Composite Structures 99 (2013) 4861

Contents lists available at SciVerse ScienceDirect

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

Finite element modelling of CFRP/steel double strap joints subjected

to dynamic tensile loadings
Haider Al-Zubaidy a,c, Riadh Al-Mahaidi b,, Xiao-Ling Zhao a
a

Department of Civil Engineering, Monash University, Clayton, VIC 3800, Australia

Faculty of Engineering and Industrial Sciences, Swinburne University of Technology, Hawthorn, VIC 3122, Australia
c
Kerbala University, Kerbala, Iraq
b

a r t i c l e

i n f o

Article history:
Available online 14 December 2012
Keywords:
CFRP sheet
Double-strap joints
Steel plate
FE modelling
Dynamic loadings

a b s t r a c t
This paper reports the numerical simulation of both CFRP/steel double strap joints with 1 and 3 CFRP layers per side at quasi-static and three dynamic tensile loading speeds of 3.35, 4.43 and 5 m/s. Simulations
are implemented using both the implicit and explicit codes respectively using non-linear nite element
(FE) package ABAQUS. In these analyses, failures of both CFRP sheet and adhesive are considered and a
cohesive element is utilised to model the interface. The developed FE models for both types of joints were
validated by comparing their quasi-static and dynamic ndings with those obtained from previous experimental program. This comparison includes four different variables such as the ultimate joint strength,
effective bond length, failure pattern and strain distribution along the bond length. It was found that
FE models proved to be able to predict all these parameters for both quasi-static and dynamic analyses
and their prediction matched well with test results.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
In recent years, the outstanding outcomes of strengthening and/
or upgrading concrete structures using the technique of adhesively
bonded carbon bre reinforced polymer (CFRP) have attracted the
engineers attention to employ the same method for metal structures. However, in general, conducting experimental tests has
many drawbacks. These include cost, time, difculties and limitations in testing full scale members and the difculties in implementing a parametric study on different variables. These
shortcomings highlight the importance of developing nite element models which are capable of predicting the behaviour of
the strengthened and/or upgraded structures. Therefore, nite element analysis (FEA) has attracted an increasing demand to analyse
adhesively bonded joints since the composite materials have become common materials of strengthening and/or upgrading.
Some numerical studies have been successfully carried out to
predict the static and dynamic behaviour and strength of adhesively bonded joints of similar and dissimilar substrates under different loading conditions. Under static tensile loading, the
behaviour and strength of CFRP composite adhesively bonded steel
plates were examined in [16]. Other studies numerically analysed
joints of similar adherends such as steel/steel [7], aluminium/aluminium [8] and composite/composite [911]. In addition, nite
Corresponding author.
E-mail address: ralmahaidi@swin.edu.au (R. Al-Mahaidi).
0263-8223/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compstruct.2012.12.003

element analysis of CFRP composite bonded to simply supported

steel beams under bending were also reported in Refs. [1214]
and analysis for continuous beams were reported in [15]. However,
the strength and behaviour of structural joints such as the single
lap joints and T-joints, which were manufactured using rigid and
elastic adhesives, were experimentally investigated and compared
under static and impact loading [16]. On the other hand, compared
to static loading, the dynamic behaviour and strength of adhesively
bonded joints attracted limited attention in numerical studies.
These investigations included joints of different substrates such
as steel/steel [17], aluminium/aluminium [18] and composite/
composite [19]. Numerical prediction of the dynamic strength
and behaviour of joints of CFRP sheet bonded to steel plates has
not been reported in the literature. To cover this gap in knowledge,
this paper aims at investigating the numerical simulation of CFRP/
double strap joints at quasi-static and the three dynamic loading
speeds of 3.35 m/s, 4.43 m/s and 5 m/s using both implicit and explicit codes in ABAQUS. Results of numerical simulations are compared with experimental ndings.
2. Summary of laboratory work
A total of 160 CFRP/steel double strap joints were prepared and
tested at quasi-static and three dynamic loading speeds of 3.35,
4.43 and 5 m/s and this number included two types of joints with
1 and 3 CFRP layers per side. These joints were formed by bonding
normal modulus CFRP sheet to steel plate using Araldite 420

49

H. Al-Zubaidy et al. / Composite Structures 99 (2013) 4861

Joint

CFRP sheet
5 mm

Adhesive layers

Steel plate
L1

L2
210mm

210mm

(a) Joints used for static tests of 3 CFRP layers and impact tests of 1 CFRP layer per side
CFRP sheet
Joint

Steel tabs
5 mm
5 mm
5 mm

75 mm

75 mm

L1 Adhesive layers

L2

210 mm

210 mm

(b) Joints used for impact tests of 3 CFRP layers per side
15 mm

25 mm

CFRP

Steel plate
G1 G2
G5 G6
G6 G7
G4 G5
G1
G2 G3
G3 G4

L2

50
mm

L1
210 mm

210 mm

(c) Specimens top face view

CFRP

Steel plate
G8

L2
210 mm

G9

50
mm

L1
210 mm

(d) Specimens bottom face view

Fig. 1. A schematic view of the specimens geometry and instrumentation used in the experimental program (not to scale).

