Accepted Manuscript

An extended cohesive damage model for simulating arbitrary damage

propagation in engineering materials

X. Li, J. Chen

PII: S0045-7825(16)31649-8
DOI: http://dx.doi.org/10.1016/j.cma.2016.11.029
Reference: CMA 11239

To appear in: Comput. Methods Appl. Mech. Engrg.

Received date: 1 October 2015

Revised date: 14 November 2016
Accepted date: 18 November 2016

Please cite this article as: X. Li, J. Chen, An extended cohesive damage model for simulating
arbitrary damage propagation in engineering materials, Comput. Methods Appl. Mech. Engrg.
(2016), http://dx.doi.org/10.1016/j.cma.2016.11.029

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.
*Manuscript
Click here to download Manuscript: CMAME-v17.docx Click here to view linked References

An extended cohesive damage model for simulating arbitrary

1
2 damage propagation in engineering materials
3
4
5 X Li and J Chen*
6
7 School of Civil Engineering and Surveying, University of Portsmouth
8 Portland Street, Portsmouth PO1 3AH, UK
9
10
11
12 Abstract
13
14 This paper introduces the extended cohesive damage model (ECDM) for simulating
15
16 arbitrary damage propagation in engineering materials. By embedding the
17
18
micromechanical cohesive damage model (CDM) into the eXtended Finite Element
19
20
21 Method (XFEM) and eliminating the enriched degree of freedoms (DoFs), the ECDM
22
23 defines the cohesive crack path at a low scale in the condensed equilibrium equations
24
25 and enables the local enrichments of approximation spaces without enriched DoFs. In
26
27 this developed ECDM, a new equivalent damage scalar as a function of strain field is
28
29 introduced to avoid the appearance of enriched DoFs, and to substitute the
30
31 conventional characterization in the approximation of displacement jump. The
32
33
34 embedment of CDM is no longer required by the ECDM, which allows discontinuities
35
36 to exist within a finite element rather than the element boundaries. This feature
37
38 enables the ECDM to simulate the reality of arbitrary cracks. Initial applications of
39
40 the ECDM in simulation of arbitrary cracks shows that the developed ECDM works
41
42 very well when compared to experiment work and XFEM analysis.
43
44
45
46 Keywords: Extended cohesive damage model; XFEM; CDM; Fracture; Arbitrary
47
48 crack propagation
49
50
51
52
53
54
55
56
57
58
59
60
61
Corresponding author: J Chen, Tel: 00 44 2392842427, Email: jiye.chen@port.ac.uk
62
63
64
65
 1 1. Introduction
2
3
4 In structural analysis, typical examples of strong or weak discontinuities are cracks or
5
6
7 interfacial behaviour between different domains or materials. Modelling arbitrary
8
9
discontinuities and their propagation is currently one of the major concerns in the
10
11
12 fracture related academic community. One of the earliest techniques used to model
13
14
15 such discontinuities is Adaptive Mesh Refining (AMR) scheme which was developed
16
17
18 between 1987 and 2000 [1, 2]. In AMR, a finite element mesh can be adaptively
19
20
21 modified according to crack propagation, and errors during numerical simulation can
22
23
24 be efficiently minimized. However, the treatment for weak discontinuities is still very
25
26
27 challenging for AMR. In addition, the modification of mesh topology in AMR is
28
29
30 tedious and brings a considerable computational burden [3]. To address all these
31
32
33
deficiencies, Babuska and Melenk [4, 5] proposed the Partition of Unity Method
34
35 (PUM), for which no mesh regeneration is needed in simulating crack growth. The
36
37
38 finite element boundary would no longer need to be the discontinuity surface, which
39
40
41 is a significant benefit to the work of modelling fracture.
42
43
44 In the past decades, based on PUM, a rapid development has been made in
45
46
47 conducting arbitrary cracking problems within continuum solids [6-10]. Among all
48
49
50 these PUM-based approaches, eXtended Finite Element Method (XFEM) was
51
52
53 originally introduced by Belytschko et al. [6] and subsequently enhanced by Moës et
54
55
56
al. [11]. XFEM is a combination of the classical Finite Element Method (FEM) and
57
58 PUM. By enriching the classical piecewise polynomial approximation basis within the
59
60
61
2
62
63
64
65
 1 FEM framework, XFEM can thoroughly capture non-smooth features independently.
2
3
4 Such features would include jumps, kinks, singularities and inhomogeneity [6, 11].
5
6
7 Unlike AMR, XFEM no longer needs remeshing or adapting finite element mesh
8
9
when discontinuities or crack propagations occur. The mesh adaption process can
10
11
12 therefore be substituted by introducing enriched or additional DoFs and partitioning
13
14
15 the domain with some triangular and quadrilateral sub-elements. The Gauss points of
16
17
18 these sub-elements would be those that were used for sub-integration of the domain of
19
20
21 elements [11, 12].
22
23
24 The appropriate treatment for the approximation space in XFEM realizes the
25
26
27 accurate characterization for non-smooth features without dependence on the mesh
28
29
30 topology. XFEM has been applied in describing discontinuity in solids for more than
31
32
33
10 years. Moës and Belytschko [13] and Mergheim et al. [14] combined the
34
35 traction-separation law, known as cohesive zone model (CZM), with XFEM to model
36
37
38 the cohesive cracks within quasi-brittle materials. Benvenuti [15] provided a unified
39
40
41 way with cohesive processes involving either a localized or finite width zone into a
42
43
44 zero-thickness interface. Unger et al. [16] carried out a discrete crack simulation for
45
46
47 concrete materials employing XFEM in conjunction with an adaptive crack growth
48
49
50 algorithm. By introducing continuum damage mechanics in the framework of XFEM,
51
52
53 characterization of failure behaviours ranging from diffuse material degradation to
54
55
56
discrete cracks can be achieved [17-19]. Hansbo et al. [20] presented a phantom-node
57
58
59
60
61
3
62
63
64
65
 1 method to model arbitrary discontinuities, which is essentially another
2
3
4 implementation of XFEM [21].
5
6
7 Although the numerical approaches mentioned above have similar or different
8
9
advantages, they in most cases also increase the computational burden and the
10
11
12 possibility of failed convergence. This is because they require additional external
13
14
15 nodes or degrees of freedom (DoFs) which would be a barrier when applying them in
16
17
18 structural level modelling. Moreover, tracking the path of crack propagation would be
19
20
21 costly for XFEM based algorithms. Attempts to incorporate such theories in large
22
23
24 scale structural calculations are confronted with the challenges of computational
25
26
27 efficiency, complexity and robustness. This is especially the case for heterogeneous
28
29
30 materials such as those used in composite structures. Oliver et al. [22] investigated an
31
32
33
embedded finite element method (E-FEM) which is an alternative approach to XFEM.
34
35 A comparative study [23] between E-FEM and XFEM demonstrated that both
36
37
38 numerical accuracy and efficiency are evidently better in E-FEM. This is achieved
39
40
41 through the implementation of elemental enrichments rather than the nodal
42
43
44 enrichments required by XFEM. Even if the mesh size is relatively coarse, E-FEM is
45
46
47 still able to predict reasonably accurate results. Yang et al. [24-26] have recently
48
49
50 carried out an interesting augmented finite element method (A-FEM) that can account
51
52
53 for path-arbitrary, multiple intra-elemental discontinuities. Their study demonstrated
54
55
56
an improvement in numerical efficiency by two orders of magnitude when compared
57
58 to XFEM [26]. Their work used a fully condensed elemental equilibrium equation
59
60
61
4
62
63
64
65
 1 with standard DoFs to simulate the discontinuities. However, as A-FEM uses four
2
3
4 internal nodes to account for the crack displacements in approximating both weak and
5
6
7 strong intra-elemental discontinuities, specified algorithms for solving the crack
8
9
displacements of internal nodes are required.
10
11
12 Besides XFEM, E-FEM and A-FEM, CZM has been extensively used as an
13
14
15 interface cohesive element in the study of localisation and fracture in engineering
16
17
18 materials and structures for more than 15 years. The application of CZM requires that
19
20
21 FEA meshes conform to the potential cracks [28]. It must be used in such a way that
22
23
24 corresponding interface cohesive elements are embedded along the crack path known
25
26
27 a priori to analyse progressive failure.
28
29
30 In the Extended Cohesive Damage Model (ECDM) presented in this paper,
31
32
33
CDM is employed to determine the micro damage initiation. The damage propagation
34
35 is calculated by a new equivalent damage scalar relating to a strain field. This
36
37
38 developed ECDM has the following specific features: enriched DoFs vanished from
39
40
41 the finally condensed equilibrium equations; corresponding cohesive properties taken
42
43
44 into account through an equivalent stiffness matrix; and a new equivalent damage
45
46
47 scalar as a function of strain field introduced to substitute for conventional
48
49
50 characterization in the approximation of the displacement jump. The ECDM allows
51
52
53 for discontinuities to exist within a finite element rather than just at the element
54
55
56
boundaries. In this way the ECDM is able to simulate discontinues much more
57
58 realistically. Unlike standard XFEM, the ECDM does not include enriched DoFs in
59
60
61
5
62
63
64
65
 1 the basic equilibrium equation. Instead a similar effect to that from enriched DoFs is
2
3
4 achieved through an equivalent stiffness matrix in the finally condensed equilibrium
5
6
7 equations of the ECDM. The derived formulation of the ECDM is presented with
8
9
standard DoFs only. However, it is influenced by an equivalent damage scalar with
10
11
12 strong and weak discontinuous characteristics within an element.
13
14
15 This paper is organized as follows: Section 1 gives relevant background
16
17
18 information in the area of computational damage mechanics. Section 2 gives detailed
19
20
21 derivation of the condensed equilibrium equations of the proposed ECDM. Section 3
22
23
24 presents examples of applications of the ECDM and the comparisons between
25
26
27 modelling simulations and experimental work. Finally, Section 4 gives a conclusion
28
29
30 to the ECDM with major highlights or achievements, and the future work.
31
32
33
2. Basic formulations of the ECDM
34
35 In section 2.1, the theoretical work starts from assuming a displacement filed with
36
37
38 enriched degree freedoms and a shifted Heaviside function to approximate a
39
40
41 discontinuity in a solid field. In section 2.2, a discrete form of equilibrium equation
42
43
44 with the standard and enriched degree freedoms is introduced first. Then the enriched
45
46
47 degree freedoms are eliminated mathematically from the condensed equilibrium
48
49
50 equations by accounting for the effects from enriched DoFs and corresponding
51
52
53 cohesive force into the equivalent stiffness matrix. A transformation matrix is also
54
55
56
introduced to simplify the calculation of cohesive force in section 2.2. Section 2.3
57
58 introduces an equivalent damage scalar as a function of strain field to avoid using a
59
60
61
6
62
63
64
65
 1 displacement gap in approximating discontinuities. At the end of section 2.4, a
2
3
4 final formulation of fully condensed equilibrium equation is presented.
5
6
7 2.1 Displacement filed and shifted Heaviside function
8
9
The theoretical derivation starts from the framework of XFEM without taking the
10
11
12 crack-tip singularity into account, which means a crack must cross an element. There
13
14
15 will be no such crack-tip singularity problems once a cohesive damage law is used
16
17
18 anywhere in the mesh in which any element can hold the crack-tip but the crack-tip
19
20
21 stress is restrained by the cohesion. A previous investigation given by Yang et al [26]
22
23
24 has reported that ignoring crack-tip singularity will not influence the result for crack
25
26
27 propagation as well as the overall response. Chen [29] has also recently investigated
28
29
30 the basic concept of combining XFEM with CDM without using a specified enriched
31
32
33
item to cope with the singularity problem at crack-tip. The XFEM test and trial
34
35 function can thus be given as:
36

