Sie sind auf Seite 1von 12

Engineering Structures 86 (2015) 213224

Contents lists available at ScienceDirect

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

Finite element modelling of debonding failures in steel beams exurally


strengthened with CFRP laminates
J.G. Teng a,, D. Fernando a,b, T. Yu a,c
a

Department of Civil and Environmental Engineering, The Hong Kong Polytechnic University, Hong Kong, China
School of Civil Engineering, The University of Queensland, QLD 4072, Australia
c
School of Civil, Mining and Environmental Engineering, Faculty of Engineering and Information Sciences, The University of Wollongong, Northelds Avenue, Wollongong,
NSW 2522, Australia
b

a r t i c l e

i n f o

Article history:
Received 25 March 2014
Revised 2 January 2015
Accepted 3 January 2015
Available online 21 January 2015
Keywords:
CFRP
Strengthening
Steel beams
Debonding
Cohesive zone modelling
Finite element analysis

a b s t r a c t
A steel beam may be strengthened in exure by bonding a carbon bre-reinforced polymer (CFRP) plate
to the tension face. Such a beam may fail by debonding of the CFRP plate that initiates at one of the plate
ends (i.e. plate end debonding) or by debonding that initiates at a local damage (e.g. a crack or concentrated yielding) away from the plate ends (intermediate debonding). This paper presents the rst nite
element (FE) approach that is capable of accurate predictions of such debonding failures, with particular
attention to plate-end debonding. In the proposed FE approach, a mixed-mode cohesive law is employed
to depict interfacial behaviour under a combination of normal stresses (i.e. mode-I loading) and shear
stresses (i.e. mode-II loading); the interfacial behaviour under pure mode-I loading or pure mode-II loading is represented by bi-linear tractionseparation models. Damage initiation is dened using a quadratic
strength criterion, and damage evolution is dened using a linear fracture energy-based criterion.
Detailed FE models of steel beams tested by previous researchers are presented, and their predictions
are shown to be in close agreement with the test results. Using the proposed FE approach, the behaviour
of CFRP-strengthened steel beams is examined, indicating that: (1) if the failure is governed by plate end
debonding, the use of a CFRP plate with a higher elastic modulus and/or a larger thickness may lead to a
lower ultimate load because plate end debonding may then occur earlier; (2) plate end debonding is more
likely to occur when a short CFRP plate is used, as is commonly expected; and (3) the failure mode may
change to intermediate debonding or other failure modes such as compression ange buckling if a longer
plate is used.
2015 Elsevier Ltd. All rights reserved.

1. Introduction
Similar to concrete beams, steel beams or steelconcrete composite beams (referred to collectively as steel beams hereafter
for simplicity) can be strengthened in exure by bonding an FRP
(generally CFRP) laminate to the tension face [111]. Such beams
are herein referred to as FRP-strengthened steel beams. The
bonded FRP laminate may be prefabricated (e.g. by pultrusion) or
formed in-situ (e.g. by the wet-layup process), and is referred to
as a plate for simplicity in this paper. CFRP is commonly preferred
to other FRPs including glass FRP (GFRP) in the strengthening of
steel structures due to the much higher stiffness of CFRP [7], so this
paper is focused on CFRP strengthening only. For simplicity of

Corresponding author. Tel.: +852 27666012.


E-mail address: cejgteng@polyu.edu.hk (J.G. Teng).
http://dx.doi.org/10.1016/j.engstruct.2015.01.003
0141-0296/ 2015 Elsevier Ltd. All rights reserved.

discussion, only simply-supported beams are explicitly considered,


so the CFRP plate is bonded to the soft of the beam.
CFRP exural strengthening can enhance both the stiffness and
the load-carrying capacity of a steel beam [7,12]. The load-carrying
capacity of such CFRP-strengthened beams may be governed by
one or a combination of the many possible failure modes [7], which
include: (a) in-plane bending failure (i.e. CFRP failure, concrete
crushing); (b) lateral buckling; (c) debonding at a plate end (i.e.
plate-end debonding); and (d) debonding away from the plate ends
induced by cracking or concentrated yielding in the steel beam (i.e.
intermediate debonding). Additional failure modes include: (e)
local buckling of the compression ange; and (f) local buckling of
the web. Among these failure modes, debonding of the CFRP plate
[failure modes (c) and (d)] has been found to be common in laboratory tests on CFRP-strengthened steel beams [1,3,6,7,13].
In a CFRP-strengthened steel beam failing by debonding of the
CFRP plate, the load-carrying capacity depends on the contribution

214

J.G. Teng et al. / Engineering Structures 86 (2015) 213224

of the CFRP at the time of debonding, which in turn depends on the


interfacial stress transfer of the adhesive layer. Therefore, accurate
simulation of the bond behaviour of CFRP-to-steel interfaces is of
particular importance in the theoretical modelling of debonding
failures.
As mentioned above, in a CFRP-strengthened steel beam, both
plate-end debonding and intermediate debonding are possible.
Intermediate debonding initiates from the presence of a defect
(e.g. a crack or concentrated yielding of the steel substrate) [3]
where the FRP plate is highly stressed, and propagates towards a
plate end where the stress in the FRP plate is lower. Very limited
research is available on intermediate debonding in CFRP-strengthened steel beams and no theoretical modelling exists so far. This
kind of debonding, similar to intermediate-crack induced debonding (IC debonding) in an FRP-strengthened concrete beam [14], is
dominated by interfacial shear stresses. Therefore, accurate simulation of intermediate debonding requires an appropriate model
for the damage behaviour of the interface in the shear direction
(i.e. under mode-II loading). Such bond-slip models have previously been developed by the authors group [15].
Compared to intermediate debonding, the modelling of plateend debonding is more involved as it is governed by both interfacial shear stresses and interfacial normal stresses [12,1618].
Therefore, the effect of interaction between mode-I loading and
mode-II loading on damage initiation and propagation within the
adhesive layer needs to be appropriately addressed. A number of
theoretical studies (e.g. [12]) using strength-based approaches
have been conducted on plate-end debonding. These theoretical
analyses have generally signicantly underestimated the performance of the strengthened beam, as the bond strength depends
on the interfacial fracture energy instead of the strength of the
adhesive [19]. De Lorenzis et al. [18] have recently presented the
only existing reliable theoretical study based on the fractureenergy approach for plate end debonding of FRP-strengthened
steel beams. However, De Lorenzis et al.s study [18], which was
conducted after the present work [15] and based on the theoretical
concepts presented in the present paper, employed a number of
assumptions to arrive at a simplied analytical solution, is only
applicable to cases where the steel beam is linear elastic and subjected to three-point bending. The analytical solution presented in
Ref. [18] was also veried using the FE method described in the
present paper. No other theoretical (numerical or analytical) studies on plate-end debonding in CFRP-strengthened steel beams have
been found in which the mixed-mode damage/fracture behaviour
of the adhesive layer is appropriately captured.
Against this background, this paper presents a nite element
(FE) approach for CFRP-strengthened steel beams, with a particular
emphasis on the accurate modelling of the bond behaviour and
debonding failures in such beams. It should be noted that the present work is based on the premise that debonding failures in FRPstrengthened steel beams occur by cohesion failure within the
adhesive layer instead of adhesion failure at the steel/adhesive
bi-material interface or the FRP/adhesive bi-material interface. Of
these two adhesion failure modes, the latter one is much less likely
and the former one needs to be avoided in practice through the
proper preparation of the steel surface [20].