adhesive (Huntsman Duxford, UK). In this experimental program,

on average, two or three CFRP/steel samples were tested for both
joints with 1 and 3 CFRP layers. Of these, 62 joints were produced
with 1 CFRP layer per side and 98 joints with 3 CFRP layers. Details
of manufacturing these joints, test procedure and experimental results can be found in the authors previous study [20]. A schematic
view of the specimens geometry and instrumentation used the
experimental program is shown in Fig. 1.
3. Finite element model
Based on dimensionality, it is well known that numerical simulation can be conducted using either 2-D or 3-D modelling and
each has certain advantages and shortcomings. Even though 2-D
modelling is much easier to simulate and the analysis does not require very powerful computers, its results are always less accurate,
particularly when analysing large-scale structures. Conversely,

more precise results are expected from 3-D modelling, although

this is more likely to pose difculties when running on normal
computers. The appearance of such difculties depends on the size
of the analysed structure. In this study, since the dimensions of the
analysed samples are not too large and the analysis can be run
using a normal PC, 3-D modelling was chosen to simulate both
the quasi-static and dynamic analyses for both types of joints (with
1 and 3 CFRP layers per side). This is in order to obtain more accurate results and to enable clear comparisons between the failure
modes for both quasi-static and dynamic loadings.
Three-dimensional models are developed in ABAQUS software
to numerically investigate the effect of increasing the test speed
on the bond between steel plate and CFRP patch using doublestrap joint samples. To clearly highlight this effect, non-linear
quasi-static and dynamic analyses have been carried out using
both ABAQUS implicit and explicit codes respectively. Due to
material and geometry symmetry conditions, only one eighth of

50

H. Al-Zubaidy et al. / Composite Structures 99 (2013) 4861

Fig. 2. Three-dimensional modelling of CFRP/steel double strap joint with its boundary conditions: (a) quasi-static model and (b) dynamic model.

the full-scale specimen is modelled. Consequently, symmetric

boundary conditions are applied to all the nodes belonging to YZ,
XZ and XY Cartesian planes. A typical geometry, nite element
mesh and boundary conditions for both the quasi-static and dynamic analyses are shown in Fig. 2. Failure of both adhesive and
CFRP patch have been included for both quasi-static and dynamic
analyses in order to adequately model the failure of double-strap
joints. For the adhesively-bonded composite technique, it is known
that loads generally transfer from the inner adherend (steel) to the
outer adherend (CFRP patch) via the adhesive layer. This emphasises the importance of the correct simulation of the connection
between the dissimilar materials. Furthermore, the adhesive layer
needs to nely mesh compared to other substrates (steel and
CFRP). For consistency with experiments, displacement is applied
as a boundary condition at the end of the samples steel plate for
the quasi-static simulation of the double strap joints (Fig. 2a),
whereas an impact tensile velocity was applied on the small steel
block, which is tied with the steel plate at the specimens end as
shown in Fig. 2b.

4. Element types
The nite element package ABAQUS has various types of elements. As the double-strap joint consists of three different materials: steel, adhesive and CFRP patch, these various materials are
modelled as three different part instances. Consequently, three
different types of elements have been used to discretize these
part instances. The adhesive layer is meshed with 8-node threedimensional cohesive elements (COH3D8) whereas the steel plate
is modelled by 8-node three dimensional reduced-integration rst
order hourglass control elements (C3D8R). The CFRP patch is
meshed with an 8-node quadrilateral in-plane general-purpose
continuum shell, with reduced integration with hourglass control,
nite membrane strains (SC8R). This type of element, which is
capable of predicting CFRP failure, has been successfully utilised
to simulate CFRP composite by other researchers [2123].
Although each of the eight nodes of this element has three degrees
of freedom, these elements acquire only transition nodal degrees of
freedom. Thus, the main feature of such types of elements is that

H. Al-Zubaidy et al. / Composite Structures 99 (2013) 4861

the continuum shell elements allow a fully three-dimensional

model while they are more computationally attractive than the
standard brick elements because they are able to capture
through-the-thickness shear stress without using one element
per layer [21,24,25]. Furthermore, their ability to capture threedimensional geometry leads to improved accuracy in resolving
contact problems. Despite their visual resemblance to ordinary
three-dimensional elements, these elements maintain constitutive
behaviour and formulation similar to conventional shell elements,
consequently allowing the usage of standard plan stress failure criteria for the composite employed for damage modelling of CFRP
layers [23].
5. Material models
5.1. CFRP sheet
For adhesively-bonded composite materials, it is always expected that failure will occur within the composite materials. This
failure may take place either in the patching material or the bonding material, or both. For more adequate and comparable numerical models, damage to composites must be numerically considered
in the analysis of bre reinforced composite materials used to
strengthen and/or repair metal structures.
The general behaviour of the unidirectional normal modulus
carbon bre reinforced polymer sheet (CFRP) is elastic-brittle
material, as clearly reported by the authors previous study [26].
Among the available material models in ABAQUS software is damage and failure for bre-reinforced composites. Consequently, utilising such a material model facilitates the implementation of
damage initiation and propagation for elastic-brittle materials
with an isotropic behaviour such as the unidirectional normal
modulus bre reinforced polymer CFRP sheet. Therefore, modelling
of the failure and damage of CFRP sheet has been achieved in this
study using this material model, which depends on continuum
damage mechanisms and employs Hashins failure criteria
[27,28]. By adopting this material model, the plasticity of CFRP
composite is always neglected and damage is detected and characterised based on the material stiffness reduction. This material
degradation can be numerically achieved based on Hashins failure
criteria [27,28] which offer numerical simulation of composite
materials damage. Therefore, in this study, to provide a more accurate validation of the numerical models with the experimental results, the CFRP composite damage has been considered during the
quasi-static and dynamic analyses of double-strap joints for both
joints with 1 CFRP layer and 3 CFRP layers per side.
5.2. Adhesive
For both quasi-static and dynamic analyses of double strap
joints, appropriate modelling of the adhesive layer is important
in order to enable correct modelling of the failure of joints. ABAQUS has a special type of element known as a cohesive element
which is more suitable to model the adhesive response and is
applicable for both types of ABAQUS analyses (implicit and explicit). It has been reported in ABAQUS [25] that the cohesive element
is more practical and suitable to model interfaces in composites
and any cases where the integrity and strength of interfaces may
be of interest as well as the behaviour of adhesively-bonded joints.
Furthermore, damage and delamination in composites can also be
successfully predicted using this type of element [22,2931]. By
adopting this element type, it is possible to model damage or crack
initiation and damage evolution leading to eventual failure at the
interface. Therefore, in this study, the adhesive layer (cohesive
zone) is modelled as cohesive elements with fracture mechanism