uh (x)   Ni  x ui   N j  x  d (x)ai

37
38 (1)
39 iI iJ
40
41 where Ni is the standard FEM shape functions associated with node i. ui and ai are the
42
43
44 nodal variables associated with standard degree freedoms and enriched degree
45
46
47 freedoms at node i and node j, respectively. The Heaviside step function   shown in
d

48
49
50 Eq. (2) can characterize the physical jump when the element is completely separated
51
52
53 (strong discontinuities).
54
55 1, x  
56   d ( x)   (2)
57 0, x  
58
59
60
61
7
62
63
64
65
 1 where, Ω+ is the one side domain of discontinuity, while the Ω— is the other side
2
3
4 domain.
5
6
7 For the approximation using Heaviside function, the Kronecker-δ property (i.e.,
8
9
NI(xJ) = δIJ) cannot be satisfied, which requires the imposition of essential boundary
10
11
12 conditions. Moreover, the total DoFs ui of an enriched node i are the summation of the
13
14
15 standard DoFs ui and the enriched contribution   (x)ai , rather than the physical
d
16
17
18 solution of the nodal displacement. Thus the interpretation of results is difficult [30].
19
20
21 A serious similar problem was raised by the phantom node method [31, 32].
22
23
24 A “shifted” enrichment is accepted in this investigation for retaining the
25
26
27 Kronecker-δ property as the enriched degree of freedom or enrichment is zero at any
28
29
30 node, which is illustrated in Fig. 1. This shifted displacement basis can also been
31
32
33
found in other researchers’ work related to XFEM [30, 33]. The introduction of
34
35 shifted Heaviside function does not alter the approximating basis while simplifies the
36
37
38 implementation attributing to that resulting enrichment vanished in elements which
39
40
41 are not cut by discontinuities. The displacement field with the “shifted” enrichment is
42
43
44 shown in Eq. (3)
45
46
47
48 iI iJ
 
uh (x)   Ni  x ui   N j  x  d (x)   d (xi ) ai (3)
49
50
Where xi is the positional coordinate for the ith node. The concept of modelling this
51
52
53 “shifted” crack is firstly illustrated in one dimension as shown in Fig. 1. Supposing
54
55
56 the left side of the interface is Ω-, and the right side is Ω+, the Heaviside function at
57
58
59 given nodes 1 and 2 are defined as :   ( x1 )  1 ,   ( x2 )  0 . In Fig. 1a, supposing the
d d
60
61
8
62
63
64
65
 1 left side of the interface is Ω-, and the right side is Ω+, the “shifted” crack modelling
2
3
4 concept in one dimension can be illuminated as shown in Fig. 1b and 1c. The graphic
5
of the standard DoF term is N1 xu1  N2 xu2 which is presented by the dash line in
6
7
8
9
Fig. 1d. The enriched term is expressed by Eq. (4).
10
11
12
13 a b
14
15
16
17
18
19 d
20 c
21
22
23
24
25
26
27 Fig. 1. The representation of the discontinuity by the shifted enrichment
28
29
30
31

