Composite Structures 145 (2016) 162–185

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

A SFEM-based evaluation of mode-I Stress Intensity Factor in composite

structures
Nicholas Fantuzzi a, Rossana Dimitri b, Francesco Tornabene a,⇑
a
DICAM Department, School of Engineering and Architecture, University of Bologna, Italy
b
Department of Innovation Engineering, University of Salento, Lecce, Italy

a r t i c l e i n f o a b s t r a c t

Article history: The Strong Formulation Finite Element Method (SFEM) based on the Generalized Differential Quadrature
Available online 2 March 2016 (GDQ) method is applied in the present work to estimate numerically the Stress Intensity Factor (SIF) for
a single edge-notched tensile specimen (SENT) with different length-to-width ratios and composite
Keywords: materials, as well as for a Center Cracked Tension (CCT) and a Double Edge-Notched Tensile (DENT) spec-
Composite materials imens, under a mode-I loading condition. The basis notions of the fracture mechanics are herein analyzed
Linear Elastic Fracture Mechanics numerically in order to evaluate the influence of possible cracks within materials and to relate the dimen-
Generalized Differential Quadrature
sions of cracks and the applied loading to the varying stress distribution. The numerical results, in terms
Stress Intensity Factor
Strong Formulation Finite Element Method
of stresses and SIFs, are straightforwardly compared to those ones given by the standard Finite Element
Method (FEM) and theoretical predictions available from the literature. This work presents a consistent
approach for the computation of the SIF using a strong form methodology. The main aim is to demon-
strate the accuracy and efficiency of the proposed methodology when treating classical plane stress prob-
lems with cracks. We show and discuss results from several numerical examples, including different
composite materials and varying geometries for a mode-I SENT, CCT and DENT specimens.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction problems, for example, have been recently solved by introducing

some expansion techniques [12–15], which have been extended
The large use of composite materials for many engineering to orthotropic and anisotropic materials [16,17]. Freese and Baratta
applications raises the necessity for design methods of joining pro- [18] used weight function and boundary collocation techniques to
cesses such as riveting or adhesive bonding. Nevertheless, the chal- re-define the geometry factor solutions for a finite plate with a sin-
lenges in modeling and characterizing adhesive joints and gle edge crack, under different loading conditions. Fett [19] pre-
composite materials is usually related to the complex stress state sented the SIF solutions for short plates in tension and bending
induced at the multi-material corners and the complex failure by using weight functions and superposition methods. Cho et al.
mechanisms featuring a sharp crack growth and a plastic or dam- [20] investigated the pure mode-I and mixed mode fracturing of
age zone development in the adhesive layers. Despite the high polysilicon for microelectromechanical systems (MEMS). A para-
stress localization introduced within a material by sharp corners, metric investigation was performed for different sizes of speci-
notches or discontinuities, most engineering components contain mens, ranging from 50 to 400 lm, with varying crack lengths and
such features, due to some simplifications during their fabrication. angles of crack inclinations.
In this sense, a realistic measure of the Stress Intensity Factors The computational problems for cracked systems, however, are
(SIFs) is required at sharp notches and cracks as a basis for engi- clearly related to the numerical technique adopted for the analy-
neering analyses, as available from the literature [1–6]. This repre- ses. Fracture mechanics, in this sense, has been extensively studied
sents a complex task to be solved numerically due to high-order with the Finite Element Method (FEM), especially after the 1960s,
gradients of the solutions in terms of displacements and stresses due to the progressive increase of the computational power avail-
at crack tips (see [7–11] among others). Stress singularity able for computational research. Chen and Wang [21] computed
the mixed mode geometry factors for a finite plate with an inclined
⇑ Corresponding author. edge crack using a weight function approach, which was then
E-mail address: [email protected] (F. Tornabene). adopted to compute the mixed-mode SIFs and T-stress for an offset
URL: http://software.dicam.unibo.it/diqumaspab-project (F. Tornabene). double edge-cracked plate. Merah and Albinmousa [22] applied the

http://dx.doi.org/10.1016/j.compstruct.2016.02.076
0263-8223/Ó 2016 Elsevier Ltd. All rights reserved.
 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185 163

FEM and photoelasticity to determine the mixed-mode crack initi- characterizes the state of stress in the vicinity of the crack tip
ation angle for a cracked polycarbonate plate with a single edge and therefore represents a key point in the application of LEFM
crack. Jogdand and Murthy [23] proposed an innovative method to further problems as fatigue life evaluation [73,74], crack initia-
for the estimation of mixed-mode SIFs based on the formation of tion evaluation [75–80], crack tip core region characterization
over-determined system of equations, which was then imple- [78–81], crack propagation simulation [82–84] and crack stabiliza-
mented in existing finite element codes. A FEM-based dislocation tion [85,86].
model was also proposed by Feng et al. [24] in order to determine The numerical results obtained with the GDQ-based Strong For-
the SIFs for an arbitrary number of through thickness cracks. mulation Finite Element Method (SFEM) are compared with the
Always based on the FEM fundamentals, Chaudhuri [25] assessed existing closed-form analytical solutions available from the stan-
numerically the effects of the stress singularity in the neighbor- dard handbooks, e.g. [1–6], as well as with numerical results
hood of the circumferential re-entrant corner lines of the elliptical obtained with the standard FEM, for validation purposes. The
and circular surface flaws, weakening plates, due to their role in remainder of this paper is organized as follows: Section 2 briefly
crack initiation. Whitcomb et al. [26] and Chaudhuri [27,28] also introduces the plane elasticity equations of the problem in an
commented on the accuracy of FEM results, based on a displace- extended and matrix form, while focusing on the mapping tech-
ment potential energy approach, in the neighborhood of stress dis- nique of a single element and boundary conditions and inter-
continuities and stress singularities. As explicitly mentioned by element connectivity conditions. Section 3 summarizes the main
Whitcomb et al. [26], the FEM allows for accurate solutions every- basics of the GDQ formulation, which is then applied to solve the
where except in the closest region around the stress discontinuity fundamental equations of the problem in a discrete form. The
or singularity. The same lack of accuracy in a FEM-based definition GDQ-based numerical evaluation of the SIF for the SENT specimen
of the stress gradients around the circumferential corner line of an is provided in Section 4 for different geometries and materials,
internal circular/elliptical cylindrical hole, was also highlighted in whose results are compared to the reference solutions available
[27,28]. Based on the above mentioned numerical limitations of in the literature and results given by standard FEM. A similar para-
FEM when approaching physical models in presence of line cracks, metric investigation is then repeated for the CCT and DENT speci-
some alternative numerical strategies have been recently devel- mens with central or lateral cracks respectively, whereas some
oped in the literature to analyze the fracture process of elastic conclusive remarks are reported in Section 5.
structures. To this end, the extended FEM [29,30], the Embedded
Mesh formulation [31,32], the Cell Method [33,34], as well as the 2. 2D theory of elasticity
Isogeometric Method [35–39] or the Generalized Differential
Quadrature (GDQ) Method [40,41], have been proposed as compu- A brief description of the 2D plane elastic theory and its basic
tationally more efficient techniques to treat fracture mechanics, mathematical formulation is provided in this section (see also
while additionally introducing some non-linear traction- Timoshenko [87]). Indeed, the structural response of the pre-
separation relationships, known as cohesive zone models, as alter- cracked SENT, CCT or DENT specimen analyzed in the following
native approach to analyze the progressive damage of materials is ruled by the plane elasticity equations which express both plane
and interfaces (see the basic notions in [42,43] and some numerical strain or plane stress conditions. In detail, when a prismatic solid in
applications in [44–48] among others). the orthogonal system Oxyz is subjected to a plane strain state in
In this study the SFEM is employed to treat the stress singular- the x
y plane, the out-of-plane strain components are equal to
ities in fracture mechanics. This advanced numerical tool, based on zero, i.e. ez ¼ cxz ¼ cyz ¼ 0. Conversely, a prismatic solid in plane
several subdomains of general shape, combines the strong formu- stress conditions is characterized by negligible out-of-plane stress
lation of a higher order collocation method and the general appli- components, i.e. ez ¼ cxz ¼ cyz ¼ 0. The plane strain state differs
cation of FEM. The GDQ approach [49–51] has been increasingly
from the plane stress one since rz –0, whereas a plane stress state
used in the literature because is a very fast and efficient methodol-
is different from the plane strain one because of the non-vanishing
ogy to solve systems of partial differential equations [52–72]. This
value of the out-of-plane deformation component ez – 0. For both
method is easily applied for regular problem domains, while
the plane stress and strain configurations, the elastic kinematic
adopting a multi-domain technique for more complex domains,
relationships can be expressed in a compact form as
including the presence of possible discontinuities within the phys-
ical domain, as usually occurs in practical applications. As far as the e ¼ Du ð1Þ
domain decomposition is concerned, two different approaches can
where u ¼ ½ u v T and e ¼ ½ ex ey cxy  are the displacement and
T
be followed. The first one, named multi-domain GDQ, is based on a
simple decomposition of the whole domain in regular sub- deformation vectors, respectively, while D is a kinematic operator
domains. The second approach, instead, is based on a coordinate defined as
transformation procedure, where the generally-shaped element " @ @
#T
@x
0 @y
in the physical coordinate system is mapped into a regular domain, D¼ ð2Þ
@ @
following exactly the same rules of the FEM. Differently from FEM, 0 @y @x
however, the GDQ method solves the strong formulation of the dif-
The constitutive equations relating the stress and strain states
ferential problem, enforcing a posteriori the boundary conditions,
can be defined as follows
which would be instead satisfied a priori in a weak (variational)
formulation. r ¼ Ce ð3Þ
It will be shown, for the first time, the possibility of computing where
SIF using the GDQ method. The numerical evaluation of the mode-I 2 3
SIF near the crack tip for a Single Edge-Notched Tensile specimen 2G þ k k 0
6 7
(SENT), a Center Cracked Tension specimen (CCT) and a Double C¼4 k 2G þ k 05 ð4Þ
Edge-Notched Tensile specimen (DENT), under a uniform tensile 0 0 G
loading, are illustrated. These tests are commonly adopted in the
literature for measuring the mode-I fracture toughness of engi- is the matrix of the Lamè’s elastic constants k and G relating the
neering materials and structural components. As previously men- Young’s modulus E and the Poisson’s ratio m (see [87]), and
tioned, in Linear Elastic Fracture Mechanics (LEFM), the SIF r ¼ ½ rx ry sxy T denotes the stress vector.
164 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185