loading on damage propagation along the interface. For (1), an


accurate bond-slip model (e.g. such as those presented in Ref.
[15]) and an accurate bond-separation model [21,22] can be
employed to predict the full-range interfacial behaviour under
pure mode-II loading and pure mode-I loading respectively. For
(2), the so-called mixed-mode cohesive law needs to be used.
Among the existing modelling techniques, a coupled cohesive zone
model appears to be the most suitable as it possesses the two
required characteristics. Cohesive zone models have been used
for simulating the fracture of ductile and brittle solids [23], the
delamination of composites [24,25] and the behaviour of adhesively bonded joints [26,27]. Bocciarelli et al. [28] presented an
FE model for CFRP-to-steel bonded joints with a cohesive zone
model and showed close predictions for results from double-shear
lap tests, but the cohesive zone model used by them only considered interfacial behaviour under pure mode-II loading. In the following section, a coupled cohesive zone model is proposed for
CFRP-to-steel interfaces, which consists of the following three
key components: a bond-slip model for mode-II loading, a bondseparation model for mode-I loading, and a mixed-mode cohesive
law. It should be noted that the model presented in the following
section is for linear adhesives (with a linear stressstrain curve
before tensile rupture) only, but the general concepts of the model
are extendable to nonlinear adhesives.
2.1. Coupled cohesive zone model
2.1.1. Bond-slip model
The bi-linear bond-slip model proposed by Fernando [15] for
linear adhesives has been shown to provide accurate predictions
for the bond behaviour of CFRP-to-steel bonded joints subjected
to mode-II loading, and is thus adopted here. The bi-linear bondslip model proposed by Fernando [15] can be written as:

8
smax dd1
>
>
<
>
>
:

if

d 6 d1

if

d1 < d 6 df

if

d > df

df d
max d d
1
f

where s is the bond shear stress, smax is the peak bond shear stress,
d is the slip, d1 is the slip at peak bond shear stress, and df is the slip
at complete failure. Based on the experimental results of CFRP-steel
bonded joints with linear adhesives, Fernando [15] derived the following expressions for the above bond-slip parameters:

smax 0:9rmax

 0:65
ta
rmax
d1 0:3
Ga
2Gf
mm
df

2
mm

smax

3
4

where rmax is the tensile strength (in MPa in Eq. (3)) of the adhesive, ta and Ga are the thickness and shear modulus of the adhesive
layer respectively, and Gf is the interfacial fracture energy given by
[15]:
2
Gf 628t0:5
a R

N=mm2 mm

2. Modelling of CFRP-to-steel interfaces

R is the tensile strain energy per unit volume of the adhesive which
is equal to the area under the uni-axial tensile stress (in MPa)
strain curve.

The successful prediction of debonding failures in CFRPstrengthened steel beams requires a model for the CFRP-to-steel
interface which has the following characteristics: (1) it accurately
predicts the behaviour of the interface subjected to pure mode-I
loading and pure mode-II loading; (2) it appropriately accounts
for the effect of interaction between mode-I loading and mode-II

2.1.2. Bond-separation model


It is common to obtain the bond-separation model and the
mode-I fracture energy using double cantilever beam tests (DCB)
[21,29]. In the absence of these test data, the bi-linear bond-separation behaviour of a linear adhesive can be closely approximated
using the tensile stressstrain data of the adhesive [30]. The peak

215

J.G. Teng et al. / Engineering Structures 86 (2015) 213224

stress of the bi-linear bond-separation model can be assumed to be


the same as the tensile strength of the adhesive, the slope of the
ascending branch can be taken to be equal to the tensile elastic
modulus divided by the adhesive thickness (Eq. (8)), and the separation at complete failure, df, can be taken as the product of the
tensile strain at complete failure and the adhesive thickness [30].
In the present study, the tensile stressstrain data of the adhesive
were used to dene the bi-linear bond-separation model, following
the approach suggested by Campilho et al. [30].
2.1.3. Mixed-mode cohesive law
In a coupled cohesive zone model, a mixed-mode cohesive law
is employed to account for the interaction between mode-I loading
and mode-II loading. As the fracture energy for mode-II loading is
often much larger than that for mode-I loading [31], a mixed-mode
law which properly accounts for this aspect is adopted, following
Xu and Needleman [32] and Hogberg [31]. For ease of discussion,
bond stresses are hereafter referred to collectively as tractions
while interfacial deformations (i.e. displacements) are referred to
collectively as separations.
The mixed-mode cohesive law adopted in the present study
considers tractions and separations in all three directions: those
normal to the interface and those parallel to the interface (i.e.
the two shear components). The normal and the two shear tractions are denoted by tn, ts and tt respectively, while the corresponding separations are denoted by dn, ds and dt respectively. With the
thickness of the cohesive element taken as the original thickness
of the adhesive layer T0, the strains in the normal (en) and the
two shear directions (cs and ct) are given by:

en

dn
;
T0

cs

ds
T0

and ct

dt
T0

2.1.3.1. Elastic behaviour. It is assumed that the interface behaves


linear-elastically until the initiation of damage [18,21,23,26]. The
interfacial behaviour before damage initiation can thus be represented by

8 9 2
K nn
>
< tn >
=
6
ts 4 0
>
: >
;
0
tt

0
K ss
0

38 9
>
< dn >
=
7
0 5 ds
>
: >
;
K tt
dt
0

Ea
T0

where Ea is the elastic modulus of the adhesive.


Kss and Ktt are assumed to be the same, and should be equal to
the initial slope of the bond-slip model presented earlier for modeII loading. From Eqs. (2) and (3):

K ss K tt 3

 0:65
Ga
T0

ht n i

rmax

Eq. (9) suggests that Kss and Ktt depend on the shear modulus Ga of
the adhesive. Therefore, the elastic stiffness in the two shear directions and that in the normal direction are inter-related through the
Poissons ratio.
2.1.3.2. Damage behaviour. Under pure mode-II loading, damage of
the interface initiates when the shear stress reaches the peak bond
shear stress [15]. Similarly, under pure mode-I loading, damage
initiates when the normal stress reaches the peak bond normal
stress. Under mixed-mode loading, a strength criterion needs to

2

ts

2

smax

tt

2

smax

10

The symbol hi is the Macaulay bracket which is used to signify that


compressive stresses do not lead to damage (i.e. when tn is negative,
htni is equal to zero). Based on Eq. (10), the damage initiation point
can be determined when the mode-mix ratio (i.e. the ratio between
the fracture energies of two different modes) is known.
After damage initiation, a scalar damage variable D is introduced. D is equal to zero at the initiation of damage and is equal
to one at complete failure of the interface. The interfacial behaviour can then be represented by the following equation:

8 9 2
1  D K nn
>
< tn >
=
6
ts 4
0
>
: >
;
tt
0

0
1  DK ss
0

38 9
>
< dn >
=
7
0
5 ds
>
: >
;
dt
1  DK tt
0

11

where means that if tn is compressive, D is equal to zero.