   
32
33 u  a1 N1  x   d ( x)   d ( x1 )  a2 N 2  x   d ( x)   d ( x2 ) (4)
34
35
36 Actually, the combination with two enriched items in Eq. (4) results a real
37
38
39 displacement with a jump u as shown in the last illustration given in Fig. 1d. This
40
41
42 shifted enrichment ensures that the value of displacement basis at any nodes is
43
44
45 single-valued for any crack geometries. It should be noticed that the enriched
46
47
48 contribution vanishes at enriched nodes, but not at integration points.
49
50
51
The proposed approach is based on the displacement approximation of XFEM
52 n
53 for a 2D cracked homogeneous domain   i , and its unit normal n, as shown in
54 i1
55
56 Fig. 2. The Dirichlet boundary conditions u were applied at u. Considering the
57
58
59
60
61
9
62
63
64
65
 1 total potential energy governing the problem, the weak form of equilibrium equation
2
3
4 can be obtained as given below:
5

    d    f   ud    f t   ud 
6 b
7   d
(5)
8
9
10
11
12
13
14
15
16
17
18
19
20
21 Fig. 2. A 2D domain with an arbitrary discontinuity d
22
23
24 Where, , , u, fb and ft are the stress tensor, strain tensor, displacement vector, body
25
26
27 forces and traction forces, respectively. d is the traction boundary with a normal m,
28
29
30 which splits the domain  into + and -.
31
32
33
2.2 Discrete form of the equilibrium equations
34
35 Using the weak form of equilibrium equation from Bubnov-Galerkin method, the
36
37
38 discrete form of equilibrium equation for static analysis can be written as shown in
39
40
41 Eq. (6).
42
43
44 K uu K ua  u  fext
u

45  au aa   
 a  (6)
46 K K   a  fext 
47
48 Where, Kuu, Kaa and Kua are the stiffness matrix associated with the standard FE
49
50
51 approximation, the enriched approximation and the coupling between the standard FE
52
53
u a
54 approximation and the enriched approximation, respectively. f ext and f ext are the
55
56 u a
57 equivalent nodal force vectors, f ext is for standard FEM freedoms while f ext is for
58
59
60 enriched freedoms. u denotes the vector including ordinary degrees of freedom while
61
10
62
63
64
65
 1 the enriched degrees of freedom is signified by a. Each detailed sub-matrix of
2
3
4 stiffness in Eq. (7) is given below.
5
6 K uu   h BT DBd 
7 

K ua   h BT DB a d 
8
9 
10
11
12
K au    B  DBd 
 h
 T (7)

   B  DB d 
13 a T
14 K aa a
 h
15
16
17 where D is the constitutive matrix, B is the deformation matrix for strain calculation.
18
19
20
In the 2D domain,
21
22 B  B1 B2 ... Bi ... Bm  (8)
23
24
25  N ix 0 
26  
27 Where, Bi   0 N iy  ,
28  N iy N ix 
29



Ba   d  x, y    d ( xi , yi ) B 
30
31 (9)
32
33
34 For solving a 2D problem, if the crack shown in Fig. 2 is a cohesive crack, the
35
36 crack
37 discontinuous boundary is a cohesive crack boundary, i.e.    coh . In Eq. (5),
d
38
39
40 the equivalent nodal force vectors without body force can be expressed as:
41
42
43
u
fext   h NT td 

44 (10)
45
a
fext   h NT ( crack ( x, y )   crack ( xi , yi ))td   fcoh
 coh coh
46
47
48
49 t is the external nodal force vector as shown in Fig. 2. Because of the existence of
50
51
52 cohesive segment, the internal nodal force vector due to cohesive traction t shown in
53
54 crack
55 Fig. 2 on the crack surface  coh can be expressed as:
56
57
58
59
60
61
11
62
63
64
65
 1 fcoh  
crack   d 
( x, y )   d  xi , yi  NT td   
coh
crack -    x, y     x , y   N
d d i i
T
(t )d 
2 coh
(11)
3  NT td 
coh
crack
4
5
6
7 where N is standard shape function applied at the enriched degree freedoms. CDM is
8
9
10 then introduced here to characterize the nonlinear cohesive segment evolution. In
11
12
13 CDM, the normal traction component tn and shearing traction component ts are used
14
15
16 to account for the damage state of cohesive segment. The cohesive tractions as a
17
18
19 function of the damage scalar d in both normal and tangential directions at crack
20
21
22
surface decrease monotonically from their initial values t0 to zero, which is
23
24 mathematically expressed by Eq. (12).
25
26
27 t n  1  d  t0n 
t   coh
s 
 t 
(12)
   1  d  t0 
28
29 t coh
30
31
32 where t0 is the cohesive traction on the cracked surface when crack propagates, which
33
34
is composed of the calculated normal and tangential stresses. It is known the tcoh is
35
36
37 given in the directions n and s at crack surface, so the transformation of coordinates
38
39
40 from local coordinate n and s to x and y (global coordinate) is necessarily required and
41
42
43 expressed by Eq. (13).
44
45
tcoh
x
 -sin cos  tcoh
n

 y 
46 global
t  s 
 (13)
tcoh   cos sin  tcoh 
47 coh
48
49
50
51 The angle  is the angle between the coordinate n-s and the coordinate x-y. There is
52
53
54 not a physical relative displacement  before a crack formed within element.
55
56
57 When the damage increases, the cohesive traction decreases, following a linear
58
59
softening damage law as shown in Fig. 3.
60
61
12
62
63
64
65
 1
2
3
4 t0 Released fracture energy
5

Cohesive traction
6
7
8   ,t 
9
10
11
12 f
13 
14 Relative displacement (crack opening)
15
16
17 Fig. 3. A micromechanical linear softening damage law
18
19
20
Substituting the expression of equivalent nodal force vector in Eq. (10) into Eq. (6)
21
22 results Eq. (14) as below.
23
24
25
26  K uu K ua  u   fuext   0 
 au     
( x, y )  H crack ( xi , yi )) td   fcoh 
27 (14)
K aa   a   crack N ( H crack
T
28 K  coh coh coh 
29
30
31 With the purpose of reaching a fully condensed equilibrium system, the additional
32
33
34 enrichment term a, is eliminated, thus the equilibrium equation with the standard
35
36
37 FEM unknown quantities can be consequently obtained as shown in Eq. (15).
38

