Construction and Building Materials 24 (2010) 956–969

Contents lists available at ScienceDirect

Construction and Building Materials

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

Structural behavior of hybrid FRP composite I-beam

Nguyen Duc Hai a,*, Hiroshi Mutsuyoshi a,*, Shingo Asamoto a, Takahiro Matsui b
a
Department of Civil and Environmental Engineering, Saitama University, Saitama 338-8570, Japan
b
Advanced Composites Technical Dept., Toray Industries, Inc., Tokyo, Japan

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

Article history: This paper presents the structural behavior of an innovative hybrid Fiber Reinforced Polymers (FRP) beam
Received 16 June 2009 consisting of carbon/glass fibers and vinyl-ester resin. The advanced feature of this hybridization is the
Received in revised form 27 September optimum use of carbon and glass fibers in the flanges to maximize structural performance while reducing
2009
the overall cost by using only glass fibers in the web section. A series of beam tests were conducted under
Accepted 19 November 2009
Available online 21 December 2009
four-point bending varying ratio of flange to web width (bf/bw) and volume content of carbon and glass
fiber in the flanges. Experimental investigations revealed that the ratio of flange to web width of hybrid
FRP I-shaped beams plays an important role in their structural behavior. Small flange beams (bf/bw = 0.43)
Keywords:
Hybrid FRP
showed stable and linear behavior under bending moment and failed in a brittle manner by delamination
Carbon fibers of the compressive flange at the interfacial layers while wide flange beams (bf/bw = 1.13) exhibited unsta-
Glass fibers ble and nonlinear behavior in the buckling and post-buckling region leading to delamination failure of the
Volume content compressive flange. The experimental and analytical results discussed in this paper emphasize on the
Best composition best composition of carbon and glass fibers for the optimum design of such hybrid beams. It is found that
the maximum strength of hybrid FRP beams can be obtained with the volume content of carbon fiber to
be 25–33%. Furthermore, the results of this study show the potential of applying hybrid FRP beams for
bridge components.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction span continuous girder pedestrian bridge as shown in Fig. 1. All

FRP composites have been increasingly used in civil infrastruc-
ture applications due to their advantageous properties such as high
specific strength/stiffness, lightweight, and corrosion resistance.
The apparent initial high cost of FRP materials has been a major
obstacle for the widespread use in bridge applications. However,
they are very competitive in term of durability.
The application of FRP in civil engineering was first introduced
in Japan about two decades ago with the development of FRP rein-
forcement and tendons. The first JSCE standard specification for the
design of FRP reinforcements was published in 1997, which has
been adopted in various other countries [1]. Since then, a wide
range of applications of FRP materials have been implemented
including strengthening of structures using FRP sheets and using
FRP rods as reinforcement instead of steel bars.
In the last decade, the research and development of all FRP
structures in civil engineering has progressed substantially in sev-
eral countries [2,3]. The first all FRP bridge in Japan was con-
structed in Okinawa prefecture in 2001 [4]. This bridge is a two
* Corresponding authors. Tel.: +84 43 773 2448 (N.D. Hai), tel.: +81 48 858 3553;
fax: +81 48 858 7374 (H. Mutsuyoshi).
E-mail addresses: hai.nguyen_2004@yahoo.com (N.D. Hai), mutuyosi@mail.sai-
tama-u.ac.jp (H. Mutsuyoshi).
the structural members have been made with Glass Fiber Rein-
forcement Polymers (GFRP). The all GFRP solution was chosen for
this bridge because it is surrounded by heavily corrosive environ-
ment. Though GFRP is cheaper than Carbon Fiber Reinforced Poly-
mers (CFRP), it has smaller stiffness and strength than CFRP. There
may be some difficulties in using all GFRP for bridges because the
ratio of deflection to span is a critical issue for designing bridges.
One possibility to overcome the above problem is to utilize GFRP
combined with high strength and stiffness of CFRP.
In this study, a hybrid FRP beam that combines GFRP and CFRP
was developed. The advantage of using this hybridization is to uti-
lize the superior strength of CFRP in the flanges while keeping the
material costs low by using GFRP in the flanges and web. By incor-
porating appropriate amount of CFRP/GFRP in hybrid composite, a
better performance in fatigue can be achieved [5,6]. However, the
most important advancements of the hybrid beam are the ability
of tailor-made/rapid construction, corrosion resistance, and reduc-
tion of carbon dioxide (CO2). These are of considerable significance
in Japan because of the large number of bridges corroded by the se-
vere environment that will need to be replaced in the near future.
Over the last decade, there has been significant growth in the re-
search and development of hybrid FRP composites. Numerous theo-
retical and experimental investigations have been conducted
worldwide regarding the combined use of carbon/glass/matrices

0950-0618/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conbuildmat.2009.11.022
 N.D. Hai et al. / Construction and Building Materials 24 (2010) 956–969 957

[7–16]. Summerscales and Short [7] have reviewed a number of

Fig. 1. Okinawa road park bridge (all GFRP bridge).
studies of hybrid carbon–glass composites in which a phenomenon
termed ‘‘hybrid effect” has been found. The essential observation of
the hybrid effect is that the failure strain of carbon fiber becomes
greater in a hybrid composite than that in an all-carbon fiber
composite. Observation of the hybrid effect was first credited to
Hayashi [8] and has subsequently been reported by several
researchers [9–15]. Although a great deal of information is available
on the coupon test results of hybrid composites, little work has been
done on the structural behavior of hybrid beam. In addition, studies
in literature cover a wide range of fiber and matrix types, fiber lay-up
and stacking sequences, etc., which result in different structural
behavior. Thus, additional investigations are required to enable
the civil engineers to have confidence in designing of hybrid beams
for real bridge application. This study focuses on the structural
behavior of hybrid FRP beams consisting of multi-layer carbon/
E-glass (woven fabric and continuous strand mat)/vinyl-ester. An
I-shaped section of FRP beam was developed as the first step of the
ongoing research project since it is easily manufactured and

CFRP +GFRP GFRP

CFRP +GFRP

Fig. 2. Dimensions of hybrid FRP beams.

