Flexural Strength of RC Beams with GFRP Laminates

Sing-Ping Chiew, M.ASCE1; Qin Sun2; and Yi Yu3

Abstract: The results of an experimental and numerical study of the flexural behavior of reinforced concrete beams strengthened with
glass-fiber-reinforced-polymer 共GFRP兲 laminates are presented in this paper. In the experimental program, ten strengthened beams and
two unstrengthened beams are tested to failure under monotonic loading. A number of external GFRP laminate layers and bond length of
GFRP laminates in shear span are taken as the test variables. Longitudinal GFRP strain development and interfacial shear stress distri-
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.

bution from the tests are examined. The experimental results generally showed that both flexural strength and stiffness of reinforced
concrete beams could be increased by such a bonding technique. In the numerical study, an eight-node interface element is developed to
simulate the interface behavior between the concrete and GFRP laminates. This element is implemented into the MARC software package
for the finite-element analyses of GFRP laminate strengthened reinforced concrete beams. Reasonably good correlations between experi-
mental and numerical results are achieved.
DOI: 10.1061/共ASCE兲1090-0268共2007兲11:5共497兲
CE Database subject headings: Finite elements; Fiber reinforced polymers; Laminates; Flexural strength; Concrete beams; Concrete,
reinforced; Numerical analysis.

Introduction tural joints, and in locations where very limited space is

available.
Structural strengthening and retrofitting of existing concrete Adequate bond strength between the existing structural member
buildings are currently major activities in the construction indus- and the external FRP composite is an important factor for suc-
try. In the past, strengthening would be carried out by casting cessful strengthening 共Smith and Teng 2001; Buyukozturk et al.
additional reinforced concrete or by dowelling in additional steel 2004兲. Although debonding by concrete cover separation was
reinforcement. Such a process is not only time consuming and commonly reported previously 共Oehlers and Seracino 2004; Teng
complicated but could also affect the structural appearance. The et al. 2002兲, the interface bond is found to be another important
development of epoxy resins in the 1960s introduced a new contributing factor for structural failure 共Shahawy et al. 1996;
strengthening method by bonding steel plates to the surfaces of Saadatmanesh and Malek 1998; Malek et al. 1998; Fanning and
the structural members. Kelly 2000; Kishi et al. 2005兲. Interfacial bond failure is a com-
In the past decade or so, fiber-reinforced-polymer 共FRP兲 com- plicated problem; many researchers believed that it initiated from
posites have emerged as a more preferable material than steel the plate end due to stress concentration at a cutoff point 共Malek
plates in the bonding technique. FRP composites have many ad- et al. 1998兲. Others believed that damage of interface bond due to
vantages over steel plates in such applications: concrete cracking played a more important role 共Leung 2001;
1. Strength/weight ratio of FRP composite is much larger than Saffi and Sayyah 2001; Yao et al. 2005兲. Full-scale investigations
that of steel; this makes site installation much easier. of interfacial stress distribution and development are very limited
2. FRP composite is noncorrosive. Therefore, FRP composite in the published literature. An experimental investigation in this
is better than steel in aggressive marine and industrial area will be helpful to understand such failure better.
environments. Large-scale testing is a very expensive and time-consuming
3. FRP composite can be used in situations where it would process and in recent years more emphasis has been placed on
be impossible or impractical to use steel plates. For in- numerical simulation to complement the test results. In the past
stance, structural members of complicated shape, struc- decade, many researchers have carried out numerical studies on
FRP composite strengthened structures. However, many of them
1
Associate Professor, School of Civil and Environmental Engineering, ignored the presence of an interface bond by connecting concrete
Nanyang Technological Univ., 50 Nanyang Ave., Singapore 639798. and FRP elements directly 共Shahawy et al. 1996; Fanning and
E-mail: cspchiew@ntu.edu.sg Kelly 2000兲. The negligence of an interface bond normally gave
2
Graduate Student, School of Civil and Environmental Engineering, higher beam stiffness and strength, and finite-element 共FE兲 mod-
Nanyang Technological Univ., 50 Nanyang Ave., Singapore 639798. els based on this assumption are unable to predict interface bond
3
Graduate Student, School of Civil and Environmental Engineering, failure. Hence, the inclusion of interface bond behavior in the
Nanyang Technological Univ., 50 Nanyang Ave., Singapore 639798. numerical simulation is a fundamental step toward the develop-
Note. Discussion open until March 1, 2008. Separate discussions must ment of accurate nonlinear analysis of such structures. In many
be submitted for individual papers. To extend the closing date by one
engineering problems, geometric discontinuities like the one be-
month, a written request must be filed with the ASCE Managing Editor.
The manuscript for this paper was submitted for review and possible tween concrete and FRP composite are common, such as the bond
publication on October 24, 2005; approved on April 9, 2007. This paper between concrete and reinforcement, interlaminate connection of
is part of the Journal of Composites for Construction, Vol. 11, No. 5, laminated composites, and soil and reinforcement interaction. In
October 1, 2007. ©ASCE, ISSN 1090-0268/2007/5-497–506/$25.00. these problems, interface elements are utilized to model the inter-

JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / SEPTEMBER/OCTOBER 2007 / 497

J. Compos. Constr. 2007.11:497-506.

 action between dissimilar materials, and encouraging results have Table 1. GFRP Laminate Configurations
been obtained 共Schellekens and de Borst 1994; Reedy et al. 1997; Laminate
Jeong and Jung 2001; Pham and Al-Mahaidi 2004兲. Interface el- thickness l2 L3
ements can also be viewed as an ideal choice to simulate the bond Beam 共mm兲 共mm兲 共mm兲 l3 / l2
between concrete and FRP composite; however, very little work
A2 1.7 800 750 0.9375
on this aspect has been done in the past.
A3 3.4 800 750 0.9375
In this study, the main objective is to check the effect of glass-
A4 5.1 800 750 0.9375
fiber-reinforced polymer 共GFRP兲 in strengthening undamaged
concrete beams instead of repairing or rehabilitating old struc- A5 1.7 800 600 0.75
tures. Although most applications of FRP in strengthening con- A6 1.7 800 450 0.5625
crete structures are applied to old structures, there are still B2 1.7 1,100 1,050 0.95
requirements to strengthen undamaged concrete structures, such B3 3.4 1,100 1,050 0.95
as underdesigned structures. Besides, there are such applications B4 5.1 1,100 1,050 0.95
in real industry to strengthen undamaged concrete beams with B5 1.7 1,100 900 0.8
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.

