See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/274392178

Computational Modeling of Aramid Fiber-Reinforced Polymer Prestressed

Girder in Composite Action with Bridge Deck

Article in ACI Structural Journal · November 2013

CITATIONS READS
4 1,555

3 authors:

Shobeir Pirayeh Gar Monique Head

Halliburton Company University of Delaware
11 PUBLICATIONS 54 CITATIONS 80 PUBLICATIONS 364 CITATIONS

SEE PROFILE SEE PROFILE

Stefan Hurlebaus
Texas A&M University
175 PUBLICATIONS 4,714 CITATIONS

SEE PROFILE

All content following this page was uploaded by Monique Head on 06 October 2016.

The user has requested enhancement of the downloaded file.

 ACI STRUCTURAL JOURNAL
Title no. 110-S78
Computational Modeling of Aramid Fiber-Reinforced Polymer
Prestressed Girder in Composite Action with Bridge Deck
by Shobeir Pirayeh Gar, Monique Hite Head, and Stefan Hurlebaus
Aramid fiber-reinforced polymer (AFRP) tendons, which are inher-
ently corrosion-resistant, can be used to replace steel prestressing
strands in bridge girders to enhance bridge sustainability. Despite
ongoing experimental research, there is a lack of uniformity and
consistency in testing procedures, definitions of material charac-
teristics, and results. Therefore, a robust computational model is
needed to perform a refined nonlinear analysis of full-scale AFRP
prestressed girders. This paper presents the development of a
computational model to numerically evaluate the flexural behavior
of an AASHTO I-girder (Type I) prestressed with AFRP tendons
in comparison to its conventional prestressing steel counterpart.
Numerical results match experimental test data with a high degree
of accuracy and reveal that an AASHTO I-girder prestressed with
AFRP meets service and strength limit states. Numerical results
also show that the deflection equation in ACI 440.4R overestimates
the maximum deflection of the AFRP prestressed girder.
Keywords: aramid fiber-reinforced polymer; composite action; compu-
tational modeling; concrete bridge girder; flexural performance;
prestressing; sustainability.
INTRODUCTION
In the 1970s, corrosion-induced deterioration of concrete
structures, particularly in bridge decks, led to a need to find
alternative design strategies that would reduce the likeli-
hood of corrosion decay in concrete structures. One alterna-
tive was to replace prestressed steel strands with prestressed
fiber-reinforced polymer (FRP) bars. Because FRPs have
high strength and are corrosion-resistant, nonmagnetic, and
lightweight, their application in construction, retrofitting,
and rehabilitation of structures has grown considerably.1 FRP
tendons are typically made from one of these three basic
fibers: glass FRP (GFRP), carbon FRP (CFRP), and aramid
FRP (AFRP). The latter is the subject of this research.
In spite of superior durability, the modulus of elasticity
of AFRP is approximately three times lower than that of
steel, leading to a substantial reduction in flexural stiffness
of AFRP prestressed girder after cracking, and resulting
in larger deflection, accordingly.2 Controlling the deflec-
tion under service loads becomes critical to meet service-
ability requirements and enhance sustainability.3 Despite
ongoing experimental research investigating the behavior of
FRP prestressed girders, there is a lack of uniformity and
consistency in testing procedures, definitions of material
characteristics, and results that raises the need for a compu-
tational model to analyze the behavior of FRP prestressed
girders.4 Kim5 investigated the flexural behavior of AFRP
prestressed rectangular beams via numerical analyses.
Different sectional properties and levels of prestressing
were studied. It was concluded that the prestressing level
typically governs the flexural performance at the service
state; however, the Ig/Icr (gross to cracked moment of inertia)
ratio controls the deflection characteristics of the AFRP
prestressed members until failure.
ACI Structural Journal/November-December 2013
TECHNICAL PAPER
Although valuable results have been gained from previous
investigations, studies have been mostly limited to rectan-
gular sections or reduced-scale beams, where the effect of
bridge deck in providing a composite section is typically
ignored.6-8 New information has been learned based on a
computational model recently developed by the authors to
gain insight into the flexural performance for the design of
full-scale girders in composite with bridge deck. Studying
the flexural response of the composite girder in different
steps of loading, including prestressing load, full dead load
of the girder and cast-in-place deck; live load when the deck
is hardened (beginning the composite action); and failure
load is required to judge if an AFRP prestressed girder meets
the service and strength limit states.
The developed computational model is capable of
performing the following: 1) nonlinear fiber element analysis
of a prestressed girder’s section with pretensioned tendons to
find the stress and strain distributions; 2) moment-curvature
analysis taking into account the effect of composite action
between the girder and slab once the concrete of the slab is
hardened; 3) refined analysis to capture cracking and failure;
4) long-term loss estimation analysis, including creep and
shrinkage of the concrete as well as relaxation of the tendons;
and 5) beam nonlinear analysis to find the load-deflection
relationship of every point along the beam until failure. In
this paper, an AASHTO I-girder (Type I) composite with a
bridge deck is designed as a fully prestressed section based
on serviceability requirements and strength demand for two
different cases: 1) pretensioned AFRP; and 2) pretensioned
steel. Then, numerical analysis is performed to determine the
stress distribution over the cross section and moment-curva-
ture and load-deflection relationships of the girder, which
will be verified by existing experimental test data. In the
final analysis, the deflection equation in ACI 440.4R9 will be
evaluated by numerical results.
RESEARCH SIGNIFICANCE
New information has been gained from experimental
