Engineering Structures 133 (2017) 24–32

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Experimental and numerical fatigue crack growth of an aluminium pipe

repaired by composite patch
H. Zarrinzadeh, M.Z. Kabir ⇑, A. Deylami
Department of Civil and Environmental Engineering, Amirkabir University of Technology, Tehran, P.O. Box: 158754413, Iran

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

Article history: In this study a cylindrical cracked aluminium pipe is considered. Fatigue crack growth behaviour of the
Received 18 June 2016 pipe is observed through experimental tests. Stress intensity factors are computed for the pipe with an
Revised 8 December 2016 inclined crack under axial tensile load. Fatigue crack trajectory and also the crack growth curves versus
Accepted 10 December 2016
number of cycles of load, are extracted. Validation of results is then achieved through the extended finite
element method (XFEM). A stand-alone MATLAB programming package is developed to study such struc-
tures with 3D degenerated shell elements. The cracked pipe is finally repaired by glass/epoxy polymer
Keywords:
composite and the effect of the patch is observed on the extension of fatigue life through experimental
Crack growth experiment
XFEM crack growth
tests and the XFEM framework.
Patch repaired pipe Ó 2016 Elsevier Ltd. All rights reserved.
Stress intensity factors

1. Introduction
Crack initiation may occur in different structures such as metal
or concrete elements. Fatigue is one of the most common kinds of
loading which causes cracking initiation and propagation in struc-
tural elements. These structural parts can be found in building area
or industry. For example, bridges which are utilized for providing
passage of vehicles or trains, are subjected to fatigue loading that
may cause damage in the bridge elements. Pipelines are another
example of strategic infrastructures which are used for energy
transferring. They may contain fluids such as gas, petroleum or
water while pressure changes of the fluid would lead to internal
fatigue loading in the pipe.
Most of the aforementioned structures can be considered as 3D
Shell elements. While this assumption takes into account the three
dimensional geometry of the parts, the computational cost will
decrease compared with 3D solid element assumption.
Different methods are proposed to analyze shell elements
which can be categorized as analytical, numerical and experimen-
tal studies. Erdogan et al. [1,2] studied on the cracked panels using
an analytical formulation for the fracture parameters such as stress
intensity factor. Other closed-form expressions for SIFs are
presented by Zahoor [3], Sanders [4] and Forman [5] for cracked
cylindrical pipes. Zárate et al. [6] presented a framework to update
and predict crack length as a function of the number of cycles in
structural elements subjected to fatigue.
⇑ Corresponding author.
E-mail address: mzkabir@aut.ac.ir (M.Z. Kabir).
Analytical methods are mostly capable to solve just simple geo-
metric problems with particular loading and boundary conditions.
Of course, it should be mentioned that the analytical procedures
are basis of the recent powerful numerical techniques. Finite ele-
ment method is one of these techniques that can solve varieties
of simple or complex engineering problems. Structures can be
modeled and analyzed under arbitrary loading and boundary con-
ditions through this framework. Lam et al. [7] studied on a cracked
steel circular tube repaired by FRP patching through FEM method.
Tong and Sun [8–11] have studied the effect of curvature existence
in elements on the adhesive stress and fracture toughness. Pavlou
et al. [12] proposed a new methodology to simulate the crack tra-
jectories under mixed-mode fatigue loading through FEM method.
The conventional finite element method remains simple until
there is no discontinuity in the model. Discontinuities in elements
would lead to singularities that make solving the problem more
difficult. For instance, in problems containing cracked parts the
generated mesh should be in a way that the crack body coincides
the element edges. Singular elements should also be used as crack
tip elements. This would lead to an irregular mashed part of the
structure. The problem becomes more complex when crack propa-
gation needs to be considered. In the case of crack propagation
such as fatigue crack growth problems the crack body and crack
tip will not be coincided with the element edges after growing
the crack, so a new mesh generation is needed for the model in
each step of solving the problem. This procedure is done in a work
by Ghaffari and Hosseini-Toudeshky [13]. They studied on crack
growth of a steel pipe under fatigue loading with and without
FRP patching through FEM method. An automated re-meshing