958 N.D. Hai et al. / Construction and Building Materials 24 (2010) 956–969

commonly used in bridge structures. A number of beam tests were Table 1

experimentally conducted under four-point bending to investigate Experimental variables.

the structural behavior of hybrid FRP I-beams and to determine Beam CFRP-0d GFRP-0/90d GFRP±45d GFRP-CSMa
the best composition of CFRP and GFRP in the beams. The test results (%) (%) (%) (%)
were discussed and compared with the theoretical calculation and Flange
three-dimensional finite element analysis. A52b, B52c 52 10 13 25
A33b, B33c 33 29 13 25
A14b 14 48 13 25
2. Experimental procedure Web
All beams – 32 42 26
All beams were manufactured by the Resin Transfer Molding process consisting
a
of carbon and E-glass/vinyl-ester. Two distinct series of FRP I-beams varying the ra- Continuous strand mat.
b
tio of flange to web width (bf/bw) were tested to examine their structural behavior, SF beams (flange width: 95 mm).
c
(i) SF: small flange (bf/bw = 0.43) and (ii) WF: wide flange (bf/bw = 1.13). The dimen- WF beams (flange width: 250 mm).
d
sions of the beams are shown in Fig. 2. The beams were simply supported and Angles of fibers to longitudinal direction.
tested in four-point bending at a span of 3000 mm with an interior loading span
of 1000 mm.
All beams were made of CFRP and GFRP in the flanges and only GFRP in the web.
A typical layer composition of the flanges and web is presented in Fig. 3. Some GFRP
Table 2
layers in the web were extended to the flanges in each beam. The angle of CFRP was
Mechanical properties of materials.
fixed to be 0° to the longitudinal direction while the angle of GFRP was 0/90/±45 or
multi-directions (continuous strand mat) to improve the mechanical behavior of Parameters Notation CFRP-0 GFRP-0/90 GFRP±45 GFRP-CSM
the hybrid FRP laminates. To investigate an appropriate beam design from the view-
points of cost and strength, three different compositions of carbon fiber with the Volume fraction Vf (%) 50 50 50 20
volume content of 52%, 33% and 14% were used in the flanges, as shown in Table Young’s modulus E11 (GPa) 113 24 11.1 10
1. The mechanical properties of carbon and E-glass fibers used in this experiment E22 (GPa) 7.4 24 11.1 10
are listed in Table 2. Shear modulus G12 (GPa) 3.2 3.5 10.909 3.8
Steel box stiffeners were installed in the web to prevent crippling and warping Poisson’s ratio m12 0.32 0.1 0.29 0.308
at the supports and local failure at the loading points. Safety rigs were installed near
the supports to prevent beams from sudden falling off in the case of lateral buck-
ling. Linear Voltage Displacement Transducers (LVDT) and laser transducers were
used to measure the deflection of the beams in mid-span section and under the
loading points. A number of strain gages were attached in flexural span, shear span was due to the crushing of fibers at the loading point followed
and near the loading points to measure the strain distributions of the beams. A by the delamination of the compressive flange and/or buckling of
high-speed camera was placed in the front of the beams to record the sudden fail- the web at the mid-span section. Delamination was observed in
ure. The test setup and instrumentation are shown schematically in Fig. 4.
all cases when the beam failed.
Unlike the stiffness, the load carrying capacity of the beam is
3. Experimental results and discussion not proportional to the volume content of CFRP in the flanges.
The ultimate load of beam A33 is the largest while the ultimate
3.1. SF beams (bf/bw = 0.43) loads of beams A52 and A14 are almost the same. An investigation
of the failure location of beam A52 suggests that the compressive
Fig. 5 shows the relationship between the load and mid-span crushing failure at the loading point is due to stress concentration.
deflection of the SF beams. It can be seen that the behavior of all In addition, in the case of beam A14, since the flange stiffness is
beams is almost linear up to the failure. The slope of load–deflec- comparatively small due to the smaller volume content of CFRP
tion curves at the mid-span section is proportional to the volume in the flange, web buckling at the mid-span section seems to occur
content of CFRP in the flanges. From Fig. 5, it is clear that an incre- before delamination failure. It was extremely difficult to visually
ment of 20% in the volume content of CFRP in the flanges can give distinguish between the delamination failure and web buckling
an increment of approximately 10% in overall beam stiffness. The at the mid-span section of the beam due to sudden failure. It is
failure modes of SF beams are shown in Figs. 6 and 7. The failure interesting to note that the load carrying capacity is not related

Flange

CF 0
GF 0/90
Continuous Strand Mat (CSM)
GF ±45

Web

Fig. 3. Layer composition of flanges and web.

 N.D. Hai et al. / Construction and Building Materials 24 (2010) 956–969 959

Hydraulic jack

Note: SG = Strain Gages Load cell

Spreader beam
Safety rig Web stiffeners Safety rig

Web stiffeners Web stiffeners

SG SG SG
Steel plate Steel plate
Support block Support block

LVDT/Laser transducers
1000 1000 1000
3000

Fig. 4. Test setup and instrumentation.

200
A33 (33% CF)
180
160 A14 (14% CF)
A52 (52% CF)
140
120
Load (kN)

100
80
A52 (Exp.)
60
A33 (Exp.)
40
A14 (Exp.)
20
0
0 10 20 30 40 50 60 70
Deflection (mm)

Fig. 5. Load–deflection of SF beams at mid-span section.

Fig. 6. Delamination of top flange and crushing in web (beam A33 and A14). Fig. 7. Crushing of fibers at loading point (beam A52).

