Construction and Building Materials 31 (2012) 74–85

Contents lists available at SciVerse ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier.com/locate/conbuildmat

Nonlinear 3D finite element modeling of RC beams strengthened

with prestressed NSM-CFRP strips
Hamid Y. Omran, Raafat El-Hacha ⇑
Dept. of Civil Eng., Univ. of Calgary, 2500 Univ. Drive N.W., Calgary, Canada T2N 1N4

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

Article history: This paper demonstrates a comprehensive 3D nonlinear Finite Element (FE) analysis of Reinforced Con-
Received 4 May 2011 crete (RC) beams strengthened with prestressed Near-Surface Mounted (NSM) Carbon Fiber Reinforced
Received in revised form 6 November 2011 Polymer (CFRP) strips. Debonding effect at the epoxy-concrete interface was considered in the model
Accepted 2 December 2011
by identification of fracture energies of the interfaces and appropriate bilinear shear stress-slip and ten-
Available online 21 January 2012
sion stress-gap models. Prestressing was applied to the CFRP strips by adopting the equivalent temper-
ature method. A constitutive confined concrete model for flexural member was generated from
Keywords:
unconfined concrete curve and assigned to the concrete materials. The results from the FE model were
Finite element modeling
Debonding
validated with experimental data available in the literature. Comparison between FE and test results con-
Prestressing firms excellent accuracy of the proposed model. An optimum prestressing level in the NSM-CFRP strips
Near-surface mounted was determined that enhances the beam performance under service and ultimate loads by maintaining
Carbon fiber reinforced polymers the amount of energy absorption in the strengthened beam equal to the un-strengthened control beam.
Fracture energy Ó 2011 Elsevier Ltd. All rights reserved.
Strengthening
Reinforced concrete beam

1. Introduction sider the local debonding in the FE model of non-prestressed

Upgrading Reinforced Concrete (RC) members using Fiber Rein-
forced Polymer (FRP) reinforcements has become one of the most
commonly used strengthening methods in recent years. FRP
strengthening methods are classified in two general techniques:
Externally Bonded (EB) and Near-Surface Mounted (NSM).
Currently, prestressing is employed the with the NSM-FRP
strengthening technique to enhance the flexural performance of
strengthened beam [1–3].
Numerical and analytical investigations of FRP strengthened
concrete members, with more focus on the EB technique, have
been extensively pursued parallel to the experimental work and
practical applications. Many researchers simulated the behavior
of EB strengthened RC flexural members using 2D/3D FE models
considering perfect bond between interfaces (concrete–epoxy
and epoxy-FRP) due to the fact that debonding failure was not ob-
served in relevant tests [4–11]. However, a few researchers consid-
ered the debonding effects in 2D FE modeling of EB strengthened
RC beam [12–15]. On the other hand, FE modeling of NSM-FRP
strengthened RC beams is rarely carried out [16–18]. Soliman
et al. [18] employed experimental shear stress-slip curve to con-
⇑ Corresponding author. Tel.: +1 403 220 4817; fax: +1 403 282 7026.
E-mail address: relhacha@ucalgary.ca (R. El-Hacha).
NSM-FRP strengthening technique. In the NSM-FRP strengthened
RC beam, the debonding occurs at the concrete–epoxy interface
which is the weakest interface and the main reason for shortage
of research in this field is the identification of appropriate bond
behavior that can be reasonably applicable to the NSM technique.
Some researchers have investigated the bond-strength behavior in
EB technique, the most well-known ones are the proposed models
by Chen and Teng [19] and Lu et al. [20]; however, these models
are derived based on the geometry of the EB plate and are not accu-
rate for NSM technique. Seracino et al. [21] derived an equation to
calculate the debonding resistance of NSM and EB plate-to-
concrete joints based on the geometry of the interface debonding
failure plane. Although FE modeling of prestressed EB-FRP
strengthened beams has rarely been performed even with consid-
ering complete bond [22], FE modeling of prestressed NSM-FRP
strengthened RC beams has never been investigated taking into ac-
count the debonding effects.
Three main objectives are outlined for this paper:
(1) To develop a 3D nonlinear FE model to simulate the exact
behavior of RC beams strengthened with prestressed-NSM-
CFRP strips. The FE model was verified with experimental
test results available in the literature [23]. The FE model
accounts for the debonding behavior at the concrete–epoxy
interface considering a bilinear shear stress-slip model and
a bilinear normal tension stress-gap model derived from

0950-0618/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conbuildmat.2011.12.054
 H.Y. Omran, R. El-Hacha / Construction and Building Materials 31 (2012) 74–85
fracture energies of the interface. From an extensive litera-
ture survey, only one reference was found which considered
both models in the FE modeling of EB FRP strengthening
technique [15], however, no work has been done with the
NSM-FRP strengthening technique by considering both
models (bilinear shear stress-slip and bilinear normal ten-
sion stress-gap). Also, the prestressing is enforced to the
CFRP strips using the equivalent temperature method.
(2) To analyze the flexural behavior of strengthened RC beams
using prestressed NSM-CFRP strips in terms of load–deflec-
tion curve, bond performance, strain profile, failure mode,
and energy absorption by considering different parameters
including the effects of prestressing and debonding.
(3) To achieve a practical optimum prestressing level in the
NSM-CFRP strips that enhances the beam performance
under service and ultimate loads by maintaining the amount
of energy absorption in the strengthened beam equal to the
un-strengthened beam.
2. Experimental program overview
Five RC beams were tested [23]: one un-strengthened control beam, and four
strengthened beam with prestressed NSM-CFRP strips. Each beam was strength-
ened using two 2  16 mm CFRP strips glued together from the width and embed-
ded in one groove pre-cut in the concrete cover on the beam’s tension soffit. The
beams were, 5000 mm long, simply supported with rectangular cross section of
200  400 mm. Various prestressing levels of 0%, 17%, 29.3%, and 51% of the ulti-
mate tensile strain of the CFRP strips were enforced to the NSM-CFRP strips (corre-
sponding to prestrain of 0, 0.0034, 0.00587, and 0.0102 equivalent to a prestressing
force of 0 kN, 28.4 kN, 49.03 kN, and 85.2 kN, respectively). The CFRP strips were
prestressed using an innovative anchorage system that consisted of two steel an-
chors bonded to the ends of the CFRP strips using epoxy and movable brackets tem-
porary mounted on the beam. More details on the system can be found in Refs.
[2,23]. The beams were tested under monotonic static loading in four-point bending
configuration [23].
Fig. 1 depicts the details of the strengthened beams. Fig. 2 represents the stress–
strain curve of the internal steel reinforcements determined from the uni-axial ten-
sion tests [23]. The concrete material possesses a Young’s modulus of 27.84 GPa, a
maximum compressive strength of 40 MPa, and a strain at ultimate strength of
0.002233, which are the average values obtained from the compression tests of con-
crete cylinders [23]. Modeling of the concrete stress–strain behavior is described in
1000
(a) Actuator Steel spreader beam
10M@200 10M@300 2-10M A 3-15M
Near-surface mounted A
End anchor
CFRP strengthening
4380
5000
5150
(b)
Fig. 1. Geometry of the tested beams: (a) elevation and (b) cross-section, grooves,
and end anchor.
75
700
600
500
Stress (MPa)
400
10M Steel Bar
300
15M Steel Bar
200
100
0
0 0.02 0.04 0.06 0.08 0.1 0.12
Strain (mm/mm)
Fig. 2. Stress–strain curves of steel bars.
Section 3.4.1. The employed CFRP strip was type Aslan 500 produced by Hughes
Brothers with a tensile strength of 2610 MPa, an ultimate strain of 0.02, and a
Young’s modulus of 130.5 GPa determined from the tension test [23] with a lin-
ear-elastic behavior up to failure. Two types of epoxy adhesives were used: Type
A (SikadureÒ 330) with a Young’s modulus of 4.5 GPa and an ultimate tensile
strength of 30 MPa [24] used in the end groove regions (around and inside the
end anchors), and epoxy Type B (SikadureÒ 30) with a Young’s modulus of
4.5 GPa and ultimate tensile strength of 24.8 MPa [25] used to fill in the intermedi-
ate groove regions between the end anchors.
3. FE analysis
3.1. Finite element model
The developed FE models are 3D and all materials including
concrete, CFRP strips, longitudinal steel reinforcements, stirrups,
epoxy adhesive, bolts, and end anchor were simulated using appro-
priate elements available in the ANSYS program library [26]. To re-
duce the computer computational time, modeling time, and
volume of the results’ file, only one quarter of the beam was mod-
eled due to the symmetry in geometry and loading conditions, as
shown in Fig. 3. Out of two mesh generation techniques: solid
modeling and direct generation, the latter was employed to gener-
ate the FE model due to intricacy of the NSM strengthened beams.
The direct generation technique has the advantage of complete
control over the geometry of every node and every element; even
if it is too time consuming for generating large-scale model and the
modeler needs to focus more on every detail of the mesh.
3.2. Debonding model
To implement debonding aspects in the FE model, the concrete–
epoxy interface, which is the weakest layer, is modeled using
contact pairs and Cohesive Zone Material (CZM) model. In most
studies, debonding was analyzed based on shear stress-slip
(bond-slip) of the interfaces [12–14,18]. In this study a normal
stress-gap model is employed in addition to the shear stress-slip
model enabling mixed-mode debonding. It should be noted, the
rational of considering both models provides an opportunity to
accurately analyze the debonding behavior; considering only the
shear stress-slip model leads to a mode of separation of the inter-
face surfaces in which relative tangential displacement (slip)
dominates the separation normal to the interface (gap); on the
other hand, considering just the tension stress-gap model leads
to a mode of failure in which the separation normal to the interface
dominates the slip tangent to the interface. Therefore, defining the
debonding based on only one of these two models means ignoring
the effect of the other and generating a difference with reality
where both models contribute to debonding. The bilinear shear
76 H.Y. Omran, R. El-Hacha / Construction and Building Materials 31 (2012) 74–85