51

constitutive denitions. It is worth noting that the cohesive zone

must be discretised with a single layer of cohesive elements
through the thickness as specied by the ABAQUS manual [25].
This requirement is based on the denition of cohesive element.
Otherwise, the utilisation of more than one cohesive element via
the adhesive thickness is not recommended because it may cause
unreliable results [32]. According to traction-separation law and
as mentioned in the ABAQUS manual [25], the separation is calculated based on the relative displacement of the top and bottom surface of the cohesive element. Thus the cohesive element thickness
(adhesive thickness) is assumed to be one, or can be calculated
from the nodal coordinates of the cohesive element.
However, generally, failure of adhesive includes crack initiation
and propagation and both can be simulated using the cohesive element. Thus, damage initiation represents the onset of degradation
in the response of the adhesive material and this starts when the
stresses and/or strains full the requirements of the adopted failure criterion. The built-in library of ABAQUS has four different failure criteria for damage initiation under the traction-separation
law. These are maximum nominal stress criterion, maximum nominal strain criterion, quadratic nominal stress criterion and quadratic nominal strain criterion. The rst two criteria assume that
adhesive damage begins only when the maximum nominal stress
or strain reaches the capacity of the adhesive, whereas the last
two criteria consider the combination effect of stresses or strains
on the damage initiation in the adhesive layer. It has been reported
by da Silva et al. [33] that adhesively-bonded joints are subjected
to complex states of stress (shear and peeling stresses) and these
stresses contribute to the adhesive failure. Therefore, in this study,
the mixed mode failure criterion, the quadratic traction damage
initiation criterion (QUADSCRT), which considers both mode I
and mode II failures, is selected. For the QUADSCRT failure criterion, adhesive damage is assumed to initiate when the following
equation is fullled:


2  2  2
tn
ts
tt

1
ton
tos
tot

where tn, ts and tt denote the stresses in three directions of the

adhesive layer (normal, rst and second shear direction). ton ; t os and
tot refer to the peak values of the nominal stresses of adhesive in
three directions layer (normal, rst and second shear direction). n,
s and t represent the directions normal, rst and second shear direction which are parallel to the interface between adhesive and
adherents.
5.3. Steel plate
For the technique of adhesively bonded joints, Hart-Smith [34]
has outlined that the theoretically-calculated effective bond length
of adhesively-bonded joints is signicantly inuenced by the ultimate tensile strength of steel. Therefore, for both quasi-static and
dynamic analyses, steel plate is modelled as elasticplastic material to accurately model double strap joints for joints with 1 CFRP
layer and 3 CFRP layers per side.
6. Material properties
Detailed information about the measured quasi-static and dynamic material properties of CFRP and adhesive and steel plate
can be found in Tables 1 and 2 respectively. The tensile properties,
which have been used in this study to dene the material properties of nite element models, involve the tensile strength, modulus
of elasticity and failure strain of each material. The shear properties of the adhesive layer (cohesive element) are also tabulated in
Table 3. In the case of joints with 3 CFRP layers, experimentally,

52

H. Al-Zubaidy et al. / Composite Structures 99 (2013) 4861

Table 1
Quasi-static and dynamic material properties of the CFRP sheet and Araldite 420 adhesive.
Property

CFRP sheet
Loading speed (m/s)

Tensile strength (MPa)

Tensile modulus (GPa)
Tensile failure strain (%)

Araldite 420
Loading speed (m/s)

3.33  10-5

3.35

4.43

3.33  10-5

3.35

4.43

1935
206.6
0.91

2420
244.2
0.99

2767
250.7
1.13

3108
261.89
1.20

29.00
1.455
9.32

93.25
2.848
4.66

96.06
2.998
4.29

99.42
3.102
4.11

Table 2
Quasi-static and dynamic tensile material properties steel plate.
Property

Loading speed (m/s)

Yield stress (MPa)

Ultimate tensile strength (MPa)
Tensile modulus (GPa)
Ultimate strain (%)

3.33  105

3.35

4.43

371.04
526.27
204.25
19.30

570.30
691.70
212.30
19.06

628.22
743.54
216.26
18.97

673.07
780.60
220.24
18.89

Table 3
Quasi-static and dynamic shear properties of the adhesive layer (cohesive element).
Loading
speed (m/s)

3.33  105
3.35
4.43
5

Shear strength
(MPa)

24.78
66.23
69.49
71.2

Stiffness of the interface (N/mm)

Normal
direction

1st shear
direction

2nd shear
direction

2745
5373
5657
5853

1017
1990
2095
2168

1017
1990
2095
2168

where r(eq)CFRPpatch, t(eq)CFRPpatch, rCFRP, tCFRP, rad and tad, and represent the equivalent tensile strength of the CFRP patch, equivalent
thickness of the CFRP patch, tensile strength of the CFRP sheet,
thickness of the CFRP sheet, tensile stress of the adhesive and thickness of the adhesive layer respectively.
The modulus of elasticity is also determined as below following
the same concept:

EeqCFRPpatch t eqCFRPpatch ECFRP t CFRP Ead t ad

ECFRP tCFRP Ead tad
EeqCFRPpatch
t eqCFRPpatch

where E(eq)CFRPpatch, t(eq)CFRPpatch, ECFRP, tCFRP, Ead and tad and refer to
equivalent modulus of elasticity of the CFRP patch, equivalent thickness of the CFRP patch, modulus of elasticity of the CFRP, thickness
of the CFRP sheet, modulus of elasticity of the adhesive and thickness of the adhesive layer respectively.
7. Comparison of results
7.1. Ultimate joint capacity

the CFRP patch is formed on each side of the double-strap joint by

adhesively bonding three CFRP layers using three adhesive layers.
Therefore, in the current FE simulations, the CFRP patch is assumed
to consist of three CFRP layers and two adhesive layers (the adhesive layers between the CFRP layers). It is also assumed that
the thickness of the two adhesive layers is identical, which is
equal to the thickness of the adhesive layer between the steel
plates and the CFRP patch. As has been mentioned in the experimental study [20], the total joint thickness was measured using a
digital measuring tool (a Mitutoyo Absolute Digimatic Caliper
500-196-20) with accuracy of 0.002 mm. The total patch thickness
is determined by subtracting the steel plate thickness from this
measurement. Consequently, the thickness of the adhesive layer
can be determined using the formula below because the thickness
of the CFRP layer is already known:

t eqCFRPpatch 3  t CFRP 2  t ad

where teq,CFRPpatch, tCFRP and tad denote to equivalent thickness of the

CFRP patch, thickness of the CFRP sheet and thickness of the adhesive layer respectively.
Thus, for simplicity, the material properties of the CFRP patch
are found using the macroscopic material properties technique
and such a method of calculation the material properties has been
also used in another study [1]. Based on this strategy, the material
properties of the CFRP patch are considered to be mainly dependent on the properties of the CFRP sheet and adhesive because
the CFRP patch consists of CFRP sheet and adhesive only. Consequently, the CFRP patch properties are calculated as equivalent
tensile strength, modulus of elasticity and strain. This can be
achieved following the formula:

reqCFRPpatch teqCFRPpatch rCFRP tCFRP rad tad

r t rad tad
reqCFRPpatch CFRP CFRP
teqCFRPpatch

Tables 411 present a clear comparison of the experimentallymeasured ultimate joint capacities and those predicted from FEA
for double-strap joints with 1 and 3 CFRP layers per side. The data
compared are the quasi-static and three impact loadings at speeds
of 3.35 m/s, 4.43 m/s and 5 m/s and for various bond lengths ranging from 10 mm to 100 mm. For joints with 1 CFRP layer, it is evident in Tables 47 that the predicted ultimate joint capacities for
all test speeds and for the different bond lengths are consistent
with the ultimate tensile loads observed from the experimental
test program reported in [20]. The correlation of the tensile failure
loads of varying bond lengths and for the four loading speeds, as
predicted by the FEA and as measured experimentally, is illustrated in Fig. 3. It can be seen that the (PFE/Avg. Pult) ratios range
from 0.846 to 1.003, 0.876 to 1.035, 0.951 to 0.992 and 0.999 to
1.030 for the quasi-static and dynamic test speeds of 3.35 m/s,
4.43 m/s and 5 m/s respectively. Thus, it can be concluded that
each of the quasi-static and dynamic joint strengths for all the
bond lengths are predicted reasonably well.

Table 4
Comparison between the quasi-static experimental and nite element analysis results
for joints with 1 CFRP layer.
Specimen
label

L1 (mm)

CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A

10
20
30
40
50
60
70
80
90
100

L2 (mm)

80
80
80
80
80
80
100
100
115
115

Experiment

Finite element analysis

Avg. Pult (kN)

PFE (kN)

PFE/Avg. Pult

19.84
37.87
45.22
44.06
47.44
46.17
46.33
48.18
45.82
46.73

19.88
32.04
44.30
44.18
44.14
44.15
44.16
44.40
44.18
44.19

1.002
0.846
0.980
1.003
0.930
0.956
0.953
0.922
0.964
0.946

53

H. Al-Zubaidy et al. / Composite Structures 99 (2013) 4861

Table 5
Comparison between the dynamic experimental and nite element analysis results
for joints with 1 CFRP layer at loading speed of 3.35 m/s.
Specimen
label

L1 (mm)

CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A

10
20
30
40
50
60
70
80
90
100

L2 (mm)

80
80
80
80
80
80
100
100
115
115

Experiment

Finite element analysis

Avg. Pult (kN)

PFE (kN)

PFE/Avg. Pult

45.66
63.49
57.77
57.37
56.21
57.17
56.37
57.16
56.65
56.99

47.24
55.62
55.12
55.32
55.00
54.72
54.84
54.76
54.4
54.08

1.035
0.876
0.954
0.964
0.978
0.957
0.973
0.958
0.960
0.949

Table 6
Comparison between the dynamic experimental and nite element analysis results
for joints with 1 CFRP layer at loading speed of 4.43 m/s.
Specimen
label
CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A

L1 (mm)

20
30
40
50
60
70

L2 (mm)

80
80
80
80
80
100

Experiment

Finite element analysis

Avg. Pult (kN)

PFE (kN)

56.21
58.73
58.94
57.26
58.18
58.09

55.74
56.26
56.08
56.12
56.18
56.06

PFE/Avg. Pult
0.992
0.958
0.951
0.980
0.966
0.965

Table 9
Comparison between the dynamic experimental and nite element analysis results
for joints with 3 CFRP layers at loading speed of 3.35 m/s.
Specimen
label

L1 (mm)

CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A

10
20
30
40
50
60
70
80
90
100

L2 (mm)

80
80
80
80
80
80
100
100
115
115

Experiment

Finite element analysis

Avg. Pult (kN)

PFE (kN)