Fig. 8 shows the relationship between load and longitudinal

with the volume content of CFRP but with the failure mode and the strain at the top and bottom of the flange at the mid-span section.
failure location. The ultimate loads of all tested beams correspond- The results indicate that both compressive and tensile strain be-
ing mid-span deflections and strains are listed in Table 3. have linearly up to the failure. The maximum compressive and ten-
960 N.D. Hai et al. / Construction and Building Materials 24 (2010) 956–969

Table 3 son, the horizontal strain in the top of the web is compressive fol-
Experimental results at failure. lowed by a tensile vertical strain. The situation at the bottom of the
Beam Load Deflection (middle Compressive strain Tensile strain web is opposite.
(kN) span) (mm) (middle span) (l) (middle span) (l) The mechanism of delamination is described in Fig. 10 based on
A52 146.9 33.50 3635 3456 visual monitoring of hybrid FRP beams during the tests. It seems
A33 194.9 56.62 6257 5981 that the stress concentration at the loading points causes local
A14 153.9 62.00 7119 7352 crushing of the fibers at those positions. In addition, the tensile
B52 239.8 30.30 3428 2334
B33 234.8 36.98 4614 3177
interlaminar stress between the interfacial layers of the top flange
and the vertical tensile stress from the web contribute to separate
fiber layers. The delamination may also be caused by the fairly low
bonding between interfacial layers. It could be initiated at the load-
sile strains are noted in beam A14 with a value of approximately
ing point and gradually develop toward the mid-span section.
7100 l, which are about 40% ultimate tensile strain and 70% ulti-
It is concluded that the load carrying capacity of the SF beam is
mate compressive strain of CFRP. It means that high strength of
governed by the stress concentration at the loading point and the
CFRP cannot be utilized effectively in the tensile flange and further
bonding strength at the interfacial layers. It is likely that the stress
development is needed in order to utilize the superior strength of
concentration increases with increases in CFRP volume content in
CFRP.
the compressive flanges.
Fig. 9 represents the vertical strains at the top and bottom of the
web at the mid-span section. The figure shows that the vertical
strain in the top of the web is tensile while the vertical strain in 3.2. WF beams (bf/bw = 1.13)
the bottom of the web is compressive. It is well known that the
horizontal and vertical strain at the same position are always in Due to the high flange to web width ratio of the WF beams, they
contrast with a positive Poisson’s ratio (ey = exmxy). For this rea- are susceptible to local flange buckling. The load–deflection curves

250
Load (kN)

A33 200 A33 (33% CF)

A14 A14 (14% CF)

A52 A52 (52% CF)
150

S3
100
A52 (S3-Exp.)
A52 (S4-Exp.)
A33 (S3-Exp.)
50 A33 (S4-Exp.)
A14 (S3-Exp.)
S4 A14 (S4-Exp.)
0
-7500 -6000 -4500 -3000 -1500 0 1500 3000 4500 6000 7500
Longitudinal Strain (µ )
Fig. 8. Load–longitudinal strain at top and bottom flanges of SF beams.

200
A33 Load (kN) A33 (33% CF)

A14 A14 (14% CF)

A52 150 A52 (52% CF)

100
P/2 P/2 A52 (S7-Exp.)
A52 (S8-Exp.)
S7 50 A33 (S7-Exp.)
A33 (S8-Exp.)
A14 (S7-Exp.)
S8
A14 (S8-Exp.)

0
-2000 -1500 -1000 -500 0 500 1000 1500 2000
Vertical Strain (µ)
Fig. 9. Load–vertical strain at top and bottom web of SF beams.
 N.D. Hai et al. / Construction and Building Materials 24 (2010) 956–969 961

P/2 P/2
Local compression failure
Direction of failure Direction of failure
Top flange

Compression stress Compression stress

Stress concentration area

Vertical tensile stress in web
Tensile interlaminar stress
Web
Fig. 10. Mechanism of delamination.

Fig. 11. Load–deflection of WF beams at mid-span section.

at the mid-span section, measured from all WF beams, are shown

in Fig. 11. The two beams, B52 and B33, varying the volume con-
tent of CFRP in the flanges showed almost linear behavior up to
200–210 kN before initial local buckling was observed in the com-
pressive flange. The differences in the strain gage data between the
top and bottom surfaces of the compressive flanges were used to
obtain the initial local buckling load. The ‘‘change of slope” method
proposed by Bank et al. [17] was modified to predict the initial
buckling load of the WF beams at which the interval change be-
tween two continuous points in the slope of the load versus the
strain difference plot exceeded the value of 10%. The initial buck-
ling loads of beams B52 and B33 obtained with this method are
210 kN and 200 kN, respectively (Fig. 11). The ultimate loads at
failure of beams B52 and B33 are almost the same at the value of
235 kN. The final failure mode of these beams was the local buck-
ling of the compressive flange leading to the separation of the
flange laminate and crushing in the web. Although the stiffnesses
of beams B52 and B33 are about 50% higher than those of beams
A52 and A33, the ultimate load carrying capacity increased only Fig. 12. Local buckling of WF beams.
39% for beam B52 and 17% for beam B33, respectively. It is there-
fore necessary to avoid buckling in order to obtain higher ultimate
load. compressive flange. It is difficult to know exactly where along
Fig. 12 shows the post-buckled deformation with two buckle the compression flange that the buckle failure initiated yet by look-
half-wavelengths in the flexural region of the beams. As can be ing at experimental photos, the failure seems to initiate at the
seen, the trough and crest of the buckle had maximum amplitude flange free edge close to the location of the crest and trough. Sim-
at the free edges and zero amplitude at the web-flange junction. By ilar observation was reported by Gilchrist [18].
a visual examination during the test, it is found that the crest and Fig. 13 represents the relationship between the load and longi-
trough of the buckle is anti-symmetric across the width of the tudinal strain at the top and bottom flange at the mid-span section
962 N.D. Hai et al. / Construction and Building Materials 24 (2010) 956–969

300

B33 B52 250 B52 (52% CF) B33 (33% CF)

200

Load (kN) S3 150

B52 (S3-Exp.)
100
B52 (S4-Exp.)