Fig. 3. The quarter of the beam to be modeled.

stress-slip and normal stress-gap models are calculated based on (a)

the appropriate fracture energies of the concrete–epoxy interface
which are identified later on.
A surface-to-surface contact pairs were applied to the interface
between concrete and adhesive to separate the fine mesh mapped
inside the groove from surrounding mesh. Fine mesh was mapped
inside the strengthening groove to increase the accuracy of the re-
sults, and it was not necessary to map fine mesh outside the groove
to decrease the modeling time and computer computational time.
The target surface (on concrete) was modeled with TARGE170 and
the contact surface (on adhesive) was modeled with CONTA173.
TARGE170 is capable to represent various 3D target surfaces for
the associated contact elements. CONTA173 is used to represent
contact and sliding between 3D target surfaces and a deformable
surface, defined by this element. The element is defined by four
nodes, considering a 2  2 set of Gaussian integration points, and (b)
is applicable to 3D structural and coupled field contact analyses.
The related shear stress-slip and tensile stress-gap model were as-
signed to the contact surface by developing a subroutine in ANSYS
command menu. The CZM is employed for bonded contacts with
the augmented Lagrangian method.
As mentioned earlier, in mixed-mode debonding the interface
separation depends on both normal and shear components. The
bilinear shear stress-slip and normal tension stress-gap models
are presented in Fig. 4. The area under each curve is equal to the
relevant fracture energy of the interface in shear or tension which
is the energy released due to debonding. In Fig. 4a and b, debond-
ing is initiated when dm = 0 and is terminated when dm = 1. Any
further separation occurs without any normal or shear contact
stress.
Fig. 4. (a) Bilinear normal tension stress-gap and (b) bilinear shear stress-slip
The equations of the shear and the normal contact stresses are models.
defined as:
st ¼ K t ut ð1  dm Þ ðMPaÞ ð1Þ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  2
un ut
Dm ¼ þ ðmm=mmÞ ð4Þ
rn ¼ K n un ð1  dm Þ ðMPaÞ ð2Þ un t
u
where st is the shear contact stress, Kt the shear contact stiffness, ut
the contact slip, rn the normal contact stress, Kn the normal contact ucn uct
stiffness, un the contact gap, and dm is the debonding parameter x¼ ¼ ðmm=mmÞ ð5Þ
ucn  u  n uct  u t
which is defined using Eqs. (3)–(5) [26].
 