investigations on FRP prestressed beams; however, there
is still a need for a computational model capable of accu-
rately evaluating the flexural behavior of full-scale FRP
prestressed girders in composite action with bridge deck.
This need is more pronounced due to a lack of uniformity in
testing procedures and high cost of experimental tests. This
paper presents the results from an extensive computational
model developed to gain insight into the flexural behavior
ACI Structural Journal, V. 110, No. 6, November-December 2013.
MS No. S-2011-358.R1 received May 24, 2012, and reviewed under Institute
publication policies. Copyright © 2013, American Concrete Institute. All rights
reserved, including the making of copies unless permission is obtained from the
copyright proprietors. Pertinent discussion including author’s closure, if any, will be
published in the September-October 2014 ACI Structural Journal if the discussion is
received by May 1, 2014.
965
Shobeir Pirayeh Gar, PhD, is a Structural Engineer at Houston Offshore Engi- of an AFRP prestressed girder compared to its conventional
neering, Houston, TX. He is a former Graduate Assistant Researcher in the Zachry steel counterpart, and to verify its acceptability based on
Department of Civil Engineering at Texas A&M University, College Station, TX. His service and strength limit states.
research interests include seismic design of structures, performance-based design,
bridge engineering, prestressed concrete, steel connections, and structural dynamics.
BRIDGE PROTOTYPE AND DESIGN
Monique Hite Head is an Assistant Professor in the Department of Civil Engineering OF PRESTRESSED GIRDER
at Morgan State University, Baltimore, MD. She is a member of Joint ACI-ASCE A bridge prototype including prestressed precast AASHTO
Committee 343, Concrete Bridge Design. Her research interests include performance- I-girders (Type I) and cast-in-place (CIP) deck is shown in
based structural design, bridge and earthquake engineering, seismic retrofitting of
bridges, structural dynamics, and experimental testing on bridge structures.
Fig. 1. Simply supported girders are spaced 6 ft (1.8 m) on
center and span 40 ft (12.2 m). The girder is designed based on
Stefan Hurlebaus is an Associate Professor in the Zachry Department of Civil Engi- AASHTO LRFD Bridge Design Specifications10 for service
neering at Texas A&M University. His research interests include smart structures, and strength limit states. The modulus of elasticity of AFRP
structural health monitoring, nondestructive testing, laser ultrasonics, active vibra-
tion control, semi-active damping, active vibration isolation, wave propagation in
is considerably less than that of steel, so once the section is
elastic solids, and vibrations. cracked, there is a substantial decline in the flexural stiffness
of the beam, leading to a larger deflection. Therefore, for
the sake of controlling the deflection of the beam, the girder
should remain uncracked under service load, and hence the
girder is designed as a fully prestressed beam. Steel strands
and AFRP tendons are assumed to be initially prestressed
up to 70% and 60% of their ultimate strength, which corre-
sponds to 189 and 121.8 ksi (1304.1 and 840.4 MPa), respec-
tively. Material properties are summarized in Table 1. The
stress-strain relationships are presented in Appendix A.
The maximum moment due to service load is 620 kft
(841 kNm) including 215 kft (292 kNm) due to the total dead
load and 405 kft (549 kNm) induced by live load (Appendix B).
For the prototype bridge, calculations show that in the case of
conventional prestressing steel, 12 seven-wire steel strands
with 0.5 in. (12.7 mm) diameter are required: 10 in the
bottom flange and two in the top flange. In this case, the area
of tensile reinforcement is As = 1.96 in.2 (1264.5 mm2). For
the AFRP case, 24 tendons are required with 0.4 in. (10 mm)
diameter: 22 in the bottom flange and two in the top flange.
In this case, the area of tensile reinforcement is Ap = 2.76 in.2
(1780.6 mm2). Similar to the balance ratio rbs commonly
used for steel prestressed beams to ensure ductile behavior, a
brittle ratio rbp is defined for FRP prestressed beams. In this
case, r ≤ rbp (under-reinforced) indicates failure by tendon
rupture, while r ≥ rbp (over-reinforced) represents concrete
compression failure.
Although concrete compression failure provides more
deformability over the tendon rupture, it is difficult to over-
reinforce most of the common prestressed shapes because
fitting in too many tendons is impractical.4 A more signifi-
cant reason for this difficulty, in the authors’ opinion, is
the effect of the topping deck that considerably raises the
Fig. 1—Bridge prototype: (a) geometry; and (b) design balance or brittle ratio. As discussed in Appendix C, for
vehicle load cases. the AFRP prestressed girder without topping deck, the area
of reinforcement corresponding to the brittle ratio is Abp =
1.21 in.2 (780.6 mm2); however, taking into account the
effect of composite action gives Abp = 8.75 in.2 (5645.2 mm2).
Table 1—Properties of material, ksi (MPa)
It is clearly seen that if the composite section was to be
Compressive strength fc′ 6 (41.4) over-reinforced, approximately 70 AFRP tendons (22 ×
Girder
Fracture modulus fcr 0.58 (4) 8.75/2.76) would have to be used, which is not feasible.
Thus, the failure mode is dominated by tendon rupture. The
Compressive strength fc′ 4 (27.6)
Slab prestressing layout of both girders prestressed with AFRP
Fracture modulus fcr 0.47 (3.24) tendons and steel strands is illustrated in Fig. 2.
Ultimate strength fpu 203 (1400)
AFRP GENERAL MOMENT-CURVATURE RELATIONSHIP
Modulus of elasticity Ep 8696 (60,000)
Figures 3(a) and (b) show the general moment-curvature
Ultimate strength fsu 270 (1863) relationships for conventional steel and AFRP prestressed
Steel strand Yield strength fy 243 (1677) AASHTO I-girder composite with bridge deck, respec-
tively. After prestressing, the dead loads of girder and slab
Modulus of elasticity Es 28,000 (193,200)
tend to reduce and counterbalance the initial camber induced