K  K ua  K aa  K au u  fext 
-K ua  K aa  f ext  K ua  K aa  f coh
39 1 1 1
uu u a
40 (15)
41
42
43 In the equation above, the nodal force on additional DoFs attributing to the
44
45
46
a
contribution of the external load is denoted as f ext . Suppose the external nodal force
47
48
49 arrays in Eq. (15) can be expressed by Eq. (16),
50
T
51
fextu   f1u    fi u f ju   f ku   f mu  (16)
52 1m
53
54 which is a m by 1 vector, and
55
56 T
57 a
fext   fi a f ja    f ka  (1 ≤ i < j < k ≤ n) (17)
58 1n
59
60
61
13
62
63
64
65
 1 which contains n items. m and n are the numbers for standard DoFs and additional
2
3
4 DoFs, respectively. Then a transformation matrix is required to be introduced in order
5
6 a u
7 to replace f ext by f ext in Eq. (15). This transformation matrix is constructed as:
8
9  m1,1  m1,i m1, j  m1,k  m1,m 
10 m  m2,i m2, j  m2,k  m2,m 
11
M
2,1
(18)
12          
13  
14  mn ,1  mn,i mn, j  mn ,k  mn ,m 
n m
15
16
17 It should be noticed that there is only one non-zero item at each row of M. According
18
19 a
20
to the array of fext expressed in Eq. (17), the non-zero items are defined as
21
22 m1,i  fi a / fi u , m2, j  f ja / f ju and mn,k  f ka / f ku . With this matrix, the required
23
24
25 transformation can be expressed by Eq. (19).
26
27
28 a
f ext  Mf ext
u
(19)
29
30
31 Substituting Eq. (19) into Eq. (15) results
32
33
34
a
f ext  Mf ext
u
(20)
35
36
37 Substituting Eq. (20) into Eq. (15) results
38
39
40 K uu
 K ua  K aa  K au u  fext
1 u
-K uu  K aa  Mf ext
u
 K ua  K aa  f coh
1 1
(21)
41
42
43 Then, the equivalent equilibrium equations with standard FEM degree freedoms can
44
45
be expressed as shown in Eq. (22).
46
47
I  K  K 
1
K   K ua  K aa  K au u
48 ua aa 1 uu 1
M
49 (22)
 
50 1
 I  K ua  K aa  M K ua  K aa  f coh
1 1
51  fext
u

52
53
54 In Eq. (22), the calculation of the equivalent nodal force due to the existence of
55
56
57 internal cohesive segment for a cracked element is given by Eq. (23).
58
59
60
61
14
62
63
64
65
  
1
fintcoh  I  K ua  K aa  M K ua  K aa  fcoh
1 1 1
2 (23)
3
4
5 2.3 Equivalent damage scalar
6
7
8 Normally, the released fracture energy as shown in Fig. 3 should be used to describe
9
10
11 the crack propagation. Therefore, the damage scalar d in Eq. (12) for the cohesive
12
13
14 behavior along the crack length can be expressed as:
15
16
1
17
 t 0  d 
18
19
d  2 crack (24)
Gc lcrack
20
21
22 In Eq. (24), Gc is fracture energy; lcrack is defined as a crack length in a failed element,
23
24
25 within which, the damage status (the status of cohesion) is assumed to be the same
26
27
28 along the crack length. In the implementation, the crack distance is calculated from
29
30
31 the start point to the end point of a crack in the failed element. This means the damage
32
33
34 status throughout one element is consistent, and crack propagates along a straight line
35
36
37 within the failed element. This can be seen from failed elements 1-4 in Fig. 4. The
38
39
40 failed elements 1-4 have crack lengths defined by the lengths of the lines AB, BC, CD
41
42
43 and DE, respectively. As shown in Eq. (24), the defined crack length lcrack is one of
44
45
46
variables to calculate the damage scalar d; the displacement jump  relating to
47
48 additional DoFs is also a variable in calculating the damage scalar d. The
49
50
51 displacement jump  is required to obtain the state of the cohesive segment by
52
53
54 calculating the damage scalar d, and determine if d=1 then the cohesive traction
55
56
57 vanishes, and the cohesive segment behaves as a strong discontinuity (element
58
59
60 separates completely).
61
15
62
63
64
65
 1
2
3
4 E
5
6
7 2 C D 4
3
8 B
9
10 1 Crack point
11
12
13 A
14 Node
15
16
17
18 Fig. 4. The illustration of the crack lengths calculated by the ECDM
19
20
21 In this proposed ECDM, a new equivalent damage scalar is used to avoid the
22
23
24 appearance of the additional DoFs related displacement gap . In the ECDM based
25
26
27 FEM modelling, the degradation of cohesive zone will present a micro behavior of
28
29
30 strain softening, so the strain energy released due to the fracture should be equal to the
31
32
33
released work done by cohesive traction. Therefore, in the utilization of the ECDM,
34
35 the damage scalar can be expressed equivalently by a released strain energy which is
36
37
38 shown in Eq. (25).
39
40
41 1 0 1
42 
 2
  d     0 d 
 2
43 d (25)
44 Gc lcrack
45
46
47 where  0 and  0 are the material strength and the initial damage strain, respectively,
48
49
50 when the damage onsets. Herein, a softening constitutive law is used for reducing the
51
cohesive traction, i.e.   1  d   . Bringing this calculation into Eq. (25), an explicit
52 0
53
54
55
56
expression of the equivalent damage scalar can be achieved as shown below.
57
58
59
60
61
16
62
63
64
65
 1 0 1
1
 
  d     0 0 d 
2
2 d 2 (26)
1
Gclcrack    0 0 d 
3
4 2
5
6
7 It should be noticed that the damage scalar d in Eq. (26) is a function of the strain 
8
9
only, which has no need to calculate the additional DoFs related displacement gap .
10
11
12 In order to simplify the problem, it is supposed that there is no distributed external
13
14
15 load applied on the cracked element, thus transformation matrix M would be a zero
16
17
18 matrix. Then using the Eqs. (11), (12), (22), (23) and (26), the nodal force given in
19
20
21 Eq. (23) with the existence of cohesive traction at the crack can be expressed as:
22

fintcoh  K ua  K aa 
23 1
24  NTenr t 0 1  d  d 
crack
coh
25
26  1 
 Gc lcrack    0 d   (27)
 K ua  K 
27 aa 1  2
28 crack NTenr t 0 
1 0 0  d
29
coh
 Gc lcrack     d  
30   2 
31
32
33
2.4 Final formulation of the fully condensed equilibrium equations
34
35 Herein, an operator is introduced for Eq. (27) to obtain the implicit expression for
36
37
38 achieving the ultimate equilibrium equations. Considering the symmetric property of
39
40
41 stiffness matrix, an operator is chosen as below:
42