Let connection between adjacent elements, where B refers to a Dirich-

" # let or Neumann boundary condition enforced at each element
@ @
0 @y
edge, while the subscripts and superscripts n and m indicate the
D ¼ @x
@ @
ð5Þ adjacent elements involved. The previously mentioned boundary
0 @y @x
conditions can be rewritten in an extended form as uðnÞ ¼ uðmÞ
(Dirichlet condition), and rn ¼ rn (Neumann condition). With
ðnÞ ðmÞ
T
be the transpose of the kinematic operator D and f ¼ ½ f x f y  the
this definition, the Dirichlet conditions can be enforced directly,
force vector, which identifies the body force; then the equilibrium
while the Neumann ones are expressed as functions of the outward
equations read T
unit normal vector n ¼ ½ nx ny  based on Eqs. (8) and (9). This
D r þ f ¼ 0 ð6Þ means that a Dirichlet condition Bm n
n or Bm is an enforced boundary

To solve the differential problem in a strong form, we have to displacement of elements XðnÞ or XðmÞ , respectively. Differently, a
n includes the stresses rn of element X
ðnÞ ðnÞ
find the fundamental system of equations in terms of the kine- Neumann condition Bm
matic quantities u and v . Thus, using Eqs. (1) and (3) we have to interacting with element X ðmÞ
whereas a Neumann condition Bnm
solve the following fundamental system of equations
includes the stresses r
ðmÞ
n of element XðmÞ interacting with element
Lu þ f ¼ 0 ð7Þ ðnÞ
X . A C -continuity is reached, in any case, at the inter-element
1


L ¼ D CD being a fundamental operator. The above system of boundaries, which is clearly higher than the C 0 order of continuity
Eq. (7) can be solved after enforcing the boundary conditions asso- of a standard FEM. Boundary conditions must be additionally guar-
ciated to the problem; i.e. a kinematic (or Dirichlet) boundary con- anteed at corner points, whose numerical implementation for
dition on the displacement vector u ¼ u
and a static (or Neumann) higher-order schemes represents a highly-discussed problem
boundary condition on the derivatives of the displacement quanti- [64–72]. Within a GDQ-based numerical integration process, a
ties @u=@n ¼ q. More specifically, for a fixed edge of the solid, it is mapping technique is applied at each irregular sub-domain in

¼ 0, while the vector q named as flux vector, is given by the
u order to transform the physical coordinates ðx; yÞ into the parent
external applied forces to the fixed physical domain, both in the coordinates ðn; gÞ. Thus, the parent element is defined as

normal and tangential directions. In order to apply the SFEM on x ¼ xðn; gÞ
sub-domains, as herein required for the numerical applications, ð10Þ
y ¼ yðn; gÞ
the elements connectivity has to be introduced, in the respect of
a C 1 -continuity at the inter-element boundary. Based on a SFEM Therefore, all degrees of freedom for the considered model,
approach, a domain is permitted to have any kind of shape, and together with their derivatives of the first and second order have
can be transformed with a mapping technique into a set of regular to be mapped into the parent domain. Using the chain rule of dif-
Cartesian parent elements. This means that the external flux ferentiation, from Eq. (10) we have that
boundary conditions must be enforced in the outward normal @ @ @
direction as [88–90] ¼ n þ g
@x @n x @ g x
ð11Þ
@ @ @
rn ¼ Nr ð8Þ ¼ n þ g
@y @n y @ g y
where where nx ¼ @n=@x, gx ¼ @ g=@x, ny ¼ @n=@y, gy ¼ @ g=@y refer to the
" # first order derivatives of the parent coordinates with respect to
n2x n2y 2nx ny the physical ones. The mapping procedure of the fundamental equa-
N¼ ð9Þ

nx ny nx ny n2x
n2y tions and boundary conditions for the in-plane problem herein con-
sidered requires the computation of the second order derivatives,
and nx , ny are the direction cosines of the outward unit normal vec- which read
tor n. Hence, the analytical problem has been implemented numer- @2 @2 @2 @2 @ @
ically in a strong form by applying a SFEM whose fundamentals are ¼ n2x 2 þ g2x 2 þ 2nx gx þ nxx þ gxx
@x 2
@n @g @n@ g @n @g
detailed in the following section.
@ @2 @2 @2 @ @
¼ n2y 2 þ g2y 2 þ 2ny gy þ nyy þ gyy
@y @n @ g @n@ g @n @ g
3. SFEM review 2 2 2   2
@ @ @ @ @ @
¼ nx ny 2 þ gx gy 2 þ nx gx þ ny gy þ nxy þ gxy
In the present work the SFEM is implemented to solve the gov- @x@y @n @g @n@ g @n @g
erning equations for the in-plane problem of a pre-cracked SENT, ð12Þ
CCT and DENT specimen. The SFEM is known to decompose an
Additional details about the coordinate transformation and the
irregular physical domain X into different sub-domains (or ele-
ðnÞ
mapping procedure applied to the differential quadrature can be
ments) X , with n ¼ 1; . . . ; ne , and ne the total number of sub- found in [64–72]. According to the GDQ approach, the spatial
domains characterizing a mesh. Based on the decomposition rule derivative of a function at a given point is approximated as a
herein adopted, each couple of sub-domains is such that their weighted linear sum of all the functional values at the same grid
intersection represents an empty set; i.e. XðnÞ \ XðmÞ ¼ £ for n–m point in the whole domain. In a one-dimensional setting, this
whereas the whole domain is the union of a collection of sets, means that the n-th order derivative of a continuous function
namely X ¼ Xð1Þ [ . . . [ Xðne Þ . For 2D plane problems, the total num- f ðxÞ with respect to the spatial dimension x at a given point xi
ber of degrees of freedom per problem is related to the kinematic can be estimated as in the following
quantities u and v per node, and computed as N n  N g  ne  2, where 
d f ðxÞ
n XN
N n and N g is the number of collocation points in the directions n n  ¼ cnij f ðxj Þ; ði ¼ 1; 2; . . . ; N n ¼ 1; 2; . . . ; N
1Þ
and g of the grid, respectively. The continuity conditions Bm n dx x¼xi
n ¼ Bm
j¼1

are enforced at the inter-element boundaries via a direct ð13Þ

 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185 165

h
y

x
a
h
b t

(a) (b) (c)

Fig. 1. Fracture mechanics problems: (a) schematic definition for the SENT specimen; (b) schematic representation of a precracked elastic body under prescribed
displacements u
and tractions r

; (c) geometrical scheme for half SENT.

where N is the total number of collocation points, and cnij are the
l
weighting coefficients which are computed recursively as (see
[91–102] among others)
!
cn
1
ij
h
cnij ¼n cn
1 1
ij c ij
for i–j; n ¼ 1; 2; . . . ; N
1 h/4
xi
xj
ð14Þ
X
N
cnii ¼
cnik for i ¼ j l/4
k¼1;k–i

while using the Lagrange interpolation polynomials L to define c1ij , (a)

i.e.
Y
N
L1 ðxi Þ ¼ ðxq
xi Þ 1
a a
q¼1;q–i
ð15Þ b
Y
N
-1 1
L1 ðxj Þ ¼ ðxq
xj Þ b c
q¼1;q–j

An important aspect to be considered for a successful applica- c

-1 a,b,c
tion of the GDQ approach is related to the selection of the grid
isoparametric
points distribution in the domain, as clearly highlighted in [40],
space
wherein, the use of a Chebyshev–Gauss–Lobatto (C–G–L) grid dis- physical space
tribution for structural mechanics computations is also recom- (b)
mended. The C–G–L sampling point rule is proven to be efficient
for numerical reasons with Lagrange interpolations, in fact, the Fig. 2. Quarter-points crack tip elements: (a) sketch of a quarter-point quadrilateral
element; (b) examples of quadrangular and a collapsed one quarter-points
approximation error of the dependent variables decreases for an
elements.
increasing number of nodes. For these reasons, a C–G–L sampling
point rule has been herein selected for our numerical computa-
tions, which is based on a harmonic normalized grid distribution
in the ðn; gÞ parent domain as follows to better comprehend the accuracy, stability and versatile of the
  GDQ method and the SFEM approach. In the articles cited above
i
1
ni ¼
cos p i ¼ 1; 2; . . . ; Nn one can found all the applications and improvements that the
Nn
1
  ð16Þ authors have done in over ten years of hard working regarding this
j
1
gj ¼
cos p j ¼ 1; 2; . . . ; Ng important topic of computational and structural mechanics.
Ng
1

where n; g 2 ½
1; 1 are the parent element coordinates involved in 4. Numerical studies
the mapping transformation, while N n ; N g is the number of grid
points inside each element along the n and g direction, respectively. This section illustrates the numerical applications provided by
An equal number of grid points N n ¼ Ng ¼ N has been here adopted the authors to show the consistent approach for the evaluation of
for the numerical applications because it has been demonstrated in the SIF using the SFEM. It will be shown in detail, the steps that
previous publications by the authors that the better accuracy is should be followed regarding the best mesh selection for the com-
reached setting such parameter. Finally, for the sake of brevity, putation of an accurate SIF and the numerical post-processing
the reader can refer to the works [103–120,97,121–138] in order technique for the correct evaluation of the SIF starting from its
166 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185

classic definition. Three geometries will be considered: Single crack tip. As already mentioned in the introduction, the linear elas-
Edge-Notched Tensile (SENT), Center Cracked Tension (CCT) and tic fracture dominating the stress and strains in the region around
Double Edge-Notched Tensile (DENT) specimens. All of these will the tip are singular, while the intensity of the elastic fields is con-
be under mode-I loading condition made of isotropic and compos- trolled by the SIF parameter K. Thus, in mode-I loading conditions,
ite materials. the near-tip fields can be completely characterized if the SIF, K I , is
known. Based on the Williams’ asymptotic solution [139], the crack
4.1. The SENT specimen: problem statement tip stress field in a linear elastic body as in Fig. 1b can be generally
expressed in the following form
The first problem considered herein is an elastic plate of finite
thickness t, width b, and height 2h, containing a though-the thick- rij ðr; #Þ ¼ C 1 r
1=2 f ð1Þ ð2Þ
ij ð#Þ þ C 2 f ij ð#Þ þ C 3 r
1=2 ð3Þ
f ij ð#Þ
ness crack of length a. The specimen, referred to as Single Edge- þ higher order terms ð17Þ
Notched Tensile specimen (SENT), is subjected to a uniform
mode-I tensile loading on the upper and lower sides, as shown in where rij is the stress tensor, r and # are the polar coordinates with
ð1Þ ð2Þ ð3Þ
Fig. 1a. The primary intent of the work is to contribute towards a origin at the crack tip, f ij , f ij , f ij are three functions of #, and C 1 ,
realistic estimation of the stress concentrations that occur at the C 2 , C 3 , are fracture parameters proportional to the externally

(a) Mesh 1 (b) Mesh 2 (c) Mesh 3 (d) Mesh 4

(e) Mesh 5 (f) Mesh 6 (g) Mesh 7 (h) Mesh 8

Fig. 3. SENT-1: mesh configurations adopted for the SFEM analyses.
 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185 167