The complete failure of the interface, when D is equal to one, is
dened based on fracture energy considerations. While a few other
criteria for the denition of complete failure are available (e.g. the
quadratic criterion or the BK criterion proposed by Benzeggagh and
Kenane [35]), the linear criterion is adopted in the present study
due to its simplicity and good performance for adhesive joints
[34,36]. The linear criterion is expressed by:

Gn Gs Gt

1
GI GII GII

12

where Gn, Gs, Gt are the works done by the tractions and their conjugate displacements in the normal and the two shear directions
respectively (Fig. 1a). GI and GII represent the interfacial fracture
energies required to cause failure when subjected to pure mode-I
loading and pure mode-II loading respectively.
To describe the evolution of damage, the denition of an effective displacement is introduced as follows:

dm

where Knn, Kss, Ktt are the elastic stiffness values of the normal and
the two shear directions respectively. It is obvious that Knn should
be equal to the initial slope of the bond-separation model for
mode-I loading and is given by

K nn

be adopted to dene the initiation of damage, considering the


interaction between mode-I and mode-II loading. Following existing studies [33,34], the following quadratic strength criterion is
adopted in the present study:

q
hdn i2 d2s d2t

13

With this denition, the displacement at complete failure, dfm ,


can be found using Eq. (12) for a certain mode-mix ratio. The damage variable D is then dened by the following equation assuming
linear softening of the interface [34]:



dfm dmax
 d0m
m

dmax
dfm  d0m
m

14

where d0m is the effective displacement at the initiation of damage


and dmax
is the maximum value of the effective displacement
m
attained in the loading process (Fig. 1b).
3. FE modelling of CFRP-strengthened steel I-beams
In this section, an FE approach for debonding failures in CFRPstrengthened steel I-beams is rst presented, in which the coupled
cohesive zone model presented above for CFRP-to-steel interfaces
is employed. Numerical results obtained with the FE approach
are then given to demonstrate the capability of the proposed FE
approach in predicting debonding failures as well as other possible
failure modes (e.g. compression ange buckling), to clarify the
effect of approximating the mode-I fracture energy on damage
propagation in the adhesive layer, and to study the effect of plate
axial stiffness on plate end debonding failure. With these aims in

216

J.G. Teng et al. / Engineering Structures 86 (2015) 213224

plate-end debonding and compression ange buckling) were


observed due to the use of CFRP plates with different lengths.
These characteristics make Deng and Lees tests [13] the most suitable for verifying the proposed FE approach, especially in terms of
the behaviour of CFRP-to-steel interfaces.

Traction
max, max
Knn,
Kss,
Ktt

(1-D)K
GI-Gn,
GII-Gs,
GII-Gf

Gn,Gs,Gt

1
1 1
n, s , t

n, s

3.1. Beam tests conducted by Deng and Lee [7]

f
n

f
s

f
t

Four of the beams tested by Deng and Lee [13] were selected for
FE simulation, including one control beam without CFRP strengthening and three beams strengthened with 3 mm thick CFRP plates
of three different lengths (i.e. 300 mm, 400 mm, and 1000 mm)
respectively. These beams were selected because they all had the
same loading conguration (i.e. three-point bending), the same
steel section, and a single continuous CFRP plate; the only variable
was the plate length to examine how the failure mode would
change with the plate length (from plate end debonding to buckling of compression ange) and whether this could be accurately
predicted by the proposed FE approach. The four selected beams
were named by Deng and Lee [13] as specimens S300 (control
beam), S303, S304 and S310 respectively, where the last two numbers represent the length of the CFRP plate and the rst number
3 indicates that the beams were subjected to three-point bending. These four steel beams all had a length of 1.2 m (with a clear
span between the supports being 1.1 m) and a cross-section of type
127x76UB13; the dimensions of the steel beams are shown in
Fig. 2. Grade 275 steel was used, which means that the steel had
a nominal yield strength of 275 MPa (with the actual yield strength
often being larger than 275 MPa) and a tensile elastic modulus of
205 GPa. The CFRP plates used all had a thickness of 3 mm, a width
of 76 mm, and an elastic modulus in the bre direction of 212 GPa.
To avoid premature ange buckling and web crushing, two 4 mm
thick steel plate stiffeners were welded to each beam at the midspan, one on each side of the web. For beams with a short CFRP
plate (i.e. 300 mm or 400 mm), plate end debonding of the CFRP
plate was observed. However, when a longer CFRP plate (i.e.
1000 mm) was used, failure was controlled by the buckling of
the compression ange of the steel section, which was the same

Separation

(a) Traction-separation curve


Traction

Gmc

0
m

max
m

f
m

Separation

(b) Linear damage evolution under mixed-mode loading


Fig. 1. Tractionseparation curves and linear damage evolution under mixed-mode
loading. (a) Tractionseparation curve. (b) Linear damage evolution under mixedmode loading.

mind, FE models were developed for four beams tested by Deng


and Lee [13], and were veried using the test results. Deng and
Lees tests [13] were selected for comparison among many other
experimental studies (e.g. [1,3,7]) as Deng and Lees tests [13]
had the following desirable characteristics: (1) debonding failures
controlled by cohesion failure occurred; (2) experimental loaddisplacement curves were reported; (3) different failure modes (e.g.

127x76UB13 steel I beam

LCFRP
1100mm
1 mm thick adhesive layer

3 mm thick CFRP plate

76mm

4mm

127mm

7.6mm
76mm
Fig. 2. Details of test specimens of Deng and Lee [13].

217

J.G. Teng et al. / Engineering Structures 86 (2015) 213224


Table 1
Details of the test beams and the FE models.

a
b

Specimen/model
name

Length of CFRP plate,


mm

Elastic modulus of CFRP,


GPa

Thickness of CFRP plate,


mm

Mode I
behaviour

Compression ange
strengthening

S300a
S303a
S304a
S310a
S300-0-000b
S303-1-212b
S303-2-212b
S303-1-330b
S304-1-212b
S310-1-212b
S310-1-212-Pb

N/A
300
400
1000
N/A
300
300
300
400
1000
1000

N/A
212
212
212
N/A
212
212
330
212
212
212

N/A
3
3
3
N/A
3
3
3
3
3
3

N/A
N/A
N/A
N/A
N/A
Model
Model
Model
Model
Model
Model

No
No
No
No
No
No
No
No
No
No
6 mm steel plate

A
B
A
A
A
A

Test beams [13].