http://dx.doi.org/10.1016/j.engstruct.2016.12.011
0141-0296/Ó 2016 Elsevier Ltd. All rights reserved.
 H. Zarrinzadeh et al. / Engineering Structures 133 (2017) 24–32 25

technique is performed through ANSYS Parametric Design Lan-

guage (APDL) to detect the new geometry of the crack in each step
of crack growth and assign adaptive mesh to the cracked panel.
All the deficiencies of the conventional finite element method
resulted to appear a new technique known as an extended finite
element method (XFEM) in which a regular mesh generation can
be used in even cracked parts. The elements in this method can
be cut in arbitrary manners and just the formulation of the ele-
ments around the crack is modified in a way that the singularity
of the problem is taken into account. Areias and Belytschko
[14–16] worked on cracked shells by XFEM method considering
the material nonlinearities without applying enrichment functions
of crack tips. Bayesteh and Mohammadi [17] have extended Areias
[14] works to study cracked shells applying Mindlin-Reissner Fig. 1. Schematic of a degenerated shell element, nodal points and coordinate
theory, while they also considered the effect of the crack tip systems.
enrichment function. Wyart et al. [18] applied extended finite ele-
ment method to analyze cracks in aircraft thin walled structures.
T
Experimental testing is another way of studying problems. It is at node i in the middle surface. fl3i m3i n3i g is related to direc-
obvious that every structural part cannot be tested experimentally tion cosine vector v 3i .
due to time and cost aspects, but the simplest experimental test
can be utilized to validate the results of the numerical method. 2.2. XFEM formulation
After a damage or crack is initiated in an element, affordable
techniques of repairing are required for the damaged parts. Tradi- In XFEM formulation, the displacement field is divided into two
tional techniques like using bolts or welding have some disadvan- parts
tages such as vulnerability of welding to fatigue loading, high
u ¼ uFE þ uENR ð2Þ
stress concentrations near bolts and increasing the total weight
of the structure after using metal repairs. Among the new tech- in which u FE
is the conventional finite element displacement.
niques of repairing, polymer composite materials are a proper n o
X
n
alternatives which cover the aforementioned deficiencies of tradi- uFE ðxÞ ¼ Ni ðn; gÞ u^ i  Ma ^i e2i
^ i e1i þ M b ð3Þ
tional retrofitting methods. Glass/epoxy polymer composite is used i¼1
in this research as a repair patch. Gandhi et al. [19] studied on
Five degrees of freedom are considered in the conventional shell
fatigue crack growth in stiffened steel tubular joints in seawater
formulation which consist of three nodal displacements u ^ i and two
environment. Kabir and Nazari [20–23] experimentally studied
local rotations a ^i with respect to local orthonormal vectors e2i
^i; b
on a cracked cylindrical steel pipe under compressive loading.
The results were also compared with FEM models for the patched and e1i .
an un-patched pipe. uENR is related to the enriched part of approximation. To model
In this study a cylindrical cracked aluminium pipe is studied by weak or strong discontinuities in XFEM framework, one need to
XFEM method and then validated by experimental tests. Fatigue incorporate two types of enrichment functions into displacement
crack growth analysis of the pipe is performed through the approximation. The first type of enrichment function which is used
extended finite element method assuming 3D degenerated shell to present the discontinuity across the crack, is the Heaviside step
elements. The effect of glass/epoxy patch repair is then observed. function. Dolbow introduced this function to simplify the repre-
The transparency of the patch makes it possible to trace the crack sentation of crack away from the tip. The Heaviside function is
 