(a) N = 9 (b) N = 11

(c) N = 13
(d) N = 15

(e) N = 17 (f) N = 19

(g) N = 21 (h) N = 25
Fig. 4. SENT-1: K I vs. the dimensionless distance from the crack tip r, for the ‘‘Mesh 8” configuration and different values of N.
168 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185

(a) d = 0.03m (b) d = 0.06 m

(c) d = 0.12 m (d) d = 0.24 m

(e) d = 0.36 m (f) d = 0.48 m

Fig. 5. SENT-1: interpolated value of K I for ‘‘Mesh 8”. N ¼ 15, q ¼ 0:05  a ¼ 0:060 m.

 
applied loads. Nearby the crack, i.e. for a distance r ! 0 from I # # 3#
f 11 ð#Þ ¼ cos 1
sin sin
the crack tip, the first term which exhibits a square-root singu- 2 2 2
 
larity prevails. Thus, the singular stress amplitude in simple # # 3#
mode-I conditions is characterized by the SIF K I as in the
I
f 22 ð#Þ ¼ cos 1 þ sin sin ð19Þ
2 2 2
following
I # # 3#
f 12 ð#Þ ¼ sin cos cos
K 2 2 2
rij ðr; #Þ ¼ pffiffiffiffiffiffiffiffi
I I
ffi f ij ð#Þ ð18Þ
2p r and K I is determined as
I pffiffiffiffiffiffi
where the angular functions f ij ð#Þ are expressed as (see [1–6]) K I ¼ Fða=bÞr pa ð20Þ
 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185 169

Table 1
SENT-1: interpolated values of K I for ‘‘Mesh 8”, as obtained from the SFEM approach.

N q [m] Distance d [m]

0.03 0.06 0.12 0.24 0.36 0.48
11 0.162 296.9783 300.1774 298.4248 298.4757 300.5540 302.8553
13 0.120 291.1617 290.5252 291.1679 292.9351 293.3133 293.7836
15 0.060 293.8538 291.1259 289.7120 290.7230 290.9203 291.1509
17 0.060 290.4615 290.0321 291.1145 292.3553 292.5410 292.7144
19 0.060 289.6566 290.5697 291.8276 292.3621 292.4374 292.5416
21 0.060 290.5151 291.8208 292.2362 292.2911 292.3227 292.3903
25 0.060 292.5313 292.3331 291.9687 291.8561 291.8773 291.9443

Table 2
SENT-1: theoretical and SFEM-based numerical values of K I for ‘‘Mesh 8”.

N q [m] Gross-Brown Tada I Tada II SFEM d ¼ 0:12 Max. Rel. error (%) Max. Abs. error

11 0.162 291.8431 292.2698 290.0982 298.4248 2.87 8.3266

13 0.120 291.8431 292.2698 290.0982 291.1679 0.37 1.0697
15 0.060 291.8431 292.2698 290.0982 289.7120
0.13
0.3862
17 0.060 291.8431 292.2698 290.0982 291.1145 0.35 1.0163
19 0.060 291.8431 292.2698 290.0982 291.8276 0.60 1.7294
21 0.060 291.8431 292.2698 290.0982 292.2362 0.74 2.1380
25 0.060 291.8431 292.2698 290.0982 291.9687 0.64 1.8705

Fða=bÞ being a geometry function varying with the crack and the
specimen widths. Among the different analytical expressions avail-
able in the handbooks for Fða=bÞ [1–6], in what follows we adopt
three different forms for the exact computation of K I , for the SENT
specimen, i.e.
– Gross, 1964 – Brown, 1966
2
Fða=bÞ ¼ 1:122
0:231ða=bÞ þ 10:550ða=bÞ
3 4

21:710ða=bÞ þ 30:382ða=bÞ ð21Þ

– Tada I, 1973 (better accuracy of 1% or 5%, for a=b < 0:2 or

a=b P 0:2 respectively)
4 0:857 þ 0:265ða=bÞ
Fða=bÞ ¼ 0:265ð1
a=bÞ þ 3=2
ð22Þ
ð1
a=bÞ
Fig. 6. SENT-1: comparative evaluation of results from SFEM, FEM and theoretical
– Tada II, 1973 (better accuracy of 0:5% independently of the a=b predictions for ‘‘Mesh 8” and d ¼ 0:12 m.
values)rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2b pa
Fða=bÞ ¼ tan
pa 2b
0:752 þ 2:02ða=bÞ þ 0:37  1
sin p2ba
3 tional element, known as crack tip element or singularity ele-
pffiffiffi
 ð23Þ ment for the mesh discretization, which accounts for the r
cos p2ba
stress behavior of the crack tip. A quarter-point element is
Once the SIF is known as a function of the crack characteris- obtained from the standard quadratic element by simply moving
tics (i.e. crack length and location), indeed, we may also deter- the mid-node coordinates three-fourths of the side towards the
mine the energy quantity known as the energy release rate G crack tip [141,142] (see Fig. 2a). In 2D finite elements, the singu-
by means of the well-known Irwin’s relationship [140] in order larities can be built from quadrilateral isoparametric elements
to express the reduction of the stored elastic energy of the sys- collapsed at a vertex to form a sort of annulus around the crack
tem with respect to the crack length under displacement control (see the triangular quarter-point element in Fig. 2b). One corner
conditions. Based on the Eq. (18), a numerical sensitivity of the of a quadratic isoparametric element is then collapsed in such a
K I parameter and the stress behavior is expected ahead of the way that three nodes are located at the same coordinate (see
crack tip within a finite element framework. This would require points a, b, c in Fig. 2b), while the mid-nodes of edges that touch
an extremely fine mesh discretization of this zone, in order to that vertex is moved to the quarter-point position.
capture well the local stress and strain singularities, which is In the following, different numerical examples of SENT speci-
here simplified with the SFEM approach. Among the different men in plane stress conditions have been solved parametrically
numerical approximation strategies adopted in the literature to for different precrack geometries and mesh discretizations in order
approach the problem and reduce the computational effort, a to demonstrate the performances of the SFEM approach and to
quarter-point element technique is also investigated with FEM search for the optimal mesh distribution to be used as discretiza-
for a numerical comparison with the SFEM solution, as largely tion when evaluating stress singularities. Within a strong formula-
adopted within the fracture mechanics community when tion context, the K I fracture parameter is then computed using the
approaching singularities. This technique applies an unconven- DiQuMASPAB software [143] whose results are compared with
170 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185

(a) Mesh 9 (b) Mesh 10 (c) Mesh 11

(d) Mesh 12 (e) Mesh 13 (f) Mesh 14

Fig. 7. SENT-2: mesh configurations adopted for the SFEM analyses.

pffiffiffiffiffi
those ones from standard FEM, as computed with the Abaqus and/ – K I;Gross
Brown ¼ 291:8431 Pa m
pffiffiffiffiffi
or Straus commercial software, as well as with the theoretical pre- – K I;Tada I ¼ 292:2698 Pa m
dictions available from the literature [1–6]. pffiffiffiffiffi
– K I;Tada II ¼ 290:0982 Pa m

4.1.1. The SENT-1 specimen In the present case two series of numerical tests have been per-
As first example we consider a SENT specimen with thickness formed. The first one aims to find the best mesh for the SIF evalu-
t ¼ 1 m, width b ¼ 4:8 m, height 2h ¼ 30 m, and a through-the ation. The first criterion here adopted considers a squared element
thickness crack of length a ¼ 1:2 m (i.e. a=b ¼ 1=4) and subjected that lays on the crack front and has the edge of the same length of
to a uniformly distributed tensional loading r ¼ 100 Pa. A homo- the crack front. ‘‘Mesh 1” and ‘‘Mesh 2”, depicted in Fig. 3(a), (b),
geneous and isotropic aluminum is considered as material for this were considered (together with many other, not reported for the
analysis, with a Young’s modulus E ¼ 69 GPa and a Poisson’s ratio sake of conciseness). These meshes were used varying the number
m ¼ 0:334. Due to the symmetry of the problem with respect to the of grid points from 9 to 25 for each element and the plot of K I vs.
x axis, only half of the geometry is considered (Fig. 2c). The exact the distance r from the crack tip was investigated. It was noted that
solution for the K I is determined through Eqs. (21)–(23), which increasing the number of grid points the curves tend to the refer-
reads ence values reported above.
 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185 171

(a) N = 9 (b) N = 11

(c) N = 13 (d) N = 15

(e) N = 17 (f) N = 19

(g) N = 21 (h) N = 25
Fig. 8. SENT-2: K I vs. the dimensionless distance from the crack tip r, for the ‘‘Mesh 9” configuration and different values of N.
172 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185

(a) N = 9 , ρ = 0.195 m (b) N = 11 , ρ = 0.2025 m

(c) N = 13 , ρ = 0.15 m (d) N = 15 , ρ = 0.075 m

(e) N = 17 , ρ = 0.075 m (f) N = 19 , ρ = 0.075 m

(g) N = 21 , ρ = 0.075 m (h) N = 25 , ρ = 0.075 m

Fig. 9. SENT-2: interpolated value of K I for ‘‘Mesh 9”, d ¼ 0:15 m, and varying distances q from the crack tip.
 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185 173

Table 3
SENT-2: interpolated values of K I for ‘‘Mesh 9”, as obtained from the SFEM approach.

N q [m] Distance d [m]

0.05 0.10 0.15 0.30 0.45 0.60
9 0.195 340.2702 340.2786 340.0712 339.8162 339.3140 339.5728
11 0.202 337.0501 347.1582 338.0055 330.7880 336.2678 340.1788
13 0.150 322.4404 321.9087 322.7242 324.6887 325.1737 325.7411
15 0.150 327.5336 324.7062 324.0855 325.1634 325.4079 325.6289
17 0.150 323.0945 323.1504 323.9893 325.3658 325.5455 325.7250
19 0.150 322.8231 324.3876 325.1041 325.7099 325.7933 325.9036
21 0.150 324.6579 325.6109 325.8049 325.8667 325.9008 325.9735
25 0.150 326.3362 326.0349 325.8683 325.7360 325.7592 325.8296

Table 4
SENT-2: theoretical and SFEM-based numerical values of K I for ‘‘Mesh 9”.