Dm  1 where u n is the contact gap at the maximum normal contact stress
dm ¼ x ðmm=mmÞ ð3Þ
Dm (tension), u t the contact slip at the maximum shear contact stress,
 H.Y. Omran, R. El-Hacha / Construction and Building Materials 31 (2012) 74–85
ucn the contact gap at the completion of debonding, and uct is the
contact slip at the completion of debonding.
In the above equations, dm = 0 for Dm 6 1 and 0 < dm 6 1 for
Dm > 1. The constraint on x (Eq. (5)) is enforced automatically by
appropriately scaling the contact stiffness values. In mixed-mode,
debonding is completed before the critical fracture energy values
are reached for the components because both normal and shear
contact stresses contribute to the total fracture energy. It is as-
sumed that no slip occurs at the interface under pure normal com-
pressive stress for mixed-mode debonding; however, slip and gap
are considered at the interface under shear and normal tensile
stresses which contribute to debonding based on the defined shear
stress-slip and tension stress-gap models. The main reasons for
this assumption are: (1) in the present work, the interface is mostly
under shear and normal tension stresses; (2) debonding is a com-
bination of shear stress and normal tension stress (as a result of
produced slip and gap) and even under compressive stress this is
the secondary effect of the compression that produces tension or
shear stresses at some regions at the interface and results in slip
or gap which are considered in the model. As mentioned earlier,
in most studies, only the shear stress-slip (bond-slip) of the inter-
faces was considered to analyze debonding and the contribution of
tensile stress was ignored [12–14,18]. The following energy crite-
rion is used to define the completion of debonding in the mixed
mode [26].
R R
rn dun rt dut
þ ¼1
Gcn Gct
where Gcn and Gct are the total values of normal and shear fracture
energies, respectively, which should be identified for the NSM
method.
3.2.1. Identification of bond-slip model
Many precise shear stress-slip models were proposed for the EB
strengthening technique [19,20], but these models are not appro-
priate for NSM technique because they are developed according
to the geometry of the FRP plate which is approximately the same
as the geometry of the interface debonding failure plane in EB tech-
nique. In the NSM technique, it is not an accurate approximation to
use the geometry of the plate/strip instead of the geometry of the
actual debonding failure plane. On the other hand, the plate strain
at debonding in the NSM technique is significantly higher than the
EB technique due to confinement of the surrounding concrete. Ser-
acino et al. [21] analyzed debonding resistance of the EB and NSM
plate-to-concrete joints and overcame the drawbacks of the previ-
ous models by considering the geometry of the debonding failure
plane; these authors proposed a fracture energy and a maximum
interface shear stress equation based on the statistical analysis of
experimental values. The considered bond-slip model (Fig. 4b) is
derived from fracture energy of the interface in NSM strengthened
specimens (Eqs. (7)–(10)).
smax ¼ ð0:802 þ 0:078uÞfc00:6 ðMPaÞ
0:976u0:526 fc00:6
Gct ¼ ðN=mmÞ
2 elastic shortening in the concrete beam reduces the strain in the
0:976u0:526
uct ¼ ðmmÞ
0:802 þ 0:078u
Groove depth þ 1 mm
u¼ ðmm=mmÞ
Groove width þ 2 mm
where smax is the maximum shear contact stress, u the aspect ratio
of the interface failure plane, fc0 the concrete compressive strength,
Gct the shear fracture energy, and uct the contact slip at the
77
completion of debonding. Using Eqs. (7)–(10), the values of
smax = 8.37 MPa and uct = 1.295 mm are assigned to the shear
stress-slip model.
3.2.2. Identification of normal stress-gap model
The tensile resistance at the concrete–epoxy interface is as-
sumed to be limited to the concrete tensile strength (Eq. (11)
[27]) [15]. Thus, when the bonding strength under tension exceeds
the concrete tensile strength, failure in the concrete occurs which
was also observed during the tests [23]. Therefore, the fracture en-
ergy of the interface under tensile stress is considered to be equal
to the fracture energy of the concrete proposed by CEB-FIP Model
Code (Eq. (12)) [28]. The contact gap (Eq. (13)) is derived using
Eqs. (11) and (12) to meet the tensile fracture energy based on
the maximum tensile stress:
qffiffiffiffi
rmax ¼ 0:6 fc0 ðMPaÞ ð11Þ
 0:7
fc0
Gcn ¼ Gfo ðN=mmÞ ð12Þ
10
pffiffiffiffiffiffi !0:2
10fc0
ucn ¼ Gfo ðmmÞ ð13Þ
24:3
where is the rmax is the maximum normal contact stress, fc0 the con-
crete compressive strength, Gcn the total value of the normal frac-
ð6Þ
ture energy, ucn the contact gap at the completion of debonding,
and Gfo is the base value of fracture energy which depends on the
maximum aggregate size. For concrete with maximum aggregate
size of 20 mm, Gfo is calculated as 0.03475 N/mm by interpolation
between different values of aggregate size reported in CEB-FIP Mod-
el Code [28]. Using Eqs. (11)–(13), = values of rmax = 3.79 MPa and
ucn ¼ 0:048 mm are assigned to the tension stress-gap model.
3.3. Modeling of prestressing
There are three methods to apply prestressing in FE modeling:
enforce initial strain, enforce initial stress, and enforce equivalent
temperature to meet the same prestrain in relevant materials.
For simplicity in the pre-processing and post-processing opera-
tions and for the considered type of selected elements as well
the assigned material properties for the CFRP strips, the equivalent
temperature method was implemented in the current research to
apply the prestressing effect. According to the FRP material data
sheet, thermal expansion of the CFRP strip, afrp, is set as
9  106 °C [29]. The equivalent temperature is calculated and
assigned to the CFRP elements to enforce the exact prestrain in
CFRP for each prestressed beam. After assigning the calculated
equivalent temperature (Dt) to the model it is found that due to
equilibrium the produced prestrain is slightly smaller than the pre-
strain value (eP). In fact, the change in prestressing is a result of the
ð7Þ straining of the beam as the self-equilibrating stress state estab-
lishes itself during an equilibrating static analysis step. Enforcing
the prestressing transfers load to the concrete beam, causing com-
ð8Þ pressive stresses in the concrete. The resulting deformation due to
CFRP strip due to prestressing. Therefore, the applied equivalent
temperature (Dtapplied) is calculated somehow to enforce the exact
ð9Þ
value of eP for each beam.
eP ¼ afrp Dt ð14Þ
ð10Þ
8
< eP ¼ 0:0034 ) Dt
> ¼ 377:78  C Dt applied ¼ 383:05  C
eP ¼ 0:005868 ) Dt ¼ 652 C
Dt applied ¼ 659:55  C
>
:
eP ¼ 0:0102 ) Dt ¼ 1133:33  C Dt applied ¼ 1148:16  C
78 H.Y. Omran, R. El-Hacha / Construction and Building Materials 31 (2012) 74–85

3.4. Modeling of materials

(a)
3.4.1. Concrete constitutive model
Eight-node solid brick element (Solid65) was employed to mod-
el the concrete. This element, considering a 2  2  2 set of Gauss-
ian Integration points, has eight nodes with three degrees of
freedom at each node, translations in the nodal x, y, and z direc-
tions with capability of plastic deformation, cracking in three
orthogonal directions and crushing, although the crushing capabil-
ity of this element is omitted in this research. Many researchers
considered this procedure in the FE models of RC beams [7–
9,11,30]. The fact is that, when the crushing capability of the So-
lid65 is turned on, the FE model prematurely fails. This happens
due to stress concentration under the loading plate which causes (b)
the crushing to start and develop within a small load and the local
stiffness decreases sharply, then, the solution diverges showing a
large displacement warning. When the crushing is omitted the sec-
ondary tensile strain produced by the Poisson’s effect causes crack-
ing and finally failure [11]. To properly model the concrete, the
considered model for concrete comprises linear and multi-linear
isotropic material properties in addition to the concrete model de-
fined in ANSYS [9,30,31]. The multi-linear isotropic material uses
the Von Mises failure criterion along with the Willam and Warnke
Fig. 5. Concrete constitutive models for: (a) compression and (b) tension.
model [32] to define the failure of the concrete. A Poisson’s ratio of
0.18 was assumed for the concrete material [31].
Most researchers do not consider the concrete stress–strain 0    ec  1
1
curve derived from experimental compression test and use avail- B 1 þ B þ n1 eo C
fc ¼ fc0 @    n A ð15Þ
able models from the literature due to difficulty in identifying the ec 1 ec
1 þ B eo þ n1 eo
descending branch of the curve which is affected by test conditions
including gage length of the measured axial deformation, loading where eo is the strain at maximum concrete compressive strength fc0
rate, type of testing machine, existence of stirrups etc. Available and ec is the strain at any concrete compressive stress fc. Eq. (15) in-
equations in the literature to estimate the unconfined concrete cludes two constants n and B that can be chosen so that the curve
behavior in compression such as Modified Hognestad [33], Todes- can match any two points on the unconfined experimental concrete
chini [33], CEB-FIP model code [28], Thorenfeldt et al. [34], Desayi stress–strain curve. From the:
and Krishnan [35], and Loov [36] result almost in similar ascending 8 9
< eo ¼ 0:002233 at fc ¼ 40 MPa
0
branch, but provide different strain at ultimate and also different > >
=
descending branch. Most of researchers employed Desayi and Krish- Point1 : ec ¼ 0:000431 and f c ¼ 12 MPa ) Bunconfined
>
: >
;
nan’s stress–strain curve in the FE models [8,9,11,12,30,31]. The Point2 : ec ¼ 0:0042 and f c ¼ 26 MPa
equations proposed by Loov [36] and Thorenfeldt et al. [34] can be
¼ 0:123056 and nunconfined ¼ 3:117772
chosen to match an experimental concrete compressive stress–
strain curve. Most researchers used the unconfined concrete Then, point 3 on the confined curve in Fig. 5a is calculated based
stress–strain curve while some did not even considered the on the strain e50h at 0:5fc0 that gives additional ductility due to the
descending branch in modeling the stress–strain curve of concrete transverse reinforcement given by the following equation:
and employed perfectly plastic behavior after ultimate compressive
sffiffiffiffiffiffi
strength [8,11]. Few researchers considered the confined concrete 00
3 b
stress–strain curve in FE modeling [7,9]. Concrete becomes confined e50h ¼ qs ð16Þ
due to lateral support of the steel stirrups, and this happens when
4 Ssh
the stresses are approaching the uni-axial strength in disturbed area
of a beam. This can improve the stress–strain of the concrete at high where qs is the ratio of volume of transverse reinforcement (one
strain locations (under loading plate). For this research, a confined stirrup) to volume of concrete core measured to outside of the stir-
concrete constitutive model in compression based on experimental rup given by the following equation:
tests was adopted in order to perform an exact FE modeling. Most of 0 0
analytical confined models for concrete confined by rectangular ties 2ðb þ d ÞAsh
qs ¼ 00 00 ð17Þ
are derived for the members under axial compression loading and ðb  d ÞSsh
consider increase in both strength and ductility [37–44]. Kemp
00
[45] derived a concrete confined model for beams by increasing where b is the width of confined core measured to outside of the
0 00
both strength and ductility of concrete based on parameters affect- stirrup, b the width to center lines of the stirrup, d the depth of
0
ing ductility. Kent and Park [46] proposed a model which considers confined core measured to outside of the stirrup, d the depth to
the enhancement in concrete ductility due to confinement by stir- center lines of the stirrup, Ash the area of one leg of the stirrup,
rups. Eq (16) for e50h [47] e50h derived by these authors results in and Ssh is the spacing of stirrups.
the additional ductility due to rectangular stirrups at stress 0:5fc0 . The strain e50c on the descending branch of the confined con-
The Loov’s equation (Eq. (15)) is employed to define an uncon- crete curve at 0:5fc0 is determined form:
fined concrete compressive stress–strain curve based on two
e50c ¼ e50u þ e50h
points (points 1 and 2 in Fig. 5a) which are obtained from the
experimental concrete stress–strain curve for the tested beams re- where e50u is the unconfined concrete strain on the descending
ported by Gaafar [23]. branch at 0:5fc0 .
 H.Y. Omran, R. El-Hacha / Construction and Building Materials 31 (2012) 74–85 79