966 ACI Structural Journal/November-December 2013

by prestressing, thereby minimizing the effects of creep on
long-term deflection. Once the concrete of the slab is hard-
ened, the girder exhibits larger flexural stiffness to bear the
live load as a result of the composite action between the
girder and the slab. This can be seen by an increase in slope
of the moment-curvature diagram before cracking.
With increasing the load, the section cracks and decom-
pression occurs in the bottom fibers of the section. From a
serviceability point of view, before prestressed steel exhibits
any inelastic stresses, the girder is in the serviceability region
as indicated by (f0, M0) in Fig. 3(a). A further increase in
load causes the prestressed steel to reach the limit state of Fig. 2—Prestressing layout of AASHTO I girder (Type I):
yielding. This stage of loading is referred to as the post- (a) pretensioned steel; and (b) pretensioned AFRP.
serviceability stage. Finally, the girder fails due to concrete
crushing at the top fibers of the section (r ≤ rbs). However,
in the case of prestressed AFRP, because tendons behave
linearly up to rupture and the reinforcement ratio is typically
less than the brittle ratio (r ≤ rbp), tendon rupture occurs
before the concrete can induce any considerable inelastic
stresses; hence, the AFRP prestressed girder exhibits low
deformability (Fig. 3(b)). This issue will be further discussed
in the subsequent sections.

FULL COMPUTATIONAL MODEL FOR

NUMERICAL ANALYSIS
Computational model development
An extensive computational model to perform refined
nonlinear analyses is developed using the fiber element
method to study the following: 1) stress distribution over the
height of the composite section; 2) moment-curvature rela-
tionship; and 3) load-deflection response of the girder up to
failure. This model starts the analysis by finding the moment
corresponding to the given curvature of the composite
section. In general, the load-deformation relationship for a
structural member can be written as

 P   EA EZ   e 0  (1)
 M  =  EZ EI   f 
    

where P is the applied axial load; M is the applied bending

moment; EA is the axial stiffness of the member; EI is the
flexural stiffness; EZ is the nonlinear coupling between
axial load and bending moment; e0 is the strain at a refer-
ence point, and f is the curvature of the section. The strain
at any fiber element of the section can be found using the
following equation, given the plane section remains plane
after bending

e( y) = e 0 + fy (2)

where y is the distance of the element from the reference Fig. 3—General moment-curvature diagram: (a) steel
point. e0 and f are the first guesses and the strain profile is prestressed girder; and (b) AFRP prestressed girder.
determined using Eq. (2). Then, the stress at each element
is computed using the stress-strain relationship for material.
Popovic’s11 equation and Menegotto and Pinto’s12 formula  ∂P ∂P 
are employed to define the nonlinear behavior of the concrete 
 dP   ∂e 0 ∂f   d e 0  de 
and prestressing steel, respectively (Appendix A). A linear
 dM  =  ∂M

  = [ D]  0  (3)
stress-strain relationship is assumed for AFRP material.   
∂M  d f   df 
Given the stress at each element is found out, P and M can be  
 ∂e 0 ∂f 
computed, subsequently. The equilibrium equations (SP = 0,
SM = 0) are used as a key to find the actual strain profile and
curvature. This can be done by an iterative analysis using a Equation (3) should be solved incrementally, which is
differential form of Eq. (1) beneficial in a highly nonlinear state of stress. dP and dM

ACI Structural Journal/November-December 2013 967

 When the moment-curvature relationship of the composite
section is numerically established, the deflection of the beam
under any load configuration can be evaluated using the
conjugate beam theory. The fibers over the cross section of
the girder are refined to the extent that cracking and failure
can be captured with the least error. The accuracy of the
program is validated by comparing the numerical results
for the steel prestressed girder and its corresponding experi-
mental test data reported by Trejo et al.13 This verification
shows that the maximum error induced by the computational
model at cracking and failure loads are 7% and 5%, respec-
tively, which indicates the high accuracy and acceptability
of the numerical analysis (Fig. 4(a)). A separate AFRP
prestressed concrete section was also experimentally tested
and compared with computational model to provide more
validation (Appendix D).

Moment-curvature analysis of section

As shown in Fig. 4(b), the cracking moment, 700 kft
(949 kNm), is higher than the moment due to the service
load, 620 kft (841 kNm), confirming that the section is
fully prestressed. It is also observed that both prestressed
AFRP and steel girders have an ultimate strength of
approximately 1200 kft (1627 kNm), which is almost
1.2 times the required strength per AASHTO LRFD
Bridge Design Specifications.10 For prestressed steel, flex-
ural failure is due to concrete crushing at top fibers of the
section, where the compressive strain reaches to 0.003. As
pointed out earlier, after cracking, decompression occurs in
bottom fibers of the section, and subsequently prestressed
steel yields due to increase in load. In the post-serviceability
region, because the ultimate strain of prestressed steel is
considerable—0.05—concrete can exhibit inelastic stresses
at top fibers of the section and finally reach its ultimate
Fig. 4—Moment-curvature diagram: (a) comparison of compressive strain before rupture of the steel strands; hence,
analytical and experimental results for prestressed steel significant flexural ductility is provided prior to failure.
girder; and (b) comparison of analytical results for steel and The numerical analysis shows the failure curvature equal
AFRP prestressed girder. to 6 × 10–4 1/in. (2.36 × 10–5 1/mm), which is 40 times
the cracking curvature. In contrast, the AFRP prestressed
girder fails in a more brittle fashion, caused by rupture of
Table 2—Results of moment-curvature analysis the tendons farthest from the neutral axis. Because the ulti-
Steel AFRP
mate strain of the AFRP—0.023—is much lower than that
of steel, rupture of the tendons occurs before the concrete
Mcr × 10–3, kft (kNm) 0.71 (0.96) 0.73 (0.99)
Cracking can provide any considerable inelastic stress, and results in
fcr × 105, 1/in. (1/mm) 1.5 (0.059) 1.58 (0.062) less deformability compared to the prestressing steel case.
M0 × 10–3, kft (kNm) 1.1 (1.49) — Numerical analysis confirms this, and shows that the failure
Yielding curvature of the AFRP prestressed girder, 2.8 × 10–4 1/in.
f0 × 10 , 1/in. (1/mm)
5
21.6 (0.85) — (1.1 × 10–5 1/mm), is almost half of that of steel; however, it
–3
Mn × 10 , kft (kNm) 1.16 (1.57) 1.26 (1.71) can provide sufficient warning before failure, as it is 18 times
Failure the cracking curvature.
fn × 10 , 1/in. (1/mm)
5
60 (2.36) 28 (1.1)
Based on experimental test data, the prestress loss at the
time of testing the steel prestressed girder was reported as
5%, most of which was due to elastic shortening.12 While
are indeed the errors found when the equilibrium equation additional experimental tests to verify the prestress loss are
is checked. In each increment, given the constant curvature, underway, loss calculations in the computational model were
a tiny change in strain is applied and the partial derivatives adjusted to result in the same loss in the prestressing force
∂P/∂e0 and and ∂M/∂e0 are computed. The same procedure, as reported from the experiment for consistency. According
however this time for the constant strain, is followed to find to what was shown in Fig. 3, results of moment-curvature
∂P/∂f and ∂M/∂f. Once the matrix D is established, the analysis are summarized in Table 2.
strain and curvature adjustments de0 and df can be computed
and added to the strain and curvature of the previous step to Stress analysis over cross section
decline the error and enhance the accuracy of the response. The computational model produces stress distribu-
This iteration should be done as many times, as the error in tions over the entire cross section from the first stage of
the equilibrium equation is found to be acceptably small. prestressing up to failure of the girder to gain a clear insight