N q [m] Gross-Brown Tada I Tada II SFEM d ¼ 0:30 Max. Rel. error Max. Abs. error

9 0.195 326.7676 326.2905 324.3397 339.8162 4.77% 15.4765

11 0.202 326.7676 326.2905 324.3397 330.7880 1.99% 6.4483
13 0.150 326.7676 326.2905 324.3397 324.6887 0.11% 0.3490
15 0.150 326.7676 326.2905 324.3397 325.1634 0.25% 0.8237
17 0.150 326.7676 326.2905 324.3397 325.3658 0.32% 1.0261
19 0.150 326.7676 326.2905 324.3397 325.7099 0.42% 1.3702
21 0.150 326.7676 326.2905 324.3397 325.8667 0.47% 1.5270
25 0.150 326.7676 326.2905 324.3397 325.7360 4.77% 15.4765

Table 5 cases have a good approximation only if the number of points is

SENT-2: interpolated values of K I for ‘‘Meshes 10–14”, as obtained from the SFEM relatively large N > 13. Starting from this preliminary result the
approach with d ¼ 0:05m. authors improved the ‘‘Mesh 1” configuration in order to get a bet-
N ‘‘Mesh 13” ‘‘Mesh 14” ‘‘Mesh 15” ‘‘Mesh 15” ‘‘Mesh 16” ter approximation. Therefore, other 6 configurations have been
L ¼ 0:300 m L ¼ 0:275 m L ¼ 0:100 m L ¼ 0:050 m L ¼ 0:025 m tested by the authors. The analyses are carried on by adopting an
9 333.7987 347.7643 363.9801 330.2022 298.5597 element of length le ¼ a=4 ¼ 0:3 m on the opposite side of the crack
11 325.1208 317.2597 327.2354 324.2662 483.1632 front, as shown in Fig. 3, wherein all the mesh depicted have an
13 325.1262 324.2819 326.5743 324.7501 –54.3478 element next to the crack tip that has an horizontal edge of size
15 326.9970 327.0471 325.1891 326.6096 325.9582
a=4. These configurations are termed ‘‘Mesh 3”-‘‘Mesh 8”. The cor-
17 325.5539 325.9504 326.3443 321.5753 330.0489
19 325.2342 325.6144 327.3602 326.0885 324.4817
responding results to these domain divisions are not reported for
21 325.3649 325.5404 325.9651 324.1363 327.7036 the sake of brevity, only the results of ‘‘Mesh 8” are reported in
25 325.3588 325.4958 325.4743 325.0114 324.6005 Figs. 4 and 5 using the same number of grid points of the previous
configurations. Choosing a small element near the crack tip
reduces the oscillatory behavior of the K I plots. Nevertheless, mesh
However, an oscillatory behavior occurred near the tip due to configurations with an irregular element at the crack front, i.e.
the strong discontinuity enforced by the geometric configuration. ‘‘Mesh 3”, ‘‘Mesh 4”, ‘‘Mesh 7”, lead to oscillating solutions all
Thus, a post-processing criterion has to be implemented in order approaching the reference theoretical value. The solution quality
to evaluate the K I systematically. It was clear to the authors that also increases and becomes more stable for an increasing dimen-
the trend with the smallest oscillation is carried out by ‘‘Mesh 1”. sion of the element beside the crack as shown for ‘‘Mesh 5”, ‘‘Mesh
‘‘Mesh 1” is characterized by an element which lays on the crack 6”. However, good quality results are obtained only for high values
front, as all the other configurations, and one element of the same of N. Upon this round of numerical tests, the best choice seems to
length next to it upon the x and y axis. ‘‘Mesh 2” has one element of be provided by ‘‘Mesh 8”, which features a square-shaped element
larger dimensions 1.5 times the one on the crack front. ‘‘Mesh 1” with the same dimension a of the crack (i.e. le  he ¼ a  a), for the
has also another advantage, that its trends tend to the reference element on the crack front. This is followed by a rectangular ele-
values increasing the number of grid points, whereas the other ment with the same height he ¼ a and length le ¼ he =4. For this

Table 6
SENT-2: maximum relative errors of K I for ‘‘Meshes 10–14” with respect to the reference solutions of Table 5 with d ¼ 0:05 m.

N ‘‘Mesh 13” L ¼ 0:300 m ‘‘Mesh 14” L ¼ 0:275 m ‘‘Mesh 15”L ¼ 0:100 m ‘‘Mesh 15” L ¼ 0:050 m ‘‘Mesh 16” L ¼ 0:025 m
(%) (%) (%) (%) (%)
9 2.92 7.22 12.22 1.81
7.95
11 0.24
2.18 0.89
0.02 48.97
13 0.24
0.02 0.69 0.13
116.63
15 0.82 0.83 0.26 0.70 0.50
17 0.37 0.50 0.62
0.85 1.76
19 0.28 0.39 0.93 0.54 0.04
21 0.32 0.37 0.50
0.06 1.04
25 0.31 0.36 0.35 0.21 0.08
174 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185

mesh, the results exhibit a good quality also for coarse grid points approach. A comparative evaluation of results is illustrated in
distributions within each subdomain (see Fig. 4). Fig. 6 for ‘‘Mesh 8”, d ¼ 0:12 m, leading to a numerical value of
The present numerical estimation of the mode-I SIF is based on K I ranging between K I;Tada II and K I;Tada I . The numerical results pro-
a linear interpolation of the global nonmonotone behavior of the K I vided by SFEM are very close to those ones found with FEM, which
as a function of r. This approach resulted to be a reasonable choice, in turn, are always comprised between the two theoretical predic-
considered also by other researches [10]. The K I value can be reck- tions K I;Tada II and K I;Tada I , as shown by Fig. 6.
oned by a linear interpolation of the points laying on a distance
d ¼ D
q of the starting branch of the curve at a distance q from 4.1.2. The SENT-2 specimen
the crack tip. Please refer to Fig. 2b for the meaning of the afore- The following example aims to analyze the ‘‘scale effect” on
mentioned symbols. In other words, the linear interpolation for the evaluation of K I on the dimension of the element next to
the K I evaluation is performed by neglecting some points near the crack front. The authors want to demonstrate the validity
the crack tip (within the distance q) and considering a few others of the present procedure, changing the geometric sizes of the
near that area (within the distance d). Fig. 5 shows straight blue problem t ¼ 1 m, width b ¼ 6 m, height 2h ¼ 30 m, and a
lines along with the corresponding numerical values computed through-the-thickness crack of length a ¼ 1:5 m, but keeping
for r ! 0 and the theoretical predictions represented by the dashed the same ratio a=b ¼ 1=4. The same homogeneous and isotropic
lines. This post-processing computation regards the choice of dif- aluminum and mechanical properties are assumed also for this
ferent radii q and lengths of interpolation d. In particular, each fig- example, together with the same uniform tensional load
ure has a fixed value for q ¼ 13:5%; 10%; 5% of a, varying the r ¼ 100 Pa acting on the upper and lower sides of the specimen.
distance d and the number of points N. It is worth noticing that, Once again, due to the symmetry of the problem with respect to
a different radius q from the crack tip has been chosen for meshes the x axis, only half of the geometry is considered (Fig. 2c). A
with lower degrees of freedom, i.e. q ¼ 0:135  a for N ¼ 11, sequence of six different meshes, depicted in Fig. 7, is adopted
q ¼ 0:1  a for N ¼ 13, q ¼ 0:05  a for N > 13, due to the for the SFEM analyses all considering the optimal shape distribu-
dependence of the point distribution for each element to the tion of subdomains of ‘‘Mesh 8” as concluded from the SENT-1
selected discretization N n  N g . The distances are tailored as example. The present parametric investigation is here performed
d ¼ 0:03; 0:06; 0:12; 0:24; 0:36; 0:48 m, however the final results varying the horizontal size of the element next to the crack front
change of a few point percentage. Table 1 summarizes the main as: le ¼ a=4 ¼ 0:375 m ‘‘Mesh 9”, le ¼ a=5 ¼ 0:300 m ‘‘Mesh 10”,
numerical results stemming from the above mentioned linear le ¼ a=6 ¼ 0:200 m ‘‘Mesh 11”, le ¼ a=10 ¼ 0:150 m ‘‘Mesh 12”,
interpolations. It can be observed that, all the results have similar le ¼ a=30 ¼ 0:050 m ‘‘Mesh 13”, le ¼ a=60 ¼ 0:025 m ‘‘Mesh 14”.
values increasing the number of grid points and changing the dis- Based on the Eqs. (21)–(23), the reference theoretical solutions
tance d. The results in Tables 1 and 2, if compared to the theoretical for K I read
predictions mentioned earlier, clearly show a very good accuracy
pffiffiffiffiffi
in the numerical estimation of K I for a distance d ranging from – K I;Gross
Brown ¼ 326:29 Pa= m
pffiffiffiffiffi
a=10 to a=5, where a is the crack length, i.e. for values of d ranging – K I;Tada I ¼ 326:77 Pa= m
0:12 m and 0:24 m (bold values in Table 1). Analogously in Table 2, pffiffiffiffiffi
– K I;Tada II ¼ 324:34 Pa= m
it can be noted that, apart from discretizations with N ¼ 11 and
N ¼ 15 where we reach the maximum and minimum relative whereas the numerical results of K I are plotted in Fig. 8 only for
and absolute error, the SFEM predictions are always comprised ‘‘Mesh 9”. The other figures cannot be presented for the sake of
between the theoretical predictions of K I . The relative error clarity. However a textual comments is reported below. The
reported is very small compared to the reference solutions listed authors observed that a decreasing length le of the first element
in Table 2. The accuracy of the proposed SFEM method can be also near the crack tip would affect significantly the stability of the
confirmed by repeating the same analyses using a classic FEM numerical response especially for coarse discretizations, causing,

Aluminium h

a
h

b t

Steel

(a) Geometry (b) DiQuMASPAB mesh (c) Abaqus mesh (d) Straus mesh
Fig. 10. SENT-3: geometrical scheme (a) and mesh configurations as adopted for SFEM (b), and FEM (c), (d).
 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185 175

at the same time, an unstable behavior also at the first points of the of interpolation. In addition, the SFEM predictions seem always
second element. This observation suggests that a larger radius q to be comprised between the theoretical predictions of K I from
from the crack tip must be selected for coarse meshes and decreas- Tada II and Gross-Brown, except for the two coarsest discretiza-
ing lengths le in order to circumvent the fluctuating values nearby tions herein considered with N ¼ 9 and N ¼ 11 respectively. Just
the crack tip and to interpolate correctly the desired value of K I . a small comment is given for the other mesh configurations (i.e.
Interpolated numerical results are obtained for the ‘‘Mesh 9” con- ‘‘Mesh 10”-‘‘Mesh 14”), whose numerical solutions maintain a
figuration, at different values of N and different lengths of interpo- good quality under a linear interpolation centered in r ¼ q, exclud-
lations d, whereas a different radius q of fluctuations has been ing the small element next to the crack and included within a dis-
specifically dropped. For brevity, a single graphical representation tance d ¼ 0:05 m. In other words, the K I can be calculated deleting
is given in Fig. 9 for d ¼ 0:15 m. Nevertheless, all these results are the values of the element indicated as 1 in the meshes of Fig. 7
summarized in Table 3. It is worth noticing that very accurate val- (b)–(f). A brief summary of the K I interpolated results based on the
ues of K I can be determined numerically by means of a linear inter- SFEM is reported in Table 5, whereas Table 6 represents the global
polation for a distance d between a=10 ¼ 0:15 m and a=5 ¼ 0:30 m relative errors in percentage for the only numerical values which
(see the values of Table 3 in bold). Based on this interpolation cri- lie within the minimum and maximum analytical predictions. The
terion, a very small error is computed both in a relative and/or quality of the solutions is clearly preserved for meshes characterized
absolute sense between the numerical and analytical estimations, by the first element near the crack having a basis length equal to 1=5
as clearly highlighted in Table 4 for the only distance d ¼ 0:30 m of the crack length, and only for a distance d ¼ 0:05 m .