FRP materials. Experiments on the GFRP strengthened concrete B6 1.7 1,100 750 0.65
beams are carried out to investigate the strengthening effect and
the stress/strain distribution in the interface between the concrete
and the GFRP layers. An interface element is developed to simu- to the length shown in Table 1. The laminate length in the shear
late debonding in the numerical analysis. The debonding of GFRP span is measured from the load point. A wet lay-up technique of
laminate is numerically simulated with the incorporation of the GFRP application is used. The laminate is saturated on site and
interface element in the FE analysis. The comparison between the glued to the soffit of the RC beam using a two-part epoxy resin
test and numerical results shows that the flexural behavior of the adhesive, TYFO-S Epoxy. The mechanical parameters of the ma-
GFRP strengthened concrete beams can be predicted with reason- terial are provided by the manufacturer. The hardened composite
able accuracy by the numerical analysis with the interface material has a Young’s modulus of 27 kN/ mm2 in the longitudi-
element. nal direction. The ultimate tensile strength of the laminate in the
longitudinal direction is 525 N / mm2. The nominal thickness of
the laminate is 1.3 mm.
Prior to bonding of the laminates, the beams are grinded using
Experimental Program
a mechanical grinder to remove the loose cement layer and to
expose the aggregate. The ground surface is then cleaned using
Test Beams water and then left in an ambient environment for two days to dry
Twelve 2.8 m long reinforced concrete 共RC兲 beams are tested to out. A layer of epoxy primer is first applied to the surface of the
failure. All beams are 200 mm wide by 350 mm deep with the beam before the GFRP laminates are applied. The epoxy saturated
same amount of internal reinforcements: two 16 mm diameter, GFRP laminates are then bonded to the beam. The laminates are
high yield steel bars with an effective depth of 317 mm are used pressed to keep out the possible air bubbles in the adhesive layer.
as tensile reinforcements; two 10 mm diameter bars at a depth of After all the laminates are attached to beams, an additional layer
30 mm are used as hanger bars. Bundles of two 10 mm mild steel of epoxy is applied to the bottom surface of the beam for protec-
bars spaced at 150 mm center-to-center are used as shear rein- tion purposes.
forcements. The average concrete cube compressive strength at
the time of testing is 41.4 N / mm2. From the uniaxial material Instrumentation and Test Procedure
tests, the Young’s modulus of the main tension reinforcement is
A special type of strain gauge 共BFLA-5-11, 5 mm gauge length兲
206 kN/ mm2, and yield and tensile stresses are 516 and
for composite is used to measure the strain developed in the
589 N / mm2, respectively. For the shear reinforcement, the
GFRP laminates. Positions of the strain gauges are shown in
Young’s modulus is 203 kN/ mm2 and the yield and the tensile
Fig. 2 where the acronym CL is short for center line. Strain
stresses are 560 and 596 N / mm2, respectively.
gauges are also attached to the tension reinforcement bars and the
All 12 beams are tested in four-point bending with a support-
surface of concrete in the compression zone near the midspan.
to-support distance of 2.6 m. The beams are divided into two
groups based on loading locations. As illustrated in Table 1 and
Fig. 1, two point loads are applied at 500 mm from the midspan
of the beams in Group A. In Group B, this distance is 200 mm.
There are six beams in each group: one control beam without
external GFRP laminates, three beams bonded with one layer, two
layers, and three layers of 2.5 m long GFRP laminates, and two
other beams strengthened with one layer of GFRP laminates at
length of 2.2 and 1.9 m. The width of the external laminates is the
same as that of the concrete beams.

External Reinforcements
TYFO S-Fibrwrap Composite System 共1997兲 共GFRP laminates
and epoxy resin adhesive兲 manufactured by FYFE Co. LLC is
used as the external tension reinforcements for flexural strength-
ening. The GFRP fabric TYFO-S Fabric SEH-51 is cut according Fig. 1. Details of test beams

498 / JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / SEPTEMBER/OCTOBER 2007

J. Compos. Constr. 2007.11:497-506.

 Fig. 2. Strain gauge position

Deflection at midspan is measured using linear variable differen-

tial transducers 共LVDTs兲. The LVDTs and the strain gauges are all
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.

connected to a computer and processed using a special program

called LabWindows. During the experiment, the load is applied
monotonically using a computer-controlled actuator with a capac-
ity of 200 kN. Fig. 4. Cracking and debonding behavior of strengthened beams:
共a兲 A2; 共b兲 A3; and 共c兲 A4