968 ACI Structural Journal/November-December 2013

 Fig. 5—Stress distribution in different steps of loading: (a) steel prestressed concrete
girder; and (b) AFRP prestressed concrete girder.
about flexural behavior of prestressed girders in composite
action with the topping slab. As illustrated in Fig. 5, four
steps of analysis were selected to study the stress distribu-
tion as follows: prestressing the precast girder, applying the
dead load of the girder and slab, applying live load up to
cracking, and increasing live load up to failure. It is seen that
from the third step, where the concrete of the slab is hard-
ened and live load is applied, composite action begins and
the slab starts to bear compressive stresses.
For the AFRP prestressed girder, induced stresses at
prestressing and at service load meet the service limit
states according to AASHTO LRFD Bridge Design Speci-
fications,10 and are similar to that of steel. For both cases,
the neutral axis lies within the slab when close to failure;
however, for the case of the AFRP prestressed girder, the
concrete compressive stresses are almost linearly distrib-
uted, indicating nonductile flexural behavior of the section
as opposed to the steel case, where the concrete slab exhibits
inelastic stresses until crushing of the top fiber.
Load-deflection analysis along girder
To find the deflection along the girder at different steps
of loading, refined nonlinear analysis is performed based
on conjugate beam theory. The results for the service load,
postcracking load, and failure load are depicted in Fig. 6(a)
through (c). The girder was designed as a fully prestressed
section such that under service load it can remain uncracked.
Therefore, the deflection of the girder under the service load
is the same for either the steel or AFRP prestressed girder.
The maximum deflection in this case is 0.28 in. (7.1 mm),
which is less than the allowable amount defined as the length
of the girder divided by 800, equal to 0.6 in. (15.2 mm),
according to AASHTO LRFD Bridge Design Specifica-
ACI Structural Journal/November-December 2013
tions.10 Once the load exceeds the cracking load, the deflec-
tion of the AFRP prestressed girder is still close to that of
steel. However, at failure, the maximum deflection of the
steel prestressed girder is larger than that of AFRP because
of the considerable ductility provided by yielding of steel
strands and compressive inelastic stresses at top fibers of the
concrete section.
The maximum deflections at failure are approximately
3 and 6 in. (76.2 and 152.4 mm) for the AFRP and steel
prestressed girder, respectively. Analytical results for
maximum deflection of the girder for both AFRP and steel
cases are compared and illustrated in Fig. 6(d). It is clearly
seen that the girder prestressed with AFRP has sufficient
cracking strength and ultimate strength. Figure 6(e) shows
the experimental data by Trejo et al.13 for midspan deflec-
tion of the steel prestressed girder compared to numerical
results. The error of the analysis in predicting the maximum
deflection at failure is 4%, representing high fidelity of the
developed computational model. It should be noted that the
Mmax, shown in Fig. 6(a) through (c), includes the applied
live load plus the dead load of the composite section.
EVALUATION OF DEFLECTION EQUATION IN
ACI 440.4R
ACI 440.4R recommends the use of the effective moment
of inertia, Ie, to calculate the deflection of FRP prestressed
concrete beams.9 In this procedure, Ie is assigned to the
entire beam and the maximum deflection is calculated using
linear elastic analysis.
3
 M cr    M 3
Ie =   β d I g + 1 -  M   I cr ≤ I g
cr
(4)
 Ma   a
 
969
 Fig. 6—Deflection of girder: (a) service load; (b) post-cracking load; (c) failure load;
(d) comparison between AFRP and steel (numerical analysis); and (e) analytical and
experimental result for steel prestressed girder.

moment of inertia. Ep and Es are the modulus of elasticity

Fig. 7—Evaluation of ACI 440.4R equation for AFRP
prestressed girder.
where Mcr is cracking moment; Ma is maximum moment in
the girder at which the deflection is being computed; Ig is
gross moment of inertia; Icr is cracked moment of inertia;
and bd = 0.5(Ep/Es + 1) is a factor to soften the effective
of the FRP tendon and steel, respectively. Figure 7 shows
that the equation underestimates the flexural stiffness of the
girder, and hence overestimates the deflection.
The authors believe that the difference is due to the fact
that this equation and also other similar formulas, which are
based on the original equation proposed by Branson14 for
steel reinforced concrete beams, assume that the flexural
behavior after cracking is linear elastic up to yielding of the
steel reinforcement (for steel-reinforced concrete) or rupture
of the tendons (for FRP-reinforced concrete). However, for
a prestressed girder that is composite with a slab, the flex-
ural behavior after cracking is not quite linear as a result of
a gradual decompression at bottom fibers (Fig. 4(b)). This
gradual decompression provides an additional postcracking
flexural stiffness that reduces the deflection. Furthermore,
investigations have shown that the ACI 440.4R equation
for predicting the effective moment of inertia is consider-
ably influenced by the Ig/Icr ratio.5 One of the main reasons
is because the Branson equation was calibrated for steel-