PFE/Avg. Pult

84.29
110.03
129.83
136.21
152.59
143.26
144.70
145.18
144.70
146.41

79.12
107.72
131.44
142.36
148.96
148.53
148.61
149.46
150.12
149.24

0.939
0.979
1.012
1.045
0.976
1.037
1.027
1.029
1.037
1.019

Table 10
Comparison between the dynamic experimental and nite element analysis results
for joints with 3 CFRP layers at loading speed of 4.43 m/s.
Specimen
label

L1 (mm)

CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A

20
30
40
50
60
70
90

L2 (mm)

80
80
80
80
80
100
115

Experiment

Finite element analysis

Avg. Pult (kN)

PFE (kN)

PFE/Avg. Pult

114.08
135.51
145.15
155.00
148.15
154.39
149.50

115.48
140.02
154.96
156.56
156.01
156.24
156.60

1.012
1.033
1.068
1.010
1.053
1.012
1.047

Table 7
Comparison between the dynamic experimental and nite element analysis results
for joints with 1 CFRP layer at loading speed of 5 m/s.
Specimen
label

L1 (mm)

L2 (mm)

CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A
CF-1-A

20
30
40
50
60
70

80
80
80
80
80
100

Experiment

Finite element analysis

Avg. Pult (kN)

PFE (kN)

PFE/Avg. Pult

57.04
58.85
59.43
57.67
58.23
57.66

58.76
59.28
59.36
59.22
59.18
59.24

1.030
1.007
0.999
1.027
1.016
1.027

Table 8
Comparison between the quasi-static experimental and nite element analysis results
for joints with 3 CFRP layers.
Specimen
label

L1 (mm)

CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A

10
20
30
40
50
60
70
80
90
100

L2 (mm)

80
80
80
80
80
80
100
100
115
115

Experiment

Finite element analysis

Avg. Pult (kN)

PFE (kN)

PFE/Avg. Pult

29.61
54.20
68.88
82.88
96.83
101.35
103.24
97.40
97.38
99.22

25.60
52.32
67.86
79.13
102.88
106.16
106.48
105.44
105.40
105.12

0.865
0.965
0.985
0.955
1.062
1.047
1.031
1.083
1.082
1.059

With regard to CFRP/steel double strap joints with 3 CFRP layers

per side, Tables 811 show a clear comparison between the numerically-predicted and experimentally-measured ultimate joint
capacities of various bond lengths and four different test speeds.
In general, it is evident that FE models predict the ultimate joint
capacities well for all test speeds and for the different bond lengths

Table 11
Comparison between the dynamic experimental and nite element analysis results
for joints with 3 CFRP layers at loading speed of 5 m/s.
Specimen
label

L1 (mm)

CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A
CF-3-A

20
30
40
50
60
70
90

L2 (mm)

80
80
80
80
80
100
115

Experiment

Finite element analysis

Avg. Pult (kN)

PFE (kN)

PFE/Avg. Pult

124.10
148.24
154.04
157.58
152.95
158.16
157.02

126.28
151.44
164.60
164.96
165.20
164.96
164.76

1.018
1.022
1.069
1.047
1.080
1.043
1.049

included in the experimental test program. This close consistency

is more obvious in Fig. 4, which depicts the correlation of the tensile joint capacities for various bond lengths and for all the four different loading velocities as calculated by the FEA and as
experimentally-measured. It shows that the ratio of PFE/Avg. Pult
ranges from 0.865 to 1.083, 0.939 to 1.045, 1.01 to 1.068 and
1.018 to 1.080 for the quasi-static and the three loading speeds
of 3.35 m/s, 4.43 m/s and 5 m/s respectively. Compared to the observed correlation ratios from joints with 1 CFRP layers, it is clear
that these ratios are slightly higher than those obtained from joints
with 1 CFRP layer. The reason for this slight difference is mainly
attributed to the slight change in the predicted failure modes compared to those realised experimentally. The CFRP delamination is
not detected by FE models for all loading speeds (quasi-static
and dynamic), whereas this failure is clearly recognised under
experimental conditions. This will be discussed further in Section 7.3. Finally, even though there is little change in the predicted
failure mode, it can be concluded that all quasi-static and dynamic
joint strengths are predicted reasonably well.

54

H. Al-Zubaidy et al. / Composite Structures 99 (2013) 4861

Fig. 5. Effect of test speed on effective bond length for joints with 1 CFRP layer
(experiment and FEA).
Fig. 3. Correlation between the experimental and predicted ultimate loads form
FEA for joint with 1 CFRP layer at different loading speeds.

Fig. 6. Effect of test speed on effective bond length for joints with 3 CFRP layers
(experiment and FEA).

Fig. 4. Correlation between the experimental and predicted ultimate loads form
FEA for joint with 3 CFRP layers at different loading speeds.

7.2. Effective bond length

The effect of increasing the test speed on the effective bond
length of CFRP/steel double-strap joints with 1 and 3 CFRP layers
per side using Araldite 420 adhesive has been experimentally
investigated and the results presented in [20]. In relation to joints
with 1 CFRP layer per side, this effect is shown again in Fig. 5 where
the predictions by the FEA are also included. It is evident that there
is a slight reduction in the effective bond length with increasing
the loading speed and this trend is well captured by FE models.
It is explicitly depicted by both the experimental results and the
FE models that the quasi-static bond length is 30 mm, whereas it
reduces insignicantly to 20 mm at dynamic loading speeds. It is
also found that it remains almost the same for all three different
loading speeds. This means that FE models capture such a phenomenon well, which indicates that increasing the loading speed beyond 3.35 m/s has not affected the effective bond length.
Concerning CFRP/steel double-strap joints with 3 CFRP layers
per side, the effect of testing speed on the effective bond length
has been numerically determined using FE models as exhibited
in Fig. 6. This graph shows a clear comparison between the
numerically-predicted and experimentally-observed effective
bond lengths for the quasi-static and loading speeds of 3.35 m/s,