 
43
r Kr 
T
1 1
 K ua  NTenr t 0 d  (28)
aa

K 
44
K  
T cohcrack
aa 1
45 ua
r r N t d
T
enr 0 u
46 crack
coh

47
48 which satisfies u = 1. Using this operator, Eq. (27) can be rewritten as Eq. (29).
49
50
51
52
53
54
55
56
57
58
59
60
61
17
62
63
64
65
  1 0 
 Gc lcrack   2   d  
1
f  K  K   crack N t 
aa 1
2
 d u
coh ua T
3 int enr 0
coh 1
4  Gc lcrack    0 0 d  
5  2 
 K ua  K aa   crack NTenr t 0
6 1 Gc lcrack
7 d u (29)
coh 1
8 Gc lcrack    0 0 d 
2
9
10 1 0
11    d
 K  K   crack N enr t 0
1 2
12
ua aa T
d u
coh 1
13 Gc lcrack    0 0 d 
14 2
15
16
17 Considering a convenient expression, the following two symbols are defined.
18
19 Gc lcrack
20 L1   NTenr t 0 d 
crack 1 0 0
Gc lcrack     d 
coh
21
2
22
23 1 0 (30)
24 2
t  d
25 L 2   crack N enr t 0
T
d 
coh 1
26 Gc lcrack    0 0 d 
27 2
28
29
30 Using Eq. (30), the nodal force vector can be simplified as shown in Eq. (31).
31
fintcoh  K ua  K aa 
1
32
33  L1  L2  u (31)
34
35 Substituting Eq. (31) into Eq. (21) for calculating the cohesive nodal force, the finally
36
37
38 condensed equilibrium equations can be expressed by Eq. (32).
39
40

 K uu  K ua  K aa  K au  K ua  K aa  u  f
41 1 1
42  L1  L 2  u
ext (32)
43
44
45 Now the basic equilibrium expression of the ECDM without the enriched or
46
47
48 additional DoFs has been obtained, in which the nonlinearity must be accounted
49
50
51 because of CDM embedded. The equivalent stiffness matrix and the right hand side
52
53
54 (i.e. the nodal force vector) are constructed at elemental level and transferred into
55
56
57 ABAQUS to assemble the system equilibrium. The nonlinear incremental iteration is
58
59
performed using a displacement control procedure. The Newton Raphson algorithm in
60
61
18
62
63
64
65
 1 conjunction with line search scheme [34] provided by ABAQUS is employed as the
2
3
4 technique to carry out nonlinear iteration. As far as authors know, this combined
5
6
7 technique is widely used in iteration procedure for solving strong nonlinearity.
8
9
Numerical iteration will end when the residual nodal force reduces to the prescribed
10
11
12 tolerance. It should be noticed that the sub-matrix Kaa in Eq. (32) is symmetric but not
13
14
15 necessarily invertible, hence the generalized inversed matrix (Moore–Penrose
16
17
18 pseudoinverse) [35] is applied to complete the calculation of the inversed matrix. The
19
20
21 ECDM formulation is a lower order equilibrium system compared to the standard
22
23
24 XFEM, which permits nodal displacement calculation of the cracked element using
25
26
27 the standard FEM DoFs only. This proposed rigorous mathematical procedure can
28
29
30 fully cover the damage evolution from a weak discontinuity to a strong discontinuity.
31
32
33
This developed novel ECDM has been implemented using the user subroutine UEL in
34
35 the commercial finite element package ABAQUS.
36
37
38 3. Numerical examples
39
40
41 In this section, four numerical examples are modelled by the ECDM to validate
42
43
44 the performance of the developed ECDM. In these examples, homogeneous
45
46
47 quasi-brittle materials are used and a cohesive strength based criteria is used to
48
49
50 characterize the damage initiation. Based on this criteria, damage starts when the
51
52
53 maximum principal stress at Gauss points of any element is beyond the cohesive
54
55
56
strength of any individual fracture mode. The perpendicular direction to the maximum
57
58 principal stress is adopted to be the crack direction within elements. A strain field
59
60
61
19
62
63
64
65
 1 related damage scalar is used to account for damage accumulation until it satisfies the
2
3
4 crack criteria determined by the fracture energy. Overlapping elements with zero
5
6
7 stiffness are placed in the FEM mesh to visualize the discretization of the ABAQUS
8
9
user elements with the ECDM. Quasi-static crack propagation is mainly considered in
10
11
12 this investigation. The accuracy, mesh-independence, robustness and capability have
13
14
15 been presented and discussed through numerical examples in simulating arbitrary
16
17
18 crack propagation using the proposed ECDM.
19
20
21 3.1 A concrete beam under three-point bending
22
23
24 One of the most popular specimens in concrete material testing, three-point
25
26
27 bending, is modelled as a preliminary verification of the proposed ECDM. The
28
29
30 geometry and boundary conditions of beam modelled can be seen from Fig. 5. This
31
32
33
simply supported beam is made of Ultra High Performance Fibre Reinforced Concrete
34
35 (UHPERC) and subjected to displacement-controlled loading [36]. The following
36
37
38 material properties for UHPERC are used in modelling simulation [36]: Young’s
39
40
41 modulus E = 47000.0MPa, Poisson’s ratio v = 0.2, tensile strength ft = 8.0MPa and
42
43
44 fracture energy release rate Gf = 30000.0N/m. In order to reduce computational work,
45
46
47 only the potential fracture zone marked by dash lines in Fig. 5 is discretized with the
48
49
50 proposed ECDM. For the purpose of investigating the mesh-independence of the
51
52
53 proposed ECDM, four sets of different mesh-sized models for modelling the potential
54
55
56
fracture zone with the area 18.5 × 100mm2, named as M1, M2, M3 and M4, are
57
58 investigated. The average element side lengths for modelling the potential fracture
59
60
61
20
62
63
64
65
 1 zone are 1.65, 1.82, 2.07 and 2.45mm, and the total numbers of elements in this zone
2
3
4 are 682, 558, 434 and 310 for M1, M2, M3 and M4, respectively. Other regions are
5
6
7 modelled by elements whose average size is 8mm. All these models are constructed
8
9
by the four-node plane strain user elements with the ECDM. It should be noted that
10
11
12 the control parameters of line search scheme are set to be 4.0, 4.0, 0.25, 0.25 and 0.15,
13
14
15 respectively for nonlinear fracture analysis in all models in this section.
16
17
18
19
20
18.5
21
22
23
24
25
26
27
28 Fig. 5. A concrete beam with a notch under three-point bending
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50 Fig. 6. Load-displacement curves given by the ECDM and experimental work
51
52
53 It can be seen from Fig. 6 that the ECDM predicted initial stiffness of the
54
55
56 notched beam agrees with test work well. The predicted failure load is slightly higher
57
58
59 than the tested one. This is because the data for fracture property and material strength
60
61
21
62
63
64
65
 1 was collected from a limited number of test specimens. In the crack propagation or
2
3
4 post failure stage, the ECMD prediction almost agrees with experimental work.
5
6
7 Because of the relatively high fracture toughness of the UHPERC, after the failure
8
9
load is reached, the response remains at a higher level in the post-failure response
10
11
12 rather than dropping down immediately. The predicted load-displacement curves are
13
14
15 plotted in terms of four models with different mesh densities shown in Fig. 6. It can
16
17
18 be seen from Fig. 6 that the load-displacement curves obtained from the meshes M1,
19
20
21 M2, M3 and M4 are almost same, only the result from mesh M3, containing 434
22
23
24 elements, is slightly higher than others.
25
26
27 Fig. 7 shows the deformation of the beam model with the finest mesh together
28
29
30 with a cracked region in the middle section presented by a maximum principal strain
31
32
33
field under applied displacement of 2.25mm. The deformation due to the macro crack
34
35 opening can be observed in the area with large deformation. The crack propagation
36
37
38 basically follows the grey band in the maximum principal strain contour shown in the
39
40
41 contours. It can be concluded that the predicted failure response is independent from
42
43
44 the mesh densities, and coincident with the experimental measurement. Therefore, the
45
46
47 ECDM based modelling provides a reliable prediction of fracture behaviour of the
48
49
50 UHPERC sample.
51
52
53
54
55
56
57
58
59
60
61
22
62
63
64
65
 Fig. 7. Maximum principal strain contour field in a concrete beam under three-point bending
1
2
(Mesh M4) simulated by the ECDM at the final state of analysis.
3
4
5 3.2 A notched concrete beam under asymmetric bending
6
7
8 A comparative study is carried out on a benchmark test using both the ECDM and the
9
10
11 standard XFEM in FEM commercial package ABAQUS (shorten as XFEM below) to
12
13
14 assess the performance of the ECDM. In the test work [37], a single notched beam as
15
16
17 shown in Fig. 8, made of plain concrete, was subjected to four-point bending. A
18
19
20
mixed failure mode with a curved cracking path was observed experimentally. The
21
22 dimensions of the beam is 440 × 100 × 100mm3; a notch with the size 5 × 20 ×
23
24
25 100mm3 is located at the top centre. The loading and boundary conditions can be seen
26
27
28 from Fig. 8. A rigid bar between two loading points is set up for the purpose of
29
30
31 imposing a proportionally distributed load P at the bottom of beam. Through the rigid
32
33
34 bar, the proportional loads applied on the left and right loading point are P/11 and
35
36
37 10P/11, respectively. Following the previous work given by references [38] and [39],
38
39
40 the material properties used in modelling are: Young's modulus E = 35000.0MPa,
41
42
43 Poisson's ratio  = 0.2, tensile strength ft = 3.0MPa and fracture energy Gf = 0.1
44
45
N/mm.
46
47
48
49
50
51
52 125
53
54
55
56
57
58
59
60
61
23
62
63
64
65
 1 Fig. 8. Configuration of the Single-notched concrete beam under asymmetric bending load.
2
3
4 Four meshes, named as M1, M2, M3 and M4 with different element sizes are used in
5
6
7 both the ECDM and XFEM for mesh sensitivity investigation. It should be noted that
8
9
only the potential fracture zone of 125 × 100mm2 marked by dash line in Fig. 8 is
10
11
12 discretised with different element sizes. The average element side lengths used in the
13
14
15 potential fracture zone are 8, 4, 2 and 1mm, respectively. The numerical simulations
16
17
18 are carried out without setting the discontinuity evolution path beforehand, so cracks
19
20
21 could propagate arbitrarily. Fig. 9 shows the deformation after failure predicted by the
22
23
24 ECDM using four different meshes. The cracked elements are presented in terms of
25
26
27 maximum principle strain (buckled zone are shown in grey), from which a curved
28
29
30 feature of the crack can be observed. The predicted crack profiles from four meshes
31
32
33
are similar. At the end of analysis, the exact crack paths predicted by the ECDM
34
35 together with the experimental envelop are shown in Fig. 10. It can be seen from Fig.
36
37
38 10 that the predicated crack paths from four different meshes are almost identical and
39
40
41 are certainly within the experimental envelop. This investigation proves the
42
43
44 mesh-independence of the ECDM in simulation of crack propagation. It should be
45
46
47 noted that each crack path given in Fig. 10 is calculated according to the location of
48
49
50 the crack tip in each failed element. In both the ECDM and XFEM modelling, the
51
52
53 crack propagations are determined using a principal stress failure criterion.
54
55
56
57
58
59
60
61
24
62
63
64
65
 1
2
3
4
5
6
7 M1
8
9
10
11
12
13
14
15
16 M2
17
18
19
20
21
22
23
24 M3
25
26
27
28
29
30
31
32 M4
33
34
35 Fig. 9. Maximum principal strain contour field in the single-notched concrete beam under
36 asymmetric bending load simulated with the ECDM at the final state of analysis.
37
38 40
Distance from notch bottom (mm)