970 ACI Structural Journal/November-December 2013

reinforced concrete beams with the reinforcement ratio at
approximately 1.65%, which is corresponding to the ratio
Ig/Icr of approximately 2.2.2 In spite of many formulas avail-
able to predict the deflection of steel- and FRP-reinforced
concrete beams,15 there is still a need for an equation with
a high degree of accuracy to predict the deflection of FRP
prestressed beams.
CONCLUSIONS
AASHTO I-girder (Type I) was designed as a fully
prestressed section conforming to AASHTO LRFD Bridge
Design Specification using two different tendons: 1) preten-
sioned AFRP; and 2) pretensioned steel. Then, a fiber element
analysis was performed using a computational model devel-
oped to study the stress distribution over the cross section,
moment-curvature, and load-deflection relationships of the
girders. Comparison between analytical and experimental
data for steel prestressed girder showed a maximum error of
5% and 7% in prediction of failure and cracking moments,
respectively, which indicates the reasonably high accuracy
of the developed computational model.
Stress and moment-curvature analyses confirmed that the
AFRP prestressed girder not only meets service limit states
but also provides sufficient flexural strength. However, in
contrast to the steel prestressed girder, whose failure mode
was crushing of the concrete at the top fibers of the section,
the failure mode of the AFRP prestressed girder was rupture
of the tendons, causing less deformability. This is mainly
due to the fact that over-reinforcing the composite girder
to reach to concrete crushing as a failure mode, rather than
ruptures of the tendons, is almost impractical because the
brittle ratio of the composite section is considerably large.
Although tendon rupture as a failure mode is less desirable,
the failure curvature was found 18 times the cracking curva-
ture that can still provide sufficient warning prior to failure.
Load-deflection analysis of the girders showed that the
deflection under service load is almost half the allowable
amount. At failure, deflection of the steel prestressed girder
was approximately twice that of the AFRP prestressed girder
due to the large ductility provided by yielding of the steel
strands and inelastic compressive stresses in concrete. It was
also shown that the deflection equation in ACI 440.4R under-
estimates the postcracking stiffness of the AFRP prestressed
girder, and consequently overestimates the deflection. This
matter becomes more crucial when the girder is partially
prestressed and deflection under service loads commonly
governs the design, and hence there is a need for an equa-
tion with a high degree of accuracy to predict the deflection
of FRP prestressed beams. The area of reinforcement in the
AFRP prestressed girder was approximately 1.5 times that
of prestressing steel, and the numerical analyses clearly indi-
cated that the AFRP prestressed girder can be successfully
designed to meet AASHTO criteria.
Although replacing the prestressing steel with AFRP
strand helps to overcome corrosion-induced deterioration
and enhance the durability of the structure, time-dependent
characteristics of AFRP bars, such as creep-rupture and
fatigue strength, should also be incorporated in design, as
they affect the serviceability of the structure. Experimental
research on creep-rupture behavior of AFRP bars has shown
a linear relationship between the logarithm of time and
creep-rupture strength, where the 50-year creep-rupture
strength can be approximately extrapolated. The ratio of
stress at failure to the ultimate strength of AFRP bars has
ACI Structural Journal/November-December 2013
been found at approximately 0.5 by Taerwe,16 0.47 by
Yamaguchi et al.,17 0.66 by Ando et al.,18 and 0.6 by Dolan
et al.,4 which almost indicates the range of 0.45 to 0.65.
ACI 440.1R imposes a safety factor of 1/0.6 to the existing
experimental data and recommends an allowable sustained
stress level equal to 0.3.19 In this paper, the AFRP bars were
assumed to be initially prestressed up to 0.6 of their ultimate
strength, which results in a sustained stress level equal to
0.48, given a 20% long-term loss in prestressing force.
Also, fatigue characteristics of AFRP bars needs to be
considered in design as the bridge deck is subjected to
dynamic load of vehicles. Research conducted by Odagiri
et al.20 shows that the maximum stress level needs to be
set between 54 to 73% of the ultimate tensile strength.
However, ACI 440.1R uses the same stress limits for fatigue
as for creep-rupture. Further experimental research seems
to be required regarding the time-dependent characteristics
of AFRP bars as well as long-term performance of AFRP
prestressed members to provide a more reliable and consis-
tent design procedure.
DISCLAIMER
The content of this paper reflects the findings of the
authors, who are responsible for the fidelity of the results
presented herein, regarding the short-term behavior of the
AFRP prestressed girder in composite action with bridge
deck. The results are not necessarily viable for the long-term
performance affected by the time-dependent characteristics
of AFRP bars, which will be investigated further by the
authors.
ACKNOWLEDGMENTS
This research was supported by a grant from the National Science Foun-
dation (NSF 0927333). The authors would like to express appreciation for
the generous support from NSF to conduct this research. The findings and
opinions presented are those of the authors and are not necessarily those of
the sponsoring agency.
NOTATION
A = cross-sectional area of girder
Abp = area of reinforcement corresponding to rbp
Ap = area of tensile reinforcement in AFRP prestressed girder
As = area of tensile reinforcement in steel prestressed girder
beff = effective width of slab in composite action with girder
c = neutral axis depth
DFM = distribution factor for moment
dM = error in bending moment when checking equilibrium equation
dP = error in axial force when checking equilibrium equation
de0 = strain adjustment in iterative solution
df = curvature adjustment in iterative solution
Ep = modulus of elasticity of AFRP
Es = modulus of elasticity of steel strand
e = eccentricity of bottom prestressed tendons from neutral axis