For the tested beams, the following is considered to determine

(a)
e50c:
( )
qs ¼ 0:008956;b00 ¼ 154:6 mm;Ssh ¼ 200 mm ) e50h ¼ 0:005905
) e50c
applying Eq:ð15Þ for unconfined curve at f c ¼ 0:5fc0 ) e50u ¼ 0:005017
¼ 0:010923

At the end, the Loov’s equation was employed considering

points 3 and 1 to define the descending branch of confined curve
up to stress 0:2 fc0 .
8 9
< eo ¼ 0:002233 at fc ¼ 40 MPa
0
> >
=
Point1 : ec ¼ 0:000431 and f c ¼ 12 MPa ) Bconfined
>
: >
;
Point3 : ec ¼ 0:010923 and f c ¼ 20 MPa
¼ 3:123426 and nconfined ¼ 1:297838
(b)
The final derived equation for the stress–strain curve of the con-
crete is presented by Eq. (18). Comparison, performed by the
authors, showed that the stress–strain curve defined by this meth-
od is very close to that defined by Desayi and Krishnan’s stress–
strain curve and the advantage is that the proposed curve is based
on the actual material properties of the concrete reported by Gaa-
far [23]:
8  
>
> 714:397834ec
fc0 1þ55:107927 ec P 0:002233
>
<  e c þ8:703247e 3:117772
c Fig. 6. Strain-strain curves: (a) CFRP strip and (b) epoxy adhesive.

fc ¼ f 0 552:665160ec
0:002233 < ec < 0:067346
> c 11398:757736ec þ9262:604615e1:297838
>
>
:
c

0:2fc0
ec 6 0:037346
ð18Þ
The stress–strain curve of the concrete in tension is presented
in Fig. 5b [26]. The cracks form in the concrete when the stress
reaches the concrete tensile strength; however, in a section situ-
ated between two cracks, bond between the concrete and the
reinforcing steel rebars restrains the elongation of the steel,
hence, the tensile force in the reinforcement at the crack is trans-
mitted to the concrete situated between the cracks. The stress and
strain in the section will be in an intermediate condition between
the un-cracked and cracked states. Therefore, the cracked con-
crete can still carry tensile stress perpendicular to the crack,
and an effective or mean value of the member stiffness must be
considered in the calculation of the elongation or curvature of
the member. The contribution of the concrete in the tension zone
to the rigidity of the beam is known as tension stiffening.
Although after cracking, the stiffening effect of tension carried
by concrete between cracks has much less significance, and the
cracked section properties can be used with little error, in this

Mesh at the end groove

Contact around the groove

Cross-section
Steel reinforcements of the beam

Fig. 7. The meshed beam.

80 H.Y. Omran, R. El-Hacha / Construction and Building Materials 31 (2012) 74–85

research, a bilinear stress–strain relation, predefined in ANSYS results and properly calculate the loads and deflections at different
program, is employed to model the tension stiffening as shown steps (i.e., at prestressing, surrounding cracking, yielding, and
in Fig. 5b.

3.4.2. Steel reinforcement (a) 160

140
Steel reinforcements in the RC members can be modeled by
employing three methods in FE: discrete model, embedded model, 120

Total Load (kN)

and smeared model. In this research, the discrete model of steel
reinforcement is used in which the link elements are connected 100
to the concrete mesh nodes. Two-node link element, Link8, was se- 80
lected to model steel reinforcements. This element is a 3D spar B0 (Experimental)
with three degrees of freedom at each node, translations in the no- 60
B0 (FE Model)
dal x, y, and z directions. Also, it is capable of plastic deformation. A 40
B1-NP (Experimental)
multi-linear material model was used to assign the actual stress-
stress curves of the steel reinforcements as presented in Fig. 2, to 20 B1-NP (FE Model)

the element along with the different real constants related to 0

cross-section of each steel rebar. 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
Mid-span Deflection (mm)
3.4.3. CFRP strip, adhesive, loading steel plate, and steel end anchors
Eight-node solid brick element (Solid45) was employed to mod- 160
el the CFRP strips, epoxy adhesive, end steel anchors and loading
(b)
140
steel plates at the supports and under the concentrated loads on
the beam. This element, considering a 2  2  2 set of Gaussian 120

Total Load (kN)

integration points, has eight nodes with three degrees of freedom 100
at each node, translations in the nodal x, y, and z directions and
is capable of assigning multi-linear elastic material model. The 80
schematic multi-linear strain–stress strain curves assigned to the 60 B1-P1 (Experimental)
CFRP strip and epoxy adhesive are presented in Fig. 6. A Poisson’s
40
ratio of 0.35 and 0.37 was assumed for CFRP and epoxy adhesive, B1-P1 (FE Model)
respectively [48]. A linear elastic material is assigned to the loading 20
plates and end anchor with a Young’s modulus of 200 GPa and a 0
Poisson’s ratio of 0.3. -10 0 10 20 30 40 50 60 70 80 90 100 110120130140
Mid-span Deflection (mm)
3.4.4. Bolts at end anchors
The Beam4 element was considered to model the steel bolts at 160
the end steel anchor. This element has two nodes with six degrees (c)
140
of freedom at each node with capabilities of tension, compression,
torsion and bending. By assigning the shear deflection constants 120
Total Load (kN)

(1.111 for circular cross-section as recommended in ANSYS pro- 100

gram library [26]) this element is also capable to take into account
the shear effects. A linear elastic material property including rele- 80
vant real constants of the bolt (diameter, cross-sectional area, and 60 B1-P2 (Experimental)
moment of inertia) were assigned to this element.
40
B1-P2 (FE Model)
3.5. Nonlinear analysis 20