39
40 20
41
42
0
43 M1
44 M2
-20 M3
45
M4
46
-40 Experimental
47
envelop
48
-60
49
50
51 -80

52
-40 -20 0 20 40 60
53
54 Distance from initial notch (mm)
55
56 Fig. 10. Predicted crack paths by the ECDM at the final state of analysis and experimental
57
58 crack envelop [37].
59
60
61
25
62
63
64
65
 1 Fig. 11 displays the sequences of predicted failure responses provided by the ECDM
2
3
4 and XFEM with progressively refined meshes. In these responses, the relative
5
6
7 difference of vertical displacements between two sides of the notch, i.e., crack mouth
8
9
sliding displacement (CMSD), is computed as the abscissa, while the reaction force
10
11
12 on the loading point is ordinate. The load - CMSD curves given by the ECDM and
13
14
15 XFEM are compared to the experimental measurements. It can be seen from Fig. 10
16
17
18 that the ECDM modelling with four different meshes give almost identical responses;
19
20
21 the predicted peak load agrees with experimental envelope very well. In the
22
23
24 post-failure softening stage, the predicted residual stiffness is slightly lower than the
25
26
27 experimental observation. This is because a residual load capacity was still recorded
28
29
30 in experimental work due to the compressive stresses around the rigid cap [37],
31
32
33
however, at the later post-failure softening stage, the ECDM prediction has same
34
35 tendency with experimental measurement. This benchmark modelling indicates that
36
37
38 the proposed ECDM is capable of predicting arbitrary crack propagation with
39
40
41 accuracy and robustness along with the change of mesh size. In contrast, the results of
42
43
44 peak load are obviously overestimated by XFEM compared to experimental
45
46
47 measurement, and is varied to some extent in different meshes. The post-failure
48
49
50 softening behaviour predicted by XFEM is also changed some extent in different
51
52
53 meshes compared to experimental observation. This implies that the outcome from
54
55
56
XFEM simulation relatively relates to the mesh size in this investigation. Moreover,
57
58
59
60
61
26
62
63
64
65
 1 XFEM simulation tends to overestimate the structural strength at the failure point and
2
3
4 the most softening stage in all four predicted load - CMSD curves.
5
6
7 60 60
Experimental envelop Experimental envelop
8 ECDM ECDM
50 50
9 XFEM XFEM
10 40 40
11 M1 M2
Load (KN)

Load (KN)
12 30 30

13
20 20
14
15 10 10
16
17 0 0
0.00 0.02 0.04 0.06 0.08 0.10 0.00 0.02 0.04 0.06 0.08 0.10
18 CMSD(mm) CSMD (mm)
19
20 60 60
Experimental envelop Experimental Envelop
21 ECDM ECDM
50 50
22 XFEM XFEM

23 40 40
24 M3 M4
Load (KN)

Load (KN)