Test Results and Discussion

The moment–deflection response for each specimen is plotted in Cracking and Debonding Behavior
Fig. 3. The strength and the stiffness of the strengthened beams
are increased with a loss of the ductility. The two control beams The cracking and debonding diagrams for beams with one, two,
fail in typical flexural manner. Failure of a typical beam is char- and three layers of GFRP laminates are shown in Fig. 4. For
acterized by yielding of the tension reinforcement followed by beams with one layer of GFRP laminate, the bond surface of the
crushing of the concrete. At the moment of failure, the beams laminate is very smooth, with very little concrete debris attached.
experience excessive deflection. All the strengthened beams fail Therefore, the bond failure is believed to initiate from the
by the debonding of the external GFRP laminates. The debonding adhesive–concrete interface. For beams with more than one layer
happens in a catastrophic manner accompanied by emission of a of laminate, small concrete blocks are detached from the bottom
very loud sound. For beams with one layer of laminate, the failure of the concrete beam. The failure is an intermediate crack induced
is believed to initiate from the interface between the adhesive and debonding. The failed specimen 共A4兲 is shown in Fig. 5. It can be
the concrete. For beams with more than one layer of laminates, seen that failure occurs in both adhesive and concrete. Most
the bond failure initiates in the concrete at the concrete-to-FRP peeled concrete blocks are attached to the laminate, whereas some
interface at the base of a flexural crack. drop off to the ground. Longitudinal cracks near the loading
points and at the midspan of the beam are also observed. The
damage degree of the beam is closely related to the number of
laminate layers. Compared with the unstrengthened beams, the
strengthened ones are finer and more uniformly distributed
cracks. This phenomenon is caused by the stress redistribution in
the concrete resulting from the confining effect of external GFRP
laminates.

Fig. 3. Experimental moment–deflection curves: 共a兲 Group A beams;

共b兲 Group B beams Fig. 5. Typical interfacial bond failure

JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / SEPTEMBER/OCTOBER 2007 / 499

J. Compos. Constr. 2007.11:497-506.

 Table 2. Experimental and Numerical Results
Ultimate moment
共kN m兲 Maximum
Increase in Maximum Average interfacial
Ratio Stiffness stiffness GFRP strain GFRP straina shear stress
Beam Test FE 共FE/test兲 S.R. 共kN m / m兲 共%兲 共10−6兲 共10−6兲 共N / mm2兲
A1 65.2 — — — 6,939.2 — — — —
A2 81.4 79.0 0.97 1.25 7,267.6 4.7 7,580 7,065 3.2
A3 87.7 86.2 0.98 1.35 7,881.6 13.6 6,791 6,135 2.5
A4 95.4 93.5 0.98 1.46 8,603.1 24.0 5,726 5,090 3.9
A5 78.4 78.1 1.00 1.20 6,942.2 0.0 9,074 6,526 2.8
A6 81.9 76.2 0.93 1.26 7,069.8 1.9 8,763 6,253 3.8
B1 67.1 — — — 7,437.4 — — — —
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.

B2 80.4 79.6 0.99 1.20 7,879.9 5.9 9,040 7,607 1.4

B3 83.6 86.4 1.03 1.25 8,984.9 20.8 6,413 5,110 2.3
B4 97.3 94.7 0.97 1.45 9,136.9 22.9 6,306 5,925 2.2
B5 79.2 78.9 1.00 1.18 8,167.3 6.8 8,093 6,457 1.7
B6 80.1 77.3 0.97 1.19 8,176.3 7.6 7,465 6,636 2.2
a
Over the constant moment zone at failure load.

Strength and Stiffness of the Strengthened Beams Longitudinal GFRP Strain

As shown in Fig. 3, both strength and stiffness of the strengthened Longitudinal GFRP strain profiles for Beam A2 are plotted in
beams are improved with the external bonding of the GFRP lami- Fig. 7 at various load levels. It is found that concrete cracking and
nates. Strengthening ratio 共S.R.兲, equaling to failure moment of yielding of tension reinforcement have a very significant effect on
the strengthened beam over that of the control beam, is calculated the strain development in the GFRP laminates. Prior to the crack-
for each beam to evaluate the effectiveness of the external ing of concrete, profile of the strain follows a trapezoidal shape
strengthening. The results are tabulated in Table 2. It can be seen which matched the moment distribution diagram very nicely. Just
that with the bonding of the GFRP laminates, the ultimate after the cracking moment at 16 kN m, GFRP strain below the
strengths of the beams are enhanced to an extent of 18–46%. The loading point increases 100–200% as compared to that of the
increases of the stiffness of the beams are noticeable—up to 24% previous step. This apparently is caused by the appearance of
of the control beam. The failures of the strengthened beams are of flexural cracks, as illustrated in Fig. 8. With the increase of load,
a brittle mode. For common concrete beams, the failure mode is strain in the constant moment zone as well as shear span increases
ductile. This is an important difference between the behaviors of gradually. After yielding of the tension reinforcement, there is
strengthened and unstrengthened concrete beams. Therefore, it is another major change of the GFRP strain profile as shown in
not reasonable to compare the strengthening ratio of stiffness Fig. 7共b兲. GFRP strain in the constant moment zone increases
共S / S0兲 directly to the strengthening ratio of maximum capacity tremendously. At failure, the average GFRP strain in the constant
共M / M 0兲 because, when considering the load carrying capacity in moment zone almost doubles the value of that at the yielding
strengthening applications, a higher design factor for strengthened load. This implies significant transfer of tension force from the
beams is needed than that for unstrengthened beams. Actually, internal tension bars to the external GFRP laminates. In compari-
this unproportional increase of stiffness and maximum capacity is son, the variation of strain in shear span, particularly in the vicin-
a common phenomenon for FRP strengthened concrete beams. ity of the plate end, is insignificant. At failure, strain distribution
The most significant parameter on strengthening ratio appears from the loading point to the plate end is not linear. Two strain
to be the number of layers of the external GFRP laminates. In gradients near the plate end and just outside the constant moment
Fig. 6, the strengthening ratio is plotted against the axial rigidity
of the GFRP laminates 共E pA p兲. E p⫽elastic modulus of the GFRP
laminate and A p⫽cross area of the GFRP laminate. In contrast,
the variation of the GFRP bond length in shear span shows very
little effect on the strengthening ratio. This can be seen by angli-
cizing Beams A2, A5, and A6. The bond lengths in shear span of
these beams are 750, 600, and 450 mm 共l3 / l2 ratios at 0.94, 0.75,
and 0.56, respectively兲 and the respective strengthening ratios are
1.25, 1.20, and 1.26. Therefore, the longer bond strength does not
show a better strengthening effect. The standard deviation of
strengthening ratios of all the 6 beams with one layer of laminate
共A2, A5, A6, B2, B5, and B6兲 is only 0.03. The stiffness increase
of each strengthened beam is listed in Table 2. The stiffness of
beams is defined as moment per unit deflection, corresponding to
the point on the moment–deflection curve where the deflection is
equal to span/500. The trend of the stiffness increase is similar to Fig. 6. Effect of axial rigidity of GFRP laminate on beam
that of the strengthening ratio. strengthening