FE models.

as the failure mode of the control beam (i.e. specimen S300). The
details of the test beams are summarized in Table 1.
3.2. FE models
FE models were created using ABAQUS [37] for the four beams,
with the exact dimensions and support conditions (i.e. simply-supported boundary conditions) as given in Fig. 2 and Table 1. The
general purpose shell element S4R with reduced integration was
adopted for both the steel section and the CFRP plate, while the
adhesive layer was modelled using the cohesive element COHD8
available in ABAQUS. The two full-depth stiffeners on the two sides
of the web in the mid-span region were included, and the three
edges of each stiffener in contact with the I beam were tied to
the top ange, the bottom ange and the web of the cross section
respectively. Similarly, the top surface and the bottom surface of
the adhesive layer were tied to the bottom surface of the steel
beam and the top surface of the CFRP plate respectively. Based
on a mesh convergence study, 2.5 mm  2.5 mm elements were
selected for the steel beam and the CFRP plate, while
2.5 mm  2.5 mm  1 mm (with 1 mm being in the thickness
direction) elements were selected for the adhesive layer. In all
the FE simulations, the analysis was terminated soon after the ultimate load had been reached.
The well-known J2 ow theory was employed to model the
material behaviour of the steel. As the experimental stressstrain
curve of the steel was not given by Deng and Lee [13], a tri-linear
(i.e. elastic-perfectly plastic-hardening) stressstrain model [38]
with a yield strength of 330 MPa (determined by a trial-and-error
process to match the linear portion of the experimental loaddisplacement curve of specimen S300) and a ultimate tensile stress
of 430 MPa (as specied in BS EN 10025-1 [39]) was adopted.
The use of such an idealized stressstrain curve for the steel is
believed to be the best pragmatic solution possible in the absence
of the experimental stressstrain curve and has only minor effects
on the predictions for the steel beam (see [15]).
The CFRP plate was treated as an orthotropic material in the FE
models. In the bre direction, the elastic modulus (i.e. E3) provided
by Deng and Lee [13] was adopted (i.e. 212 GPa based on a nominal
thickness of 3 mm). The elastic modulus in the other two directions

(i.e. E1, E2), the Poissons ratios and the shear moduli were assumed
the following values respectively based on the values reported in
Deng et al. [16]: E1 = E2 = 10 GPa, m12 = 0.3, m13 = m23 = 0.0058,
G12 = 3.7 GPa and G13 = G23 = 26.5 GPa.
The coupled cohesive zone model presented earlier was
adopted to model the constitutive behaviour of the adhesive layer.
Deng and Lee [13] provided only the tensile strength (29.7 MPa),
the tensile elastic modulus (8 GPa) and the shear modulus
(2.6 GPa) for the adhesive. Considering that the adhesive used by
Deng and Lee [13] was a linear adhesive, the strain energy was calculated by assuming a bi-linear stressstrain curve with the slope
of the ascending branch being equal to the elastic modulus (i.e.
8 GPa), the peak stress being equal to the tensile strength (i.e.
29.7 MPa) and the ultimate strain being equal to 4% which is the
value provided by the manufacturer [13]. With the above parameters, the bond-slip model for mode-II loading can be determined
using Eqs. (1)(4), and the key parameters of the so-obtained
bond-slip model are given in Table 2. The bond-separation model
for mode-I loading can also be determined using the assumed bilinear stressstrain curve; the key parameters of the so-obtained
bond-separation model are given in Table 2 as model A. Besides
model A whose mode-I fracture energy is 0.059 N/mm, another
bond-separation model (i.e. model B, see Table 2) was also used,
whose mode-I fracture energy (i.e. 0.11 N/mm) is twice the elastic
energy of model A. The use of two different bond-separation models for mode-I loading was to explore the effect of mode-I fracture
energy on damage propagation in the adhesive layer.
As failure of specimens S300 and S310 was controlled by compression ange buckling, their behaviour may be affected by geometric imperfections such as those specied in Section 14.4.3 of
AS4100 [40]. As no measured geometric imperfections were
reported by Deng and Lee [13] for the test beams, the out-ofsquare imperfection, which was found to be the most inuential
for ange compression buckling among the three types of imperfections specied in AS4100 [40] (see [15]), was chosen for inclusion in the FE models; a magnitude of 1.3 mm was selected by a
trial-and-error process to match the ultimate load of the control
beam (i.e. specimen S300). Residual stresses, as described in Pi
and Trahair [41], were also included in the FE models [15]. It
should be noted that although the geometric imperfection and

Table 2
Key parameters for tractionseparation models.
Loading mode

Peak bond stress (MPa)

Displacement at peak bond stress, d1 (mm)

Interfacial fracture energy, Gf (N/mm)

Mode I (model A)
Mode I (model B)
Mode II

29.7
29.7
26.7

0.00371
0.00371
0.0526

0.0594
0.110
1.59

J.G. Teng et al. / Engineering Structures 86 (2015) 213224

the residual stresses adopted in the FE models may not be exactly


the same as those in the test beams, their effects on the predictions
are very limited: the geometric imperfection has little effect on the
loaddisplacement curve before the ultimate load which is controlled by the buckling of the steel section, and the residual stresses only have some effects on the slope of the curve close to the
yield load (i.e. the load at which the loaddisplacement becomes
nonlinear) (see [15]).
In total six FE models were developed (Table 1). The control
beam FE model is referred to as S300-0-000, where the rst 4 digits
indicate the test specimen conguration, the middle digit indicates
the model employed for the bond-separation behaviour under
mode-I loading (0 for the control beam as no CFRP strengthening
was provided), and the last three digits represent the elastic modulus of the CFRP plate in the bre direction given in GPa (000 as
no CFRP strengthening was provided in the control beam). The
same naming method is used in the paper for the other FE models.
Two FE models were developed for specimen S303 tested by
Deng and Lee [13], and they differ only in the bond-separation
model for mode-I loading. These two FE models are referred to as
models S303-1-212 and S303-2-212 respectively, where the numbers 1 and 2 in the middle indicate respectively the use of the
model A and the model B bond-separation laws for mode-I loading.
Besides these two FE models, an additional FE model (model S3031-330) was also created, with all the details being the same as
model S303-1-212 except that the elastic modulus of the CFRP
plate in the bre direction was increased to 330 GPa in model
S303-1-330. This additional model was created to investigate the
effect of axial stiffness of CFRP plate.
FE models S304-1-212 and S310-1-212 were respectively created for beams S304 and S310 tested by Deng and Lee [13]. To
examine the possibility of intermediate debonding, an additional

140
120

Load (kN)

218

100
80
S300- Experimental
(Deng and Lee 2007)

60
40

S300-0-000

20
0
0

10

30

Fig. 3. Loaddisplacement curves of a bare steel I beam.