Aluminium

Steel

Fig. 11. SENT-3: tensional stresses ry vs. r along the straight paths 1 and 2, as given by SFEM and FEM for N ¼ 9.
176 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185

Aluminium

Steel

Fig. 12. SENT-3: tensional stresses ry vs. r along the straight paths 3 and 4, as given by SFEM and FEM for N ¼ 9.

Table 7
SENT-3: theoretical, FEM- and SFEM-based numerical values of K I .

Gross-Brown Tada I Tada II N ¼ 11 N ¼ 15 N ¼ 17 N ¼ 19 N ¼ 21 N ¼ 25 Abaqus

KI 326.29 326.76 324.34 324.74 323.48 325.19 326.24 325.28 326.69 326.91

4.1.3. The SENT-3 specimen from a micro- and macroscopic point of view, through the combi-
The accuracy of the SFEM approach is also benchmarked against nation of different materials with different stiffness, strength,
the analytical predictions for a SENT specimen made of composite weight, thermal properties, strength to corrosion or fatigue etc.
materials, as usually encountered in bonded joints or composite In what follows we consider a composite SENT specimen made of
structures in many fields of engineering and applied science. It is a double material, i.e. Aluminum 1050-O and Steel B450C. The Alu-
noted for the sake of clarity, that the geometric dimensions are minum 1050-O is modeled as an isotropic material with Young’s
kept equal to the previous case (SENT-2) and the mesh configura- modulus E ¼ 69 GPa and Poisson’s ratio m ¼ 0:334, whereas the
tion has one squared element on the crack front and another ele- Steel B450C is modeled as an isotropic material with E ¼ 210 GPa
ment next to it of width a=4 and height a. Composite materials, and m ¼ 0:300. Beyond the analytical-SFEM comparison of K I
indeed, are well known to increase many engineering properties results, the accuracy of the SFEM is now investigated against
 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185 177

FEM predictions as given by two different commercial software, i.e.

Straus and Abaqus, for the same total number of degrees of freedom.
Fig. 10 shows the geometrical scheme of the problem along with the
mesh here adopted for numerical analyses with DiQuMASPAB, Aba-
qus and Straus, respectively. The tensional stresses ry is first evalu-
ated with SFEM along some selected straight paths, i.e.

 Path 1 ½ð0; 0Þ; ððb
aÞ; 0Þ

 Path 2 ½ð0; 0Þ; ððb
aÞ; aÞ
 Path 3 ½ð0; 0Þ; ððb
aÞ; ðb
aÞÞ
 Path 4 ½ð0; 0Þ; ð0; hÞ

A systematic investigation of the problem for a varying number

of degrees of freedom, i.e. for a different number of grid points N
Fig. 13. SENT-3: comparison between the theoretical and numerical solutions for K I has been carried out. Nevertheless Figs. 11 and 12 present the
as given by FEM and SFEM approaches. tensional stresses by the SFEM and FEM along all the previously

(a) N = 11 (b) N = 15

(c) N = 17 (d) N = 19

(e) N = 21 (f) N = 25
Fig. 14. SENT-3: SFEM-based K I vs. r varying the number of grid points N.
178 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185

mentioned paths only for N ¼ 9. It is underlined by the authors for our discretizations are not sufficient to model the crack tip
that increasing the number of points the performance of the pro- and are clearly unable to study typical singularities of LEFM. Addi-
posed algorithm are proven when treating singularities in fracture tional results of the SFEM-based tensional stresses cannot be
mechanics. Despite the coarseness of the mesh here adopted, a reported for the sake of conciseness to investigate in more detail
smooth curve can be correctly captured near the crack tip by using the effect of the mesh discretization on the final performances of
the SFEM approach, compared to the FEM approach. In the last the proposed algorithm. However, the convergence of the present
case, the solution given by the Abaqus code is likely to experience case has been investigated varying the number of point per ele-
the singular near-tip behavior, due to the application of the ment from N ¼ 9 to N ¼ 25 for all the selected paths 1–4. As
quarter-points collapsed elements as implemented in the numeri- expected, it was observed that increasing the number of points
cal code. Conversely, the numerical stresses given by the Straus the stress value at the tip increases. Moreover, due to the compat-
code are much lower than expected due to the coarseness of the ibility conditions it should be noted that the stress field computed
mesh together with the type of elements implemented in this soft- at the Aluminum or Steel sides are the same (first and third plots
ware. The quadratic elements of the type QUAD8, herein adopted on the left of Fig. 11). Table 7 summarizes the values computed

(a) N = 11 , d = 0.9 m (b) N = 15 , d = 0.9 m

(c) N = 17 , d = 0.9 m (d) N = 19 , d = 0.9 m

(e) N = 21 , d = 0.75 m (f) N = 25 , d = 0.75 m

Fig. 15. SENT-3: interpolated value of K I varying the number of grid points N.
 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185 179

σ σ tip for N ¼ 11; 15; 17; 19, or an interpolation distance

d ¼ 0:35 m for finer meshes with N ¼ 21; 25 (see the plots in
Fig. 15).

Steel Steel
B450C h B450C h 4.2. The CCT and DENT specimens: problem statements

Two additional benchmarks are now selected among mode-I

standard specimens, in order to check the flexibility of the pro-
2a a a posed model and its performances in fracture mechanics also
h h for different geometries. The first example considers a standard
2b 2b configuration of CCT specimen of length 2b ¼ 12 m, height
t t 2h ¼ 30 m, with a central crack of length 2a ¼ 3 m as shown in
Cast iron ASTM A-197 Cast iron ASTM A-197
Fig. 16a, while the second example proposes a DENT specimen
of length 2b ¼ 10 m, height 2h ¼ 30 m, with a double lateral crack
σ σ of length a ¼ 1 m as shown in Fig. 16b. A composite specimen is
considered in both cases, which is made of a Steel B450C with
(a) (b) E ¼ 210 GPa and m ¼ 0:300 (upper adherent), and a cast iron
Fig. 16. Geometrical scheme for the CCT (a) and the DENT (b) specimens. ASTM A-197 with E ¼ 172 GPa and m ¼ 0:280 (lower adherent).
The accuracy of the SFEM approach is checked by comparing
the SFEM-based numerical solutions of K I with the closed-form
predictions available in the open literature [1–6], as well as with
for K I in the present case, where a very good agreement is observed the numerical results from FEM, as given by Abaqus. The software
changing the number of grid points. A graphical representation of Straus has been now omitted for comparison purposes due to its
the same results is depicted in Fig. 13, wherein the validity of limitations in presence of singularities related to the type of
the present SFEM solutions are better underlined. In this case, a quadratic element implemented in the software, as previously
correct computation of K I has required a linear interpolation of mentioned.
results within the second element, after neglecting the oscillating
values pertaining to the first element (see the plots in Fig. 14). 4.2.1. The CCT specimen
Thus, an interpolation radius of q ¼ 0:40 m has been chosen For the CCT specimen the exact value of K I is determined by
together with an interpolation distance d ¼ 0:5 m from the crack means of the following expressions for Fða=bÞ [2]

(a) N = 11 (b) N = 15

(c) N = 19 (d) N = 25
Fig. 17. CCT: SFEM-based K I vs. r varying the number of grid points N.
180 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185

(a) N = 11 (b) N = 15

(c) N = 19 (d) N = 25
Fig. 18. CCT: interpolated value of K I varying the number of grid points N with q ¼ 0:4 m, d ¼ 0:3 m.

Table 8
CCT: theoretical, FEM- and SFEM-based numerical values of K I .

Irwin Brown Feddersen Koiter Tada I Tada II N ¼ 11 N ¼ 15 N ¼ 19 N ¼ 25 Abaqus

KI 222.94 214.95 225.846 224.44 224.26 225.55 215.27 219.85 222.55 224.75 225.97

– Isida 1962, 1956, 1973 – Irwin 1957 (approximation by periodic crack solution)
h irffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pa
ffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
Fða=bÞ ¼ 1
0:1ða=bÞ þ 0:96ða=bÞ
4 2b pa
sec Fða=bÞ ¼ tan
b ð24Þ pa b ð25Þ
for a b ) f ða=bÞ 1 for ða=bÞ 6 0:5 ) precision better than 5%
– Brown, 1966 (least squares fitting to Isida’s results)
a a2 a3
Fða=bÞ ¼ 1 þ 0:128
0:288 þ 1:525
b b b ð26Þ
for ða=bÞ 6 0:7 ) precision of 0:5%
– Feddersen, 1966 (guess based on Isida’s results)
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pa
Fða=bÞ ¼ sec
2b
 ð27Þ
ða=bÞ 6 0:7 ) precision of 0:3%
for
ða=bÞ ¼ 0:8 ) precision of 1%
– Koiter, 1956 (asymptotic approximation)
a 2
1
0:5 a
þ 0:326
Fða=bÞ ¼ p
b
ffiffiffiffiffiffiffiffiffiffiaffi b
1
b ð28Þ
8ða=bÞ ) precision of 1%
Fig. 19. CCT: comparison between the theoretical and numerical solutions for K I as
given by FEM and SFEM approaches. – Tada I, 1973 (modification of Koiter’s formula)
 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185 181