500 / JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / SEPTEMBER/OCTOBER 2007

J. Compos. Constr. 2007.11:497-506.

Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 9. Effect of axial rigidity of GFRP laminate on maximum GFRP

strain

Maximum GFRP Strain

Due to the presence of concrete cracks, GFRP strain in the con-
stant moment zone is not uniform. In the study of maximum
GFRP strain, both the maximum single GFRP strain and the av-
erage value of GFRP strain in the constant moment zone are
presented in Table 2. For both cases, GFRP strain value seems to
reduce with the increase of the number of GFRP laminate layers.
However, for beams with one layer of GFRP laminate but with
different l3 / l2 ratios, the maximum GFRP strain shows significant
Fig. 7. Longitudinal GFRP strain distribution 共Beam A2兲: 共a兲
inconsistency. In Fig. 9, where the maximum GFRP strain is plot-
Cracking load; 共b兲 yielding load and failure load
ted against axial rigidity of the external GFRP laminates, the re-
lationship is established through linear regression analysis.

Interfacial Shear Stress

zone are apparent. The inconsistency in the strain in the GFRP is
caused by the partial bond failure or the concrete cover separation Shear stress along the GFRP and concrete beam interface is cal-
from the beam. The strain is higher where the local bond failure culated based on force equilibrium, and the expression for inter-
or concrete peel-off happens. For instance, when the moment is facial shear stress is
16 kN m, the local failure happens at the bond or bottom concrete
just in vertical line with the loading points. At the same time, the ␶ = E pt p共d␧兲/共dl兲 共1兲
stress in the whole structure redistributes and the GFRP in this where ␶⫽interfacial shear stress 共N / mm 兲; t p⫽thickness of the
2
area shares more stress transferred from the beam to keep the external GFRP laminates 共mm兲; and 共d␧兲 / 共dl兲⫽strain gradient
force of the whole structure balanced. Therefore, the strain dis- between two adjacent strain gauge readings.
tributed in this area is higher. In Fig. 10, interfacial shear stress distributions for Beam A4
before and after the concrete cracking are plotted. Two stress
concentration zones can be observed in Fig. 10: At the end of the
laminate and just outside the constant moment zone. The stress
concentration at laminate end is a well-known phenomenon,
which has been studied by many researchers in the past. It is
caused by the sudden change of beam cross section due to termi-
nation of GFRP laminate. Interfacial shear concentration outside
the constant moment zone is caused by cracking of concrete and
drop of tensile force in the GFRP laminate. Prior to cracking,
concrete behaves like elastic material. Stress concentration adja-
cent to the constant moment zone is smaller than that at the lami-
nate end, as shown by the lowest curve at M = 16 kN m.
After cracking, interfacial shear stress beside the constant mo-
ment zone increases tremendously. It triples the value in the pre-
vious step as shown by the second lowest curve at M = 22 kN m.
The drastic increase is undoubtedly caused by localized concrete
cracking. However, the increase of load shows much less effect
on the stress concentration at the laminate end. It increases merely
0.03 N / mm2 from 0.07 to 0.1 N / mm2. It can also be observed
from the curves that, at a moment level of 16 kN m, only the bond
Fig. 8. Effect of concrete cracks on strain measurement in the region between 600 and 125 mm from midspan is effective.

JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / SEPTEMBER/OCTOBER 2007 / 501

J. Compos. Constr. 2007.11:497-506.

 Fig. 11. Eight-node interface element
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.