trajectory in the lower panel. Besides developing a stand-alone 1 abov e the crack
XFEM package in the MATLAB programming software, similar test hðxÞ ¼ ð4Þ
1 below the crack
specimens are provided to experimentally validate the numerical
results. The Heaviside function comes in the degenerated shell formula-
tion as below [17]
2. Numerical formulation X
n n
uHeav iside ðxÞ ¼ Ni ðn; gÞ ðHð/ðxÞÞ  Hð/ðxi ÞÞÞ
2.1. Degenerated shell element i¼1
 o
 ai  Maai  e1i þ Mabi  e2i ð5Þ
Using the so-called degenerated shell element formulation,
computations are done similarly in a planar scheme, while the where ai is the vector of enriched displacement degrees of freedom
3D properties of element are kept. Fig. 1 shows schematic of a at the mid-surface of the shell and aai and abi are rotations with
degenerated shell element, nodal points and coordinate systems. respect to e2 and e1 respectively.
The geometry of an element is described as The second type of functions called crack-tip enrichment func-
8 9 88 9 8 99 tions, is usually derived from the asymptotic analytical solution.
>
<x> = X8 >
<>
< xi >
= 1 >
< l3i >
=>
= They are used to represent the singularity of stress field near the
y ¼ Ni ðn; gÞ yi þ M i ðfÞ m3i ð1Þ
>
: > ; i¼1 >
:>
: > ; 2 >
: >
;>
; crack tip. The crack tip enrichment functions are used in a similar
z zi n3i formulation as shown in Eq. (5) for shell elements, which consist of
in-plane and out-of-plane enrichment functions as below [17]
where n; g are the two curvilinear coordinates in the middle of
The enrichment functions for in-plane degrees of freedom are
plane of shell and f is a linear coordinate in the thickness direction.
         
N i ðn; gÞ are shape functions in element plane directions, pffiffiffi h pffiffiffi h pffiffiffi h pffiffiffi h
Fðr; hÞ ¼ r sin ; r cos ; r sin sinðhÞ; r cos sinðhÞ
fxi yi zi gT is the Cartesian coordinate. Mi ðfÞ ¼ f t2i is the one 2 2 2 2
dimensional shape function in f direction and t i is the thickness ð6Þ
26 H. Zarrinzadeh et al. / Engineering Structures 133 (2017) 24–32

The above functions are also used to enrich the rotational the analysis computations for millions of cycles, after calculating
degrees of freedom. SIFs for one step of loading, a predefined crack growth increment
          Da is assumed, then the number of cycles DN related to this Da is
pffiffiffi h pffiffiffi h pffiffiffi h pffiffiffi h
Rðr; hÞ ¼ r sin ; r cos ; r sin sinðhÞ; r cos sinðhÞ determined with respect to material constants c; m.
2 2 2 2
ð7Þ

The other degree of freedom related to out-of-plane displacement 3. Experiments

u3 is enriched with the following function
  
pffiffiffi h
Gðr; hÞ ¼ r sin
2 A cylindrical pipe made of 6063 aluminium alloy with a
where r; h are the polar coordinates in the local crack tip coordinate
system. The crack tip is the origin of the system and h ¼ 0 is parallel
to the crack.
Fig. 2 shows the enriched nodes with the Heaviside and crack-
tip enrichment functions around the crack body.
Stress intensity factors (SIFs) in the current shell model are
calculated through the use of J-Integral method [24]. To perform
a fatigue crack growth analysis, the Paris law is used
Da
¼ cðDKÞm
DN
where Da is crack growth increment, DN the increment for number
of cycles, DK the range of SIF due to maximum and minimum load
value and c; m are material constants for Paris equation. To facilitate
3.1. Specimens’ geometry
ð8Þ
through-the-thickness crack is considered. The pipe has a length
of 400 mm with internal and external diameters of 86.5 mm and
90.2 mm, respectively. Maximum tensile stress of 60 Mpa with a
load ratio R ¼ 0:1 applies to pipe. The initial crack is placed circum-
ferentially in the middle of the pipe. As the wire-cut technique
could not be used to make such a crack in the pipe, Electrical Dis-
charge Machining (EDM) technique is utilized as an alternative
method. In this technique pure copper electrode with thickness
of 0.18 mm has been prepared to be used in the spark procedure.
ð9Þ To maximize the accuracy of inserting a proper initial crack, the
copper electrode was modeled in a computer designing program
and then this shape was extracted from a copper sheet by a CNC
laser cutting machine. Fig. 3 shows the copper electrode and the
aluminium pipe in the EDM machine. The initial length of the crack

