is carried out by using α instead of α. Variable ∆D is the difference between the damage
variable for current time step and one for previous time step. If this variable ∆D is equal
AC

to zero and damage variable Di−1 is not equal to zero at an element, there is no difference
of damage variables for the element with time increment and its means that damage was
initiated at the previous time step but damage evolution has stopped at the current time
step. The equation ∆D > 0 is satisfied at the elements that damage evolution has been
performed in current time step.
In applying CDM theory to simulate fracture process for structures it can be seen that
damage begins simultaneously at some potential crack elements which are placed in front
of the crack owing to high stress at crack tip. Once damage variable reaches up to 1 for
any element of them during damage evolution, this element loses the load-capacity with

10
 ACCEPTED MANUSCRIPT

deleting itself and then the crack tip is moved. At the moment, damage evolution stops
at the elements placed in front of the previous crack tip. These potential crack elements at
which damage variable is greater than zero but damage evolution had stopped with satisfying
the following equation δe < δe∗ are choose and their mechanical properties are restored to its
initial state. This operation conform to the actual physical phenomenon and can reduce
computation cost.
The effective stress σe and effective displacement δe are calculated from stress-strain re-
sults using Eq. (2) and (3), and then damage variable D at all damaged elements and crack
angle α at a newly damaged elements are calculated based on Eq.(5) and Eq. (14) respec-

T
tively in every time step. It is noted that comparison between damage variable distribution

IP
for the previous time step and one for the current time step is performed and the bigger
value is choose in every damaged element, for damage variable can not be decrease accord-

CR
ing to the second principle of thermodynamics that requires the mechanical dissipation to
be non-negative. The damaged elasticity matrix is upgraded with the new damage variable
and crack angle and the loop is repeated.

3.2. Implementation in ANSYS software

US
The present ACDM model and algorithm in the previous sections can be implemented
in several commercial FE analysis software. In this section, brief summarizations are per-
AN
formed about some techniques that are incorporated into ANSYS by ANSYS Parameter
Design Language(APDL). The initial crack was formed by killing element technique or by
reflecting in geometry model. Considering large deformation, the nonlinear analysis is per-
formed in the simulation, while 2-D 4-Node Plane182 Structural Solid element is employed
M

in plane problem. Anisotropic elastic(ANEL) modulus matrix is used to set the material
property. According to ACDM algorithm, crack direction parameter and damage variable
are calculated by using element table and updated with the time step by ANSYS APDL.
ED

Upgrading of damage variable is only performed for the elements at which the equation
δe > δe∗ is satisfied in every time step and compared with the previous time step, while crack
direction parameters are calculated for the elements at which D is equal to zero and δe is
greater than δe∗ in every time step before damage variable calculation. Damaged stiffness
PT

matrices are calculated at all damaged elements, saved to element tables and updated with
the time step.
CE

4. Results and discussion

To check out the validity of our ACDM method based on continuum medium model,
AC

several numerical simulations are performed and results are compared with experiments and
some discontinuum damage models such as XFEM. Crack propagation problem under 3-point
bending was considered in three cases.

4.1. The straight crack propagation with mode I failure

A three-point bending test is considered, whereby a simply supported beam is loaded by
an imposed displacement at the centre of the top edge [30]. The geometry and the loading
conditions are shown in Fig. 3.

11
 ACCEPTED MANUSCRIPT

The material properties are as follows: E = 100N/mm2 , ν = 0, ft = 0.5N/mm2 , Gf =

0.01N/mm. The imposed displacement is as follows: u0 = 0.18mm.
In this problem, a crack had been propagating along the direction of initial crack and the
crack direction was not changed during the fracture process such as mode I failure problem.
The simulation results by ACDM theory are compared with the ones when using CZM and
XFEM.
By CZM, both initiation and propagation of crack can be exactly predicted in the problem
which has the pre-known crack propagation path, for the crack can be propagated along with
the interface cohesive elements. The deformed shape, contour plots of cohesive separation

T
and displacement distribution was shown in Fig. 4 when using CZM. The separation has

IP
the highest value at the element of the bottom center edge and the regular development of
crack is archived as expected.