Finite-Element Analysis

Finite-Element Modeling
An eight-node interface element is implemented into a general-
purpose FE package MARC 共1997兲 to carry out numerical analy-
sis of the GFRP laminate strengthened RC beams. The schematic
view of the element is shown in Fig. 11. The development details
of this interface element are shown in the Appendix. In the analy-
sis, double symmetry of load and geometry is made use of, and
Fig. 10. Interfacial stress distribution 共Beam A4兲: 共a兲 before only a quarter of the beam is modeled. Appropriate boundary
yielding; 共b兲 after yielding conditions are imposed.
In the model, Hognestad’s parabola and linear descending
branch are used to simulate behavior of the concrete in compres-
With further increase of the load, the stress concentration caused sion. Its uniaxial compressive stress–strain curve is assumed to
by concrete cracking becomes more significant than that at the follow the curve in Fig. 12共a兲, in which f cu⫽compressive strength
laminate end, as depicted by the two interfacial shear stress of concrete 共N / mm2兲. Strain ␧0 is taken as 0.002. The equation of
profiles at M = 38.5 and 55 kN m. At these two load levels, con- the parabola is given as
centrated stresses induced by concrete cracking are normally
three times of that at the laminate end. The increase of load not f c = f cu关2␧c/␧0 − 共␧c/␧0兲2兴 共2兲
only aggravates the level of interfacial stress concentration, but
also activates more interfacial debonding toward the laminate Crushing strain ␧cu of the concrete is taken as 0.0035; the equa-
end. At a moment of 38.5 kN m, the span of effective bond region tion for the linear descending branch is given as
increases from 600 to 1,050 mm from midspan. Along with the
activation of interfacial debonding, the position of the maximum 共0.85f cu − f cu兲/共␧cu − ␧0兲 = 共f c − f cu兲/共␧c − ␧0兲 共3兲
interface shear stress also shifts toward the support.
In MARC, low tensile capacity material like concrete is mod-
Fig. 10共b兲 shows four interfacial shear stress profiles: one prior
eled using the smeared crack approach by activating the CRACK
to yielding of the tension steel at 55 kN m, two after yielding of
DATA option. This approach makes use of the drastic material
the steel at 71.5 and 90 kN m, and one at failure load at
property changes caused by cracking as a way to simulate the
97.4 kN m. It is noticed that at Point A in the curve, negative
discontinuity. Such changes are achieved by reducing the stiffness
interfacial stress occurs. It is caused by the appearance of a
in the direction orthogonal to the crack without introducing any
concrete crack near the strain gauge as illustrated in Fig. 8. As
gap in the initial mesh. As the material properties are evaluated
the interfacial shear stress is calculated based on the strain gradi-
only at specific integration points in an element, the alteration of
ent between two adjacent strain gauges, the appearance of
the material properties due to cracking consequently affects the
the crack near the strain gauge results in high strain, and thus
contributing region from which these properties are evaluated,
leads to the adversely distributed interfacial shear stresses on both
hence smearing the effect of cracking over the whole region.
sides of the crack. After yielding of the tension reinforcement,
By activating the CRACK DATA option, several material prop-
the interfacial shear stress in the shear span keeps increasing with
erties are needed in modeling of the concrete: Critical cracking
the increase of load. The interfacial debonding starting from the
stress f tu, modulus of linear strain softening Et, strain at which
crack toward the laminate end is gradually activated. Debonding
crushing occurs ␧cu, and shear retention factor ␤. The critical
occurs when the interfacial shear stress reaches 2.2 N / mm2 at
cracking stress f tu⫽tensile strength of the concrete from material
40 mm from the laminate end, as indicated by the curve at
test or calculated based on the formula given in the CEB-FIP
97.4 kN m.
model code 共CEB-FIP 1993兲
The calculated maximum interfacial shear stresses 共␶max兲 of all
the ten strengthened beams are tabulated in Table 2. The results
f tu = 0.3f 2/3 共4兲
show significant scatter, with the lowest at 1.4 N / mm2 and high- cu

est at 3.9 N / mm2. Variations in GFRP thickness and bond length The tension softening modulus of the concrete Et can also be
show no clear effect on ␶max. calculated based on the CEB-FIP model code 共CEB-FIP 1993兲

502 / JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / SEPTEMBER/OCTOBER 2007

J. Compos. Constr. 2007.11:497-506.

Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 13. Numerical and experimental moment–deflection curves 共B2兲

conducted by Horiguchi and Saeke 共1997兲. The relative slip be-

tween the concrete and the GFRP laminate is very difficult to be
determined through experimental investigation. Based on the re-
port of Nakaba et al. 共2001兲, the slip is believed to be in the range
of 10– 100 ␮m. Lu et al. 共2005a兲 also proposed bond–slip models
by using the mesoscale FE modeling technique 共Lu et al. 2005b兲.
In the analysis carried out here, the slip at maximum bond stress
Fig. 12. Uniaxial material models: 共a兲 concrete; 共b兲 steel; 共c兲 GFRP level is thus taken as 10 ␮m in the normal direction and 50 ␮m in
laminate; 共d兲 interface in tangential direction; and 共e兲 interface in the tangential direction, which is a little smaller than that has been
normal direction numerically predicted by Lu et al. 共2005a兲.

Et = 2G f /L 共5兲 Numerical Results

where L⫽element size; G f ⫽fracture energy of the concrete, and it The strengthened beam models with interface elements are loaded
can be calculated by at the locations according to the actual test setup using displace-
ment control. The analysis stops when a prescribed displacement
G f = G f0共f cu/f cu0兲0.7 共6兲
is exceeded or the equilibrium is attained during a given load
2 increment.
where f cu0 = 10 N / mm ; and G f0 is related to the maximum ag-
gregate size based on Table 3. As shown from Figs. 13–15, the numerical and experimental
Shear retention factor 共␤兲 is taken as 0.1 throughout the analy- moment–deflection curves of three beams, B2, A3, and A4, are
sis based on the suggestion given by Kotsovos and Pavlovic compared. It is evident that the behaviors of the beams are
共1995兲. predicted fairly well. Cracking moment, yielding moment, and
Tension reinforcement steel is idealized as an isotropic elastic– ultimate failure moment from the numerical analysis are close to
perfectly plastic material. The elastic modulus and the yielding those from the experiments. As shown in Table 2, the error be-
stress are taken as 200 kN/ mm2 and 516 N / mm2, respectively, tween the numerical and the experimental results falls into the
based on the coupon tests. The Poisson’s ratio is taken as 0.3. The range from −7 to 3%. The numerical results cannot agree per-
idealized stress–strain curve is shown in Fig. 12共b兲. The GFRP fectly with those of the experiments because the bond strength of
laminate is idealized as an isotropic elastic material. Its elastic the interface element comes from the average interfacial shear
modulus and ultimate strength are taken as 27 kN/ mm2 and stress of the tested beams. This is why there is a discrepancy
525 N / mm2, respectively, as provided by the supplier. The stress– between the maximum experimental and calculated deflections in
strain curve of the GFRP material is shown in Fig. 12共c兲.
A linear cutoff bond–slip constitutive law is assumed to gov-
ern its behavior of the interface element in both tangential and
normal directions, as shown in Figs. 12共d and e兲. The bond
strength in the tangential direction is taken as 2.6 N / mm2, the
average value of the calculated interfacial shear stress from the
experiment. The bond strength in the normal direction is taken as
1.6 times that in the tangential direction based on the experiment