FE model was built, where all the details are exactly the same as
those of model S310-1-212 except that an additional 6 mm thick
steel plate identical in material properties to the steel section
was added (using tied nodes on the plate edges in the FE model)
to the top ange of the steel section, so that the buckling of the
top ange can be suppressed. This FE model is referred to as
S310-1-212-P where P indicates the addition of a steel plate on
the top ange.
4. Results and discussions
4.1. Accuracy of assumed properties for the steel beam
The FE results are compared with the experimental loaddisplacement curve of the control beam (i.e. specimen S300) in
Fig. 3. As explained earlier, both the material stressstrain curve

Plate end debonding

Plate end debonding

Adhesive layer
Dark blue color: zero stress region
Red color: high stress region

(a) Deformed shape of S303-1-212 at failure

20

Mid-span deflection (mm)

Adhesive layer
Dark blue color: zero stress region
Red color: high stress region

(b) Deformed shape of model S304-1-212 at failure

Compression flange buckling

(c) Deformed shape of model S310-1-212 at failure


Fig. 4. Deformed shapes from FE models for CFRP-strengthened specimens.

219

Load (kN)

J.G. Teng et al. / Engineering Structures 86 (2015) 213224

debonding initiation
of S303-1-212

130
120
110
100
90
80
70
60
50
40
30
20
10
0

debonding
initiation of
S303-1-330

Debonding
initiation

debonding initiation of
S303-2-212
S303-Experimental
(Deng and Lee 2007)
S303-1-212

damage initiation of
S303-1-212 and
S330-2-212

300mm
Midspan

S303-2-212
damage initiation
of S303-1-330

S303-1-330

10

12

Mid-span deflection (mm)


(a) Load-deflection curves for specimen S303

Load (kN)

damage initiation of
S310-1-212-P

200
180
160
140
120
100
80
60
40
20
0

Debonding
initiation

Z
76mm

debonding initiation of
S310-1-212-P

(a) 92.6 kN
S304- Experimental
(Deng and Lee 2007)
S310- Experimental
(Deng and Lee 2007)
S304-1-212

debonding initiation of
S304-1-212

10

Fig. 6. Longitudinal shear stresses in the adhesive from S303-1-212. (a) 92.6 kN. (b)
118.3 kN.

S310-1-212

damage initiation of
S304-1-212

(b) 118.3 kN

Red colour: high stress region; Green colour:


intermediate stress region; Blue colour: low stress region

S310-1-212-P

20

30

4.2. Plate end debonding failures

Mid-span deflection (mm)


(b) Load-deflection curves for specimens S304 and S310
Fig. 5. Loaddeection curves. (a) Loaddeection curves for specimen S303. (b)
Loaddeection curves for specimens S304 and S310.

and the geometric imperfections of the steel beam employed in the


FE model were obtained through a trial-and-error process to
achieve a close prediction of the experimental loaddisplacement
curve of specimen S300. With these calibrated input data, the FE
results are seen to agree closely with the test results (Fig. 3). The
material and geometric properties adopted in the FE model for
the control beam are thus believed to approximate the experimental values well, and any errors arising from these input data are
believed to have negligible effects on the predicted response of
CFRP-strengthened steel beams.

In Deng and Lees tests [13], specimens S303 and S304 were
found to fail by the plate end debonding of the CFRP plate. The
same failure mode was also predicted by all the three FE models
of S303 (i.e. S303-1-212, S303-2-212, and S303-1-330) and the
FE model of S304 (i.e. S304-1-212). The failure mode (i.e. deformed
shape at ultimate load) obtained from model S303-1-212 is shown
in Fig. 4a while those for models S303-2-212 and S303-1-330 are
similar; the failure mode from model S304-1-212 is shown in
Fig. 4b.
The loaddeection curves obtained from these FE models are
compared with the experimental curve in Fig. 5; other key results
are summarized in Table 3. To further examine the FE results for
beams failing by debonding (i.e. S303, S304 and S310-1-212-P),
two key points are marked on each of the predicted loaddisplacement curves in Fig. 5: (1) the point when damage initiates in the

Table 3
Experimental and FE results.
FE model

S300-0000a
S303-1212b
S303-2212b
S303-1330b
S304-1212b
S310-1212a
S310-1212-Pc
a
b
c

Experimental results

FE results

Ultimate load,
Pu (kN)

Deection at ultimate
load, Du (mm)

Ultimate load,
Pu-FE (kN)

Deection at ultimate
load, Du-FE (mm)

Load at debonding
initiation, Pd-FE (kN)

Deection at debonding
initiation, Dp-FE (mm)

123

20.7

120

21.0

N/A

N/A

120

5.12

125

7.05

118

5.08

120

5.12

125

7.07

122

5.78

N/A

N/A

123

6.17

115

4.60

135

7.00

136

11.0

132

8.00

160

20.1

158

20.8

N/A

N/A

N/A

N/A

188

27.3

188

27.3

Compression ange buckling.


Plate end debonding.
Intermediate debonding.

J.G. Teng et al. / Engineering Structures 86 (2015) 213224

0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

Normal stress-150mm from


the mid-span

Normalized stress

Normalized stress

220

Longitudinal shear stress150mm from the mid-span


Normal stress-147.5mm from
the mid-span
Longituninal shear stress147.5mm from the mid-span

0.05

0.1

0.15

0.8

Longitudinal shearS303-1-212

0.7

Normal-S303-1-212

0.6
0.5

Longitudinal shearS303-2-212

0.4

Normal-S303-2-212

0.3
0.2

Longitudinal shearS303-1-330

0.1

Normal-S303-1-330

Strain

0
Fig. 7. Interfacial stressstrain behaviour at the plate end from model S303-1-212.

20

40

60

80

100

120

140

Load (kN)
Fig. 8. Normalized interfacial stresses at the plate end for specimen S303.

adhesive layer (i.e. the interfacial stresses start to decrease with


the bond displacements); (2) the point when debonding initiates
(i.e. when complete damage occurs at a certain location and the
interfacial stresses there reduce to zero). For the specimens failing
by plate end debonding (i.e. S303 and S304), the predicted loads at
the damage initiation point are seen to be much lower than the
corresponding ultimate loads achieved in the tests (Fig. 5). Considering that damage initiates when the strength of the adhesive is
reached, this observation demonstrates that the strength-based
approach (e.g. [2,12]) can substantially underestimate the load at
plate-end debonding, as found by Colombi and Poggi [1]. It is also
interesting to note that the predicted loads by FE models S303-1212 and S304-1-212 at the debonding initiation point (i.e.
118.3 kN and 132.0 kN) are both very close to their experimental
ultimate loads respectively (i.e. 120.0 kN and 135.0 kN); the

corresponding displacements predicted by the two FE models


(i.e. 5.08 mm and 8.00 mm) are also close to their respective experimental displacements at the ultimate load (i.e. 5.12 mm and
7.00 mm). This observation suggests that if failure of the strengthened beam is assumed to occur at the debonding initiation point,
these FE models can closely predict both the ultimate load and
the loaddisplacement curve up to the ultimate load. Such an
assumption is regarded to be reasonable as after the initiation of
debonding at a CFRP plate end, a sudden energy release can be
expected as the debonding propagation is a dynamic process driven by both the interfacial normal stresses and the interfacial shear
stresses. During this process, idealistic debonding propagation predicted by a static analysis (i.e. the part after the debonding initiation point on the predicted loaddisplacement curve) cannot