CR
XFEM can capture the displacement field of discontinuity in front of the crack tip and
in this approach the discrete crack can be propagated within the element. The deformed
model and displacement distribution by XFEM in ANSYS are shown in Fig. 5 and it can be
notified that it is very similar with the ones shown in Fig. 4. The maximum displacement

CZM is 0.2090mm. US
predicted by XFEM is 0.2108mm at the centre of the bottom edge, while one predicted by

The simulation results for the discontinuum crack propagation process by ACDM theory
AN
are shown in Fig. 6.
Deformed shape was shown in Fig.6a with assuming αmax = 45o and it can be seen that
the high deformation in x direction appeared at the damaged elements. The parameter αmax
plays a key role in the crack direction correction through the maximum crack inclined angle
M

criteria. In general, it can be chosen a value below 60o in order to avoid sharp changes of
the crack direction. Fig. 6b shows the contour plot of damage variable. The red parts are
the fully damaged elements with the damage variable of nearly one in the crack direction,
ED

because of full crack development. As can be seen in Fig. 6, the damaged elements seem
to be deformed like discontinuum medium and the contour plot of crack propagation has a
small difference with the ones in Fig. 4 by CZM model. The displacement distribution was
PT

shown in Fig. 6c and it can be seen that the displacement at an initial cracked element were
nearly the same in three cases. The numerical results with respect to element size from 0.2
to 0.01 were compared (as shown in Fig. 7). When the element size is 0.02mm, the relative
difference between 0.4350N of maximum bearing capacity predicted by ACDM approach
CE

and 0.4334N predicted by CZM approach is 0.37%, while one between 0.4350N predicted by
ACDM approach and 0.4369N predicted by XFEM approach is 0.44%. It can be seen that
the values of maximum bearing capacity predicted by ACDM approach are quite similar to
AC

the ones by CZM and XFEM approach, even if the ACDM model slightly underestimates
the strength predicted by XFEM and slightly overestimates the strength predicted by CZM.
The numerical results by XFEM increase with respect to fine mesh and converge to the
desired value, while the ones by CZM also increase with respect to fine mesh and converge to
almost the same value by XFEM. From Fig. 7 we can easily find that the numerical results
by ACDM also converge to the same value and the results are coincidental in both cases
with element size 0.02 and 0.01 for all numerical methods.
The load-displacement curves by several numerical methods such as CZM, XFEM and
ACDM, were compared and it can be seen that the load-displacement curve with using

12
 ACCEPTED MANUSCRIPT

ACDM is in good agreement with the ones using CZM and XFEM (as shown in Fig. 8). All
the computations have been carried out by a standard PC equipped with a quad-core Intel
CPU at 3.2 GHz and 16 GB Ram- processor. The computing times for CZM, XFEM and
ACDM approach are equal to 324s, 402s and 254s respectively. ACDM approach requires
the lowest computational cost among three methods, while XFEM approach requires the
most expensive computational cost among three methods.
It is clear that ACDM can predict the crack propagation process and load bearing capa-
bility for damaged structures with high accuracy.

T
4.2. The curved crack propagation with mixed mode failure

IP
Firstly the curved crack propagation process under mixed mode load conditions is aimed
at checking the validation of ACDM finite element model.

CR
The crack is initiated at the bottom edge of beam with 0.8mm offset, not at the centre
of bottom edge (as shown in Fig. 9). The material properties and geometries are the same
as in the previous case, with only difference in the position of initial crack. The crack is
propagated in a curved path from the initial crack position towards the centre of upper edge.

US
The deformed model from the FE model using XFEM in ANSYS is shown in Fig. 10a
and it shows that the crack propagation process is as same as the expected one.
The displacement distribution by XFEM in ANSYS are shown in Fig. 10b and it can be
AN
seen that XFEM can numerically simulate the curved crack propagation process with high
computational cost.
The deformed shape and crack propagation path simulated by isotropic damage model
(Eq. (1)) are shown in Fig. 11 and it is clear that this model can not be predicted from the
M

crack propagation path any more.

Fig. 12a showed deformed shape for αmax = 30o and from this shape we can confirm that
the curved crack path was also properly predicted by ACDM. Fig. 12b showed the contour
ED

plot of the damage variable plotted by FE model using ACDM and Fig. 12c showed the
displacement distribution by FE model using ACDM . The areas with red color are the fully
damaged elements and the line with white color is the predicted crack propagation path from
PT

ACDM model. The crack propagation path can be plotted considering the crack direction
angles at the fully damaged elements. The crack propagation path using ACDM FE model
has a small difference from ones using XFEM model as can be seen from Fig. 10 and Fig.
CE

12, and the trend is coincidental with the same geometry.