Table 3. Coefficients for Determination of the Fracture Energy

dmax G f0
共mm兲 共J / mm2兲
8 25
16 30 Fig. 14. Numerical and experimental moment–deflection curves
32 38 共A3兲

JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / SEPTEMBER/OCTOBER 2007 / 503

J. Compos. Constr. 2007.11:497-506.

Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 15. Numerical and experimental moment–deflection curves

共A4兲

Figs. 13 and 14. All the strengthened beams in the FE modeling

fail by the debonding of external GFRP laminates from beam
soffit, corresponding to the considerable drop of the moment in
the moment–deflection curves. The interfacial bond failure ini-
tiates from the loading points. This is consistent with experimen-
tal observation. In the numerical analysis, debonding failure
occurs when the slips of all the interface elements reach their
maximum values. Once the GFRP laminate debonds, the behavior
of the beam restores to that of the plain unstrengthened beam. Fig. 16. Interfacial stress distribution 共Beam A2兲
This is evident in Figs. 3共a兲 and 15.

nate in the shear span is gradually activated after cracking of the

Interfacial Stress Development from Numerical
concrete from the loading points toward the beam supports; and
Analysis
共3兲 at failure, stress concentration level at the laminate end is very
Interfacial stress distribution and development along the beam are high. The interfacial shear stress development diagram also re-
presented in Fig. 16. As can be seen from Fig. 16共a兲, prior to the veals that the debonding initiates near the loading point and then
concrete cracking the interfacial stress is relatively small: propagates toward the plate end. This agrees with the recent ob-
0.1 N / mm2 at the plate end and 0.35 N / mm2 below the loading servation made by Saffi and Sayyah 共2001兲.
point. At this stage, the interfacial stresses in the interface ele- Though the interfacial shear stress profile from the FE analy-
ments are still below the ultimate bond strength and none of the ses does not match the experimental data exactly, the numerical
interface elements are damaged. model portrays the effect of concrete cracking, as well as interfa-
When the stress in the concrete element had reached the speci- cial stress development. With the brittle nature of interfacial bond
fied cracking strength, the stiffness matrix of the element is up- in mind, the model can be deemed as adequate in studying the
dated with the given tension softening modulus Ect and shear behavior of the GFRP laminate strengthened RC beams.
retention factor ␤. The cracking of the concrete led to damage of
the interface element at the crack location and a drastic increase
Effect of Bond Strength
of the interfacial shear stress near the crack is induced. After
cracking, the interfacial shear stress below the loading point in- The numerical models are also used to study the effect of the
creases from 0.35 to 1.44 N / mm2, whereas the stress at the plate variation in bond strength on the beam behavior. FE analyses
remains almost unchanged. Following the damage of the first in- based on Beam A3 with bond strengths of 1.5, 3, 4.5, and
terface element near the loading point, an interfacial crack be- 7.5 N / mm2 are performed. The slip at the maximum stress is kept
tween concrete and GFRP laminate is produced. It propagates unchanged. The perfect bond model is also considered, in which
toward the end of the laminate and the midspan of the beam with the concrete elements are directly connected with the GFRP ele-
a continuous increase of the load. When the load reaches ments. Results from these models are shown in Fig. 17. The ex-
63.3 kN m, all the interface elements in the constant moment perimental moment–deflection curves from the perfect bond
zone are damaged and only those in the shear span remain effec- model are also plotted for comparison.
tive. Following that, the interface elements are damaged gradually At relatively low moment levels, the curves from the analyses
from the interfacial crack tip toward the end of the laminate. The with and without debonding almost coincide with each other.
beam reaches its ultimate capacity when the last interface element They start to develop differently when the interface elements in
connecting the laminate and the concrete is damaged. the constant moment region are all damaged. From Fig. 17, the
The curves of interfacial shear stress distribution and develop- beneficial effect of higher bond strength is evident in terms of
ment based on the numerical analyses show similarities with ex- beam strength, stiffness, and ductility. It can also be seen that with
perimental observations in three aspects: 共1兲 Flexural cracks have the increase of bond strength, the moment–deflection curve of the
significant effects on the interfacial stress development; 共2兲 the model with debonding moves toward that of the model with a
interfacial debonding between the concrete and the GFRP lami- perfect bond. However, even when the bond strength is increased

504 / JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / SEPTEMBER/OCTOBER 2007

J. Compos. Constr. 2007.11:497-506.

 4 4 4

x= 兺
i=1
N ix i, y= 兺
i=1
N i y i, z= 兺
i=1
N iz i 共7兲

in which 共xi , y i , zi兲⫽coordinates of node i and Ni⫽its shape func-

tion which can be expressed as follows:

Ni = 共1 + ␰i␰兲共1 + ␩i␩兲/4 再 i = 1,2,3,4

i = 5,6,7,8
for lower surface
for upper surface
共8兲
The relative displacement of the two surfaces in the axial di-
rection may be expressed as
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 17. Effect of bond strength on beam behavior 兵⌬其 = 关⌬u ⌬v ⌬w 兴T = 关N兴兵␦其 共9兲

where 关N兴 and ␦⫽shape function matrix and the nodal displace-
to 7.5 N / mm2, the ultimate capacity of the beam is only three ment vector of the interface element, respectively
quarters of that of the model with the perfect bond. This suggests
the beam strength could be greatly overestimated if debonding is 关N兴 = 关− 关N1兴 ¯ − 关N4兴关N1兴 ¯ 关N4兴兴 共10兲
not considered in the modeling.
The previous study shows that the bond strength plays an im-

冤 冥
portant role in the behavior of GFRP laminate strengthened RC Ni 0 0
beams. The assurance of good bond quality is vital for the success 关Ni兴 = 0 Ni 0 共11兲
of such a strengthening technique. In the analysis of GFRP lami-
nate strengthened beams, the inclusion of debonding behavior 0 0 Ni
produces more realistic and sensible results.
兵␦其 = 关兵␦1其T兵␦2其T ¯ 兵␦8其T兴T 共12兲

Conclusions
兵␦i其T = 关ui vi wi 兴 共13兲
By bonding GFRP laminates to the tension face of flexural RC
beams, both strength and stiffness of the beams can be increased. The traction force at the interface can be assumed to vary propor-
The strengthening ratio increases linearly with the increase of the tionally with the relative displacements of the upper and lower
axial rigidity of the external GFRP laminates. In contrast, the faces of the interface element, i.e.
variation of bond length in the shear span has little effect as long
as the l3 / l2 ratio is larger than 0.56. The interfacial debonding is 兵t其 = 关␭兴兵⌬其 共14兲
progressively activated with the increase of the external load from
below the loading point toward the end of the laminate. The in- where 兵t其 = 关tx ty tz 兴T⫽traction vector and 关␭兴⫽stiffness coeffi-
terfacial shear stress concentration due to the cutoff effect is less cient matrix in the global coordinate system that can be obtained
significant than that caused by flexural cracking. Debonding fail- from 关␭⬘兴, the stiffness coefficient matrix in the local coordinate
ure occurs when the interfacial bond in the shear span is fully system via transforming matrix 关L兴:
utilized.
The FE analysis with the eight-node interface element can 关␭兴 = 关L兴T关␭⬘兴关L兴 共15兲
predict the moment–deflection relationship of the strengthened
beams fairly well. The nonlinear FE model captures the gradual in which
activation of the interfacial debonding. The numerical study on

冤 冥
the effect of the variation in the bond strength shows that higher ␭s 0 0
bond strength could lead to a higher strengthening ratio, greater
关␭⬘兴 = 0 ␭s 0 ,
stiffness, and better ductility. The omission of the interfacial de-
bonding in the analysis would lead to overestimation of the beam 0 0 ␭n
strength as well as stiffness.
␭s⫽stiffness coefficient in the tangential direction of the interface
and ␭n⫽stiffness coefficient in the normal direction of the inter-
face; and
Appendix. Development of Interface Element

冤 冥
k1 k2 k3
An eight-node interface element as shown in Fig. 11 is developed
to simulate the interface behavior between concrete and GFRP 关L兴 = m1 m2 m3
laminate. Although the corresponding nodes, e.g., Nodes 1 and 5, n1 n2 n3
on the two surfaces have the same coordinate, they have different
node numbers and degrees of freedom. Each surface is expressed ⫽transforming matrix of the local coordinate system, in which ki,
by its nodes and corresponding shape functions. For example, the mi, ni⫽cosines of the local coordinate axis xi⬘ with respect to the
lower surface is expressed by global coordinate axes X1, X2, X3, respectively.

JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / SEPTEMBER/OCTOBER 2007 / 505

J. Compos. Constr. 2007.11:497-506.

 The element stiffness matrix can thus be calculated as follows: Horiguchi, T., and Saeke, N. 共1997兲. “Effect of test methods and quality

冕冕
+1 +1 of concrete on bond strength of CFRP sheet.” Proc., 3rd Int. Symp.,
关K兴e = 关N兴T关L兴T关␭兴关L兴关N兴兩J兩d␰d␩ 共16兲 Nonmetallic (FRP) Reinforcement for Concrete Structures, Vol. 1,
−1 −1 Tokyo, 265–270.
Jeong, Y. J., and Jung, K. H. 共2001兲. “Nonlinear analysis with slip
in which and composite action in composite structure of sandwich system.”
兩J兩 = 兵关共⳵ x/⳵ ␰兲共⳵ y/⳵ ␩兲 − 共⳵ y/⳵ ␰兲共⳵ x/⳵ ␩兲兴2 1st Int. Conf. on Steel and Composite Structures, Pusan, Korea,
1505–1514.
+ 关共⳵ y/⳵ ␰兲共⳵ z/⳵ ␩兲 − 共⳵ z/⳵ ␰兲共⳵ y/⳵ ␩兲兴2 Kishi, N., Zhang, G., and Mikami, H. 共2005兲. “Numerical cracking and
debonding analysis of RC beams reinforced with FRP sheet.” J. Com-
+ 关共⳵ z/⳵ ␰兲共⳵ x/⳵ ␩兲 − 共⳵ x/⳵ ␰兲共⳵ z/⳵ ␩兲兴2其1/2 共17兲 pos. Constr., 9共6兲, 507–514.
Kotsovos, M. D., and Pavlovic, M. N. 共1995兲. Structural concrete: Finite-
Notation element analysis for limit-state design, Thomas Telford Services Ltd.,
London.
Downloaded from ascelibrary.org by New York University on 05/10/15. Copyright ASCE. For personal use only; all rights reserved.