B33 (S3-Exp.)
50
S4 B33 (S4-Exp.)
0
-5000 -4000 -3000 -2000 -1000 0 1000 2000 3000 4000 5000
Longitudinal Strain (µ)
Fig. 13. Load–longitudinal strain at top and bottom flanges of WF beams.

300
B52 Load (kN)
B52 (52% CF)
250

200
B33 (33% CF)
B33 150 P/2 P/2

B52 (S7-Exp.)
100 S7
B52 (S8-Exp.)

B33 (S7-Exp.)
50 S8
B33 (S8-Exp.)

0
-2000 -1000 0 1000 2000 3000 4000 5000 6000
Vertical Strain (µ)
Fig. 14. Load–vertical strain at top and bottom web of WF beams.

of WF beams. It can be seen that the compressive strains in the top
flange behaved nonlinearly when the load was over 200–210 kN,
while the tensile strains in the bottom flange responded linearly
up to the failure. Moreover, neither the ultimate compressive
strain nor the tensile strain of the beams reaches the ultimate
strain of FRP materials.
The distribution of vertical strains at the top and bottom web at
the mid-span section of beams B52 and B33 are shown in Fig. 14.
The tensile vertical strains increased significantly in the post-buck-
led region while the increase in compressive strains in the bottom
web was insignificant (the ultimate tensile strain at the failure was
approximately 5000 l while the ultimate compressive strain at the
failure was only 1300 l). This implies that buckling in top of the
web at mid-span section is more critical than that in the bottom
of the web. Web buckling may initiate after the buckle of the com-
pressive flange.
3.3. Effect of stiffeners
Since the local buckling of WF beams results in a reduction of
the load carrying capacity, it is necessary to avoid buckling in order
to obtain higher ultimate load. In the case of beam B33, four addi-
tional stiffeners were attached to both sides of the flexural span by
adhesive bonding to prevent local buckling of the top flange. The
extra stiffeners successfully prevented local buckling in the
flexural span of the top flange, yet beam B33 failed near the
loading point due to the appearance of local buckling in the shear
span.
Similarly, eight additional stiffeners were installed in flexural
and shear span in the case of beam B52 so as to fully eliminate local
buckling. The stiffeners effectively avoided buckling, however the
beam failed because of load concentration at the loading point,
similar to the failure of SF beam A52. The load–deflection curve
of the stiffened beams is shown in Fig. 15. It can be seen that the
responses of the beams are almost linear because there was no
buckling in flexural span.
It is concluded that local buckling can be avoided through the
appropriate installation of stiffeners. However, the failure of the
WF beams sufficiently reinforced with stiffeners is not associated
with the in-plane compressive and tensile strengths of the materi-
als but instead is dominated by the out-plane concentrated stress
at the loading point. If such out-plane stress concentration can
be avoided by using an appropriate loading plate material, a higher
load carrying capacity could be achieved.
 N.D. Hai et al. / Construction and Building Materials 24 (2010) 956–969 963

300 P/2 P/2

1000 1000 1000

250
Additional stiffeners in flexural and shear span

200 B52 (52% CF)

Load (kN)
P/2 P/2
1000 1000 1000
150 B33 (33% CF)

Additional stiffeners in flexural span

100
P/2 P/2
1000 1000 1000 B52
50 B52 (Stiffened)
B33
Without additional stiffeners B33 (Stiffened)
0
0 10 20 30 40 50
Deflection (mm)

Fig. 15. Effect of stiffeners to load–defection curve.

4. Theoretical calculation where ECFRP, EGFRP-0/90, EGFRP-45 and EGFRP-CSM are Young’s moduli of
corresponding materials CFRP, GFRP-0/90, GFRP-45 and GFRP-CSM,
4.1. Prediction of beam stiffness respectively. With the transformed section, the equivalent moment
of inertia (Ieth ) for the beams can be calculated as follows:
Prediction of beam stiffness is required to evaluate the flexural Ieth ¼ If þ Iw ð4Þ
behavior of beams. The beam stiffness is determined by two fac- X m
tors: Young’s modulus (E) and moment of inertia (I). The beam If ¼ 2 Iif ð5Þ
stiffness is expressed as EI. There are two methods to calculate i¼1

the theoretical beam stiffness, the equivalent method and the lam- X
n
Iw ¼ Ijw ð6Þ
ination method. In the equivalent method, the beam stiffness EIeth is j¼1
directly computed based on the transformed section technique.
Note that If and Iw are the moment of inertia about the neutral
The transformed section is obtained by replacing layers of CFRP,
axis of the flanges and web (assume that the neutral axis passes
GFRP-45 and GFRP-CSM with an equivalent amount of GFRP-0/90
through the center of gravity of the beam section); m is total layers
(Fig. 16a). The transformation factor is determined by the ratio,
of one flange and n is total layers of a web.
n, of the elastic moduli.
 P 2
t 3if h mi¼1 t if
ECFRP Iif ¼ bif þ bif  tif ð7Þ
n1 ¼ ð1Þ 12 2
EGFRP-0=90 3
ðh  2tf Þ
EGFRP-45 Ijw ¼ bjw ð8Þ
n2 ¼ ð2Þ 12
EGFRP-0=90
where i = (1, m), j = (1, n), bf = flange width, tf = total flange thick-
EGFRP-CSM ness, tw = total web thickness, and tif = thickness of layer i in the
n3 ¼ ð3Þ
EGFRP-0=90 flange.

bf bf

GFRP-45 GFRP-CSM
CFRP GFRP-0/90
E iiflange
t1f
tf t if tf
t mf
GFRP-0/90
GFRP-45 E iiweb
GFRP-CSM
h h - 2t f
t 1w t nw tw
t jw
tw flange
E ii
tf tf

(a) Equivalent method (b) Lamination method

Fig. 16. Transformed section technique to calculate moment of inertia.
964 N.D. Hai et al. / Construction and Building Materials 24 (2010) 956–969