Fig. 13 shows comparison between the load-displacement curves predicted by ACDM and
ones by XFEM. It is clear from Fig. 13 that both the force-displacement curves predicted by
ACDM and the ones by XFEM are very similar in shape except for the peak load by XFEM
AC

model which is about 5.62% higher than that obtained from ACDM model. It can be seen
that the force decreases with crack propagation and is approached to zero in this problem.
In the following, the experiment results of curved crack path under mixed mode bend-
ing[11] are compared with simulation ones by ACDM finite element model to check the
validation of our method.
The geometry of the problem and the loading conditions applied to the concrete beam are
shown in Fig. 14 [11]. The material properties are assumed as follows: E = 38N/mm2 , ν =
0.2, ft = 3N/mm2 , GI = 0.069N/mm. A refined mesh using an element length of 2.0mm
was used in the simulation of a specimen and contained a total of 59975 elements, which

13
 ACCEPTED MANUSCRIPT

changed a little by calculation results with further mesh refining (as shown in Fig. 15). It is
noted that the ratio between the element size and the dimension of the structure is kept in
these two examples.
Deformed shape, contour plot of the damage variable and crack propagation path by FE
model using ACDM for αmax = 36o were shown in Fig. 16. The crack propagation process
was predicted well by ACDM under mixed mode bending.
All simulations were performed with setting shear relation parameter β as 1. Research
[17] conducted, shows that shear stiffness G degrades with increasing matrix cracking and
considers the effect of parameter β in anisotropic damage of the transverse crack. The

T
effect on the interaction of damage in the shear direction on the crack propagation curve is

IP
assessed with varying parameter β. We performed comparison of the crack propagation path
predicted numerically by ACDM model with different values of parameter β which is ranged

CR
from 0 to 1 with the experiment results (as shown in Fig. 17). From Fig. 17 it can be seen
that the crack paths by ACDM model with setting β as 0.8 and 1.0 and experiment ones
by [11] are similar, whilst the crack path predicted by ACDM model change with varying of
parameter β.

US
Fig. 18 showed comparison between the load-CMOD curves predicted by ACDM and
experiment results. It is clear from Fig. 18 that the load-CMOD curve predicted by ACDM
are quite similar to the experiment ones. In this problem the force increases with CMOD,
AN
once it gets to the value of the maximum force, it decreases to become near to zero. It can
be seen that the numerical model underestimates the strength less than experiments in the
last part of the simulation.
M

5. Conclusions
In this paper the theory and implementation for ACDM approach are presented, which
ED

can correctly predict the crack propagation and fracture behavior in continuum model. The
anisotropic damage concept is introduced for the discrete crack problem in isotropic materi-
als and damage mechanics model which can predict the discrete crack propagation direction
and path was formulated on the local coordinates placed on the crack surface. The MS
PT

criterion is used to predict the damage initiation and crack propagation direction, and the
ACDM with adding the crack direction parameter used to predict the damage evolution
process. The algorithm for the application of our ACDM theory in the numerical simulation
CE

is described and the FE model is established using ANSYS 17.1 commercial software. A
three-point bending test is considered, whereby a simply supported beam is loaded by an
imposed displacement at the top edge. The straight crack propagation process of the beam
AC

was simulated under three-point bending load by ACDM model and the results by ACDM
model were compared with the ones by CZM and XFEM model, while the curve crack prop-
agation process for beam was simulated by ACDM model and the results were compared
with the experiment results. The influence of shear relation parameter β is considered and
good results are obtained.

Acknowledgment

14
 ACCEPTED MANUSCRIPT

This work was supported by the National Natural Science Foundation of China [grant
numbers 11272096, 11472086, 11532013].

References
[1] Rybicki E F, Kanninen M F. A finite element calculation of stress intensity factors by a
modified crack closure integral[J]. Engineering Fracture Mechanics, 1977, 9(4):931-938.

T
[2] Krueger R. Virtual crack closure technique: History, approach, and applications[J].
Applied Mechanics Reviews, 2004, 57(2):109-143.

IP
[3] Carvalho N V D, Chen B Y, Pinho S T, et al. Modeling delamination migration in

CR
cross-ply tape laminates[J]. Composites Part A Applied Science and Manufacturing,
2015, 71:192-203.