0
The nonlinear solution was operated using displacement con- -10 0 10 20 30 40 50 60 70 80 90 100 110120130140
trol method which is done by applying the non-zero displacement Mid-span Deflection (mm)
constraints to the loading plate in the model and calculating the re-
lated reaction forces related to that non-zero displacement. In this 160
manner the reaction load–deflection curve is the same as the ap- (d)
plied load–deflection curve in the test. The advantages of displace- 140
ment control versus load control are to overcome both the 120
Total Load (kN)

convergence difficulties and the rigid body modes when two

100
bodies are disconnected in contact pairs, and follow/obtain the
descending branch of the stress–strain curve/load-deflection curve. 80
Each prestressed strengthened beam was solved by defining seven 60 B1-P3 (Experimental)
load-steps. In fact, the number of load-steps in a FE analysis is arbi-
trary selected by the modeler based on the type of the problem, 40
B1-P3 (FE Model)
convergence behavior of the FE system, structure’s load-response 20
curve, and the cost in term of computer computational time. The
0
main reasons for using seven load-steps instead of one were to: -10 0 10 20 30 40 50 60 70 80 90 100 110120130140
(i) assign the effect of prestressing to the model, and (ii) apply
Mid-span Deflection (mm)
the load gradually by assigning different maximum and minimum
numbers of the sub-steps defined for each load-step. The latter Fig. 8. Comparison between experimental and numerical load–deflection curves:
leads to the following advantages: to increase the accuracy of the (a) B0 and B1-NP, (b) B1-P1, (c) B1-P2 and (d) B1-P3.
 H.Y. Omran, R. El-Hacha / Construction and Building Materials 31 (2012) 74–85 81

Table 1
Key points of load–deflection curves: Comparison of test results with the FE analysis from ANSYS.

Beam ID# Prestrain in CFRP Results Do (mm) Pcr (kN) Dcr (mm) Py (kN) Dy (mm) Pu (kN) Du (mm) l u (kNmm) Failure mode
B0 N.A. FE 0 16 1.18 81.9 24.29 92.8 107.49 4.43 8446.6 CC
Testa 0 12.5 1.25 78.9 25.13 83.8 109.89 4.37 8050.2 CC
Error (%) 0 28 5.6 3.8 3.3 10.7 2.2 1.4 4.9
B1-NP 0 FE 0 15.4 1.14 93.6 25.42 136.8 111.9 4.4 11816.2 FR
Testa 0 16.8 1.55 90.8 25.83 135.1 118.79 4.6 12357.4 CCS
Error (%) 0 8.3 26.5 3.1 1.6 1.3 5.8 4.3 4.4
B1-P1 0.0034 FE 0.48 22.7 1.2 102.1 25.04 140.4 88.26 3.52 9501.8 FR
Testa 0.47 22.1 1.24 103 24.12 148 103.65 4.3 11828.7 FR
Error% 2.1 2.7 3.2 0.9 3.8 5.1 14.8 18.1 19.7
B1-P2 0.00587 FE 0.82 28.1 1.27 111.3 25.81 144.3 78.27 3.03 8582 FR
Testa 0.93 30.1 1.69 105.8 23.62 148.2 77.96 3.3 8813.1 FR
Error% 11.8 6.6 24.9 5.2 9.3 2.6 0.4 8.2 2.6
B1-P3 0.0102 FE 1.42 37.1 1.33 123.3 25.84 147.3 56.41 2.18 6223.4 FR
Testa 1.6 42.1 2.64 122.8 25.77 149.2 58.13 2.26 6529 FR
Error% 11.3 11.9 49.6 0.4 0.3 1.3 3.0 3.5 4.7

Pcr and Dcr = load and deflection at cracking, Py and Dy = load and deflection at yielding, Pu and Du = load and deflection at ultimate, Do = camber due to prestressing,
CC = concrete crushing, l = ductility index = Du/Dy, CCS = concrete cover spalling, u = area under P-D curve up to Pu, FR = CFRP rupture.
a
Gaafar [23].

ultimate), to control computer computational time so as to mini-
mize the number of Newton–Raphson equilibrium iterations re-
quired, and to assist in convergence of the nonlinear problem.
These seven load-steps were enforced as follow: (1) from zero load
and zero prestressing force up to the camber induced by the re-
quired prestressing level, (2) up to a displacement of 1.2 mm be-
fore cracking, (3) up to a displacement of 4.0 mm after cracking,
(4) up to a displacement of 20.0 mm before yielding, (5) up to a dis-
placement of 26.0 mm after yielding, (6) up to a displacement of
62.0 mm before ultimate, and (7) up to displacement of
120.0 mm after ultimate. The selected values of the displacements
for the load-steps were determined after initial analysis of the
beams to capture the exact values of the loads and deflections at
cracking, yielding, and ultimate. After several trials, a maximum
displacement convergence tolerance of 0.05 is applied to the model
based on the convergence sensitivity analysis. To take into account
the effect of the self-weight of the beam, a preliminary analysis
was performed on the model by applying density and inertia and
the deflection, Dsw, was calculated at the point load. Then, another
analyzes was conducted by applying the same deflection Dsw as a
boundary condition to the model without the effect of self-weight
and the reaction forces which are equivalent to the reactions due to
self-weight are calculated, called Psw. Then, Psw and Dsw were sub-
tracted from the reaction forces and deflections obtained from the
nonlinear solution of each beam. In this manner, the applied load
versus deflection is plotted which is the same as the experiment
that included the self-weight. The meshed beam is presented in
Fig. 7. At the location of the end anchors, the groove width is
25 mm to account for the embedded anchors and in between along
the length of the CFRP strip the groove width is 16 mm which leads
to using some relatively high aspect ratio elements at the right side
of the groove in Fig. 7.
4. Numerical results, validation, and discussion
4.1. Load–deflection curve
Fig. 8 depicts the comparison between experimental and
numerical load–deflection curves comprising the un-strengthened
control beam (B0) and strengthened beams with non-prestressed
(B1-NP) and prestressed (B1-P1, B1-P2, and B1-P3) NSM-CFRP
strip. The numerical analyses of the beams were terminated after
CFRP rupture or concrete crushing whichever occurs first accompa-
nied by a message from the software showing non-convergence.
The load–deflection curves include the negative camber due to pre-
stressing, initiation of flexural cracks, yielding of tensile steel rebar,
local debonding (which causes small fluctuations at large deflec-
tion), and CFRP rupture which causes a large drop of total load at
ultimate stage.
Table 1 represents the key points of the load–deflection curves
obtained from the tests versus the FE analysis in addition to type of
failure, ductility index (the ratio of the ultimate deflection to the
deflection at yielding), and energy absorption (the area under
load–deflection curve up to the peak load). Also, percentage of dif-
ference between corresponding experimental and numerical value
is presented in Table 1. A relatively large percentage of error is ob-
served at cracking stage which might be due to presence of the mi-
cro cracks in the large-scale beams before testing; results of the
beam B0 at cracking confirm this statement. On the other hand,
the cracking load and corresponding deflection in the tested beams
were obtained based on the visual inspections and in some cases
might be overestimated. At yielding stage the differences between
FE and experimental values are negligible. At ultimate stage, the FE
loads are almost the same as those from experiment; however, the
predicted ultimate deflections are less than the experimental val-
ues. This could be due to the fact that the CFRP strip rupture in
FEM occurs when the first fiber in the CFRP strip reaches its ulti-
mate tensile strain, while during the test this might not be the case.
So during the test there is a potential that not all the fibers were
ruptured at the same time and the strips would bear the load car-
rying capacity until the entire strip ruptured. The recognition of
this difference would not be of any importance, since for design
purposes the initiation of failure should be considered as failure.
On the other hand, the material characteristic values (i.e., CFRP
ultimate strain) are not absolutely constant values and could be
a little more or less than the specified values which would result
in difference at ultimate deflection when CFRP rupture governs.
The type of failure from the FE analysis compared well with the
tested beams which is concrete crushing for un-strengthened con-
trol beam (B0) and CFRP rupture for the prestressed strengthened
beams except beam B1-NP strengthened with non-prestressed
CFRP strips which failed due to concrete cover spalling. The fluctu-
ation of the FE curve of B1-NP in Fig. 8a which happens due to local
debonding initiated from top face of the groove (shown in Fig. 11a)
could be a warning for this type of failure. Comparison indicates
that the load–deflection curves from the FE models are perfectly
matched with those from the experiment.
4.2. Strain distribution
Typical, strain profiles along the length of the NSM-CFRP strip
and strain distribution across the depth at mid-span are portrayed
82 H.Y. Omran, R. El-Hacha / Construction and Building Materials 31 (2012) 74–85