High
shear
stress

Midspan

Softening
near midspan

1000mm

Z
76mm
(a) 102 kN

(b) 140.5 kN

(c) 159.7 kN (peak load)

Red colour: high stress region; Green colour: intermediate stress region; Blue colour: low
stress region
Fig. 9. Longitudinal shear stresses in the adhesive from model S310-1-212. (a) 102 kN. (b) 140.5 kN. (c) 159.7 kN (peak load).

J.G. Teng et al. / Engineering Structures 86 (2015) 213224

Normalized stress (MPa)

occur; sudden debonding propagation is likely to happen instead


which means no further load increases can occur in a test.
The interfacial stress distributions over the adhesive layer in
model S303-1-212 are shown in Fig. 6 for two different load levels.
The interfacial stress distributions in the other two S303 FE models
and the S304 FE model are similar and are thus not provided here.
Damage is seen to initiate at the two plate ends and propagates
towards the mid-span (Fig. 6). Fig. 7 compares the normalized normal stress (i.e. normalized by the tensile strength)normal strain
curves and the normalized shear stress (i.e. normalized by the peak
bond shear stress)shear strain curves for the adhesive at the end
of the plate (i.e. 150 mm from the mid-span) and those for the
adhesive at 2.5 mm away from the plate end (i.e. 147.5 mm from
the mid-span). From Fig. 7 it can be seen that, after the initiation
of damage at a plate end, both the interfacial normal stress and
the longitudinal shear stress at the very end of the plate decrease,
but the shear stress in the region nearby starts to increase. Fig. 7
also shows that the maximum normal stress and the maximum
longitudinal shear stress are equally high at the very end of the
plate, but the former is signicantly lower than the latter at a location 2.5 mm away from the plate end, suggesting that the signicant effect of the normal stress on damage propagation of the
interface is limited only to a small region close to the plate end.

0.25
0.2
0.15
0.1
0.05
0
-0.05 0
-0.1
-0.15
-0.2

Normal
stress

100

200

300

400

500

Longitudinal
shear stress

Distance from the mid span (mm)


(a) 102 kN
Normalized stress (MPa)

0.8
0.6
0.4

Normal stress

0.2
0
-0.2

100

200

300

400

500
Longitudinal
shear stress

-0.4
-0.6

Distance from the mid span (mm)

Normalized stress (MPa)

(b) 140.5 kN
1
0.8
0.6
0.4
0.2
0
-0.2 0
-0.4
-0.6
-0.8
-1

Normal stress

100

200

300

400

500

Longitudinal
shear stress

Distance from the mid span (mm)


(c) 159.5 kN (peak load)
Fig. 10. Interfacial stress distributions along section YY in Fig. 9a, from model
S310-1-212. (a) 102 kN. (b) 140.5 kN. (c) 159.5 kN (peak load).

221

4.3. Effect of mode-I fracture energy and CFRP plate stiffness


Fig. 5a shows that FE models S330-1-212 and S303-2303-2-212
predict very similar loaddeection curves despite the use of different bond-separation models. The loaddeection curve predicted by model S303-1-330 is slightly higher than the other two
after the initial linear portion because of the use of a stiffer CFRP
plate, but ends at a lower ultimate load. It is also seen that the
curves predicted by FE models S303-1-212 and S303-2-212, for
which the CFRP properties adopted are the same as those of the
test specimen, are very close to the experimental curve except that
they both have a higher ultimate load (Fig. 5a).
The interfacial normal stresses and the interfacial longitudinal
shear stresses at the plate end (i.e. 150 mm from the mid span)
from the three models (i.e. S303-1-212, S303-2-212 and S303-1330) are shown in Fig. 8. The magnitudes of the interfacial transverse shear stresses are very small, so they are not shown in this
gure. It is clear that damage initiates at a load of 83 kN in FE models S303-1-212 and S303-2-212, but initiates at a smaller load (i.e.
71 kN) in model S303-1-330. Before damage initiation, the normalized normal stress and the normalized shear stress are similar in FE
models S303-1-212 and S303-2-212, and are lower than those in
model S303-1-330, indicating that a stiffer CFRP plate leads to larger interfacial stresses. In addition, the load at the debonding initiation point is smaller from model S303-1-330 (i.e. 115 kN) than
those from the other two models (i.e. around 120 kN), suggesting
that steel beams strengthened with a stiffer CFRP plate are likely
to fail at a smaller load by plate end debonding.
It is also interesting to note that although quite different bondseparation models for mode-I loading were employed, the predictions of FE models S303-1-212 and S303-2-212 are very similar
(Fig. 5a). The predictions are exactly the same before the initiation
of damage at the plate end (Fig. 8). After damage initiation, the
interfacial stresses in model S303-1-212 are seen to decrease
slightly more rapidly with the load than those in model S303-2212, as the mode-I interfacial fracture energy adopted in the former is smaller which leads to a smaller total interfacial fracture
energy at failure. However, as the mode-II fracture energy (i.e.
1.59 N/mm) adopted by both FE models is much larger than the
mode-I fracture energy (i.e. 0.059 N/mm for model S303-1-212
and 0.11 N/mm for model S303-2-212), the use of a larger modeI fracture energy in model S303-2-212 has only a small effect on
the total fracture energy at failure for mixed-mode loading. This
explains the very similar predictions of the two models. It should
be noted that for linear adhesives commonly used in CFRP-to-steel
bonded joints, the mode-II fracture energy is often much larger
than the mode-I fracture energy [31], so the debonding of such
joints under mixed-mode loading is often governed by the modeII fracture energy. This also suggests that the method adopted in
the present study for estimating the mode-I fracture energy (i.e.
the method used for deriving bond-separation model A in Table 2)
can work well for common linear adhesives.

4.4. Compression ange buckling failures


Deng and Lee [13] indicated that beam S310 failed by the buckling of the compression ange of the steel section. The same failure
mode was also predicted by model S310-1-212. The deformed
shape at failure obtained from model S310-1-212 is shown in
Fig. 4c. The loaddeection curve predicted by model S310-1-212
is seen to compare very well with the experimental results
(Fig. 5b). It should also be noted that, while S300-0-000 and
S310-1-212 both failed due to compression ange buckling, the
latter achieved a higher load carrying capacity than the former
due to the contribution of the bonded FRP plate.