Fig. 2. Definition of enriched nodes and illustration of level-set functions.

Fig. 3. (a) Copper electrode used to insert crack in the spark procedure. (b) The aluminium pipe in the EDM machine.
 H. Zarrinzadeh et al. / Engineering Structures 133 (2017) 24–32 27

Fig. 4. (a) DARTEC 9600 testing machine with the installed fixture. (b) Schematic of the lower part of the designed fixture to apply tensile load on the pipe.

after applying a pre-cracking to provide a sharpened crack tip is in which k is a crack length parameter equal to ðk ¼ a=2eÞ and C is
a0 = 30 mm. 8 9
p k2  0:0293k3
< 1 þ 16 k 6 1=
C ¼  pffiffi 0:5
0:885
3.2. Test preparation : 2 2k þ 0:179 k > 1;
p k

The specimens are tested with a DARTEC 9600 fatigue testing The other parameters e and gðaÞ are defined as below
machine. As there was no prefabricated fixture to perform a tensile  
t 1=2
fatigue loading test on a cylindrical pipe, a particular fixture shown e2 ¼ ½12ð1  m2 Þ
R
in Fig. 4 has been designed and built. To apply a uniform tensile
stress, inside of the pipe is filled with a solid circular steel filler 0 12
at both ends. Outside surfaces of both ends are then gripped with pffiffiffi 1  a cot a
gðaÞ ¼ 2 2@1 þ pffiffiffi h pffiffiffiiA
two other pieces which hold the pipe edges fixed during the load- 2a cot a þ 2a cot ðp  aÞ= 2
ing procedure. The whole pieces are connected to the machine
with two wedges at both ends. in which t and R are the thickness and radius of the pipe,
respectively. The comparison of results is presented in Fig. 5. By this
4. Results and discussion work, besides validating the prepared XFEM program, values
of DK in Paris equation are computed so the material constants c
4.1. Fatigue crack growth in a pipe with a circumferential crack and m can be predicted by a regression technique like the
least square method. The computed material constants are
To extract the Paris constants c and m, experimental results c ¼ 1:4125  109 ; m ¼ 2:3.
of the cracked pipe with known magnitudes of Da; DN and DK Fig. 6 shows the crack path in the test specimen while the XFEM
are required in different steps of fatigue crack growth. To extract modeled crack growth in the pipe is depicted in Fig. 7.
values of DK for each step of crack growth, analytical formula-
tions or numerical modeling techniques can be applied. Here 4.2. Fatigue crack growth in a pipe with an inclined crack
the specimens are modeled in the prepared MATLAB code (called
as Shell MXFEM), to simulate the fatigue crack growth through The cylindrical aluminium pipe with the same geometry
the XFEM framework. The results of SIFs are also verified by described in the previous section is tested with an inclined crack