e′ = eccentricity of top prestressed tendons from neutral axis
fb = normal stress at bottom fiber of section due to service load
(D + 0.8L)
fbp = compressive stress due to prestressing at bottom fiber of section
fc′ = compressive strength of concrete
fcr = fracture modulus of concrete
fpu = ultimate strength of AFRP
fsu = ultimate strength of steel strand
fT = normal stress at top fiber of girder at transfer
ft = allowable tensile stress at service load
fy = yield strength of steel strand
Icr = cracked moment of inertia
Ie = effective moment of inertia
Ig = gross moment of inertia
IM = impact factor
k = slope controller factor in Popovic’s equation
M0 = moment corresponding to initiation of yielding in steel strands
971
Ma = maximum moment in girder at which deflection is being
computed
Mcr = cracking moment
Mn = failure moment
n = curve-fitting factor in Popovic’s equation
Q = model parameter in Mengotto and Pinto’s equation
R = model parameter in Mengotto and Pinto’s equation
Sb = section modulus for bottom fiber of girder
Sbc = section modulus for bottom fiber of composite section
ST = section modulus for top fiber of girder
bd = factor to soften effective moment of inertia
e0 = strain at reference point
ef = strain capacity remained in tendon for flexure
ep0 = prestressing strain in tendon after loss
f = curvature of section
f0 = curvature corresponding to initiation of yielding in steel strands
fcr = cracking curvature
fn = failure curvature
r = reinforcement ratio
rbp = brittle ratio for AFRP prestressed girder
rbs = balance ratio for steel prestressed girder
REFERENCES
1. Trejo, D.; Aguiniga, F.; Buth, C. E.; Yuan, R.; James, R.; and
Keating, P., “FRP Reinforcing Bars in Bridge Decks: State of the Art
Review,” Report No. FHWA-01/1520-2, Texas A&M Transportation Insti-
tute, College Station, TX, 2000, pp. 1-143.
2. Bischoff, P. H., “Reevaluation of Deflection Prediction for Concrete
Beams Reinforced with Steel and Fiber Reinforced Polymer Bars,” Journal
of Structural Engineering, ASCE, V. 131, No. 5, 2005, pp. 752-767.
3. Bischoff, P. H., “Deflection Calculation of FRP Reinforced Concrete
Beams Based on Modifications to the Exisiting Branson Equation,” Journal
of Composites for Construction, ASCE, V. 11, No. 1, 2007, pp. 4-14.
4. Dolan, C. W.; Nanni, A.; Hamilton, H. R.; and Bakis, C. E., “Design
Recommendations for Concrete Structures Prestressed with FRP Tendons,”
V. 1, Cont. DTFH61-96-C-00019, Federal Highway Administration, Wash-
ington, DC, 2001, pp. 1-113.
5. Kim, Y. J., “Flexural Response of Concrete Beams Prestressed with
AFRP Tendons: Numerical Investigation,” Journal of Composites for
Construction, ASCE, V. 14, No. 6, 2010, pp. 647-658.
6. Pisani, M. A., “A Numerical Survey on the Behavior of Beams Pre-
stressed with FRP Cables,” Construction and Building Materials, V. 12,
1998, pp. 221-232.
7. Toutanji, H., and Saafi, M., “Performance of Concrete Beams
Prestressed with Aramid Fiber-Reinforced Polymer Tendons,” Composite
Structures, V. 44, 1999, pp. 63-70.
8. Rafi, M. M., and Najdai, A., “Evaluation of ACI 440 Deflection Model
for Fiber-Reinforced Polymer Reinforced Concrete Beams and Suggested
Modification,” ACI Structural Journal, V. 106, No. 6, Nov.-Dec. 2009,
pp. 762-771.
9. ACI Committee 440, “Prestressing Concrete Structures with FRP
Tendons (440.4R-04),” American Concrete Institute, Farmington Hills, MI,
2004, 35 pp.
10. AASHTO, “LRFD Bridge Design Specifications,” fifth edition,
American Association of State Highway and Transportation Officials,
Washington, DC, 2010.
11. Popovic, S., “Numerical Approach to the Complete Stress-Strain
Relation for Concrete,” Cement and Concrete Research, V. 3, No. 5, 1973,
pp. 583-599.
12. Menegotto, M., and Pinto, P., “Method of Analysis for Cyclically
Loaded Reinforced Concrete Plane Frames Including Changes in Geom-
etry and Non-Elastic Behavior of Elements under Combined Normal Force
and Bending,” IABSE Symposium on the Resistance and Ultimate Deform-
ability of Structures Acted on by Well-Defined Repeated Loads, Lisbon,
Portugal, 1973, V. 13:15-22.
13. Trejo, D.; Hueste, M. B.; Kim, Y. H.; and Atahan, H., “Character-
ization of Self-Consolidating Concrete for Design of Precast, Prestressed
Bridge Girders,” Report No. FHWA/TX-09/0-5134-2, Texas A&M Trans-
portation Institute, College Station, TX, 2008, pp. 1-364.
14. Branson, D. E., “Instantaneous and Time-Dependent Deflections
of Simple and Continuous Reinforced Beams,” HPR Report No. 7, Part 1,
Alabama Highway Department, Bureau of Public Roads, Montgomery, AL,
1965, pp. 1-78.
15. Mota, C.; Alminar, S.; and Svecova, D., “Critical Review of Deflec-
tion Formulas for FRP-RC Members,” Journal of Composites for Construc-
tion, ASCE, V. 10, No. 3, 2006, pp. 183-194.
972
16. Taerwe, L., “Non-Metalic (FRP) Reinforcement for Concrete
Structures,” Proceedings of the Second International RILEM Symposium
(FRPRCS-2), London, UK, 1995, pp. 1-714.
17. Yamaguchi, T.; Kato, Y.; Nishimura, T.; and Uomoto, T., “Creep
Rupture of FRP Rods Made of Aramid, Carbon and Glass Fibers,” Proceed-
ings of the Third International Symposium on Non-Metallic (FRP) Reinforce-
ment for Concrete Structures (FRPRCS-3), V. 2, Japan Concrete Institute,
Tokyo, Japan, 1997, pp. 179-186.
18. Ando, N.; Matsukawa, H.; Hattori, A.; and Mashima, A., “Experi-
mental Studies on the Long-Term Tensile Properties of FRP Tendons,”
Proceedings of the Third International Symposium on Non-Metallic (FRP)
Reinforcement for Concrete Structures (FRPRCS-3), V. 2, Japan Concrete
Institute, Tokyo, Japan, 1997, pp. 203-210.
19. ACI Committee 440, “Guide for the Design and Construction of
Structural Concrete Reinforced with FRP Bars (ACI 440.1R-06),” Amer-
ican Concrete Institute, Farmington Hills, MI, 2006, 44 pp.
20. Odagiri, T.; Matsumoto, K.; and Nakai, H., “Fatigue and Relaxation
Characteristics of Continuous Aramid Fiber Reinforced Plastic Rods,”
Proceedings of the Third International Symposium on Non-Metallic (FRP)
Reinforcement for Concrete Structures (FRPRCS-3), V. 2, Japan Concrete
Institute, Tokyo, Japan, 1997, pp. 227-234.
APPENDIX A—STRESS-STRAIN RELATIONSHIP
OF MATERIAL
Stress-strain behavior of the concrete is defined using the
Popovic’s equation10
n  e c / e c′ 
 