8
> n1  bf if layer i is CFRP where P = applied load, L = span length, EI = beam stiffness,
>
>
< n  b if layer i is GFRP-45 k = shear correction factor, G = shear modulus, A = shear area
2 f
bif ¼ (cross-sectional area of web), and kGA = shear stiffness.
> n3  bf if layer i is GFRP-CSM
>
>
:
bf if layer i is GFRP-0=90
8 5. Finite element analysis
< n2  t jw if layer j is GFRP-45
>
bjw ¼ n3  t jw if layer j is GFRP-CSM 5.1. Finite element (FE) modeling and comparisons
>
:
t jw if layer j is GFRP-0=90
Since delamination and local buckling of the compressive flange
As a result, were the main failure modes leading to premature failure of the
EIeth ¼ EGFRP-0=90  Ieth ð9Þ hybrid FRP beams, it is necessary to fully understand their mechan-
ical behavior by means of numerical analysis. This part focuses on
In the lamination method, the beam stiffness is determined EIlth clarifying (1) stiffness and deflection of beams, (2) the effect of
based on the Classical Lamination Theory (CLT) and transformed CFRP volume content on the failure strength of beams, and (3)
section technique: the effect of stress concentration. Discussions will emphasize on
EIlth ¼ Eii  Ilth ð10Þ determining the best composition of CFRP and GFRP for optimum
design of hybrid FRP beams.
A11 A22  A212 Nonlinear three-dimensional FE analysis is used to investigate
E11 ¼ ð11Þ
tA22 the structural behavior of hybrid FRP beams since it can solve pre-
A11 A22  A212 cisely with the multi-layer composition of the hybrid FRP lami-
E22 ¼ ð12Þ nates. The hybrid FRP beam is modeled using MSC.Marc code
tA11
[22]. The flanges and web of beams used 8-node composite brick
where Eii is determined based on CLT [19], E11 and E22 are equiva-
elements (Element 149) with 46 layers of CFRP, GFRP-0/90,
lent Young’s moduli of the flange or web in the longitudinal and
GFRP-45 and GFRP-CSM in each flange and 32 layers of GFRP-0/
transverse direction, respectively, Aij is the laminate extensional
90, GFRP-45 and GFRP-CSM in the web. The steel plate under the
stiffness, and t is the thickness of the flange or web.
loading points/supports and the box stiffeners were modeled with
Note that Eqs. (11) and (12) are the reduced forms of the gen-
8-node brick element (Element 7). It was assumed that full com-
eral equation since the flange and web of the beam are assumed
posite actions between the layers are present. Fig. 17 shows the fi-
to be balanced symmetric laminates.
nite element geometry and mesh with the local coordinate systems
After obtaining the Eii value, the moment of inertia Ilth is deter-
11, 22 and 33. The input data for the FE analysis is listed in Table 5.
mined based on the transformed section technique (Fig. 16b):
The accuracy of the model is critically examined by comparing re-
"  2 #
bf  t 3f h  tf ðh  2t f Þ3 sults with experiments.
Ilth ¼2 þ bf  t f þ kt w ð13Þ The load–deflection and load–strain curves obtained from the
12 2 12
experiment and FE analysis/theoretical calculation were compared
flange for the SF beams. The results from this comparison are presented in
where k ¼ Eweb
ii =Eii is the transformation factor of the web to the
flanges (the web section is transformed by an equivalent amount of Figs. 18–21. It can be seen that the deflection and flange strain
the flange section). curves obtained from the FE analysis give slightly stiffer behavior
Table 4 shows the predicted values of Young’s modulus E, shear than those obtained from the experiments for all SF beams. The
modulus G and Poisson’s ratio m in the flanges and web of the SF load–deflection curves obtained from the equivalent method show
beams based on the CLT. almost the same behavior as the experiment while those obtained
from the lamination method show somewhat lower stiffness than
4.2. Prediction of deflections those from the experiment.

Since unidirectional composite materials have a very high ratio

of longitudinal modulus to shear modulus, the deformation of the
beam due to shear will increase when this anisotropic ratio in- 33
creases. Mottram [20] emphasized that there is a potential danger
in the analysis and design of FRP beams without including shear 22
deformation. The deflection of beams is therefore computed based
on Timoshenko’s beam theory, which takes into account shear 11 Applied load
deformation. The Timoshenko equation [21] for a four-point load-
ing test with the loads applied at the third points is Ste
el
p
23PL3 PL lat
es Applied load
dmax ¼ dflexure þ dshear ¼ þ ð14Þ
1296EI 6kGA

Fib
Table 4
er
d ire
Summary of predicted values (E, G, m) of SF beams. cti
o n
Composition E11 E22 G12 m12 m21 Stiffeners
(GPa) (GPa) (GPa) (–) (–) (steel boxes)
Flange 52% CFRP (A52) 65.89 11.48 3.91 0.270 0.047
33% CFRP (A33) 48.25 14.14 4.41 0.249 0.073
14% CFRP (A14) 31.04 16.66 4.89 0.234 0.126
Web All beams 15.72 15.72 6.68 0.328 0.328
Fig. 17. Finite element geometry and mesh.
 N.D. Hai et al. / Construction and Building Materials 24 (2010) 956–969 965

Table 5
Input data for finite element analysis.