the analytical expression proposed by Forman [5] who proposed of angle a ¼ 45 and initial length of a0 ¼ 25:5 mm. The fatigue
the following formulation to calculate the stress intensity factor crack trajectory and also the a-N curves are extracted from the test
for a cylindrical pipe subjected to a tension axial loading with a specimen, while they are also validated with XFEM results.
circumferential crack of length 2a and central angle of 2a for the Fig. 8 shows the calculated values of equivalent stress intensity
crack factors of the XFEM model. It should be noted that an inclined
 1=2 initial crack in the pipe would lead to happen a mixed-mode frac-
I pffiffiffiffiffiffi
KI ¼ r pa ture condition, so both KI and KII values should be considered to
2pa calculate DK into Paris equation. The proposed criterion by
Considering r as the axial tension stress applied at the ends of Richard[25] has been used to extract the equivalent stress intensity
the pipe. I is defined as factor as below:
h pffiffiffii qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
KI 1
I ¼ e1 a2 gðaÞ þ pk1 C 2  2 2 K eq ¼ þ K 2I þ 4ðaK II Þ2
2 2
28 H. Zarrinzadeh et al. / Engineering Structures 133 (2017) 24–32

Fig. 5. XFEM and analytical values of SIFs versus different crack lengths.

Fig. 6. Circumferential crack in the pipe (a) Initial crack shape. (b) Crack growth under tensile axial stress.

The same expression is suggested to be used by Ghaffari and
Hosseini-Toudeshky [13] who modeled fatigue crack growth in a
repaired pipe by FEM method.
After Computing SIFs in each step of crack growth, the a-N curve
from the test experiments can be compared with the XFEM output.
Fig. 9 shows the half crack length versus the number of load cycles
for the pipe with the inclined crack. Reconsidering the logged data
file of applied load cycles for the test specimen, it is observed that
the maximum applied load has randomly exceeded the input value
due to a difficulty in the hydraulic system of the testing machine so
the experimental results of this test may prone to errors. Such a
difficulty would lead to faster crack growth in the experimental
test so the numerical predicted life is above the experimental
result.
A crack trajectory in the pipe with an inclined initial flaw is
another issue in the experimental test that can be compared with
XFEM models. Fig. 10 depicts the crack path in the test specimen
and the predicted crack trajectory by the numerical method is
shown by a dashed line in the figure. The results of the XFEM
model are in a very good agreement with the experiment. The
inclined crack in the pipe behaves as a mixed-mode fracture prob-
lem, so a sudden turning occurs in the crack growth path. This
turning happens at the first step of crack growth both in experi-
mental test and XFEM model as shown in Fig. 10. By increasing
the crack, since the load is in the axial direction, the first fracture
mode becomes dominant and the crack tends to grow in a perpen-
dicular path to the loading direction. So a horizontal extension of
the crack can be observed in this figure.
4.3. Fatigue crack growth in a patch repaired pipe
The pipe with a circumferential crack is now repaired with
glass/epoxy polymer composite. The transparency of the glass/
epoxy patch makes it possible to trace the crack trajectory in the
aluminium pipe. The pipe is wrapped in a length of 100 mm with
the patch. The surface of the pipe is prepared using the ASTM
D2651 specification. It should be noted the surface is degreased
by non-etching alkaline cleaner. The cracked aluminium pipe is
then wrapped by 2 layers of unidirectional fiberglass using epoxy
resin. The fibers are aligned in the longitudinal axis of the pipe
 H. Zarrinzadeh et al. / Engineering Structures 133 (2017) 24–32 29

Fig. 7. Schematic of the XFEM model of the circumferentially cracked pipe.

Fig. 8. Values of equivalent stress intensity factor versus different crack lengths of the pipe with an inclined crack.

which is perpendicular to the crack direction. The specimen is
finally cured under curing temperature of 50 °C for at least 15 h.
Fig. 11 shows the installed repaired pipe in the testing machine.
Mechanical properties of the polymer composites are measured
according to the ASTM D7205 standard as below:
E11 ¼ 31 GPa; E22 ¼ 4 GPa; G12 ¼ 3:6 GPa; m12 ¼ 0:26;
total thickness ¼ 0:6 mm
Fig. 12 compares the values of the SIF for the circumferential
cracked pipe with and without patch repair.
As shown in this figure SIF values are reduced after repairing
the pipe with the polymer composite. This reduction is more sig-
nificant for a longer length of crack which means that the presence
of the patch would be more effective as more as the crack grows.
A comparison between the results of crack length versus num-
ber of cycles for the cracked pipe with and without patch repair is
presented in Fig. 13. A considerable fatigue life extension is
observed for the repaired pipe which is due to the reduction in
SIF values in each step of crack growth after using the patch repair.
This figure also exhibits the results of XFEM model of the pipe. As
30 H. Zarrinzadeh et al. / Engineering Structures 133 (2017) 24–32