2 a 3
1
0:5ðbaÞ þ 0:370 ba
0:044 finer meshes. Nevertheless, it is worth noticing a good agreement
f ða=bÞ ¼ pffiffiffiffiffiffiffiffiffiffiaffi b
between the SFEM- and FEM-based results as shown in Table 8
1
b ð29Þ
and Fig. 19. The numerical values of K I provided by the Abaqus code
8ða=bÞ ) precision of 0:3% correspond almost to the upper bound limit based on the theoretical
– Tada II, 1973 (modification of Feddersen’s formula) previsions of Feddersen, whereas results based on the SFEM formu-
a4 rffiffiffiffiffiffiffiffiffiffiffiffiffiffi
lation fall always within the theoretical previsions of Brown and Fed-
a2 pa
Fða=bÞ ¼ f1
0:025 þ 0:06 g sec dersen, 1966, independently from the mesh refinement.
b b 2b ð30Þ
8ða=bÞ ) precision of 1% 4.2.2. The DENT specimen
For the last example analyzed in this work, (i.e. for the DENT
specimen) we refer to the following theoretical expressions for
Based on the previous relations, the exact solution of K I for the Fða=bÞ [2] to compute the exact value of K I , i.e.
first geometry of CCT specimen gives
pffiffiffiffiffi – Irwin 1957 (approximation by periodic crack solution)
– K I;Irwin ¼ 222:9476 Pa m rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi a
– K I;Brown ¼ 214:9469 Pa m 2b pa
pffiffiffiffiffi F ¼ tan
– K I;Feddersen ¼ 225:8463 Pa m b pa b ð31Þ
pffiffiffiffiffi a
– K I;Koiter ¼ 224:4372 Pa m for > 0:4 ) precision better than 5%
pffiffiffiffiffi b
– K I;Tada I ¼ 224:2649 Pa m
pffiffiffiffiffi – Brown, 1966 (least squares fitting to Isida’s results)
– K I;Tada II ¼ 225:5463 Pa m
a a a2 a3
F ¼ 1:12 þ 0:203
1:197 þ 1:930
The numerical solution is analyzed in terms of K I for four differ- b b b b
ent mesh discretizations, as shown in Fig. 17. The resulting behav- a
for 6 0:7 ) precision of 2%
ior is shown to deliver oscillating results within the first element, b
next to the crack tip, as well as in the transition zone between ð32Þ
the first and the second element near the crack tip. As done earlier – Benthem, 1972 (asymptotic approximation)
for the other examples, the final computation of K I is related to the 3
preliminary suppression of its oscillating value and its consecutive a 1:122
0:561 a

0:015 a 2
þ 0:091 a
¼ b
pffiffiffiffiffiffiffiffiffiffiaffi b b
linear interpolation within a radius q ¼ 0:4 m and for a distance b 1
b ð33Þ
d ¼ 0:3 m (see Fig. 18). The final numerical values seem to be quite a
accurate, if compared to the theoretical predictions, especially for 8 ) precision better than 2%
b

(a) N = 7 (b) N = 9

(c) N = 21 (d) N = 25
Fig. 20. DENT: SFEM-based K I vs. r varying the number of grid points N.
182 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185

(a) N = 11 (b) N = 15

(c) N = 19 (d) N = 25
Fig. 21. DENT: interpolated value of K I varying the number of grid points N with q ¼ 1:5 m, d ¼ 0:1 m.

Table 9
DENT: Theoretical, FEM- and SFEM-based numerical values of K I .

Irwin Brown Benthem Tada I Tada II N¼7 N¼9 N ¼ 21 N ¼ 25 Abaqus

KI 180.25 199.96 219.55 198.25 218.59 197.58 197.07 217.86 211.86 198.98

– Tada I, 1973 (modification of Irwin’s formula) The numerical response based on the SFEM approach is given in
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Fig. 20 in terms of K I , for different levels of refinements. As visible
a  pa 2b pa
F ¼ 1 þ 0:122 cos4 tan from Fig. 21, it seems that results obtained from coarser meshes
b
a
2b p a 2b ð34Þ feature a better quality than finer ones. The final values of K I are
8 ) precision of 0:5% finally computed through a linear interpolation within a radius of
b q ¼ 1:5 m and for an interpolation distance d ¼ 0:1 m (see
– Tada II, 1973 (modification of Benthem’s formula) Fig. 21). In addition, the good agreement between the numerical
4 and theoretical results summarized in Table 9 demonstrates, once
a 1:122
0:561 a 2

0:205 ab þ 0:471 a 3

0:190 a
again, the performance of the proposed algorithm for fracture
F ¼ b
pffiffiffiffiffiffiffiffiffiffia b b
b 1
b mechanics studies.
a
8 ) precisionof 0:5%
b 5. Conclusions
ð35Þ
This paper proposes a SFEM approach based on the GDQ
approximation to study the singularity-driven fracture mechanics.
Based on the previous relations, the exact solution of K I for the
Moreover, a systematic and detailed estimation of the Stress Inten-
DENT specimen gives
sity Factor (SIF) for some standard specimens subjected to a mode-
pffiffiffiffiffi I loading, such as the Single Edge-Notched Tensile specimen
– K I;Irwin ¼ 180:2553 Pa m (SENT), the Center Cracked Tension specimen (CCT) or the
pffiffiffiffiffi
– K I;Brown ¼ 199:9612 Pa m Double-Edge-Notched Tensile specimen (DENT) specimen made
pffiffiffiffiffi
– K I;Benthem ¼ 219:5541 Pa m of single or composite materials. By using the GDQ technique,
pffiffiffiffiffi
– K I;Tada I ¼ 198:2470 Pa m the numerical statement of the problem does not pass through
pffiffiffiffiffi
– K I;Tada II ¼ 218:5902 Pa m any variational formulation, but deals directly with the governing
equations in a strong form. A wide numerical investigation is
 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185 183

carried out for different geometries, materials and mesh discretiza- [25] Chaudhuri RA. A combined FEM and three-dimensional asymptotic solution
based analysis on stress concentration/intensity around elliptical/circular
tions, in order to demonstrate the performance of the proposed
cylinder shaped surface flaws in stretched plates. Appl Math Model
algorithm. In particular, the best way of discretizing such problems 2015;39:5341–53.
has been shown and demonstrated in detail. The SIFs obtained [26] Whitcomb JD, Raju IS, Goree JG. Reliability of the finite element method for
from the GDQ method are based on a linear interpolation of the calculating free edge stresses in composite laminates. Comput Struct
1982;15:23–7.
stresses near the crack tip within a certain distance of interpola- [27] Chaudhuri RA. Computation of transverse shear stresses in the vicinity of the
tion, and agree very well with those computed by using a FEM circumferential re-entrant corner line of an internal part-though hole
approach, or with theoretical predictions available in the literature. weakening an edge-loaded plate. Compos Struct 2009;89:315–20.
[28] Chaudhuri RA. Transverse shear stress distribution through thickness near an
This proves the accuracy and efficiency of the proposed method for internal part-through elliptical hole in a stretched plate. Compos Struct
fracturing problems, also with the adoption of a limited number of 2010;92:818–25.
grid distributions in space near the crack tip. [29] Campilho RDSG, Banea MD, Pinto AMG, da Silva LFM, de Jesus AMP. Strength
prediction of single-and double-lap joints by standard and extended finite
element modelling. Int J Adhes Adhes 2011;31:363–72.
Acknowledgments [30] Benvenuti E, Tralli A. Simulation of finite-width process zone in concrete-like
materials by means of a regularized extended finite element model. Comput
Mech 2012;50(4):479–97.
The second author acknowledges the support from the Euro- [31] Fang X, Charalambides PG. The fracture mechanics of cantilever beams with
pean Research Council under the European Union’s Seventh Frame- an embedded sharp crack under end force loading. Eng Fract Mech
work Programme (FP7/2007-2013), ERC Starting Researcher Grant 2015;149:1–17.
[32] Dolbow J, Harari I. An efficient finite element method for embedded interface
INTERFACES (Grant agreement n° 279439). This research topic is
problems. Int J Numer Methods Eng 2009;78(2):229–52.
one of the subjects of the Centre of Study and Research for the [33] Ferretti E. Crack propagation modeling by remeshing using the Cell Method
Identification of Materials and Structures (CIMEST)-‘‘M. Capurso” (CM). CMES-Comput Model Eng 2003;4:51–72.
of the University of Bologna (Italy). [34] Ferretti E. A discrete nonlocal formulation using local constitutive laws. Int J
Fract 2004;130(3):175–82.
[35] Verhoosel CV, Scott MA, de Borst R, Hughes TJR. An Isogeometric approach to
References cohesive zone modeling. Int J Numer Methods Eng 2011;87(1–5):336–60.
[36] Nguyen VP, Nguyen-Xuan H. High-order B-splines based finite elements for
[1] Sih GC. Handbook of stress intensity factors for researchers and delamination analysis of laminated composites. Compos Struct 2013;102:
engineering. Bethlem: Lehigh University, Institute for Fracture and Solid 261–75.
Mechanics; 1973. [37] Dimitri R, De Lorenzis L, Wriggers P, Zavarise G. NURBS- and T-spline-based
[2] Tada H, Paris PC, Irwin GR. Stress analysis of cracks handbook. American isogeometric cohesive zone modeling of interface debonding. Comput Mech
Society of Mechanical Engineers; 2000. 2014;54:369–88.
[3] Pilkey WD. Formulas for stress, strain and structural matrices. John Wiley and [38] Dimitri R. Isogeometric treatment of large deformation contact and
Sons; 2005. debonding problems with T-splines: a review. Curved Layer Struct 2015;2:
[4] Pilkey WD, Pilkey DF. Peterson’s stress concentration factors. Wiley; 2008. 59–90.
[5] Unger DJ. Analytical fracture mechanics. Academic Press; 1995. [39] Dimitri R, Zavarise G. T-splines discretizations for large deformation contact
[6] Perez N. Fracture mechanics. Kluwer Academic Publishers; 2004. problems. PAMM Proc Appl Math Mech 2015;15:183–4.
[7] Viola E, Marzani A. Crack effect on dynamic stability of beams under [40] Tornabene F, Fantuzzi N, Ubertini F, Viola E. Strong formulation finite element
conservative and nonconservative forces. Eng Fract Mech 2004;71(4– method based on differential quadrature: a survey. Appl Mech Rev
6):699–718. 2015;67:1–55.
[8] Lin KY, Mar JW. Finite element analysis of stress intensity factors for cracks at [41] Viola E, Tornabene F, Ferretti E, Fantuzzi N. GDQFEM numerical simulations of
a bi-material interface. Int J Fact 1976;12:521–31. continuous media with cracks and discontinuities. CMES-Comput Model Eng
[9] Ricci P, Viola E. Stress intensity factors for cracked T-sections and dynamic 2013;94(4):331–69.
behaviour of T-beams. Eng Fract Mech 2006;73(1):91–111. [42] Barenblatt GI. The mathematical theory of equilibrium cracks in brittle
[10] Wang L, Chen J-S, Hu H-Y. Subdomain radial basis collocation method for fracture. Adv Appl Mech 1962;7:55–129.
fracture mechanics. Int J Numer Methods Eng 2010;83:851–76. [43] Dugdale DS. Yielding of steel sheets containing slits. J Mech Phys Solids
[11] Osman T, Matbuly MS, Mohamed SA, Nassar M. Analysis of cracked plates 1960;8:100–4.
using localized multi-domain differential quadrature method. Appl Comput [44] Banks-Sills L, Bortman Y. A mixed-mode fracture specimen: analysis and
Math 2013;2:109–14. testing. Int J Fract 1986;30(3):181–201.
[12] Chaudhuri RA, Xie M. Free-edge stress singularity in a biomaterial laminate. [45] Moës N, Belytschko T. Extended finite element method for cohesive crack
Compos Struct 1998;40:129–36. growth. Engng Fract Mech 2002;69(7):813–33.
[13] Xie M, Chaudhuri RA. Three-dimensional stress singularity at a bimaterial [46] Camanho PP, Dávila CG, De Moura MF. Numerical simulation of mixed-mode
interface crack front. Compos Struct 1998;40:137–47. progressive delamination in composite materials. J Compos Mater 2003;37
[14] Chaudhuri RA, Xie M. A tale of two saints: St. Venant and St. Nick – does St. (16):1415–38.
Venant’s principle apply to bimaterial straight edge and wedge singularity [47] Dimitri R, Trullo M, De Lorenzis L, Zavarise G. A consistency assessment of
problems? Compos Sci Technol 2000;60:2503–15. coupled cohesive zone models for mixed-mode debonding problems. Frattura
[15] Barroso A, París F, Mantič V. Representativity of the singular stress state in the ed Integrità Strutturale 2014;29:266–83.
failure of adhesively bonded joints between metals and composites. Compos [48] Dimitri R, Trullo M, De Lorenzis L, Zavarise G. Coupled cohesive zone models
Part B Eng 2009;69:1746–55. for mixed-mode fracture: a comparative study. Eng Fract Mech 2015;148:
[16] Chaudhuri RA. Three-dimensional singular stress field at the front of a crack 145–79.
in a unidirectional fiber reinforced composite plate. Compos Struct [49] Shu C, Richards BE. Parallel simulation of incompressible viscous flows by
2011;93:513–27. generalized differential quadrature. Comput Syst Eng 1992;3:271–81.
[17] Chaudhuri RA. On three-dimensional singular stress field at the front of a [50] Shu C, Richards BE. Application of generalized differential quadrature to solve
planar rigid inclusion (anticrack) in an orthorhombic mono-crystalline plate. two-dimensional incompressible Navier-Stokes equations. Int J Numer
Int J Fract 2012;174:103–26. Methods Fluids 1992;15:791–8.
[18] Freese CE, Baratta FI. Single edge-crack stress intensity factor solutions. Eng [51] Shu C. Differential quadrature and its application in engineering. Springer;
Fract Mech 2006;73:616–25. 2000.
[19] Fett T. Stress intensity factors for edge-cracked plates under arbitrary loading. [52] Ng TY, Li H, Lam KY, Loy CT. Parametric instability of conical shells by the
Fatigue Fract Eng Mater Struct 1998;22:301–5. generalized differential quadrature method. Int J Numer Methods Eng
[20] Cho SW, Jonnalagadda K, Chasiotis I. Mode I and mixed mode fracture of 1999;44:819–37.
polysilicon for MEMS. Fatigue Fract Eng Mater Struct 2007;30:21–31. [53] Ng TY, Li H, Lam KY. Generalized differential quadrature method for the free
[21] Chen CH, Wang CL. Stress intensity factors and T-stresses for offset double vibration of rotating composite laminated conical shell with various
edge-cracked plates under mixed-mode loadings. Int J Fract 2008;152: boundary conditions. Int J Mech Sci 2003;45:567–87.
149–62. [54] Ng CHW, Zhao YB, Wei GW. Comparison of discrete convolution and
[22] Merah N, Albinmousa J. Experimental and numerical determination of mixed generalized differential quadrature for the vibration analysis of rectangular
mode extension angle. ASTM J Test Eval 2008;37(2):95–107. plates. Comput Methods Appl Mech Eng 2004;193:2483–506.
[23] Jogdand PV, Murthy KSRK. A finite element based interior collocation method [55] Civalek Ö. Application of differential quadrature (DQ) and harmonic
for the computation of stress intensity factors and T-stresses. Eng Fract Mech differential quadrature (HDQ) for buckling analysis of thin isotropic plates
2010;77:1116–27. and elastic columns. Eng Struct 2004;26:171–86.
[24] Feng XQ, Shi YF, Wang XY, Li B, Yu SW, Yang Q. Dislocation-based semi- [56] Civalek Ö. Geometrically nonlinear analysis of doubly curved isotropic shells
analytical method for calculating stress intensity factors of cracks: two- resting on elastic foundation by a combination of harmonic differential
dimensional cases. Eng Fract Mech 2010;77:3521–31. quadrature-finite difference methods. Int J Press Vessels Pip 2005;82:470–9.
184 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185