[4] Zhao L, Zhi J, Zhang J, et al. XFEM simulation of delamination in composite lami-

US
nates[J]. Composites Part A Applied Science and Manufacturing, 2016, 80:61-71.

[5] Jin Z H, Sun C T. Cohesive zone modeling of interface fracture in elastic bi-materials[J].
Engineering Fracture Mechanics, 2005, 72(12):1805-1817.
AN
[6] Harper P W, Hallett S R. Cohesive zone length in numerical simulations of composite
delamination[J]. Engineering Fracture Mechanics, 2008, 75(16):4774-4792.
M

[7] Chen X, Deng X, Sutton M A, et al. An Inverse Analysis of Cohesive Zone Model
Parameter Values for Ductile Crack Growth Simulations[J]. International Journal of
Mechanical Sciences, 2014, 79(1):206-215.
ED

[8] Dugdale D S. Yielding of steel sheets containing slits[J]. Journal of the Mechanics and
Physics of Solids, 1960, 8(2):100-104.
PT

[9] Barenblatt G I. The Mathematical Theory of Equilibrium Cracks in Brittle Fracture[J].

Advances in Applied Mechanics, 1962, 7:55-129.

[10] Turon A, Dávila C G, Camanho P P, et al. An engineering solution for mesh size effects
CE

in the simulation of delamination using cohesive zone models[J]. Engineering Fracture

Mechanics, 2007, 74(10):1665-1682.
AC

[11] Gálvez J C, Elices M, Guinea G V, et al. Mixed Mode Fracture of Concrete under
Proportional and Nonproportional Loading[J]. International Journal of Fracture, 1998,
94(3):267-284.

[12] Dolbow J, Belytschko T. A finite element method for crack growth without remeshing[J].
International Journal for Numerical Methods in Engineering, 1999, 46(1):131-150.

[13] Motamedi D, Mohammadi S. Fracture analysis of composites by time independent

moving-crack orthotropic XFEM[J]. International Journal of Mechanical Sciences, 2012,
54(1):2037.

15
 ACCEPTED MANUSCRIPT

[14] Sosa J L C, Karapurath N. Delamination modelling of GLARE using the extended finite
element method[J]. Composites Science and Technology, 2012, 72(7):788-791.

[15] 3.2. XFEM-based crack analysis and crack-growth simulation, Fractrue Analysis Guide,
Mechanical APDL, ANSYS 17.1 User’s Manual, ANSYS Inc., 2016.

[16] Li X, Chen J. The implementation of the extended cohesive damage model for multicrack
evolution in laminated composites[J]. Composite Structures, 2016, 139:6876.

[17] Schipperen J H A. An anisotropic damage model for the description of transverse matrix

T
cracking in a graphiteepoxy laminate[J]. Composite Structures, 2001, 53(3):295-299.

IP
[18] Wang L, Zheng C, Luo H, Wei S, Wei Z. Continuum damage modelling and progressive
failure analysis of carbon fiber/epoxy composite pressure vessel[J]. Heavy Castings and

CR
Forgings, 2015, 134:475-482.

[19] Shen W, Peng L H, Tang C Y. An anisotropic damage-based plastic yield criterion

and its application to analysis of metal forming process[J]. International Journal of

US
Mechanical Sciences, 2005, 47(12):1897-1922.

[20] Kuna-Ciska H, Skrzypek J J. CDM based modelling of damage and fracture mechanisms
AN
in concrete under tension and compression [J]. Engineering Fracture Mechanics, 2004,
71(46):681-698.

[21] Maimı́ P, Camanho P P, Mayugo J A, et al. A continuum damage model for composite
M

laminates: Part I Constitutive model[J]. Mechanics of Materials, 2007, 39(10):897-908.

[22] Kachanov LM. On the time to failure under creep condition,Izv Akad Nauk USSR, Otd
ED

Techn Nauk 1958, 8: 26-31 [In Russian].

[23] Rabotnov Y N. On the equation of state of creep[J]. 1963, 178(31):117-122.

[24] Salavatian M, Smith L V. An investigation of matrix damage in composite laminates

PT

using continuum damage mechanics[J]. Composite Structures, 2015, 131(17):565-573.

[25] Erdogan F, Sih G C. On the Crack Extension in Plates Under Plane Loading and
CE

Transverse Shear[J]. Journal of Basic Engineering, 1963, 85(4):527.