fc = nk fc ′ (A1)
n - 1 +  e c / e c′ 
 
where ec′ is the strain corresponding to fc′; n is a curve-
fitting factor; and k is a factor that controls the slopes of the
ascending and descending parts of the stress-strain curve.
For concrete with fc′ = 6 ksi (41.4 MPa), n and k are equal
to 2.77 and 1.33, respectively. Formulas for the factors n
and k can be found in the reference. Figure A1(a) shows the
assumed behavior for the concrete of the girder based on the
Popovic’s equation. It is seen that the stress-strain relation-
ship is almost linear up to 0.7fc′, which is equal to 4.2 ksi
(29 MPa). Stress-strain behavior for prestressing steel strand
is defined using Menegotto and Pinto’s formula11
 
 1-Q 
f s = Es e s  Q + 1/ R
 (A2)
 1 + e E / f  
R
  s s y  
Parameters Es, fy, Q, and R are determined experimentally.
For prestressing steel (Grade 270), these parameters are equal
to Es = 28,000 ksi (193,200 MPa), fy = 243 ksi (1656 MPa),
Q = 0.03, and R = 6. Further discussion can be found in
the reference. The assumed behavior for prestressing steel
strands and AFRP tendons is depicted in Fig. A1(b).
APPENDIX B—FLEXURAL DESIGN OF AFRP
PRESTRESSED GIRDER
Maximum dead load moments induced by each compo-
nent of the bridge prototype are summarized in Table B1,
including the dead load of the girder, slab, barrier, and future
wearing. Maximum live load moments, per lane, due to the
truck load and lane load, are calculated as ML-Truck = 450 kft
(610 kNm), and ML-Lane = 128 kft (174 kNm). It should be
noted that, for the girder shown in Fig. 1, tandem load results
in pretty much the same maximum moment as the truck load.
ACI Structural Journal/November-December 2013
 Fig. B1—Dimensions, neutral axis location, and CM of
tendons for AFRP prestressed girder.

Table B1—Maximum deal load moments, kft (kNm)

MD-Beam MD-Slab MD-Barrier MD-Wearing
55 (75) 110 (149) 24 (33) 26 (35)
S =165 (224) S = 50 (68)
MD = 215 (292)

Fig. A1—Stress-strain relationships: (a) concrete (girder);

– 0.46 ≈ 2 ksi (13.8 MPa). This compressive stress can be
and (b) prestressing steel and AFRP.
written as

The distribution factor for moment is computed as DFM =
0.56 based on AASHTO 4.6.2.2.2b-1. Assuming the impact
factor as IM = 1.33, the maximum live load moment can be
computed from the following equation
M L = DFM ( M L -Truck × IM + M L - Lane )
which gives ML = 405 kft (550 kNm). According to AASHTO
3.4, the service load for deflection control (D + L) is Ms =
620 kft (841 kNm) and the ultimate load for strength design
(1.25D + 1.75L) is Mu = 978 kft (1326 kNm). To compute
the required number of AFRP tendons, the service load for
controlling the tension in prestressed concrete (D + 0.8L) is
considered as follows
 M D - Beam + M D - Slab 
fb =  +
 Sb  (B2)

(
 M D - Barrier + M D -Wearing + 0.8 M L -Truck + M L - Lane
 Sbc
Given the section modulus for the bottom fiber of the girder
and composite section are Sb = 1807 in.3 (0.029 m3) and Sbc =
3693 in.3 (0.061 m3), respectively, the tensile normal stress
at the bottom fiber of the section is computed as fb = 2.4 ksi
(16.56 MPa). Based on AASHTO 5.9.4.2.2-1, the allowable
tensile stress is ft = 0.19√fc′ = 0.46 ksi (3.2 MPa). Therefore,
the compressive stress due to prestressing must be fbp = 2.4
P Pe
fbp = + (B3)
A Sb
where the cross-sectional area of the girder is A = 276 in.2
(0.18 m2). As shown in Fig. B1, the prestressing eccentricity
is e = 7.5 in. (190.5 mm), so the required prestressing force
equals P = 265 kip (1179 kN). AFRP tendons are prestressed
up to 60% of their ultimate strength and it can be conser-
(B1) vatively assumed that the total loss during the lifetime of
the structure is 20%. Thus, the required number of AFRP
tendons in the bottom flange is
265
N= ≈ 22 (B4)
0.126 × 203 × 0.6 × (1 - 0.2)
To control the stress at transfer (before loss) close to the
support, two more prestressed AFRP tendons are used at the
top flange. The normal tensile stress at the top fiber can be
calculated as
 22 + 2 22 × e 2 × e ′ 
f T = - + T - T  × 0.126 × 203 × 0.6 (B5)
)  A S S 