@ Cracking Load, 28.1 kN, FEM-ANSYS @ Cracking Load, 30.1 kN, Experimental and the ductility index by 1%; on the other hand, increases the
@ Yielding Load, 102.8 kN, FEM-ANSYS @ Yielding Load, 105.8 kN, Experimental yielding and ultimate load by 14% and 47%, respectively, and in-
@ Ultimate Load, 144.3 kN, FEM-ANSYS @ Ultimate Load, 148.2 kN, Experimental
creases the amount of energy absorption at yielding and ultimate
(a) 0.02 loads by 19% and 40%, respectively. Fig. 10a shows linear changes
of the energy absorption at yielding and ultimate loads, the crack-
ing, yielding and ultimate loads, and the ductility index as the pre-
0.015
stressing level increases. Applying prestressing up to 51% enhances
the cracking, yielding and ultimate loads by 132%, 51% and 59%,
Strain

0.01 and the amount of energy absorption at yielding by 93%; however,

it decreases the ductility index by 49% and the energy absorption
by 26%.
0.005

0
0 1000 2000 3000 4000 5000 (a) 150 Energy Absorption@Yielding
150

Energy Absorption@Ultimate
Distance from the support (mm) 120 120
Cracking Load
Yielding Load

% of change w.r.t B0
@ Cracking Load, 28.1 kN, FEM-ANSYS @ Cracking Load, 30.1 kN, Experimental 90 Ultimate Load 90
@ Yielding Load, 102.8 kN, FEM-ANSYS @ Yielding Load, 105.8 kN, Experimental Ductility Index
@ Ultimate Load, 144.3 kN, FEM-ANSYS @ Ultimate Load, 148.2 kN, Experimental 60 60
Top Fibre of the Beam
0
(b) Top Steel Level 30 30
-50
0 0
Section Depth (mm)

-100

-150 -30 -30

-200
-60 -60
-250 0 10 20 30 40 50
Prestressing level (% of CFRP ultimate strength)
-300
Bottom Steel Level
-350
CFRP Strips Level (b) 150 Energy Absorption@Yielding
150
-400 Energy Absorption@Ultimate
-0.005 0 0.005 0.01 0.015 0.02 120 120
Cracking Load
% of change w.r.t B1-NP

Strain Yielding Load

90 90
Ultimate Load
Fig. 9. Comparison between experimental and numerical strain values for beam
Ductility Index
B1-P2: (a) strain profile along the CFRP strip, (b) strain distribution across the depth 60 60
at mid-span.
30 30

for beam B1-P2 in Fig. 9 at three different load levels: cracking, 0 0

yielding, and ultimate. The experimental values are based on the
-30 -30
reading of the strains in the CFRP strip, steel rebars, and in the con-
crete at extreme compression fiber. The predicted strain profiles -60 -60
along the CFRP strip are greatly correlated with the experimental 0 10 20 30 40 50
values at all steps. It can be seen from the strain profile at ultimate Prestressing level (% of CFRP ultimate strength)
that the maximum strain in CFRP strip happens somewhere be-
tween the applied point load to mid-span of the beam (from (c) 14000
2000 mm to 2500 mm on the x axis of the graph) which could
Area under load-deflection curve

not be monitored due to the location of the installed strain gauges. 12000
The FE strain distributions at mid-span in Fig. 9b are calculated at
the same location of the experimental measured strains and line- 10000
arly connected together. The distribution shows perfect correlation
(kN.mm)

at cracking and yielding and with insignificant difference at ulti- 8000

mate. It should be noticed that the strain distribution at mid-span
section is not linear due to the effect of prestrain applied to the 6000
Energy absorption (Experimental)
CFRP strip.
Energy absorption (B0-Experimental)
4000
Energy absorption (FE Model)
2000 Energy absorption (B0-FE Model)
4.3. Effects of prestressing