222

J.G. Teng et al. / Engineering Structures 86 (2015) 213224

1000mm
High
longitudinal
shear
stresses

Midspan

Z
76mm

(a) 33.7kN

Midspan

Debonding

(b) 159.5kN

Debonding

initiation

(c) 188.05kN (peak load)

(d) 186.7kN (post peak curve)

Red colour: high stress region; Green colour: intermediate stress region; Blue colour: low
stress region
Fig. 11. Damage propagation in the adhesive layer in model S310-1-212-P. (a) 33.7 kN. (b) 159.5 kN. (c) 188.05 kN (peak load). (d) 186.7 kN (post peak curve).

The interfacial longitudinal shear stress patterns over the adhesive layer at different load levels are shown in Fig. 9, while the normalized interfacial stress distributions along section ZZ at
different load levels are shown in Fig. 10. Before the load reaches
102 kN, both the normal stress and the longitudinal shear stress
are relatively low, and the maximum interfacial stresses occur at
the plate end (Fig. 10a). As the load increases, the longitudinal
shear stress at the mid-span becomes higher than those at the
plate ends (Fig. 10b). At the ultimate load, softening in the region

close to the mid-span has already begun (Fig. 10c), but no debonding occurs before the buckling of the compression ange.
4.5. Intermediate debonding failures
As expected, the failure mode predicted by model S310-1-212-P
is the intermediate debonding of the CFRP plate initiating from
near the mid-span (Fig. 11c). The loaddeection curve predicted
by model S310-1-212-P is shown in Fig. 5b. The interfacial

J.G. Teng et al. / Engineering Structures 86 (2015) 213224

Normalized stress (MPa)

0.8
0.6
0.4

Normal stress

seen to have dropped signicantly (Fig. 12c). The longitudinal


shear stress distribution at a post-peak state (Fig. 12d) clearly indicates the existence of two debonded portions of the interface on
the two sides of the mid-span.

0.2
0
-0.2

100

200

300

400

500

Longitudinal
shear stress

-0.4
-0.6

Distance from the mid span (mm)


(a) 176.8 kN
Normalized stress (MPa)

223

1.2
1
0.8
0.6
0.4

Normal stress

0.2
0
-0.2 0

100

200

300

400

500

-0.4

Longitudinal
shear stress

-0.6

Distance from the mid span (mm)

4.6. Possible failure modes of CFRP-strengthened steel beams


The discussions above indicate that the proposed FE approach
can provide accurate predictions for the response of CFRPstrengthened steel beams failing in different failure modes, in
terms of both the ultimate load and the loaddisplacement curve.
In CFRP-strengthened steel beams, the contribution of the CFRP
plate relies on the stress transfer function of the adhesive layer, so
the ultimate load of such beams is often governed by the failure of
the adhesive layer. When a short CFRP plate is used, the interfacial
stresses (i.e. both the normal stress and the shear stress) at the
plate end are large, and the failure is often governed by plate end
debonding. When a longer CFRP plate is used, the interfacial stresses at the plate end are smaller and the critical region of the adhesive layer may move to the mid-span region. In such cases, the
failure mode may change from plate end debonding to intermediate debonding, buckling of the compression ange of the steel section, and tensile rupture of the FRP plate.

(b) 184.8 kN
Normalized stress (MPa)

5. Conclusions
1
0.8
0.6
0.4
0.2
0
-0.2 0
-0.4
-0.6
-0.8
-1

Normal stress

100

200

300

400

500

Longitudinal
shear stress

Distance from the mid span (mm)

Normalized stress (MPa)

(c) 188.05 kN (peak load)


1
0.8
0.6
0.4
0.2
0
-0.2 0
-0.4
-0.6
-0.8
-1

Normal stress

100

200

300

400

500

Longitudinal
shear stress

Distance from the mid span (mm)


(d) 186.7 kN (a peak-peak state)
Fig. 12. Interfacial stress distributions along section ZZ in Fig. 11a, from model
S310-1-212-P. (a) 176.8 kN. (b) 184.8 kN. (c) 188.05 kN (peak load). (d) 186.7 kN (a
peakpeak state).

longitudinal shear stress patterns over the adhesive layer at different load levels are shown in Fig. 11, while Fig. 12 shows the
normalized interfacial stress distributions along section ZZ
(Fig. 11a) at different load levels. The interfacial stress distributions
from model S310-1-212-P at low load levels are similar to those
from model S310-1-212. However, as the load increases, a large
increase in the longitudinal shear stress near the mid-span is seen,
which also means a higher contribution of the CFRP plate. At the
ultimate load, the longitudinal shear stress near the mid-span is

This paper has been concerned with the accurate prediction of


debonding failures in simply-supported steel beams strengthened
in exure with a bonded CFRP soft plate using the FE method.
An FE approach has been presented in the paper, in which bi-linear tractionseparation models are employed to represent pure
mode-I and pure mode-II responses of the interface for a linear
adhesive; a mixed-mode cohesive law is employed to consider
interactions between mode-I loading and mode-II loading. Damage initiation is dened using a quadratic strength criterion,
and damage evolution is dened using a linear fracture energybased criterion, both of which take due account of mixed-mode
loading. The proposed FE approach represents a signicant
advancement in the modelling of debonding failures in CFRPstrengthened steel structures.
Predictions from the FE approach have been shown to compare
well with the test results reported by Deng and Lee [13] for CFRPstrengthened beams failing by either the plate-end debonding of
the CFRP plate or the compression ange buckling of the steel section. It was also concluded from the study that when a static FE
analysis is conducted, the ultimate load of a beam failing by plate
end debonding should be taken as the load at which debonding initiates at the plate end.
Using the proposed FE approach, the behaviour of CFRPstrengthened steel beams was examined and it was found that:
(1) a CFRP plate with a higher elastic modulus and/or a larger
thickness leads to a lower ultimate load by plate end debonding;
(2) plate end debonding is more likely to occur when a short CFRP
plate is used, as is commonly expected; and (3) the failure mode
may change to intermediate debonding or other modes such as
compression ange buckling if sufciently long CFRP plate is used.

Acknowledgements
The authors are grateful for the nancial support received from
The Hong Kong Polytechnic University provided through an International Postgraduate Scholarship for PhD Studies to the second
author and through a Postdoctoral Fellowship to the third author.