where positive and negative signs indicate tension and
compression, respectively. Substituting the section modulus
for the top fiber of the girder, ST= 1476 in.3 (0.024 m3), gives
fT= 0.1 ksi (0.69 MPa), which is less than the allowable
tension at transfer, fti = 0.24√fci′ = 0.52 ksi (3.6 MPa) based
on AASHTO 5.9.4.1.2-1. Referring to Fig. 5, the computa-
tional model showed the prestressing stress (after loss) at the
top fiber of the section equal to f T= 0.083 ksi (0.57 MPa),
which is in a good agreement with the aforementioned hand

ACI Structural Journal/November-December 2013 973

 Fig. C1—Balance failure of AFRP prestressed girder: (a) single section; and (b)
composite section.

calculations. As discussed, numerical analysis showed the

sufficient strength of the composite section and acceptable
deflection under service load.

APPENDIX C—EFFECT OF TOPPING DECK ON

BRITTLE RATIO OF AFRP PRESTRESSED GIRDER
The brittle ratio of the AFRP prestressed girder can be
calculated using the basic equations of equilibrium and
strain compatibility for the girder section. As illustrated in
Fig. C1(a), the depth of the neutral axis can be determined
from the strain compatibility equation

c 0.003 (C1)
=
d 0.003 + e f

Assuming 60% prestressing and 5% as loss in prestressing

force gives ep0 = 0.013 and ef = 0.01. ep0 is the strain due
to prestressing (after loss) and ef is what is left for flexure.
Referring to Fig. C1(a), d ≈ 23 in. (584.2 mm), and based on
the aforementioned equation, c = 5.3 in. (134.6 mm). The
area of reinforcement corresponding to the brittle ratio can
be found from the following equilibrium equation

( )
C = e p 0 + e f E p Abp (C2)

which results in Abp = 1.21 in.2 (780.6 mm2). Now, if the

effect of topping deck is taken into account (Fig. C1(b)), d
= 31 in. (787.4 mm) and c = 7.15 in. (181.6 mm), which
means that the neutral axis is within the slab. Given b = beff,
the equilibrium equation gives Abp = 8.75 in.2 (5645.2 mm2),
which is approximately seven times that of the case without
the topping slab. If the area of reinforcement in AFRP
prestressed girder, Ap = 2.76 in.2 (1780.6 mm2), is compared
to Abp = 1.21 in.2 (780.6 mm2), it might be concluded that
the flexural failure is due to the crushing of the concrete.
However, accounting for the effect of composite action
between the girder and topping deck shows that the rein-
forcement ratio is approximately one-third of the actual
brittle ratio (2.76/8.75), confirming the tendon rupture
as the failure mode. Put another way, to change the
Fig. D1—Experimental testing of AFRP prestressed section:
(a) test setup; (b) moment-curvature response.
failure mode from tendon rupture to concrete crushing,
the area of reinforcement should be tripled, meaning that
approximately 70 AFRP tendons have to be accommodated
in the bottom flange of the girder, which is impractical.
APPENDIX D—EXPERIMENTAL TEST OF AFRP
PRESTRESSED SECTION
An AFRP prestressed section was constructed and
tested under flexure at High-Bay Structural and Mate-
rials Testing Laboratory of Texas A&M University.

974 ACI Structural Journal/November-December 2013

The 8 x 12 in. (203 x 305 mm) concrete section was concen- response are in good agreement with experimental results,
trically prestressed with two AFRP bars with the same prop- which show the high fidelity of the developed computational
erties as assumed for the I-girder, and was tested under four- model. The moment and normalized curvature, curvature
point loading (Fig. D1). The AFRP bars were prestressed up multiplied the height of the section, for cracking and failure
to 60% of the ultimate tensile strength—14.5 kip (65 kN)— stages are shown in Table D1.
and the concrete was poured after 72 hours. The total loss Table D1—Comparison between experimental and
until the testing date was evaluated approximately 5%, analytical results, kft (kNm)
which results in 14 kip (62 kN) as the net prestressing force
per each bar. The compressive strength of the concrete was AFRP prestressed section
found fc′ = 7 ksi (48.3 MPa). Two linear variable displace- Experiment Computational model
ment transducers (LVDTs), each 12 in. (305 mm) long, were Mcr 9.9 (13.4) 9.4 (12.7)
used at top and bottom fibers of the midspan to measure
the strain and induced curvature, accordingly. One string Mu 15.4 (20.9) 15.6 (21.2)
pot (STP) was used right at the midspan to measure the (fh)cr 0.0005 0.0004
maximum deflection. Flexural cracks were first observed (fh)u 0.02 0.023
at bottom fibers of the midspan at M = 9.6 kft (13 kNm)
efu* — >0.02
and resulted in a significant drop in flexural stiffness of the
section (Icr/Ig = 0.015). With increasing the load, flexural ecu† — –0.0027
cracks propagated toward the top fibers and the specimen Failure mode Tendon rupture Tendon rupture
eventually failed due to rupture of the tendons at M = 15.4 kft *
efu is maximum strain at AFRP bar when specimen fails.
(21 kNm). The numerical results for moment-curvature †
ecu is maximum compressive stress at concrete when specimen fails.

ACI Structural Journal/November-December 2013 975

 Notes:

976 ACI Structural Journal/November-December 2013

View publication stats