0
The effects of prestressing with respect to the FE results of the 0 10 20 30 40 50
un-strengthened control beam (B0) and the strengthened beam
Prestressing level (% of CFRP ultimate strength)
with non-prestressed CFRP strip (B1-NP) are presented in
Fig. 10a and b, respectively. With respect to beam B0, strengthen- Fig. 10. Effects of prestressing: (a) w.r.t. beam B0 based on FE results, (b) w.r.t.
ing with non-prestressed CFRP decreases the cracking load by 4% beam B1-NP based on FE results and (c) w.r.t. energy absorption.
 H.Y. Omran, R. El-Hacha / Construction and Building Materials 31 (2012) 74–85
(a)
HI: Horizontal Interface between Epoxy and Concrete
VI: Vertical Interface between Epoxy and Concrete
CFRP strips
HI
VI
Concrete
Epoxy
.983916
B1-NP- Debonding Parameter (dm) Contour
(c)
HI: Horizontal Interface between Epoxy and Concrete
VI: Vertical Interface between Epoxy and Concrete
CFRP strips
HI
VI
Concrete
Epoxy
B1-P2- Debonding Parameter (dm) Contour
Fig. 11. Debonding Parameter (dm) contour at the interfaces around the groove in the model: (a) B1-NP at initiation of debonding (load = 130.4 kN, deflection = 80.14 mm), (b)
B1-NP at ultimate, (c) B1-P2 at initiation of debonding (load = 140.4 kN, deflection = 68.32 mm), (d) B1-P2 at ultimate.
With respect to beam B1-NP, increasing the prestressing level
up to 51% enhances the cracking load significantly up to 141%,
yielding load up to 32%, ultimate load up to 8%, and the amount
of energy absorption at yielding load up to 63% (Fig. 10b); on the
other hand, the ductility index decreases by 50% and the amount
of energy absorption at ultimate load decreases by 47%.
Overall, a linear relationship exists between the prestressing le-
vel in CFRP strip and the percentage of change in cracking, yielding,
and ultimate loads; ductility index; and energy absorption. In this
research, the optimum prestressing level is defined as the pre-
stressing level in the NSM-CFRP strips (taken as a percentage of
the ultimate tensile strength of the CFRP strips) that enhances
the beam performance under service and ultimate loads and main-
tain the amount of energy absorption (up to the peak load) in the
strengthened beam equal to the un-strengthened control beam.
The practical optimum prestressing level is defined by the authors
and has never been taken into account by other researchers. Proce-
dure to determine the optimum prestressing level (intersection of
the ductility curve of the strengthened beams with the un-
strengthened beam) is plotted in Fig. 10c. The FEM results yield
83
(b)
HI: Horizontal Interface between Epoxy and Concrete
VI: Vertical Interface between Epoxy and Concrete
CFRP strips
HI
VI
Concrete
Epoxy
B1-NP- Debonding Parameter (dm) Contour
(d)
HI: Horizontal Interface between Epoxy and Concrete
VI: Vertical Interface between Epoxy and Concrete
CFRP strips
HI
VI
Concrete
Epoxy
B1-P2- Debonding Parameter (dm) Contour
an optimum prestressing level of 30.5% versus 36.5% obtained from
experimental results. For design purposes, the CFRP material
creep-rupture stress limit shall be considered as the maximum
prestressing level which can be enforced. A creep-rupture stress
limit of 65% of the ultimate tensile strength of the CFRP reinforce-
ment is recommended by Canadian Highway Bridge Design Code
[49].
4.4. Debonding aspects
To examine the effect of debonding occurrence, the initiation of
debonding and its propagation at ultimate load for two models
(B1-NP and B1-P2) are plotted in Fig. 11 as obtained from the FE
results. The contour in the figure shows the debonding parameter
values (dm) ranging from 0 to 1. Debonding is initiated when dm = 0
and approaches to unity by further propagation. The termination of
debonding is defined by Eq. (6) when the total fracture energy is
dissipated. The results show that debonding happens at the hori-
zontal interface between the concrete and the epoxy on the top
surface of the groove and debonding propagation at ultimate load
84 H.Y. Omran, R. El-Hacha / Construction and Building Materials 31 (2012) 74–85
is less for the prestressed beam. In fact, debonding occurs due to
high interfacial shear and tensile stresses which are mainly caused
by high deflection and crack opening and because prestressing de-
creases the ductility of the beam it avoids the debonding to hap-
pen. Results showed no debonding occurs for beam B1-P3 up to
failure. For beam B1-P1, debonding occurred before failure and
was much less than beam B1-NP. Thus, with increasing the pre-
stressing level the parts that are debonded at failure decrease,
i.e., less debonding occurs.
5. Conclusions
A comprehensive nonlinear 3D finite element model was devel-
oped to investigate the behavior of RC beams strengthened in flex-
ure with prestressed NSM-CFRP strip, considering the effect of
debonding. The model was validated with the experimental re-
sults. According to this research, the following conclusions can be
drawn:
1. The proposed 3D FE model was validated with the experimental
results and a very good correlation was observed. The validated
FE model properly estimated the behavior of the NSM-CFRP
strengthened RC beams. The FEM procedure and selected ele-
ments along with the assigned material constitutive models
and considered mixed-mode debonding model (normal tension
stress-gap and shear stress-slip models) can be employed to
generate a FE model for future studies.
2. Using the developed FE model, the optimum prestressing level
(percentage of the ultimate tensile strength of the CFRP rein-
forcement) can be determined in order to maintain the beam’s
original ductility before strengthening and at the same time
taking advantage of the prestressing effect applied to the sys-
tem. The optimum prestressing level shall be checked with
the design code criteria. For design purposes, the CFRP creep-
rupture stress limit specified in the design code and guidelines
shall be considered as the upper bound of the optimum pre-
stressing level.
3. Considering the percentage change of loads due to prestressing
with respect to the non-prestressed strengthened beam, the
increase in the cracking load was the most followed by the
yielding load than the ultimate load.
4. Considering the percentage change of loads of the non-pre-
stressed strengthened beam with respect to the un-strength-
ened beam, the increase in the ultimate load was more
significant than the yield load.
5. Comparing the prestressed beams with the non-prestressed
beam and the un-strengthened beam, increasing the prestress-
ing level resulted in almost same decrease in the ductility index.
The decrease in the energy absorption at ultimate was higher
when comparing the prestressed beams with the non-pre-
stressed beam. The increase in the energy absorption at yielding
was more significant when comparing the prestressed beams
with the un-strengthened beam than with the non-prestressed
beam. Comparing the non-prestressed strengthened beam with
the un-strengthened beam, the percentage of increase in energy
absorption at ultimate was higher than at yielding, and the
cracking load as well the ductility index almost remained
constant.
6. When a strengthened beam is tested under four-point bending
configuration, the maximum strain in CFRP strip does not hap-
pen at mid-span section, but occurs at a location somewhere
from the point load to mid-span of the beam which could not
be possibly captured due to the infinite location of the installed
strain gauges along the length of the CFRP strips. Also, the strain
distribution at mid-span section is not linear due to the effect of
prestrain applied to the CFRP strip.
7. The results show that debonding propagation at ultimate load
which is mainly caused by high deflection and crack opening
is less for the prestressed beam.
This paper demonstrated the applicability of the developed 3D
FE model to confidently simulate RC beams strengthened with pre-
stressed or non-prestressed FRP. Original practical recommenda-
tions are provided in the paper for FE modeling including: (1) the
bond by considering mixed-mode debonding to enhances the accu-
racy, and identification of appropriate shear stress-slip and tension
stress-gap models; (2) accounting the prestressing effect in the
NSM-CFRP strips using the equivalent temperature method; and
(3) considering a more rational confined concrete compressive
stress–strain model based on experimental data. The reliability of
the model was confirmed by comparing several experimental test
results with the FE results showing excellent correlation between
prediction and reality. The proposed model can be employed with
confidence as a predictive tool in future researches. Furthermore, a