[57] Tornabene F, Liverani A, Caligiana G. FGM and laminated doubly-curved [89] Zhong H, Yu T. A weak form quadrature element method for plane elasticity
shells and panels of revolution with a free-form meridian: a 2-D GDQ solution problems. Appl Math Model 2009;33(10):3801–14.
for free vibrations. Int J Mech Sci 2011;53:446–70. [90] Xing Y, Liu B, Liu G. A differential quadrature finite element method. Int J Appl
[58] Tornabene F, Liverani A, Caligiana G. Laminated composite rectangular and Mech 2010;2:207–27.
annular plates: a GDQ solution for static analysis with a posteriori shear and [91] Tornabene F, Fantuzzi N, Viola E, Batra RC. Stress and strain recovery for
normal stress recovery. Compos Part B Eng 2012;43:1847–72. functionally graded free-form and doubly-curved sandwich shells using
[59] Tornabene F, Liverani A, Caligiana G. Static analysis of laminated composite higher-order equivalent single layer theory. Compos Struct 2015;119:67–89.
curved shells and panels of revolution with a posteriori shear and normal [92] Tornabene F, Fantuzzi N, Bacciocchi M, Dimitri R. Dynamic analysis of thick
stress recovery using generalized differential quadrature method. Int J Mech and thin elliptic shell structures made of laminated composite materials.
Sci 2012;61:71–87. Compos Struct 2015;133:278–99.
[60] Tornabene F, Reddy JN. FGM and laminated doubly-curved and degenerate [93] Tornabene F, Fantuzzi N, Bacciocchi M, Dimitri R. Free vibrations of composite
shells resting on nonlinear elastic foundations: a GDQ solution for static oval and elliptic cylinders by the generalized differential quadrature method.
analysis with a posteriori stress and strain recovery. J Indian Inst Sci Thin Wall Struct 2015;97:114–29.
2013;93:635–88. [94] Eftekhari SA. Document A simple and accurate mixed Ritz-DQM formulation
[61] Tornabene F, Viola E. Static analysis of functionally graded doubly-curved for free vibration of rectangular plates involving free corners. Ain Shams Eng
shells and panels of revolution. Meccanica 2013;48:901–30. J, in press. doi: http://dx.doi.org/10.1016/j.asej.2015.05.016.
[62] Viola E, Tornabene F, Fantuzzi N. General higher-order shear deformation [95] A note on mathematical treatment of the Dirac-delta function in the
theories for the free vibration analysis of completely doubly-curved differential quadrature bending and forced vibration analysis of beams and
laminated shells and panels. Compos Struct 2013;95:639–66. rectangular plates subjected to concentrated loads Appl Math Model
[63] Viola E, Tornabene F, Fantuzzi N. Static analysis of completely doubly-curved 2015;39:6223–42.
laminated shells and panels using general higher-order shear deformation [96] Zong Z, Zhang Y. Advanced differential quadrature methods. Chapman & Hall/
theories. Compos Struct 2013;101:59–93. CRC applied mathematics & nonlinear science. Taylor & Francis; 2009.
[64] Viola E, Tornabene F, Fantuzzi N. Generalized differential quadrature finite [97] Ferreira AJM, Carrera E, Cinefra M, Viola E, Tornabene F, Fantuzzi N, et al.
element method for cracked composite structures of arbitrary shape. Compos Analysis of thick isotropic and cross-ply laminated plates by generalized
Struct 2013;106:815–34. differential quadrature method and a unified formulation. Compos Part B Eng
[65] Viola E, Tornabene F, Ferretti E, Fantuzzi N. Soft core plane state structures 2014;58:544–52.
under static loads using gdqfem and cell method. CMES-Comput Model Eng [98] Tornabene F, Viola E. Vibration analysis of spherical structural elements using
2013;94(4):301–29. the GDQ method. Comput Math Appl 2007;53(10):1538–60.
[66] Viola E, Tornabene F, Ferretti E, Fantuzzi N. On static analysis of plane state [99] Tornabene F, Viola E. 2-D solution for free vibrations of parabolic shells using
structures via GDQFEM and cell method. CMES-Comput Model Eng generalized differential quadrature method. Eur J Mech A-Solid 2008;27
2013;94:419–56. (6):1001–25.
[67] Fantuzzi N, Tornabene F. Strong formulation finite element method for [100] Tornabene F, Viola E. Free vibration analysis of functionally graded panels
arbitrarily shaped laminated plates – I theoretical analysis. Adv Aircr Spacecr and shells of revolution. Meccanica 2009;44(3):255–81.
Sci 2014;1:124–42. [101] Tornabene F, Viola E. Free vibrations of four-parameter functionally graded
[68] Fantuzzi N, Tornabene F. Strong formulation finite element method for parabolic panels and shell of revolution. Eur J Mech A-Solid 2009;28
arbitrarily shaped laminated plates – II numerical analysis. Adv Aircr Spacecr (5):991–1013.
Sci 2014;1:143–73. [102] Tornabene F, Fantuzzi N, Bacciocchi M. The strong formulation finite element
[69] Fantuzzi N. New insights into the strong formulation finite element method method: stability and accuracy. Fract Struct Integr 2014;29:251–65.
for solving elastostatic and elastodynamic problems. Curved Layer Struct [103] Viola E, Tornabene F. Vibration analysis of damaged circular arches with
2014;1:93–126. varying cross-section. Struct Integ Dur 2005;1:155–69.
[70] Fantuzzi N, Tornabene F, Viola E. Four-parameter functionally graded cracked [104] Viola E, Tornabene F. Vibration analysis of conical shell structures using GDQ
plates of arbitrary shape: a GDQFEM solution for free vibrations. Mech Adv method. Far East J Appl Math 2006;25:23–39.
Mater Struct 2016;23:89–107. [105] Viola E, Dilena M, Tornabene F. Analytical and numerical results for vibration
[71] Fantuzzi N, Bacciocchi M, Tornabene F, Viola E, Ferreira AJM. Radial basis analysis of multi-stepped and multi-damaged circular arches. J Sound Vib
functions based on differential quadrature method for the free vibration of 2007;299:143–63.
laminated composite arbitrary shaped plates. Compos Part B Eng [106] Marzani A, Tornabene F, Viola E. Nonconservative stability problems via
2015;78:65–78. generalized differential quadrature method. J Sound Vib 2008;315:176–96.
[72] Fantuzzi N, Tornabene F, Viola E, Ferreira AJM. A strong formulation finite [107] Tornabene F. Free vibration analysis of functionally graded conical,
element method (SFEM) based on RBF and GDQ techniques for the static and cylindrical shell and annular plate structures with a four-parameter power-
dynamic analyses of laminated plates of arbitrary shape. Meccanica law distribution. Comput Methods Appl Mech Eng 2009;198:2911–35.
2014;49:2503–42. [108] Tornabene F, Viola E, Inman DJ. 2-D differential quadrature solution for
[73] Mohanty JR, Verma BB, Ray PK. Evaluation of overload-induced fatigue crack vibration analysis of functionally graded conical, cylindrical shell and annular
growth retardation parameters using an exponential model. Eng Fract Mech plate structures. J Sound Vib 2009;328:259–90.
2008;75:3941–51. [109] Viola E, Tornabene F. Free vibrations of three parameter functionally graded
[74] Mohanty JR, Verma BB, Ray PK. Prediction of fatigue life with interspersed parabolic panels of revolution. Mech Res Commun 2009;36:587–94.
mode-I and mixed-mode (I and II) overloads by an exponential model: [110] Tornabene F, Marzani A, Viola E, Elishakoff I. Critical flow speeds of pipes
extensions and improvements. Eng Fract Mech 2009;76:454–68. conveying fluid by the generalized differential quadrature method. Adv Theor
[75] Erdogan F, Sih GC. On the crack extension in plates under plane loading and Appl Mech 2010;3:121–38.
transverse shear. J Basic Eng 1963;85:519–27. [111] Tornabene F. 2-D GDQ solution for free vibrations of anisotropic doubly-
[76] Sih GC. Strain-energy-density factor applied to mixed mode crack problems. curved shells and panels of revolution. Compos Struct 2011;93:1854–76.
Int J Fract 1974;10(3):305–21. [112] Tornabene F. Free vibrations of laminated composite doubly-curved shells
[77] Kong XM, Schluter N, Dahl W. Effect of triaxial stress on mixed-mode and panels of revolution via the GDQ method. Comput Methods Appl Mech
fracture. Eng Fract Mech 1995;52(2):379–88. Eng 2011;200:931–52.
[78] Theocaris PS, Andrianopoulos NP. The T-criterion applied to ductile fracture. [113] Tornabene F. Free vibrations of anisotropic doubly-curved shells and panels
Int J Fract 1982;20:R125–30. of revolution with a free-form meridian resting on Winkler-Pasternak elastic
[79] Ukadgaonker VG, Awasare PJ. A new criterion for fracture initiation. Eng Fract foundations. Compos Struct 2011;94:186–206.
Mech 1995;51(2):265–74. [114] Tornabene F, Liverani A, Caligiana G. General anisotropic doubly-curved shell
[80] Khan SMA, Khraisheh MK. A new criterion for mixed mode fracture initiation theory: a differential quadrature solution for free vibrations of shells and
based on the crack tip plastic core region. Int J Plast 2004;20:55–84. panels of revolution with a free-form meridian. J Sound Vib 2012;331:
[81] Theocaris PS, Andrianopoulos NP. The mises elastic-plastic boundary as the 4848–69.
core region in fracture criteria. Eng Fract Mech 1982;16(3):425–32. [115] Ferreira AJM, Viola E, Tornabene F, Fantuzzi N, Zenkour AM. Analysis of
[82] Plank R, Kuhn G. Fatigue crack propagation under non-proportional mixed sandwich plates by generalized differential quadrature method. Math Prob
mode loading. Eng Fract Mech 1999;62:203–29. Eng 2013:1–22. ID 964367.
[83] Bhattachary A. Analysis of crack growth rate based on rotation at the crack tip [116] Tornabene F, Ceruti A. Free-form laminated doubly-curved shells and panels
plastic zone. Int J Fract 1997;83:159–65. of revolution resting on Winkler–Pasternak elastic foundations: a 2-D GDQ
[84] Weber W, Willner K, Kuhn G. Numerical analysis of the influence of crack solution for static and free vibration analysis. World J Mech 2013;3:1–25.
surface roughness on the crack path. Eng Fract Mech 2010;77:1708–20. [117] Tornabene F, Ceruti A. Mixed static and dynamic optimization of four-
[85] Joyce JA, Link RE, Roe C, Sobotka JC. Dynamic and static characterization of parameter functionally graded completely doubly-curved and degenerate
compact crack arrest tests of navy and nuclear steels. Eng Fract Mech shells and panels using GDQ method. Math Prob Eng 2013:1–33. ID 867079.
2010;77:337–47. [118] Tornabene F, Fantuzzi N, Viola E, Ferreira AJM. Radial basis function method
[86] Zaikin AD. Crack stabilization in a brittle body using stiffeners. J Appl Mech applied to doubly-curved laminated composite shells and panels with a
Technol Phys 2006;47(6):886–91. general higher-order equivalent single layer theory. Compos Part B Eng
[87] Timoshenko S. Theory of elasticity. engineering societies 2013;55:642–59.
monographs. McGraw-Hill book Company; 1934. [119] Tornabene F, Viola E, Fantuzzi N. General higher-order equivalent single layer
[88] Zhong H, He Y. Solution of poisson and laplace equations by quadrilateral theory for free vibrations of doubly-curved laminated composite shells and
quadrature element. Int J Solids Struct 1998;35(21):2805–19. panels. Compos Struct 2013;104:94–117.
 N. Fantuzzi et al. / Composite Structures 145 (2016) 162–185 185