4.43 m/s and 5 m/s. Again, similar to the double-strap joints with
1 CFRP layer, it is obvious that the developed FE models determine
the effective bond length for all loading speeds quite well, and the
results conrm the general trend, which is gradually decreasing
with increasing impact velocity, as experimentally observed.
This means that the numerical modelling is quite capable of
determining the effective bond lengths for all loading speeds. It
is believed that the uctuations of the strength values determined
experimentally for bond lengths in excess of effective bond length
is attributed to the presence of CFRP delamination through the
CFRP layers. This type of failure is not detected by FE models
because the three CFRP layers are modelled as one patch of known
properties, and as a consequence, CFRP delamination failure cannot
be simulated.
As mentioned in the previous investigation [20] and clearly
illustrated in Fig. 6 by both the experimental results and the FE
models, the quasi-static effective bond length is 50 mm. However,
for the three dynamic tests, it can be seen that although it remains
50 mm at a test speed of 3.35 m/s like the quasi-static bond length,
it slightly reduces to 40 mm for loading speeds of 4.43 m/s and
5 m/s. Thus, this excellent agreement between FE models and
experimental results highlights the adequacy of the FEA analyses.
7.3. Failure pattern
In the case of CFRP/steel double-strap joints with 1 CFRP layer
per side, the FEA results of all the various bond lengths and different loading speeds have indicated almost similar failure modes to

H. Al-Zubaidy et al. / Composite Structures 99 (2013) 4861

those observed in the experiments. As reported in the authors previous study [20], the prevalent failure pattern in the experiments is
CFRP failure (bre breakage and some CFRP delamination). Fig. 7a
d shows the failure modes predicted by FE models for the quasistatic and three impact loading speeds. Inspection of these gures
clearly illustrates that FE models are able to detect CFRP breakage,
while the experimentally-observed slight CFRP delamination cannot be clearly realised. It is believed that this is because the CFRP
layer is modelled as one layer of known properties, whereas in
reality it comprises many carbon bre bundles and these bundles
are impregnated by epoxy during the manufacture of double strap
joints to form a CFRP patch with 1 CFRP layer. It is also important
to note that the type of failure determined in this study is based on

55

checking both the deformation of the failed samples at failure and

the satisfaction of the adopted failure criteria of both CFRP and
adhesive.
Concerning the failure mechanisms of CFRP/steel double-strap
joints with 3 CFRP layers per side, it has been shown in [20] that
the experimentally-observed quasi-static failure modes were debonding (steel and adhesive interface failure) and CFRP delamination, whereas CFRP delamination was the prevailing dynamic
failure pattern for all dynamic tests. However, some differences exist between the predicted and experimentally-observed failure
modes. The quasi-static FE model indicates one failure mode similar to those observed in the experiments. This is debonding failure,
as shown in Fig. 8a. The other failure mode, CFRP delamination, is

Fig. 7. Predicted failure patterns from FEA for joints with 1 CFRP layer per side (a) 3.34  105 m/s, (b) 3.35 m/s, (c) 4.43 m/s, (d) 5 m/s.

56

H. Al-Zubaidy et al. / Composite Structures 99 (2013) 4861

Fig. 7. (continued)

not detected. This is attributed to the fact that the three CFRP layers per side are simulated as an equivalent laminate with known
properties, as calculated in Section 6. Thus, the FE models are unable to predict CFRP delamination which occurs experimentally
via the bre bundles for joints with 1 CFRP layer and within subsequent layers for joints with 3 CFRP layers.
On the other hand, in relation to the dynamic failure patterns,
the numerically-predicted failure modes have also highlighted little variance in comparison with those observed experimentally.
Fig. 8bd illustrates that the failure mode commonly predicted
by the explicit FE models for all the impact loading speeds is CFRP
breakage instead of CFRP delamination, and these predictions are

similar to those modes predicted numerically for CFRP/steel

double-strap joints with 1 CFRP layers per side. Thus, CFRP
delamination disappears again for the reason mentioned above. It
is worth noting that although CFRP delamination is not predicted
numerically, the detection of CFRP breakage conrms the key
experimental nding of this study, which is that no debonding failure (steel and adhesive interface failure) occurs under any of the
dynamic loading speeds. This is attributed to the shear strength
enhancement of the epoxy between the CFRP patch and the steel
plate under dynamic loading, as explained in [20].
Overall, although there is some difference between the numerically-predicted and experimentally-observed failure patterns, the

H. Al-Zubaidy et al. / Composite Structures 99 (2013) 4861

57

Fig. 8. Predicted failure patterns from FEA for joints with 3 CFRP layers per side (a) 3.34  105 m/s, (b) 3.35 m/s, (c) 4.43 m/s, (d) 5 m/s.

failure mode predictions of FE models are still considered reasonably high, especially when compared with the predictions of other
parameters (ultimate joint strength, effective bond length and
strain distributions along the CFRP bond length).
7.4. Strain distribution along the bond length
The experimentally-measured and numerically-predicted strain
distributions along the bond length of CFRP/steel double strap
joints with 1 and 3 CFRP layers per side at quasi-static and three
dynamic loading speeds of 3.35 m/s, 4.43 m/s and 5 m/s are compared in Figs. 9ad and 10ad respectively. The comparison is

made at three load levels. For comparison purposes, the load levels
of both experimental and numerical strain distributions are chosen
to be as close as possible to each other. As clearly depicted in the
authors previous investigation [20], strain values were experimentally recorded using several foil strain gauges mounted at xed distances (15 mm) starting from the mid-joint towards the end of the
CFRP patch. Thus, the strain monitoring points in the FE models are
selected as close as possible to the experimental locations where
the strain gauges were placed.
For CFRP/steel double-strap joints with 1 CFRP layer, in general,
it can be seen in Fig. 9ad that the FE models simulate the strain
distributions along the bond length for both the quasi-static and