practical definition of the optimum prestressing level is presented
to maintain the amount of energy absorption in the strengthened
beam equal to the un-strengthened beam.
References
[1] Nordin H, Täljsten B. Concrete beams strengthened with prestressed near
surface mounted CFRP. J Compos Constr 2006;10(1):60–8.
[2] El-Hacha R, Gaafar M. Flexural strengthening of reinforced concrete beams
using prestressed near-surface mounted CFRP bars. PCI J Fall 2011:134–51.
[3] Badawi M, Soudki K. Flexural strengthening of RC beams with prestressed NSM
CFRP rods – experimental and analytical investigation. Constr Build Mater
2009;23(10):3292–300.
[4] Rafi MM, Nadjai A, Ali F. Analytical modeling of concrete beams reinforced
with carbon FRP bars. J Compos Mater 2007;41(22):2675–90.
[5] Camata G, Spacone E, Zarnic R. Experimental and nonlinear finite element
studies of RC beams strengthened with FRP plates. Compos Part B: Eng
2007;38(2):277–88.
[6] Supaviriyakit T, Pornpongsaroj P, Pimanmas A. Finite element analysis of FRP-
strengthened RC beams. Songklanakarin J Sci Technol 2004;26(4):497–507.
[7] Chansawat K. Nonlinear finite element analysis of reinforced concrete
structures strengthened with FRP laminates. PhD thesis. Oregon State
University, Oregon, USA; 2003.
[8] Jia M. Finite element modeling of reinforced concrete beams strengthened
with FRP composite sheets. MSc thesis. The University of Alabama in
Huntsville, Alabama, USA; 2003.
[9] Chansawat K, Yim SCS, Miller TH. Nonlinear finite element analysis of a FRP-
strengthened reinforced concrete bridge. J Bridge Eng 2006;11(1):21–32.
[10] Nour A, Massicotte B, Yildiz E, Koval V. Finite element modeling of concrete
structures reinforced with internal and external fibre-reinforced polymers (1).
Can J Civil Eng 2007;34(3):340–54.
[11] Kachlakev D, Miller T, Yim S, Chansawat K, Potisuk T. Finite element modeling
of reinforced concrete structures strengthened with FRP laminates – final
report SPR 316. Oregon, USA: Oregon Department of Transportation Research
Group; 2001.
[12] Coronado CA, Lopez MM. Sensitivity analysis of reinforced concrete beams
strengthened with FRP laminates. Cem Concr Compos 2006;28(1):102–14.
[13] Pham HB, Al-Mahaidi R. Finite element modelling of RC beams retrofitted with
CFRP fabrics. In: Proceeding of the 7th international symposium on FRP
reinforcement for concrete structures (FRPRCS-7). November 6–9; Kansas City,
MO, USA; 2005. p. 499–514.
[14] Buyle-Bodin F, David E, Ragneau E. Finite element modelling of flexural
behaviour of externally bonded CFRP reinforced concrete structures. Eng
Struct 2002;24(11):1423–9.
[15] Kishi N, Zhang G, Mikami H. Numerical cracking and debonding analysis of RC
beams reinforced with FRP sheet. J Compos Constr 2005;9(6):507–14.
[16] Kang JY, Park YH, Park JS, You YJ, Jung WT. Analytical evaluation of RC beams
strengthened with near surface mounted CFRP laminates. In: Proceeding of the
7th international symposium on FRP reinforcement for concrete structures
(FRPRCS-7). November 6–9. Kansas City, MO, USA; 2005. p. 779–94.
[17] Omran HY, El-Hacha R. Finite element modelling of RC beams strengthened in
flexure with prestressed near surface mounted CFRP rebars. In: Proceeding of
the 8th international conference on short and medium span bridge (SMSB
2010). August 3–6; Niagara Falls, Ontario, Canada: CD-ROM; 2010. p. 10.
[18] Soliman SM, El-Salakawy E, Benmokrane B. Flexural behaviour of concrete
beams strengthened with near surface mounted fibre reinforced polymer bars.
Can J Civil Eng 2010;37(10):1371–82.
[19] Chen JF, Teng JG. Anchorage strength models for FRP and steel plates bonded to
concrete. J Struct Eng 2001;127(7):784–91.
[20] Lu XZ, Teng JG, Ye LP, Jiang JJ. Bond-slip models for FRP sheets/plates bonded to
concrete. Eng Struct 2005;27(6):920–37.
 H.Y. Omran, R. El-Hacha / Construction and Building Materials 31 (2012) 74–85
[21] Seracino R, Saifulnaz MRR, Oehlers DJ. Generic debonding resistance of EB and
NSM plate-to-concrete joints. J Compos Constr 2007;11(1):62–70.
[22] Kim YJ, Green MF, Wight RG. Flexural behaviour of reinforced or prestressed
concrete beams including strengthening with prestressed carbon fibre
reinforced polymer sheets: application of a fracture mechanics approach.
Can J Civil Eng 2007;34(5):664–77.
[23] Gaafar M. Strengthening reinforced concrete beams with prestressed near
surface mounted fibre reinforced polymers. MSc thesis. University of Calgary,
Canada; 2007.
[24] ‘‘SikadureÒ 330’’. Product data sheet, impregnation resin for fabric
reinforcement [Internet]. 06.2010/v1. <http://can.sika.com/content/canada/
main/en/solutions_products/document_download/Structural_Strengthening.
html>.
[25] ‘‘SikadureÒ 30’’. Product data sheet, high modulus, high strength, structural
epoxy paste adhesive for use with sikaÒ CarboDurÒ reinforcement [Internet].
06.2010/v1. <http://can.sika.com/content/canada/main/en/solutions_ products/
document_download/Structural_Strengthening.html>.
[26] SAS. ANSYS 12 finite element analysis system, SAS IP, inc.; 2009.
[27] CSA A23.3-04. Design of concrete structures, Canadian standards association;
2004.
[28] CEB-FIP Model Code 1990. Design code; 1993.
[29] ‘‘Aslan 500 CFRP tape’’. Product data sheet, Aslan 500 CFRP tape for near
surface mount (NSM) structural strengthening (flexure & shear) of existing
concrete, masonry or wood members. <http://aslanfrp.com/Aslan500/
Aslan500_CFRP_Tape.html>.
[30] Wolanski AJ. Flexural behaviour of reinforced and prestressed concrete beams
using finite element analysis. MSc thesis. Marquette University, Milwaukee,
Wisconsin, USA; 2004.
[31] Özcan DM, Bayraktar A, Sßahin A, Haktanir T, Türker T. Experimental and finite
element analysis on the steel fiber-reinforced concrete (SFRC) beams ultimate
behavior. Constr Build Mater 2009;23(2):1064–77.
[32] Willam KJ, Warnke ED. Constitutive model for the triaxial behavior of concrete.
In: Proceeding of the international association for bridge and structural
engineering. Bergamo, Italy. ISMES, vol. 19; 1975. p. 174–90.
[33] Wight JK, MacGregor JG. Reinforced concrete mechanics and design. fifth
ed. New Jersey, USA: Pearson Prentice Hall; 2009.
[34] Thorenfeldt E, Tomaszewicz A, Jensen JJ. Mechanical properties of high
strength concrete and application to design. In: Proceeding of the
symposium: Utilization of high-strength concrete. June, Stavanger, Norway:
Tapir, Trondheim; 1987. p. 149–59.
85
[35] Desayi P, Krishnan S. Equation for stress–strain curve of the concrete. ACI J
1964;61(3):345–50.
[36] Loov RE. A general stress–strain curve for concrete implications for high
strength concrete columns. In: Proceeding of the annual conference of the
Canadian society for civil engineering. May 29–31, Vancouver, BC, Canada;
1991. p. 302–11.
[37] Sheikh SA, Uzameri SM. Mechanism of confinement in tied columns.
Proceeding of the 7th world conference on earthquake engineering.
September, Istanbul, Turkey; 1980. p. 71–8.
[38] Chan WWL. The ultimate strength and deformation of plastic hinges in
reinforced concrete frameworks. Mag Concr Res (London) 1955;7(21):
121–32.
[39] Roy HEH, Sozen MA. Ductility of concrete. Detroit,USA: Flexural Mechanics of
Reinforced Concrete, SP-12, American Concrete Inistitute/American Society of
Civil Engineers; 1965.
[40] Soliman MTM, Yu CW. The flexural stress–strain relationship of concrete
confined by rectangular transverse reinforcement. Mag Concr Res
1967;19(61):223–38.
[41] Sargin M. Stress–strain relationships for concrete and the analysis of structural
concrete sections. Ontario, Canada: Solid Mechanics Division, University of
Waterloo. Report No. Study No. 4; 1971.
[42] Vallenas J, Bertero VV, Popov EP. Concrete confined by rectangular hoops and
subjected to axial loads. USA: Earthquake Engineering Research Center,
University of California, Berkeley. Report No.: UCB/EERC-77/13; 1977.
[43] Park R, Priestley MJN, Gill WD. Ductility of square confined concrete columns. J
Struct Div (ASCE) 1982;108(4):929–50.
[44] Mander JB, Priestley MJN, Park R. Theoretical stress–strain model for confined
concrete. J Struct Eng 1988;114(8):1804–26.
[45] Kemp AR. The achievement of ductility in reinforced concrete beams. Mag
Concr Res 1998;50(2):123–32.
[46] Kent DC, Park R. Flexural members with confined concrete. J Struct Div (ASCE)
1971;97(7):1969–90.
[47] Park R, Paulay T. Reinforced concrete structures. New York, USA: John Wiley
and Sons; 1975.
[48] Kabir MZ, Hojatkashani A. A comparison between finite element and analytical
solutions of interfacial stress distribution in a RC beam retrofitted with FRP
composites. Int J Sci Technol 2008;19(68):55–65.
[49] Canadian Standard Association. Canadian highway bridge design code. CAN/
CSA S6-06, Ontario, Canada; 2006.