25 30 30

26
20 20
27
28 10 10
29
30 0 0
0.00 0.02 0.04 0.06 0.08 0.10 0.00 0.02 0.04 0.06 0.08 0.10
31 CSMD (mm) CSMD (mm)
32
33
34
Fig. 11. Load - CMSD curves obtained from the ECDM, XFEM and experimental work [37].
35
36
37
38
39 3.3 A steel plate specimen with two holes and two notches under tension
40
41
42 This example is investigated to evaluate the capability of the ECDM in
43
44
45 simulating multiple crack propagations. The geometry, boundary conditions and
46
47
48 material properties of a rectangular steel plate with two holes and two initial cracks
49
50
51 are taken from previous work [33] as shown in Fig.12a. The thickness of the plate is 1
52
53
54 mm. The Young’s modulus and Poisson’s ratio are, respectively, E = 210000MPa and
55
56
57 v=0.3. Material tensile strength ft = 235.0MPa. Plane stress material model is used in
58
59
the investigation. Displacement is prescribed on the upper side of the specimen shown
60
61
27
62
63
64
65
 1 in Fig. 12a. It should be noted that this analysis focuses on multiple crack
2
3
4 propagations, only the crack paths are presented in this paper, and the cohesive effect
5
6
7 is ignored in simulation for quick tracking of the crack paths.
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23 Fig.12. (a) A steel plate with two holes and two initial cracks under tension; (b) The ECDM
24 solution presented by principal strain contour at the final state of analysis.
25
26
27
28
29 In this modelling work, two cracks are tracked with the assistance of a common
30
31
32 data block COMXYC in coding, by which the crack tip information can be shared by
33
34
35 different elements and updated at each loading increment. Fig. 12b shows the
36
37
38 simulated two-crack propagations using a mesh with 17414 elements. It is expected
39
40
41 that the two simulated crack paths are symmetric and look similar. The simulated
42
43
44 crack paths by the ECDM reflect the reality of the actual cracking pattern in this steel
45
46
plate. This example basically confirms the capability of the ECDM in tracking the
47
48
49 multiple crack propagations, although there is no experimental work reported in
50
51
52 literatures for comparison.
53
54
55
56
57
4. Conclusions and future work
58
59
60
61
28
62
63
64
65
 1 The finite element framework and detailed rigorous derivation in forming the
2
3
4 condensed equilibrium equilibriums of the ECDM have been presented in this paper.
5
6
7 This is a new model for simulating arbitrary crack propagation within a continuum
8
9
solid. Based on the fundamental concept of XFEM, the ECDM introduces a
10
11
12 displacement field with a shifted Heaviside function and enriched DoFs to
13
14
15 approximate discontinuities at the beginning. In the final condensed equilibrium
16
17
18 equation of the ECMD, the additionally enriched DoFs are eliminated at the element
19
20
21 level. This enables ECDM to improve the overall convergence in numerical
22
23
24 calculation, and to be capable of carrying out nonlinear fracture analysis with large
25
26
27 computing work at structural level. Considering the length of this paper, detailed
28
29
30 investigation of the computing efficiency of the ECDM will be discussed in different
31
32
33
papers.
34
35 The ECDM embeds the micro-mechanical cohesive damage model CDM into
36
37
38 the macro-mechanical finite element formula. Unlike the classic CDM which requires
39
40
41 the displacement gap to qualify the state of cohesive segment, the ECDM employs a
42
43
44 new equivalent damage scalar as a function of strain field. This feature allows it to
45
46
47 avoid the requirement of explicit displacement gap and to make it possible to embed
48
49
50 the CDM into the ECDM. With this new model, the crack nucleation as well as crack
51
52
53 propagation can be characterized without any re-meshing effort. In the numerical
54
55
56
implementation via user subroutine UEL in ABAQUS, the maximum principal stress
57
58 based criteria is employed for the judgement of damage initiation and the
59
60
61
29
62
63
64
65
 1 determination of the crack direction. The introduced equivalent damage scalar and
2
3
4 fracture energy criteria are used for the judgement of crack propagation. More details
5
6
7 for determining the crack direction as well as the integration scheme of stiffness
8
9
matrix of the ECDM based user element will be given in different papers.
10
11
12 Through all investigated examples in this paper, it has been validated that the
13
14
15 developed ECDM has the capability of capturing multiple crack propagations in
16
17
18 solids with sufficient accuracy, efficiency and stability. The mesh-independence of
19
20
21 the ECDM has been verified by a number of models with different meshes which
22
23
24 produce identical solutions. This developed ECDM supplies a promising numerical
25
26
27 approach for simulating arbitrary damage propagation in engineering materials. In the
28
29
30 future, the developed ECDM will be applied in simulating multiple crack
31
32
33
propagations in heterogeneous materials such as fibre composites. This will include
34
35 multiple layered delamination, multiple matrix crack and fibre breakage to fully
36
37
38 validate the capability of the ECDM in simulating the reality of arbitrary damage
39
40
41 propagation.
42
43
44
45
References
46
47
[1] Zienkiewicz, O. C., & Zhu, J. Z., A simple error estimator and adaptive procedure for practical
48
49 engineering analysis. International Journal for Numerical Methods in Engineering 1987; 24(2),
50 337-357.
51
52 [2] Murthy, K. S. R. K., & Mukhopadhyay, M., Adaptive finite element analysis of mixed‐mode
53 crack problems with automatic mesh generator. International Journal for Numerical Methods in
54
55 Engineering 2000; 49(8), 1087-1100.
56 [3] Khoei, A. R., Azadi, H., & Moslemi, H., Modeling of crack propagation via an automatic adaptive
57
58 mesh refinement based on modified superconvergent patch recovery technique. Engineering
59 Fracture Mechanics 2008; 75(10), 2921-2945.
60
61
30
62
63
64
65
 [4] Melenk, J. M., & Babuška, I., The partition of unity finite element method: basic theory and
1
applications. Computer methods in applied mechanics and engineering 1996; 139(1), 289-314.
2
3 [5] Babuška, I., & Melenk, J. M, The partition of unity method, International Journal for Numerical
4 Methods in Engineering 1997; 40, 727–758.
5
6 [6] Belytschko, T. & Black, T., Elastic crack growth in finite elements with minimal remeshing.
7 International Journal for Numerical Methods in Engineering 1999; 45: 601-620.
8
9 [7] Duarte, C. A., Hamzeh, O. N., Liszka, T. J., & Tworzydlo, W. W., A generalized finite element
10 method for the simulation of three-dimensional dynamic crack propagation. Computer Methods in
11
12 Applied Mechanics and Engineering 2001; 190(15), 2227-2262.