58

H. Al-Zubaidy et al. / Composite Structures 99 (2013) 4861

Fig. 8. (continued)

all the three dynamic speeds reasonably well. Very close matching
is realised between the measured and predicted strain values for
all load levels and this is the case for all four test speeds. This
can be attributed to the fact that the observed failure mode of both
the numerical analyses and experiments is very similar (bre
breakage).
Careful inspection of Fig. 10ad reveals that the predicted strain
values along the CFRP patch of CFRP/steel double-strap joints with
3 CFRP layers per side for all the quasi-static and dynamic speeds

are consistent with the experimental results, with the exception

of the strain reading at the mid-joint location. The measured strain
at this position represents the strain captured by strain gauge 1
(G1) as depicted in [20]. It is evident that there is a pronounced difference between the predicted and measured strain values at this
location, as the former is much higher than the latter. Thus, the ratio of the predicted strain/measured strain is found to be approximately 1.18, 1.56, 1.57 and 1.51 for the quasi-static and for the
dynamic test speeds, 3.35 m/s, 4.43 m/s and 5 m/s respectively. It

H. Al-Zubaidy et al. / Composite Structures 99 (2013) 4861

Fig. 9. Comparison of the predicted and measured strain distribution at quasi-static rate for joints with 1 CFRP layer (experiment and FE).

Fig. 10. Comparison of the predicted and measured strain distribution at quasi-static rate for joints with 3 CFRP layers (experiment and FE).

59

60

H. Al-Zubaidy et al. / Composite Structures 99 (2013) 4861

is clear that the quasi-static ratio is much lower than the dynamic
ratios. This is because the occurrence of CFRP delamination accelerates with increasing test speed, as discussed in [20]. However,
it is believed that this marked difference between the predicted
and measured strain is related to the slight change between the
numerically-predicted and experimentally-observed failure mechanisms. While the experimentally-observed quasi-static failure
modes were debonding (steel and adhesive interface failure) and
CFRP delamination, CFRP delamination was shown to be the dominant dynamic failure for all dynamic speeds. Conversely, CFRP
breakage rather than CFRP-delamination is numerically predicted.
The reasons for this slight change in failure mode are explained in
Section 7.3.
Compared to the strain proles of joints with 1 CFRP layer, a
clear difference between the predicted and measured strain value
appears specically at the mid-joint of the joints with 3 CFRP layers although generally there is a very close matching between the
predicted and measured strain distribution at other locations along
the bond length. This is because the probability of occurrence of
CFRP delamination for joints with 1 CFRP layer is much lower than
that for joints with 3 CFRP layers. This is evident when the experimentally-observed failure modes of both types of CFRP/steel double-strap joints with 1 and 3 CFRP layers are compared.
8. Concluding remarks
In this paper, three-dimensional nite element method is utilised in the numerical analysis of both CFRP/steel double-strap
joints with 1 and 3 CFRP layers per sides using ABAQUS program.
Steel plate has been modelled as elasticplastic material and CFRP
and adhesive failures are taken into account in these simulations.
Continuum shell elements are utilised to simulate the CFRP patch
whereas the adhesive layer is modelled as a cohesive element.
The ndings from this work are:
 Overall, the developed FE models reasonably predicted the
quasi-static and dynamic behaviour of both CFRP/steel double-strap joints with 1 and 3 CFRP layer per side. This is proven
through the sufcient prediction of the peak load, effective
bond length, failure patterns and strain distribution along the
bond length of types of joints.
 It is found that, generally, there is a good correlation between
the predicted and experimental ultimate joint strength of both
types of joints with 1 and 3 CFRP layers. However, the predicted
joint strength for joints with 3 CFRP layers is slightly higher
than the experimental one, specically for bond lengths equal
to or greater than the effective bond length and at higher loading speed (5 m/s). This is due to the inability of the FE model to
simulate CFRP-delamination through the layers which is the
experimentally common failure mode.
 The numerical simulations predict well the effective bond
length for both joints with 1 and 3 CFRP layers and are in excellent agreement with those determined experimentally. This
proved to be the case for all loading speeds.
 Little difference between the experimentally observed and the
numerically predicted failure patterns is realised for both types
of joints especially for joints with 3 CFRP layers. Delamination
through CFRP layers is the dominant failure mode experimentally, however, it does not realise numerically. This is attributed
to the fact that the three CFRP layers per side are simulated as
an equivalent laminate with macroscopic properties.
 There is an excellent match between the predicted and the
experimentally measured strain proles along the bond length
for all loading speeds and at all load levels particularly for joints
with 1 CFRP layer. However, compared to joints with 3 CFRP

layers, a clear difference between the measured and predicted

strain value is detected at the mid-joint whereas the other
strain values along the bond length are in good agreement.