Parameter Units Notation CFRP GFRP-0/90 GFRP-CSM GFRP-45

Young’s modulus MPa E11 113,000 24,000 10,000 11,100
E22 = E33 7400 24,000 10,000 11,100
Poisson’s ratio – m12 0.32 0.1 0.308 0.29
m23 = m31 0.05 0.1 0.308 0.29
Shear modulus MPa G12 = G23 = G31 3200 3500 3800 10,909
Max. tensile stress, X MPa Xt 2100 230 70 165
Max. tensile stress, Y, Z MPa Yt = Zt 60 210 70 150
Max. compressive stress, X MPa Xc 1200 210 137 150
Max. compressive stress, Y, Z MPa Yc = Zc 170 230 137 165
Max. shear stress, XY, YZ, ZX MPa S12 = S23 = S13 70 50 55 65
Interactive strength, XY (1/MPa)2 F12 3.1E06 1E05 5.2E05 2E05
Interactive strength, YZ (1/MPa)2 F23 4.9E05 1E05 5.2E05 2E05
Interactive strength, XZ (1/MPa)2 F13 3.1E06 1E05 5.2E05 2E05

200
180
160 A52 (52% CF)
140
120
Load (kN)

100
80
A52 (Experiment)
60
A52 (Theory-Equivalent Method)
40 A52 (Theory-Lamination Method)
20 A52 (Finite Element Analysis)
0
0 5 10 15 20 25 30 35 40
Deflection (mm)

Fig. 18. Load–deflection curve (beam A52).

200
A33 (33% CF)
180
160
140
120
Load (kN)

100
80
A33 (Experiment)
60 A33 (Theory-Equivalent Method)
40 A33 (Theory-Lamination Method)
20 A33 (Finite Element Analysis)

0
0 10 20 30 40 50 60 70
Deflection (mm)

Fig. 19. Load–deflection curve (beam A33).

5.2. Effect of carbon volume content on the failure strength of beams
To determine the optimum CFRP volume content in the flanges,
a parametric FE study of the SF beams varying the CFRP volume
content from 5% to 60% was conducted. A total of seven beams
named A5, A14, A25, A33, A45, A52 and A60 (each number indicat-
ing the percentage of CFRP in the flanges) were analyzed. The
stacking sequence and ply orientation in the flanges and web are
given in Fig. 3.
The relationship between the failure load and CFRP volume con-
tent is shown in Fig. 22. It can be seen that the maximum load in-
creases with increases in CFRP volume content from 5% to 25%. The
966 N.D. Hai et al. / Construction and Building Materials 24 (2010) 956–969

200
180
160 A14 (14% CF)

140
120
Load (kN) 100
80
A14 (Experiment)
60
A14 (Theory-Equivalent Method)
40 A14 (Theory-Lamination Method)
20 A14 (Finite Element Analysis)
0
0 10 20 30 40 50 60 70
Deflection (mm)

Fig. 20. Load–deflection curve (beam A14).

250
Load (kN)

A33 200 A33 (33% CF)

A14 A14 (14% CF)

A52 A52 (52% CF)
150

S3
100
A52 (S3-Exp.) A52 (S4-Exp.)
A33 (S3-Exp.) A33 (S4-Exp.)
50 A14 (S3-Exp.) A14 (S4-Exp.)
A52 (S3-FEM) A52 (S4-FEM)
A33 (S3-FEM) A33 (S4-FEM)
S4 A14 (S3-FEM) A14 (S4-FEM)
0
-7500 -6000 -4500 -3000 -1500 0 1500 3000 4500 6000 7500
Longitudinal Strain (µ)
Fig. 21. Load–longitudinal strain at top and bottom flanges of SF beams.

Fig. 22. Effect of CFRP volume content in flanges.

 N.D. Hai et al. / Construction and Building Materials 24 (2010) 956–969 967

failure mode of beams under this range is governed by the crushing elastic theory). This explains the premature failure in the web of
of fiber in the web at the mid-span section. The failure load and the beams with low CFRP volume content. The results of FE analy-
failure mode of these beams were predicted by the Tsai-Wu failure sis showed that beams A5, A14 and A25 failed at almost similar
criterion [21], given by failure stress of approximately 132 MPa (Table 6).
      In contrast with beams A5, A14 and A25, the failure load of
1 1 1 1 1 1 r21 beams A33, A45, A52 and A60 decreased when the CFRP volume
F¼  r1 þ  r2 þ  r3 þ
Xt Xc Yt Yc Zt Zc Xt Xc content increased from 33% to 60% (Table 6). Since the maximum
r22 r23 s212 s223 s213 Tsai-Wu failure indices of these beams in the web region are much
þ þ þ þ þ þ 2F 12 r1 r2 þ 2F 23 r2 r3 less than 1, the failure mode of these beams was predicted to be
YtYc Zt Zc S212 S223 S213
governed by the delamination of the top flange between the inter-
þ 2F 13 r1 r3 P 1 ð15Þ
facial layers. The delamination was caused by the concentration of
the tensile interlaminar normal stress at the top layers of the top
where Xt, Xc, Yt, Yc, Zt, Zc, S12, S23, S13, F12, F23 and F13 are as listed in
flange. Fig. 23 shows the relationship between the interlaminar
Table 5, r1, r2, r3 are principal stresses, and s12, s23, s13 are shear
normal stress and the flange thickness of all layers at the mid-span
stresses.
section of beams A14, A33 and A52. The figure shows that the total
The failure of these beams was assumed to occur when the fail-
interlaminar normal stresses (compressive or tensile) in the top
ure index F satisfies Eq. (15). Indeed, the maximum Tsai-Wu failure
flange of beams are proportional to the CFRP volume content.
indices obtained from the FE analysis of beams A5, A14 and A25
The higher volume content of CFRP tends to have larger total ten-
reached the value of 1 in the region of web-flange junction, as pre-
sile stresses at the interfacial layers. The total tensile interlaminar
sented in Table 6. Since the geometry of the investigated I-beams is
normal stress would cause separation of the layers, while total
symmetric, the neutral axis always passes through the center of
compressive interlaminar normal stress would tend to keep the
gravity of the cross section. It is a fact that a smaller volume con-
layers together. Therefore, the delamination of beams with higher
tent of CFRP in the flanges will result in a smaller equivalent flange
volume content of CFRP seems to be caused by the large total ten-
modulus. Thus, with the same web modulus and under equal ap-
sile normal stresses (this supports the above conclusion for the
plied load, the beam with the smallest CFRP volume content in
failure of beams A33 to A60).
the flanges will cause the highest stress in the web (based on the