13 [8] Oden, J. T., Duarte, C. A. M., & Zienkiewicz, O. C., A new cloud-based hp finite element method.
14 Computer methods in applied mechanics and engineering 1998; 153(1), 117-126.
15
16 [9] Glowinski, R., He, J., Rappaz, J., & Wagner, J., Approximation of multi-scale elliptic problems
17 using patches of finite elements. Comptes Rendus Mathematique 2003; 337(10), 679-684.
18
19 [10] Strouboulis, T., Babuška, I., & Hidajat, R., The generalized finite element method for Helmholtz
20 equation: theory, computation, and open problems. Computer Methods in Applied Mechanics and
21
22 Engineering 2006; 195(37), 4711-4731.
23 [11] Moёs, N, Dolbow, J, & Belytschko, T., A finite element method for crack growth without
24
remeshing. International Journal for Numerical Methods in Engineering 1999; 46(1), 131-150.
25
26 [12] Belytschko, T., Gracie, R., & Ventura, G. A review of extended/generalized finite element
27 methods for material modeling. Modelling and Simulation in Materials Science and Engineering
28
29 2009; 17(4), 043001.
30 [13] Moës, N., & Belytschko, T., Extended finite element method for cohesive crack growth.
31
32 Engineering fracture mechanics 2002; 69(7), 813-833.
33 [14] Mergheim, J., Kuhl, E., & Steinmann, P., A finite element method for the computational
34
35 modelling of cohesive cracks. International Journal for Numerical Methods in Engineering 2005;
36 63(2), 276-289.
37
[15] Benvenuti, E. A regularized XFEM framework for embedded cohesive interfaces. Computer
38
39 Methods in Applied Mechanics and Engineering 2008; 197(49), 4367-4378.
40 [16] Unger, J. F., Eckardt, S., & Könke, C. (2007). Modelling of cohesive crack growth in concrete
41
42 structures with the extended finite element method. Computer Methods in Applied Mechanics and
43 Engineering 2007; 196(41), 4087-4100.
44
45 [17] Comi, C., Mariani, S., & Perego, U. An extended FE strategy for transition from continuum
46 damage to mode I cohesive crack propagation. International Journal for Numerical and
47
Analytical Methods in Geomechanics 2007; 31(2), 213-238.
48
49 [18] Simone, A., Wells, G. N., & Sluys, L. J. From continuous to discontinuous failure in a
50 gradient-enhanced continuum damage model. Computer Methods in Applied Mechanics and
51
52 Engineering 2003; 192(41), 4581-4607.
53 [19] Wang, Y., & Waisman, H. From diffuse damage to sharp cohesive cracks: A coupled XFEM
54
55 framework for failure analysis of quasi-brittle materials. Computer Methods in Applied Mechanics
56 and Engineering 2016; 299, 57-89.
57
58
59
60
61
31
62
63
64
65
 [20] Hansbo, A., & Hansbo, P., A finite element method for the simulation of strong and weak
1
discontinuities in solid mechanics. Computer methods in applied mechanics and engineering 2004;
2
3 193(33), 3523-3540.
4 [21] Areias, P., & Belytschko, T., A comment on the article “A finite element method for simulation of
5
6 strong and weak discontinuities in solid mechanics” by A. Hansbo and P. Hansbo [Comput.
7 Methods Appl. Mech. Engrg. 193 (2004) 3523–3540]. Computer Methods in Applied Mechanics
8
9 and Engineering 2006; 195(9), 1275-1276.
10 [22] Oliver, J., Huespe, A. E., & Samaniego, E. A study on finite elements for capturing strong
11
12 discontinuities. International Journal for Numerical Methods in Engineering 2003; 56(14),
13 2135-2161.
14 [23] Oliver, J., Huespe, A. E., & Sanchez, P. J. A comparative study on finite elements for capturing
15
16 strong discontinuities: E-FEM vs X-FEM. Computer methods in applied mechanics and
17 engineering 2006; 195(37), 4732-4752.
18
19 [24] Fang, X. J., Yang, Q. D., Cox, B. N., & Zhou, Z. Q., An augmented cohesive zone element for
20 arbitrary crack coalescence and bifurcation in heterogeneous materials. International Journal for
21
22 Numerical Methods in Engineering 2011; 88(9), 841-861.
23 [25] Liu, W., Yang, Q. D., Mohammadizadeh, S., Su, X. Y., & Ling, D. S., An accurate and efficient
24
augmented finite element method for arbitrary crack interactions. Journal of Applied Mechanics
25
26 2013; 80(4), 041033.
27 [26] Liu, W., Yang, Q. D., Mohammadizadeh, S., & Su, X. Y., An efficient augmented finite element
28
29 method for arbitrary cracking and crack interaction in solids. International Journal for Numerical
30 Methods in Engineering 2014; 99(6), 438-468.
31
32 [27] Park, K., & Paulino, G. H., Cohesive zone models: a critical review of traction-separation
33 relationships across fracture surfaces. Applied Mechanics Reviews 2011; 64(6), 060802.
34
35 [28] Elices, M., Guinea, G. V., Gomez, J., & Planas, J. The cohesive zone model: advantages,
36 limitations and challenges. Engineering fracture mechanics 2002; 69(2), 137-163.
37
[29] J. Chen, An extended cohesive damage model with a length scale in fracture analysis of
38
39 adhesively bonded joints, Engineering Fracture Mechanics 2014; 119, 202–213.
40 [30] Rabczuk, T., Zi, G., Gerstenberger, A., & Wall, W. A., A new crack tip element for the phantom‐
41
42 node method with arbitrary cohesive cracks. International Journal for Numerical Methods in
43 Engineering 2008; 75(5), 577-599.
44
45 [31] Song, J. H., Areias, P., & Belytschko, T., A method for dynamic crack and shear band propagation
46 with phantom nodes. International Journal for Numerical Methods in Engineering 2006; 67(6),
47
868-893.
48
49 [32] Giner, E., Sukumar, N., Tarancon, J. E., & Fuenmayor, F. J., An Abaqus implementation of the
50 extended finite element method. Engineering fracture mechanics 2009; 76(3), 347-368.
51
52 [33] Fries, T. P., & Baydoun, M., Crack propagation with the extended finite element method and a
53 hybrid explicit–implicit crack description. International Journal for numerical methods in
54
55 engineering 2012; 89(12), 1527-1558.
56 [34] Bertsekas DP. Nonlinear Programming (2nd edn). Athena Scientific: Belmont, MA, 2003.
57
58 [35] Wang, G., Wei, Y., Qiao, S., Lin, P., & Chen, Y., Generalized inverses: theory and computations
59 2004; 37-75, Beijing Science Press.
60
61
32
62
63
64
65
 [36] K. Awinda, J. Chen, S. Barnett and D. Fox, Modelling behaviour of ultra high performance fibre
1
reinforced concrete, Advances in Applied Ceramics, 113, 8, 502–508, 2014
2
3 [37] Nooru-Mohamed, Mohamed Buhary. Mixed-mode fracture of concrete: an experimental
4 approach. TU Delft, Delft University of Technology, 1992.
5
6 [38] Wells, G. N., & Sluys, L. J., A new method for modelling cohesive cracks using finite
7 elements. International Journal for Numerical Methods in Engineering, 2001; 50(12),
8
9 2667-2682.
10 [39] Wu, J. Y., Li, F. B., & Xu, S. L., Extended embedded finite elements with continuous
11
12 displacement jumps for the modeling of localized failure in solids. Computer Methods in
13 Applied Mechanics and Engineering, 2015; 285, 346-378.
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
33
62
63
64
65