References
[1] Fawzia S, Al-Mahaidi R, Zhao X-L. Experimental and nite element analysis of a
double strap joint between steel plates and normal modulus CFRP. Compos
Struct 2006;75:15662.
[2] Colombi P, Poggi C. Strengthening of tensile steel members and bolted joints
using adhesively bonded CFRP plates. Constr Build Mater 2006;20:2233.
[3] Ban C-S, Lee Y-H, Choi J-H, Kweon J-H. Strength prediction of adhesive joints
using the modied damage zone theory. Compos Struct 2008;86:96100.
[4] Wu C, Zhao X, HuiDuan W, Al-Mahaidi R. Bond characteristics between ultra
high modulus CFRP laminates and steel. Thin-Walled Struct 2012;51:
14757.
[5] Bocciarelli M, Colombi P. Elasto-plastic debonding strength of tensile steel/
CFRP joints. Eng Fract Mech 2012;85:5972.
[6] Haghani R. Analysis of adhesive joints used to bond FRP laminates to steel
members a numerical and experimental study. Constr Build Mater
2010;24:224351.
[7] Zgoul M, Crocombe AD. Numerical modelling of lap joints bonded with a ratedependent adhesive. Int J Adhes Adhes 2004;24:35566.
[8] Campilho RDSG, Banea MD, Pinto AMG, da Silva LFM, de Jesus AMP. Strength
prediction of single- and double-lap joints by standard and extended nite
element modelling. Int J Adhes Adhes 2011;31:36372.
[9] Kumar SB, Sridhar I, Sivashanker S, Osiyemi SO, Bag A. Tensile failure of
adhesively bonded CFRP composite scarf joints. Mater Sci Eng: B 2006;132:
11320.
[10] Campilho RDSG, de Moura MFSF, Domingues JJMS. Using a cohesive damage
model to predict the tensile behaviour of CFRP single-strap repairs. Int J Solids
Struct 2008;45:1497512.
[11] Pinto AMG, Campilho RDSG, de Moura MFSF, Mendes IR. Numerical evaluation
of three-dimensional scarf repairs in carbon-epoxy structures. Int J Adhes
Adhes 2010;30:32937.
[12] Colombi P, Poggi C. An experimental, analytical and numerical study of the
static behavior of steel beams reinforced by pultruded CFRP strips. Compos
Part B: Eng 2006;37:6473.
[13] Haghani R, Al-Emrani M, Kliger R. Interfacial stress analysis of geometrically
modied adhesive joints in steel beams strengthened with FRP laminates.
Constr Build Mater 2009;23:141322.
[14] Narmashiri K, Jumaat MZ. Reinforced steel I-beams: a comparison between 2D
and 3D simulation. Simul Modell Pract Theory 2011;19:56485.
[15] Kadhim MMA. Effect of CFRP plate length strengthening continuous steel
beam. Constr Build Mater 2012;28:64852.
[16] Loureiro A, Da Silva LFM, Sato C, Figueiredo M. Comparison of the mechanical
behaviour between stiff and exible adhesive joints for the automotive
industry. J Adhes 2010;86:76587.
[17] Crocombe AD. Modelling and predicting the effects of test speed on the
strength of joints made with FM73 adhesive. Int J Adhes Adhes 1995;15:217.
[18] Harris JA, Adams RA. Strength prediction of bonded single lap joints by nonlinear nite element methods. Int J Adhes Adhes 1984;4:6578.
[19] Zhou DW, Louca LA, Saunders M. Numerical simulation of sandwich T-joints
under dynamic loading. Compos Part B: Eng 2008;39:97385.
[20] Al-Zubaidy HA, Zhao X-L, Al-Mahaidi R. Dynamic bond strength between CFRP
sheet and steel. Compos Struct 2012;94:325870.
[21] Faggiani A, Falzon BG. Predicting low-velocity impact damage on a stiffened
composite panel. Compos Part A: Appl Sci Manuf 2010;41:73749.
[22] Mustapha F, Sim N, Shahrjerdi A. Finite element analysis (FEA) modeling
on adhesive joint for composite fuselage model. Int J Phys Sci 2011;6:
515365.
[23] Smojver I, Ivancevic D. Bird strike damage analysis in aircraft structures using
Abaqus/Explicit and coupled Eulerian Lagrangian approach. Compos Sci
Technol 2011;71:48998.
[24] Falzon BG, Hitchings D, Besant T. Fracture mechanics using a 3D composite
element. Compos Struct 1999;45:2939.
[25] ABAQUS. ABAQUS 6.8 Documentation. Dessault systems; 2008.
[26] Al-Zubaidy H, Zhao X-L, Al-Mahaidi R. Mechanical characterisation of the
dynamic tensile properties of CFRP sheet and adhesive at medium strain rates.
Compos Struct, in press.
[27] Hashin Z, Rotem A. A fatigue criterion for bre-reinforced materials. J Compos
Mater 1973;7:44864.
[28] Hashin Z. Failure criterion for unidirectional bre composites. J Appl Mech
1980;47:32934.
[29] de Moura MFSF, Gonalves JPM. Modelling the interaction between matrix
cracking and delamination in carbonepoxy laminates under low velocity
impact. Compos Sci Technol 2004;64:10217.
[30] Heru Utomo BD, van der Meer BJ, Ernst LJ, Rixen DJ. High-speed impact
modelling and testing of dyneema composite. In: The Proceeding of the 11th
international conference on mechanics and technology of composite materials,
October 24, Soa, Bulgaria; 2006.
[31] Ullah H, Harland AR, Lucas T, Price D, Silberschmidt VV. Analysis of nonlinear
deformations and damage in CFRP textile laminates. In: 9th International

H. Al-Zubaidy et al. / Composite Structures 99 (2013) 4861

conference on damage assessment of structures (DAMAS 2011) Published by
the Journal of Physics: Conference series 305; 2011.
[32] Chou S-P. Finite element application for strength analysis of scarf-patchrepaired composite laminates. MSc Thesis, Department of Aerospace
Engineering and the faculty of Graduate School, Wichita State University,
USA; 2006.

61

[33] da Silva LFM, Rodrigues T, Figueiredo M, De Moura M, Chousal J. Effect of

adhesive type and thickness on the lap shear strength. J Adhes
2006;82:1091115.
[34] Hart-Smith L. Adhesive-bonded double-lap joints. In: Technical reports NASA
CR-112235 Douglas Aircraft Company: Long Bench, California, USA; 1973.