[26] Lemaitre J. How to use damage mechanics [J]. Nuclear Engineering and Design, 1984,
80(2):233-245.
AC

[27] Lemaitre J. A continuous damage mechanics model for ductile fracture[J]. Journal of
Engineering Materials and Technology, 1985, 107(107):83-89.

[28] Pineda E J, Waas A M. Numerical implementation of a multiple-ISV

thermodynamically-based work potential theory for modeling progressive damage
and failure in fiber-reinforced laminates[J]. International Journal of Fracture, 2013,
182(1):93-122.

16
 ACCEPTED MANUSCRIPT

[29] Cervera M, Pelà L, Clemente R, et al. A crack-tracking technique for localized damage
in quasi-brittle materials[J]. Engineering Fracture Mechanics, 2010, 77(13):2431-2450.

[30] Mergheim J, Kuhl E, Steinmann P. A finite element method for the computational mod-
elling of cohesive cracks[J]. International Journal for Numerical Methods in Engineering,
2005, 63(2):276-289.

[31] 4.24. Material damage, Material Reference, Mechanical APDL, ANSYS 17.1 User’s Man-
ual, ANSYS Inc., 2016.

T
[32] Xin S H, Wen H M. A progressive damage model for fiber reinforced plastic compos-

IP
ites subjected to impact loading[J]. International Journal of Impact Engineering, 2015,
75:40-52.

CR
[33] Chow C L, Wang J. An anisotropic theory of elasticity for continuum damage mechan-
ics[J]. International Journal of Fracture, 1987, 33(1):3-16.

[34] Hägglund R, Isaksson P. Analysis of localized failure in low-basis-weight paper[J]. In-

US
ternational Journal of Solids and Structures, 2006, 43(1819):5581-5592.

[35] Murakami S, Kamiya K. Constitutive and damage evolution equations of elastic-brittle

AN
materials based on irreversible thermodynamics[J]. International Journal of Mechanical
Sciences, 1997, 39(4):473-486.

[36] Ismail J, Zaı̈ri F, Naı̈t-Abdelaziz M, et al. How cracks affect the contact characteris-
M

tics during impact of solid particles on glass surfaces: A computational study using
anisotropic continuum damage mechanics[J]. International Journal of Impact Engineer-
ing, 2012, s 4041(2):10-15.
ED

[37] Sun X, Khaleel M A. Modeling of Glass Fracture Damage Using Continuum Damage
Mechanics - Static Spherical Indentation[J]. International Journal of Damage Mechanics,
2004, 13(3):263-285.
PT

[38] Özden U A, Bezold A, Broeckmann C. Numerical Simulation of Fatigue Crack Prop-

agation in WC/Co based on a Continuum Damage Mechanics Approach [J]. Procedia
CE

Materials Science, 2014, 3:1518-1523.

[39] Maimı́ P, Camanho P P, Dávila C G. A Thermodynamically Consistent Damage Model

for Advanced Composites[J]. Recercat Principal, 2006.
AC

[40] Han F, Lubineau G, Yan A. Adaptive coupling between damage mechanics and peri-
dynamics: A route for objective simulation of material degradation up to complete
failure[J]. Journal of the Mechanics and Physics of Solids, 2016, 94:453-472.

[41] Han F, Lubineau G, Yan A, et al. A morphing approach to couple state-based peridy-
namics with classical continuum mechanics[J]. Computer Methods in Applied Mechanics
and Engineering, 2016, 301:336-358.

17
 ACCEPTED MANUSCRIPT

[42] Yan A, Han F, Lubineau G. The morphing method as a flexible tool for adap-
tive local/non-local simulation of static fracture[J]. Computational Mechanics, 2014,
54(3):711-722.

[43] Lubineau G, Yan A, Han F, et al. A morphing strategy to couple non-local to local con-
tinuum mechanics[J]. Journal of the Mechanics and Physics of Solids, 2012, 60(6):1088-
1102.

[44] Lu M, Zhang H, Zheng Y, et al. A multiscale finite element method with embedded

T
strong discontinuity model for the simulation of cohesive cracks in solids[J]. Computer
Methods in Applied Mechanics and Engineering, 2016, 311:576-598.

IP
CR
Figure Captions

Fig.1. Local coordinates in an element with crack.

Fig.2. The algorithm for ACDM approach.