Table 6
Prediction of failure load using finite 1 element analysis.
P
Beam CFRP (%) Failure load Principal stress, (r1) Tsai-Wu Failure Index rza (N/mm2) Failure mode
(in top web at mid-span)
Exp. FEM Exp. (N/mm2) FEM (N/mm2) FEM FEM FEM Exp.
A60 60 – 124.8 – 51.79 0.20 105.64 F-Delb
A52 52 147 144 60.04 65.5 0.30 108.77 F-Delb F-Delb
A45 45 – 153.6 – 76.39 0.39 104.87 F-Delb
A33 33 194 192 117.42 113.95 0.78 108.18 F-Delb F-Delb
W-Cruc
A25 25 – 192 – 132.16 0.99 88.6 W-Cruc
A14 14 154 153.6 130.08 132.28 1.01 45.84 W-Cruc W-Cruc
F-Delc
A5 5 – 124.8 – 135.59 1.07 22.53 W-Cruc
a P
rz: summation of interlaminar tensile normal stress in top layers of top flange at mid-span section.
b
F-Del: flange-Delamination.
c
W-Cru: web-crushing.

-60 -50 -40 -30 -20 -10 0 10 20 30

0
0 Upper flange
Flange thickness

CF 0
2
Flange Thickness (mm)

GF 0/90
CSM
4
GF ±45
14 Lower flange 6

A14 A33 A52 8

10

12

14
Interlaminar Normal Stress (N/mm 2 )

Fig. 23. Distribution of interlaminar normal stress in each layer of the compressive flange.
968 N.D. Hai et al. / Construction and Building Materials 24 (2010) 956–969
200
180
160
140
120
Load (kN)
100
80
60
40
20
0
0 20
Total Tensile Interlaminar Normal Stress (N/mm2)
Fig. 24. Load-total tensile interlaminar normal stress at mid-span and loading point.
It was assumed that all of these beams failed when the summa-
tion of the interlaminar tensile normal stress in the top layers of
the top flange at the mid-span section reached the value of
105 MPa (this value was determined experimentally using the
measured data of the interlaminar strain) and the analytical results
could well predict the final failure load of beams (Table 6). It is
interesting to note that beams A25 and A33 have almost the same
failure load although there is an 8% difference in CFRP volume con-
tent between these two beams. The failure mode of beam A25 is
dominated by the crushing of fibers in the web at the mid-span
section while beam A33 failed due to the large tensile interlaminar
normal stress leading to the delamination of the top flange. When
the CFRP volume content increases from 25% to 33%, the failure
load will not increase but the failure mode may change from
web crushing to delamination. This suggests that the optimum
CFRP volume content is in the range of 25–33% and this range is
recommended for designing hybrid FRP beams for bridge
applications.
Comparisons of failure load and stresses between the FE analy-
sis and experiments are shown in Table 6. As can be seen, the aver-
age differences in failure load between the experiments and FE
analysis is less than 2%. When comparing the principal stress in
the web and total interlaminar tensile normal stress in the com-
pressive flange of hybrid FRP beams at failure, these differences
are smaller than 4%. Thus, it is concluded that the results of FE
analysis are important and reliable to predict the strength of hy-
brid FRP beams.
5.3. Effect of stress concentration
Since stress concentration at the loading point causes the pre-
mature failure of the hybrid FRP beams, clarification by FE analysis
is needed. Fig. 24 shows the relationship between the load and to-
tal tensile interlaminar normal stress of beams A33 to A60 at the
mid-span section and loading points. It can be seen that the total
tensile interlaminar normal stress at the loading point of all beams
is approximately 32% higher than that at the mid-span section.
This indicates that the beams with higher volume content of CFRP
will have larger total tensile interlaminar stresses concentrated at
the loading point. A similar phenomenon for the in-plane stress
concentration is observed for all beams. However, the difference
in the in-plane stress between the mid-span section and loading
point in this case is only 12% (experimental data show about 10%
difference in cases of beams A33 and A52). It seems that the final
Δσz 32% A33
A45
A52
A60
A33 (MS)
A33 (LP)
A45 (MS)
A45 (LP)
MS: Mid-Span A52 (MS)
LP: Loading Point A52 (LP)
A60 (MS)
A60 (LP)
40 60 80 100 120 140 160 180
failure load of beams A33 to A60 is governed by the concentration
of tensile interlaminar stress at the loading point because of the
high load concentration factor compared to the case of in-plane
stress.
6. Conclusions
In this paper, experiments and analyses of the structural behav-
ior of an innovative hybrid FRP I-beams is described. The following
main conclusions can be drawn:
(1) The ratio of flange to web width (bf/bw) of hybrid FRP I-
shaped beams plays an important role in their structural
behavior. SF beams (bf/bw = 0.43) showed stable and linear
behavior under bending moment and failed in a brittle man-
ner by delamination of the compressive flange at the interfa-
cial layers while WF beams (bf/bw = 1.13) exhibited unstable
and nonlinear behavior in the buckling and post-buckling
region leading to delamination failure of the compressive
flange.
(2) The failure strength and failure mode of hybrid FRP beams
are dependent on the carbon volume content in the flanges.
It seems that when the higher carbon volume content is uti-
lized, the extent of delamination is greater resulting in the
lower failure load of beams. The optimum carbon volume
content in the flanges for the best hybridization of FRP
beams was experimentally and numerically determined to
be 25–33%.
(3) Good agreement is obtained in this study between the theo-
retical calculations, FE analysis and experiments. Classical
lamination theory (CLT) and Timoshenko beam theory can
effectively be used to compute the deflections and stiffness
of hybrid beams. The FE analysis is fully capable of accu-
rately predicting and designing the strength of hybrid FRP
beams.
(4) The results of this study show the potential of applying