The following symbols are used in this paper: Leung, C. K. Y. 共2001兲. “Delamination failure in concrete beams retrofit-
dmax ⫽ maximum aggregate size; ted with a bonded plate.” J. Mater. Civ. Eng., 13共2兲, 106–113.
f c ⫽ compressive stress in concrete; Lu, X. Z., Teng, J. G., Ye, L. P., and Jiang, J. J. 共2005a兲. “Bond-slip
models for FRP sheets/plates bonded to concrete.” Eng. Struct.,
f cu0 ⫽ critical compressive strength of concrete;
27共6兲, 920–937.
f y ⫽ yield strength of steel;
Lu, X. Z., Ye, L. P., Teng, J. G., and Jiang, J. J. 共2005b兲. “Mesoscale finite
G f0 ⫽ critical fracture energy of concrete; element model for FRP plates/sheets bonded to concrete.” Eng.
J ⫽ Jacobian matrix; Struct., 27共4兲, 564–575.
关K兴 ⫽ element stiffness matrix; Malek, A. M., Saadatmanesh, H., and Ehsani, M. R. 共1998兲. “Prediction
l1 ⫽ length between two loading points; of failure load of RC beams strengthened with FRP plate due to stress
l2 ⫽ length from loading point to support; concentration at the plate end.” ACI Struct. J., 95共1兲, 142–152.
l3 ⫽ length from loading point to laminate end; MARC Analysis Research Corporation. 共1997兲. Volume D: User
M ⫽ moment; subroutine/special routines, Santa Ana, Calif.
兵t其 ⫽ traction vector; Nakaba, K., Kanakubo, T., Furuta, T., and Yoshizawa, H. 共2001兲. “Bond
ti ⫽ traction in the direction Xi; behaviour between fibre-reinforced polymer laminates and concrete.”
ts ⫽ tangential traction of the interface; ACI Struct. J., 98共3兲, 359–367.
ui , vi , wi ⫽ displacement of node i in the X , Y, and Z Oehlers, D. J., and Seracino, R. 共2004兲. Design of FRP and steel plated
directions, respectively; RC structures: Retrofitting beams and slabs for strength, stiffness, and
us ⫽ tangential relative displacement of the interface; ductility, Elsevier, Amsterdam, The Netherlands.
x , y , z ⫽ coordinates in global coordinate system; Pham, H. B., and Al-Mahaidi, R. 共2004兲. “Assessment of available
␦nm ⫽ maximum allowable slip in normal direction; prediction models for the strength of FRP retrofitted RC beams.”
␦sm ⫽ maximum allowable slip in tangential direction; Compos. Struct., 66共1–4兲, 601–610.
␧c ⫽ strain at outmost compressive fiber of concrete; Reedy, E. D., Mello, F. J., and Guess, T. R. 共1997兲. “Modeling the
␧s ⫽ strain in steel; initiation and growth of delaminations in composite structures.”
J. Compos. Mater., 31共8兲, 812–831.
␧y ⫽ yield strain of steel;
Saadatmanesh, H., and Malek, A. M. 共1998兲. “Design guidelines for flex-
␰ , ␩ ⫽ local coordinate axes of the interface element;
ural strengthening of RC beams with FRP plates.” J. Compos. Constr.,
␰i , ␩i ⫽ coordinates of the node i of the interface 2共4兲, 158–164.
element; Saffi, M., and Sayyah, T. 共2001兲. “Health monitoring of concrete
␴nm ⫽ normal bond strength; structures strengthened with advanced composite materials using
␶max ⫽ maximum interfacial shear stress 共bond strength兲; piezoelectric transducers.” Composites, Part B, 32共4兲, 333–342.
and Schellekens, J. C. J., and de Borst, R. 共1994兲. “Free edge delamination in
␶sm ⫽ tangential bond strength. carbon-epoxy laminates: A novel numerical/experimental approach.”
Compos. Struct., 28共4兲, 357–373.
Shahawy, M. A., Arockiasamy, M., Beitelman, T., and Sowrirajan, R.
References 共1996兲. “Reinforced concrete rectangular beams strengthened with
CFRP laminates.” Composites, Part B, 27B, 225–233.
Buyukozturk, O., Gunes, O., and Karaca, E. 共2004兲. “Progress on under- Smith, S. T., and Teng, J. G. 共2001兲. “FRP-strengthened RC beams. I:
standing debonding problems in reinforced concrete and steel mem- Review of debonding strength models.” Eng. Struct., 24共4兲, 385–395.
bers strengthened using FRP composites.” Constr. Build. Mater., Teng, J. G., Chen, J. F., Smith, S. T., and Lam, L. 共2002兲. FRP-
18共1兲, 9–19. strengthened RC structures, Wiley, Hoboken, N.J.
Comité Euro-International du Béton 共CEB-FIP兲. 共1993兲. CEB-FIP model TYFO S-Fibrwrap Composite System. 共1997兲. Technical data for TYFO
code 1990—Design code, Thomas Telford, London, CEB Bulletin SEH-51 composite system using TYFO-S epoxy, FYFE Co., LLC,
d’Information No. 213/214. San Diego.
Fanning, P., and Kelly, O. 共2000兲. “Smeared crack models of RC Yao, J., Teng, J. G., and Lam, L. 共2005兲. “Experimental study on inter-
beams with externally bonded CFRP plates.” Comput. Mech., 26共4兲, mediate crack debonding in FRP-strengthened RC flexural members.”
325–332. Adv. Struct. Eng., 8共4兲, 365–395.

506 / JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / SEPTEMBER/OCTOBER 2007

J. Compos. Constr. 2007.11:497-506.