224

J.G. Teng et al. / Engineering Structures 86 (2015) 213224

References
[1] Colombi P, Poggi C. An experimental, analytical and numerical study of the
static behavior of steel beams reinforced by pultruded CFRP strips. Compos B:
Eng 2006;37(1):6473.
[2] Lenwari A, Thepchatri T, Albrecht P. Debonding strength of steel beams
strengthened with CFRP plates. J Compos Constr 2006;10(1):6978.
[3] Sallam HEM, Ahmad SSE, Badawy AAM, Mamdouh W. Evaluation of steel Ibeams strengthened by various plating methods. Adv Struct Eng
2006;9(4):53544.
[4] Youssef MA. Analytical prediction of the linear and nonlinear behaviour of
steel I beams rehabilitated using FRP sheets. Eng Struct 2006;28(6):90311.
[5] Benachour A, Benyoucef S, Tounsi A, Adda bedia EA. Interfacial stress analysis
of steel beams reinforced with bonded prestressed FRP plate. Eng Struct
2008;30(11):330515.
[6] Linghoff D, Haghani R, Al-Emrani M. Carbon-bre composites for
strengthening steel structures. Thin-Wall Struct 2009;47(10):104858.
[7] Teng JG, Yu T, Fernando D. Strengthening of steel structures with bre
reinforced polymer composites. J Constr Steel Res 2012;78:13143.
[8] Nozaka K, Shield CK, Hajjar JF. Effective bond length of carbon ber reinforced
polymer strips bonded to fatigued steel bridge I-girders. J Bridge Eng
2005;9(4):30412.
[9] Nozaka K, Shield CK, Hajjar JF. Design of a test specimen to assess the effective
bond length of carbon ber reinforced polymer strips bonded to fatigued steel
bridge girders. J Compos Constr 2005;10(2):195205.
[10] Dawood M, Rizkalla S, Sumner E. Fatigue and overloading behavior of steelconcrete composite exural members strengthened with high modulus CFRP
materials. J Compos Constr 2007;11(6):65969.
[11] Schnerch D, Rizkalla S. Flexural strengthening of steel bridges with high
modulus CFRP strips. J Bridge Eng 2008;13(2):192201.
[12] Schnerch D, Dawood M, Rizkalla S, Sumner E. Proposed design guidelines for
strengthening of steel bridges with FRP materials. Constr Build Mater
2007;21(5):100110.
[13] Deng J, Lee MMK. Behaviour under static loading of metallic beams reinforced
with a bonded CFRP plate. Compos Struct 2007;78(2):23242.
[14] Teng JG, Smith ST, Yao J, Chen JF. Intermediate crack-induced debonding in RC
beams and slabs. Constr Build Mater 2003;17(67):44762.
[15] Fernando ND. Bond behaviour and debonding failures in CFRP-strengthened
steel members. PhD thesis, The Hong Kong Polytechnic University, Hong Kong,
China; 2010.
[16] Deng J, Lee MMK, Moy SSJ. Stress analysis of steel beams reinforced with a
bonded CFRP plate. Compos Struct 2004;65(2):20515.
[17] Zhang L, Teng JG. Finite element prediction of interfacial stresses in structural
members bonded with a thin plate. Eng Struct 2010;32(2):45971.
[18] De Lorenzis L, Fernando D, Teng JG. Coupled mixed-mode cohesive zone
modeling of interfacial stresses in plated beams. Int J Solids Struct
2013;50(1415):247794.
[19] Yu T, Fernando D, Teng JG, Zhao XL. Experimental study on CFRP-to-steel
bonded interfaces. Compos B: Eng 2012;43(5):227989.
[20] Fernando D, Teng JG, Yu T, Zhao XL. Preparation and characterization of steel
surfaces for adhesive bonding. J Compos Constr 2013;17(6):04013012.

[21] De Moura MFSF, Chousal JAG. Cohesive and continuum damage models
applied to fracture characterization of bonded joints. Int J Mech Sci
2006;48(5):493503.
[22] Yuan H, Xu Y. Computational fracture mechanics assessment of adhesive
joints. Comput Mater Sci 2008;43(1):14656.
[23] De Lorenzis L, Zavarise G. Cohesive zone modeling of interfacial stresses in
plated beams. Int J Solids Struct 2009;46(24):418191.
[24] Sorensen BF. Cohesive law and notch sensitivity of adhesive joints. Acta Mater
2002;50(5):105361.
[25] Li S, Thouless MD, Waas AM, Schroeder JA, Zavattieri PD. Mixed-mode
cohesive-zone models for fracture of an adhesively bonded polymer-matrix
composite. Eng Fract Mech 2006;73(1):6478.
[26] Goncalves JPM, de Moura MFSF, Magalhaes AG, de Castro PMST. Application of
interface nite elements to three-dimensional progressive failure analysis of
adhesive joints. Fatigue Fract Eng Mater Struct 2003;26(5):47986.
[27] Liljedahl CDM, Crocombe AD, Wahab MA, Ashcroft IA. Damage modelling of
adhesively bonded joints. Int J Fract 2006;141(12):14761.
[28] Bocciarelli M, Colombi P, Fava G, Poggi C. Interaction of interface delamination
and plasticity in tensile steel members reinforced by CFRP plates. Int J Fract
2007;146(12):7992.
[29] Pardoen T, Ferracin T, Landis CM, Delannay F. Constraint effects in adhesive
joint fracture. J Mech Phys Solids 2005;53(9):195183.
[30] 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(5):1497512.
[31] Hogberg JL. Mixed mode cohesive law. Int J Fract 2006;141(34):54959.
[32] Xu XP, Needleman A. Void nucleation by inclusion debonding in a crystal
matrix. Modell Simul Mater Sci Eng 1993;1(2):11132.
[33] Da vila CG, Johnson ER. Analysis of delamination initiation in post buckled
dropped-ply laminates. AIAA J 1993;31(4):7217.
[34] Camanho PP, Davila CG, de Moura MF. Numerical simulation of mixed-mode
progressive delamination in composite materials. J Compos Mater
2003;37(16):141538.
[35] Benzeggagh ML, Kenane M. Measurement of mixed-mode delamination
fracture toughness of unidirectional glass/epoxy composites with mixedmode bending apparatus. Compos Sci Technol 1996;56(4):43949.
[36] Xie D, Chung J, Wass AM, Shahwan KW, Schroeder JA, Boeman RG, Kunc V,
Klett LB. Failure analysis of adhesively bonded structures: from coupon level to
structural level predictions and verication. Int J Fract 2005;134:23150.
[37] ABAQUS Users Manual. Version 6.5; Inc., Rising Sun Mills, 166 Valley Street,
Providence, RI 02909-2499, USA; 2004.
[38] Bayeld MP, Davies JM, Dhanalakshmi M. Calculation of the strain hardening
behavior of steel structures based on mill tests. J Constr Steel Res
2005;61(2):13350.
[39] BS EN 10025-1. Hot rolled products of non-alloy structural steels. Technical
delivery conditions. British Standards Institution; 2004.
[40] AS4100. Steel structures. Standards Australia, NSW 2142, Australia; 1998.
[41] Pi YL, Trahair NS. Inelastic bending and torsion of steel I-beams. J Struct Eng
1994;120(12):3397417.

Das könnte Ihnen auch gefallen