hybrid FRP beams for bridge components. Although delam-
ination and local buckling in the compressive flange are
observed failure modes of hybrid beams subjected to flex-
ural loading, it can be avoided when the hybrid FRP beam
is connected with a deck system to form a complete
bridge. Hence further development in composite action
between hybrid FRP beam and deck is needed to apply
 N.D. Hai et al. / Construction and Building Materials 24 (2010) 956–969
the research of hybrid FRP composite for real bridge
application.
7. Further developments
Since delamination and local buckling of the compressive flange
was the main failure mode of the hybrid FRP beams subjected to
the bending moment, it is necessary to prevent such failures in or-
der to utilize the superior characteristics of FRP materials. The idea
is to set a neutral axis of the hybrid FRP beam section towards the
top to reduce the compressive stress in the top flange. An effective
way to do this is providing a stiffer material on the top flange of the
hybrid FRP beams that can form a part of the bridge deck. We have
conducted a number of trial tests using hybrid FRP beam and or-
dinary concrete deck to investigate theirs composite behavior.
Stainless steel bolts and epoxy resin were used for the shear con-
nection between hybrid FRP beam and the concrete deck. The re-
sults showed a significant increase in stiffness and load carrying
capacity of the hybrid FRP-concrete beam compared to those of hy-
brid FRP beam alone. In addition, a significant decrease in compres-
sive strain in the top flange of the hybrid FRP-concrete beam is
observed, thereby preventing delamination. However the ordinary
concrete deck has some disadvantages such as its low compres-
sive/crack strength, low durability and heavy weight which may
result in reduction of structures’ service life. Ultra high strength Fi-
ber reinforced Concrete (UFC) is found to be a better selection for
the combined use with the hybrid FRP beam since it has superior
strength and durability to those of ordinary concrete. UFC deck is
therefore developed to overcome the weaknesses of ordinary con-
crete. Currently more suitable shear connection is being investi-
gated to form a composite action between the UFC deck and
hybrid FRP beam and the details will be published in the near fu-
ture. We believe that with such further development, the system
of hybrid FRP beam and UFC deck can be applied for the construc-
tion of pedestrian/highway bridges in Japan.
Acknowledgements
This research was funded by a Ministry of Land, Infrastructure
and Transport of Japan (MLIT) Grant-in-Aid for scientific research
of construction technology, which is gratefully acknowledged.
969
References
[1] Machida A. Recommendation for design and construction of concrete
structures using continuous fiber reinforcing materials. Concrete Engineering
Series, 23. Tokyo: Japan Society of Civil Engineers; 1997.
[2] Bakis CE, Bank LC, Brown VL, Cosenza E, Davalos JF, Lesko JJ, et al. Fiber-
reinforced polymer composites for construction. ASCE J Compos Construct
1997;6(2):73–87 [150th Anniversary Paper].
[3] Keller T. Use of fiber reinforced polymers in bridge construction. Int Assoc
Bridge Struct Eng 2003;34:76 [Zurich].
[4] Nayomon Uno, Nobuhiko Kitayama. Design, fabrication and erection of the
pedestrian bridge in the road-park of Ikei-Tairagawa in Okinawa. IHI Eng Rev
2003;36(2):35–9.
[5] Shan Ying, Liao Kin. Environmental fatigue behavior and life prediction of
unidirectional glass–carbon/epoxy hybrid composites. Int J Fatigue
2002;24:847–59.
[6] Dickson RF, Fernando G, Adam T, Reiter H, Harris B. Fatigue behavior of hybrid
composites. Part 2: carbon–glass hybrids. J Mater Sci 1989;24:227–33.
[7] Summerscales J, Short D. Carbon fiber and glass fiber hybrid reinforced
plastics. Composites 1978;9:157–66.
[8] Hayashi T, Koyama K, Yamazaki A, Kihira M. Development of new material
properties by hybrid composition. Compos Mater 1972.
[9] Zweben Carl. Tensile strength of hybrid composites. J Mater Sci
1977;12:1325–37.
[10] Chou TW, Kelly A. Mechanical properties of composites. Annu Rev Mater Sci
1980;10:229–59.
[11] Manders PW, Bader MG. The strength of hybrid glass/carbon fiber composites.
Part 1: failure strain enhancement and failure mode. J Mater Sci
1981;16:2233–45.
[12] Manders PW, Bader MG. The strength of hybrid glass/carbon fiber composites.
Part 2: a statistical model. J Mater Sci 1981;16:2246–56.
[13] Fukuda Hiroshi. An advanced theory of the strength of hybrid composites. J
Mater Sci 1983;19:974–82.
[14] Khatri Subhash C, Koczak Michael J. Thick-section AS4-graphite/E-glass/PPS
hybrid composites. Part I: tensile behavior. Compos Sci Technol
1996;56:181–92.
[15] Khatri Subhash C, Koczak Michael J. Thick-section AS4-graphite/E-glass/PPS
hybrid composites. Part II: flexural response. Compos Sci Technol
1996;56:473–82.
[16] Piggott MR, Harris B. Compression strength of hybrid fiber-reinforced plastics.
J Mater Sci 1981;16:687–93.
[17] Bank LC, Yin J, Nadipelli M. Local buckling of pultruded beams: nonlinearity,
anisotropy and inhomogeneity. J Construct Build Mater 1995;9(6):325–31.
[18] Gilchrist MD, Kinloch AJ, Matthews FL, Osiyemi SO. Mechanical performance of
carbon-fiber and glass-fiber-reinforced epoxy I-beams: I. Mech behav, Compos
Sci Technol 1996;56:37–53.
[19] Gibson Ronald F. Principles of composite material mechanics. McGraw-Hill,
Inc.; 1994.
[20] Mottram JT. Lateral-torsional buckling of a pultruded I-beam. J Compos
1992;23(2):81–92.
[21] Hayes Michael D, Lesko John J. Measurement of the Timoshenko shear
stiffness. Part I: effect of warping. J Compos Construct – ASCE
2007;11(3):336–42.
[22] MARC/Mentat User’s Guide.