Fig. 9. Experimental and XFEM results of crack growth versus number of cycles for a pipe with inclined crack.

Fig. 10. (a) Schematic of crack deformation in the pipe with inclined crack. (b) Experimental and XFEM crack growth path.

shown in the figure a discrepancy is observed between the

experimental and XFEM results. This difference may be explained
by debonding of the patch in the experimental test which is not
considered in XFEM simulation. The numerical results have a
good agreement with the experimental output until no debonding
has occurred between the patch and the pipe. At the half crack
length of about a = 20 mm the debonding effect can be observed
in the trend of the experimental curve but in numerical method
the patch is assumed to be tied to the pipe surface. In future works
numerical models can be developed which consider the debonding
effect. Such a model should apply cohesive elements between
the patch and the pipe which leads to a nonlinear and iterative
model.

5. Conclusion

In the current study experimental tests has been performed to

Fig. 11. Cracked pipe repaired by glass/epoxy polymer composite installed in the
testing machine.
study the behaviour of crack growth in a cracked aluminium pipe
under tensile fatigue loading. As the wire-cut technique could
not be used to make such a crack in the pipe, Electrical Discharge
Machining (EDM) technique is utilized as an alternative method
to insert the initial crack in the pipe. The specimens are tested with
a DARTEC 9600 fatigue testing machine. As there was no prefabri-
cated fixture to perform a tensile fatigue loading test on a cylindri-
cal pipe, a particular fixture has been designed and built.
 H. Zarrinzadeh et al. / Engineering Structures 133 (2017) 24–32
Fig. 12. Comparison of stress intensity factors for repairs and unrepaired pipe.
Fig. 13. Experimental and XFEM results of crack growth versus number of cycles for a the repaired and unrepaired pipe.
First a circumferentially cracked aluminium specimen is tested
to extract the material constants required in the Paris equation for
fatigue crack growth modeling. The pipe with an inclined crack is
then tested under mixed-mode fracture condition. The results of
the crack growth versus the number of applied cycles of load and
also the fatigue crack trajectories are extracted from the experi-
mental tests. The results are also validated by a XFEM code devel-
oped in MATLAB software. 3D degenerated shell elements are
utilized to model such a cracked structure. It is shown that the
numerical method can model the crack growth behaviour of the
pipe in a very good agreement with experimental results. To
increase the strength of the structure, the cracked pipe is wrapped
by glass/epoxy polymer composite. The transparency of the patch
makes it possible to trace the crack trajectory in the lower alu-
minium panel. It is observed that the stress intensity factor
decreases after repairing the pipe with polymer composites and
the fatigue life of the repaired pipe is extended considerably.
References
[1] Erdogan F, Sih G. On the crack extension in plates under plane loading and
transverse shear. J Basic Eng 1963;85:519–25.
[2] Erdogan F, Kibler J. Cylindrical and spherical shells with cracks. Int J FractMech
1969;5:229–37.
31
[3] Zahoor A. Closed form expressions for fracture mechanics analysis of cracked
pipes. J Pressure Vessel Technol 1985;107:203–5.