[120] Fantuzzi N, Tornabene F, Viola E. Generalized differential quadrature finite [132] Bacciocchi M, Eisenberger M, Fantuzzi N, Tornabene F, Viola E. Vibration
element method for vibration analysis of arbitrarily shaped membranes. Int J analysis of variable thickness plates and shells by the generalized differential
Mech Sci 2014;79:216–51. quadrature method. Compos Struct 2016;89:187–218.
[121] Tornabene F, Fantuzzi N, Bacciocchi M. The local GDQ method applied to [133] Tornabene F, Fantuzzi N, Bacciocchi M. Higher-order structural theories for
general higher-order theories of doubly-curved laminated composite shells the static analysis of doubly-curved laminated composite panels reinforced
and panels: the free vibration analysis. Compos Struct 2014;116:637–60. by curvilinear fibers. Thin Wall Struct 2016;102:222–45.
[122] Tornabene F, Fantuzzi N, Bacciocchi M. Free vibrations of free-form doubly- [134] Brischetto S, Tornabene F, Fantuzzi N, Viola E. 3D exact and 2D generalized
curved shells made of functionally graded materials using higher-order differential quadrature models for free vibration analysis of functionally
equivalent single layer theories. Compos Part B Eng 2014;67:490–509. graded plates and cylinders, Meccanica 2016, in press. doi: http://dx.doi.org/
[123] Tornabene F, Fantuzzi N, Viola E, Carrera E. Static analysis of doubly-curved 10.1007/s11012-016-0361-y.
anisotropic shells and panels using CUF approach, differential geometry and [135] Tornabene F. General higher order layer-wise theory for free vibrations of
differential quadrature method. Compos Struct 2014;107:675–97. doubly-curved laminated composite shells and panels, Mech Adv Mater
[124] Tornabene F, Fantuzzi N, Viola E, Reddy JN. Winkler-Pasternak foundation Struct 2016, in press. doi: http://dx.doi.org/10.1080/15376494.2015.
effect on the static and dynamic analyses of laminated doubly-curved and 1121522.
degenerate shells and panels. Compos Part B Eng 2014;57:269–96. [136] Tornabene F, Fantuzzi N, Viola E. Inter-laminar stress recovery procedure for
[125] Viola E, Rossetti L, Fantuzzi N, Tornabene F. Static analysis of functionally doubly-curved, singly-curved, revolution shells with variable radii of
graded conical shells and panels using the generalized unconstrained third curvature and plates using generalized higher-order theories and the local
order theory coupled with the stress recovery. Compos Struct 2014;112: GDQ method, Mech Adv Mater Struct 2016, in press. doi: http://dx.doi.org/10.
44–65. 1080/15376494.2015.1121521.
[126] Brischetto S, Tornabene F, Fantuzzi N, Bacciocchi M. Refined 2D and exact 3D [137] Tornabene F, Fantuzzi N, Bacciocchi M. The local GDQ method for the natural
shell models for the free vibration analysis of single- and double-walled frequencies of doubly-curved shells with variable thickness: a general
carbon nanotubes. Technologies 2015;3:259–84. formulation, Compos Part B Eng 2016, in press. doi: http://dx.doi.org/10.
[127] Tornabene F, Brischetto S, Fantuzzi N, Viola E. Numerical and exact models 1016/j.compositesb.2016.02.010.
for free vibration analysis of cylindrical and spherical shell panels. Compos [138] Viola E, Rossetti L, Fantuzzi N, Tornabene F. Generalized stress-strain
Part B Eng 2015;81:231–50. recovery formulation applied to functionally graded spherical shells and
[128] Tornabene F, Fantuzzi N, Bacciocchi M, Viola E. Accurate inter-laminar panels under static loading, Compos Struct 2016, in press. doi: http://dx.doi.
recovery for plates and doubly-curved shells with variable radii of curvature org/10.1016/j.compstruct.2015.12.060.
using layer-wise theories. Compos Struct 2015;124:368–93. [139] Williams ML. On the stress distribution at the base of a stationary crack. J
[129] Tornabene F, Fantuzzi N, Bacciocchi M, Viola E. A new approach for treating Appl Mech Trans ASME 1957;24(1):109–14.
concentrated loads in doubly-curved composite deep shells with variable [140] Irwin G. Analysis of stresses and strains near the end of a crack traversing a
radii of curvature. Compos Struct 2015;131:433–52. plate. J Appl Mech 1957;24:361–4.
[130] Tornabene F, Fantuzzi N, Bacciocchi M, Viola E. Higher-order theories for the [141] Henshell RD, Shaw KG. Crack tip finite elements are unnecessary. Int J Numer
free vibration of doubly-curved laminated panels with curvilinear reinforcing Methods Eng 1975;9:495–507.
fibers by means of a local version of the GDQ method. Compos Part B Eng [142] Barsoum RS. On the use of isoparametric finite elements in linear fracture
2015;81:196–230. mechanics. Int J Numer Methods Eng 1976;10:25–37.
[131] Tornabene F, Fantuzzi N, Bacciocchi M, Viola E. Effect of agglomeration on the [143] Viola E, Tornabene F, Fantuzzi N. DiQuMASPAB Software. DICAM Department,
natural frequencies of functionally graded carbon nanotube-reinforced Alma Mater Studiorum University of Bologna. <http://software.dicam.unibo.
laminated composite doubly-curved shells. Compos Part B Eng 2015;89: it/diqumaspabproject>.
187–218.