US
Fig.3. The three-point bending problem with initial crack.
Fig.4. Simulation results for crack propagation by CZM: (a) deformed shape; (b) contour
plot of cohesive separation; (c) displacement distribution
AN
Fig.5. Simulation results for crack propagation by XFEM: (a) deformed shape; (b) displace-
ment distribution
Fig.6. Simulation results for crack propagation by ACDM: : (a) deformed shape; (b) con-
tour plot of damage variable; (c) displacement distribution
M

Fig.7. Numerical results with element size by ACDM,CZM and XFEM.

Fig.8. Comparison with load-displacement curves by ACDM, CZM and XFEM.
Fig.9. Three-point bending problem with the curved crack propagation path.
ED

Fig.10. Simulation results for curved crack propagation problem by XFEM: (a) deformed
shape; (b) displacement distribution.
Fig.11. Numerical results by isotropic damage model: (a) deformed shape; (b) contour plot
PT

of damage variable.
Fig.12. Numerical results for curved crack propagation problem by ACDM: (a) deformed
shape; (b) contour plot of damage variable; (c) displacement distribution.
CE

Fig.13. Comparison between results by ACDM and XFEM.

Fig.14. Mixed bending test for concrete specimen.
Fig.15. Finite element mesh.
AC

Fig.16. Numerical results by ACDM: (a) deformed shape; (b) contour plot of damage vari-
able; (c) crack path by removing the killed elements.
Fig.17. Crack propagation paths according to shear relationship parameter β.
Fig.18. Comparison with load-CMOD curves by ACDM and experiment result[11].

18
 ACCEPTED MANUSCRIPT

T
IP
CR
US
AN
M
ED
PT
CE

Fig.1. Local coordinates in an element with crack.

AC

19
 ACCEPTED MANUSCRIPT

T
IP
CR
US
AN
M
ED
PT
CE
AC

Fig.2. The algorithm for ACDM approach.

20
 ACCEPTED MANUSCRIPT

T
IP
Fig.3. The three-point bending problem with initial crack.

CR
US
AN
M
ED
PT
CE
AC

Fig.4. Simulation results for crack propagation by CZM: (a) deformed shape(×5); (b) contour
plot of cohesive separation(×5); (c) displacement distribution(×5).

21
 ACCEPTED MANUSCRIPT

T
IP
CR
US
AN
M
ED

Fig.5. Simulation results for crack propagation by XFEM: (a) deformed shape(×5); (b)
PT

displacement distribution(×5).
CE
AC

22
 ACCEPTED MANUSCRIPT

T
IP
CR
US
AN
M
ED

Fig.6. Simulation results for crack propagation by ACDM: (a) deformed shape(×5); (b) contour
plot of damage variable(×5); (c) displacement distribution(×5).
PT
CE
AC

Fig.7. Numerical results with element size by ACDM,CZM and XFEM.

23
 ACCEPTED MANUSCRIPT

T
IP
CR
Fig.8. Comparison with load-displacement curves by ACDM, CZM and XFEM.

US
AN
M

Fig.9. Three-point bending problem with the curved crack propagation path.
ED
PT
CE
AC

Fig.10. Simulation results for curved crack propagation problem by XFEM: (a) deformed
shape(×5); (b) displacement distribution(×5).

24
 ACCEPTED MANUSCRIPT

T
IP
CR
Fig.11. Numerical results by isotropic damage model: (a) deformed shape(×5); (b) contour plot

US
of damage variable(×5).
AN
M
ED
PT
CE
AC

Fig.12. Numerical results for curved crack propagation problem by ACDM: (a) deformed
shape(×5); (b) contour plot of damage variable(×5); (c) displacement distribution(×5).

25
 ACCEPTED MANUSCRIPT

T
IP
CR
Fig.13. Comparison between results by ACDM and XFEM.

US
AN
M
ED

Fig.14. Mixed bending test for concrete specimen.

PT
CE
AC

Fig.15. Finite element mesh.

26
 ACCEPTED MANUSCRIPT

T
IP
CR
US
AN

Fig.16. Numerical results by ACDM: (a) deformed shape(×100); (b) contour plot of damage
M

variable(×1); (c) crack path by removing the killed elements(×1).

ED
PT
CE
AC

Fig.17. Crack propagation paths according to shear relationship parameter β.

27
 ACCEPTED MANUSCRIPT

T
IP
CR
US
AN
M
ED
PT

Fig.18. Comparison with load-CMOD curves by ACDM and experiment result[11].

CE
AC

28