[4] Sanders JL. Circumferential through-cracks in cylindrical shells under tension. J
Appl Mech 1982;49:103–7.
[5] Forman R, Hickman J, Shivaskumar V. Stress intensity factors for
circumferential through cracks in hollow cylinders subjected to combined
tension and bending loads. Eng Fract Mech 1985;21:563–71.
[6] Zárate BA, Caicedo JM, Yu J, Ziehl P. Bayesian model updating and prognosis of
fatigue crack growth. Eng Struct 2012;45:53–61.
[7] Lam C, Cheng J, Yam C. Finite element study of cracked steel circular tube
repaired by FRP patching. Procedia Eng 2011;14:1106–13.
[8] Tong L, Sun X. Shape optimization of bonded patch to cylindrical shell
structures. Int J Numer Meth Eng 2003;58:793–820.
[9] Tong L, Sun X. Optimization of ply drop-offs in bonded patch to cylindrical
shell structures. J Compos Mater 2003;37:1289–314.
[10] Tong L, Sun X. Nonlinear stress analysis for bonded patch to curved thin-
walled structures. Int J Adhes Adhes 2003;23:349–64.
[11] Sun X, Tong L. Fracture toughness analysis of inclined crack in cylindrical shell
repaired with bonded composite patch. Compos Struct 2004;66:639–45.
[12] Pavlou D, Labeas G, Vlachakis N, Pavlou F. Fatigue crack propagation
trajectories under mixed-mode cyclic loading. Eng Struct 2003;25:869–75.
[13] Ghaffari MA, Hosseini-Toudeshky H. Fatigue crack propagation analysis of
repaired pipes with composite patch under cyclic pressure. J Pressure Vessel
Technol 2013;135:031402.
[14] Areias P, Belytschko T. Non-linear analysis of shells with arbitrary evolving
cracks using XFEM. Int J Numer Meth Eng 2005;62:384–415.
[15] Areias PM, Song J, Belytschko T. Analysis of fracture in thin shells by
overlapping paired elements. Comput Methods Appl Mech Eng 2006;195:
5343–60.
[16] Areias PM, Belytschko T. Analysis of finite strain anisotropic elastoplastic
fracture in thin plates and shells. J Aerospace Eng 2006;19:259–70.
32 H. Zarrinzadeh et al. / Engineering Structures 133 (2017) 24–32
[17] Bayesteh H, Mohammadi S. XFEM fracture analysis of shells: the effect of crack
tip enrichments. Comput Mater Sci 2011;50:2793–813.
[18] Wyart E, Coulon D, Pardoen T, Remacle J-F, Lani F. Application of the
substructured finite element/extended finite element method (S-FE/XFE) to
the analysis of cracks in aircraft thin walled structures. Eng Fract Mech
2009;76:44–58.
[19] Gandhi P, Murthy DR, Raghava G, Rao AM. Fatigue crack growth in stiffened
steel tubular joints in seawater environment. Eng Struct 2000;22:1390–401.
[20] Kabir MZ, Nazari A. The study of ultimate strength in notched cylinders
subjected to axial compression. J Constr Steel Res 2011;67:1442–52.
[21] Kabir MZ, Nazari A. Numerical Study on Reinforcing of Thin Walled Cracked
Metal Cylindrical Columns Using FRP Patch. Sci Iran Trans A, Civ Eng
2010;17:407.
[22] Kabir MZ, Nazari A. Experimental and numerical study on the nonlinear
response of notched cylinders under compressive loading. Sci Iran
2012;19:355–65.
[23] Kabir MZ, Nazari A. Enhancing ultimate compressive strength of notch
embedded steel cylinders using overwrap CFRP patch. Appl Compos Mater
2012;19:723–38.
[24] Chau-Dinh T, Zi G, Lee P-S, Rabczuk T, Song J-H. Phantom-node method for
shell models with arbitrary cracks. Comput Struct 2012;92:242–56.
[25] Richard HA, Buchholz FG, Kullmer G, Schöllmann M. 2D-and 3D-mixed mode
fracture criteria. Key Engineering Materials. Trans Tech Publ; 2003. p. 251–60.