Critical Review of Deflection Formulas for FRP-RC Members

Carlos Mota1; Sandee Alminar2; and Dagmar Svecova3

Abstract: The design of fiber-reinforced polymer reinforced concrete 共FRP-RC兲 is typically governed by serviceability limit state
requirements rather than ultimate limit state requirements as conventional reinforced concrete is. Thus, a method is needed that can predict
the expected service load deflections of fiber-reinforced polymer 共FRP兲 reinforced members with a reasonably high degree of accuracy.
Nine methods of deflection calculation, including methods used in ACI 440.1R-03, and a proposed new formula in the next issue of this
design guide, CSA S806-02 and ISIS M03-01, are compared to the experimental deflection of 197 beams and slabs tested by other
investigators. These members are reinforced with aramid FRP, glass FRP, or carbon FRP bars, have different reinforcement ratios,
geometric and material properties. All members were tested under monotonically applied load in four point bending configuration. The
objective of the analysis in this paper is to determine a method of deflection calculation for FRP RC members, which is the most suitable
for serviceability criteria. The analysis revealed that both the modulus of elasticity of FRP and the relative reinforcement ratio play an
important role in the accuracy of the formulas.
DOI: 10.1061/共ASCE兲1090-0268共2006兲10:3共183兲
CE Database subject headings: Concrete, reinforced; Fiber-reinforced polymers; Deflection; Curvature; Codes; Serviceability;
Statistics.

Introduction Effective Moment of Inertia Approach

Fiber-reinforced polymer 共FRP兲 reinforcing bars are currently ACI 318 共ACI 1999兲 and CSA A23.3-94 共CSA 1998兲 recommend
available as a substitute for steel reinforcement in concrete struc- the use of the effective moment of inertia, Ie, to calculate the
tures that may be vulnerable to attack by aggressive corrosive deflection of cracked steel reinforced concrete members. The pro-
agents. In addition to superior durability, FRP reinforcing bars cedure entails the calculation of a uniform moment of inertia
have a much higher strength than conventional mild steel. How- throughout the beam length, and use of deflection equations
ever, the modulus of elasticity of FRP is typically much lower derived from linear elastic analysis.
than that of steel. This leads to a substantial decrease in the stiff- The effective moment of inertia, Ie, is based on semiempirical
ness of FRP reinforced beams after cracking. Since deflections are considerations, and despite some doubt about its applicability to
inversely proportional to the flexural stiffness of the beam, even conventional reinforced concrete members subjected to complex
loading and boundary conditions, it has yielded satisfactory
some FRP over-reinforced beams are susceptible to unacceptable
results in most practical applications over the years. In North
levels of deflection under service conditions. Hence, the design of
American codes, deflection calculation of flexural members are
FRP reinforced concrete 共FRP-RC兲 is typically governed by ser-
mainly based on equations derived from linear elastic analysis,
viceability requirements and a method is needed that can calculate
using the effective moment of inertia, Ie, given by Branson’s
the expected service load deflections of FRP reinforced members
formula 共1965兲

冉 冊 冋 冉 冊册
with a reasonable degree of accuracy.
The objective of this paper is to point out the inconsistencies M cr 3 M cr 3

in existing deflection formulas and to demonstrate that all of Ie = Ig + 1 − Icr 艋 Ig 共1兲

Ma Ma
the current equations have limitations for deflection calculation
of FRP-RC members. Only instantaneous deflections will be M cr⫽cracking moment; Ig⫽moment of inertia of the gross
discussed in this paper. section; Icr⫽moment of inertia of the cracked section transformed
to concrete; and Ie⫽effective moment of inertia.
1 Research by Benmokrane et al. 共1996兲 suggested that in order
Research Assistant, Dept. of Civil Engineering, Univ. of Manitoba,
to improve the performance of the original equation, Eq. 共1兲 will
Winnipeg MB, Canada R3T 5V6.
2
Research Assistant, Dept. of Civil Engineering, Univ. of Manitoba, need to be further modified. Constants to modify the equation
Winnipeg MB, Canada R3T 5V6. were developed through a comprehensive experimental program.
3
Associate Professor, Dept. of Civil Engineering, Univ. of Manitoba, The effective moment of inertia was defined according to Eq. 共2兲
Winnipeg MB, Canada R3T 5V6 共corresponding author兲. E-mail: if the reinforcement was FRP

冉 冊 冋 冉 冊册
[email protected] 3 3
Note. Discussion open until November 1, 2006. Separate discussions M cr Ig M cr
Ie = + 0.84 1 − Icr 艋 Ig 共2兲
must be submitted for individual papers. To extend the closing date by Ma 7 Ma
one month, a written request must be filed with the ASCE Managing
Editor. The manuscript for this paper was submitted for review and pos- Further research has been done in order to define an effective
sible publication on November 2, 2004; approved on July 21, 2005. This moment of inertia equation which is similar to that of Eq. 共1兲, and
paper is part of the Journal of Composites for Construction, Vol. 10, No. converges to the cracked moment of inertia quicker than the cubic
3, June 1, 2006. ©ASCE, ISSN 1090-0268/2006/3-183–194/$25.00. equation. Many researchers 共Benmokrane et al. 1996; Brown and

JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / MAY/JUNE 2006 / 183

Bartholomew 1996; Toutanji and Saafi 2000兲 argue that the basic
form of the effective moment of inertia equation should remain
as close to the original Branson’s equation as possible, because
Ie = 冉 冊 冋 冉 冊册
M cr 3
Ma
␤Ig + 1 −
M cr
Ma
3
Icr 艋 Ig 共7兲

it is easy to use and designers are familiar with it. Brown and where
Bartholomew 共1996兲 proposed that the basic form of Eq. 共1兲
could be used with reasonable accuracy to find the service deflec-
tions of FRP-RC beams if a fifth order equation was used rather
than a cubic. The modified equation is presented in the following
␤=␣ 冉 冊 EF
Es
+1 共8兲

equation: where ␤⫽reduction coefficient; ␣⫽bond dependent coefficient

共until more data become available, ␣⫽0.5兲; and EF⫽modulus of

Ie = 冉 冊 冋 冉 冊册
M cr 5
Ma
Ig + 1 −
M cr
Ma
5
Icr 艋 Ig 共3兲
elasticity of the FRP reinforcement.
Upon finding that the ACI 440.1R-03 共ACI 2003兲 equation
often underpredicted the service load deflection of FRP reinforced
A further investigation of the effective moment of inertia was concrete members, several attempts have been made in order to
performed by Toutanji and Saafi 共2000兲. It was found that the modify Eq. 共7兲. For instance, Yost et al. 共2003兲 claimed that the
order of the equation depends on both the modulus of elasticity accuracy of Eq. 共7兲 primarily relied on the reinforcement ratio of
of the FRP, as well as the reinforcement ratio. Based on their the member. It was concluded that the formula could be of the
research, Toutanji and Saafi 共2000兲 have recommended that the same form, but that the bond dependent coefficient, ␣, had to be
following equations be used to calculate the deflection of FRP modified. A modification factor, ␣, was proposed in the following
reinforced concrete members: form:

Ie = 冉 冊 冋 冉 冊册
M cr
Ma
m
Ig + 1 −
M cr
Ma
m
Icr 艋 Ig 共4兲 ␣ = 0.064 冉 冊 ␳FRP
␳bal
+ 0.13 共9兲

where where ␳bal⫽balanced reinforcement ratio.

The ACI 440 Committee 共ACI 2004兲 has also proposed
EFRP revisions to the design equation in ACI 440.1R-03 共ACI 2003兲.
m = 6 − 10 ␳FRP
ES The moment of inertia equation has retained the same familiar
form as that of Eq. 共7兲 in these revisions. However, the form of
if the reduction coefficient, ␤, to be used in place of Eq. 共8兲 was
modified. The new reduction coefficient has changed the key
EFRP variable in the equation from the modulus of elasticity to the
␳FRP ⬎ 0.3
ES relative reinforcement ratio as shown in the following equation:

otherwise 共5兲 ␤= 冉 冊
1 ␳FRP
5 ␳bal
艋 1.0 共10兲

m=3
where ␳FRP⫽reinforcement ratio; EFRP⫽modulus of elasticity Moment–Curvature Approach
of FRP reinforcement; and ES⫽modulus of elasticity of steel
reinforcement. The moment–curvature approach for deflection calculation is
The ISIS Design Manual M03-01 共Rizkalla and Mufti 2001兲 based on the first principles of structural analysis. When a
has suggested the use of an effective moment of inertia which is moment–curvature diagram is known, the virtual work method
quite different in form compared to the previous equations. It can be used to calculate the deflection of structural members
suggests using the modified effective moment of inertia equation under any load as
defined by the following equation to be adopted for future use:

冕
L
M
ITIcr ␦= m dx 共11兲

冋 冉 冊册
Ie = 2
艋 Ig 共6兲 0 EI
M cr
Icr + 1 − 0.5 共IT − Icr兲 where L⫽simply supported length of the section;
Ma
M / EI⫽curvature of the section; and m⫽bending moment due to
where IT⫽uncracked moment of inertia of the section transformed a unit load applied at the point where the deflection is to be
to concrete. calculated.
Eq. 共6兲 is derived from equations given by the CEB-FIP A moment–curvature approach was taken by Faza and
MC-90 共CEB-FIP 1990兲. Ghali et al. 共2001兲 have verified that Ie GangaRao 共1992兲, who defined the midspan deflection for four-
calculated by Eq. 共6兲 gives good agreement with experimental point bending through the integration of an assumed moment
deflection of numerous beams reinforced with different types of curvature diagram. Faza and GangaRao 共1992兲 made the assump-
FRP materials. tion that for four-point bending, the member would be fully
According to ACI 440.1R-03 共ACI 2003兲, the moment of cracked between the load points and partially cracked everywhere
inertia equation for FRP-RC is dependent on the modulus of elas- else. A deflection equation could thus be derived by assuming that
ticity of the FRP and the following expression for Ie is proposed the moment of inertia between the load points was the cracked
to calculate the deflection of FRP reinforced beams: moment of inertia, and the moment of inertia elsewhere was the

184 / JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / MAY/JUNE 2006

effective moment of inertia defined by Eq. 共1兲. Through the Table 1. Range of Properties Present in Database
integration of the moment curvature diagram proposed by Faza Minimum Property Maximum
and GangaRao 共1992兲, the deflection for four-point loading is
defined according to the following equation: 520 f fu 共MPa兲 2,640
26.2 EF 共GPa兲 174
25.5 f c⬘ 共MPa兲 97.4
Pa 1.82 a/d 14.40
␦max = 共8a2Icr + 3L2Ie − 12a2Ie兲 共12兲
24EcIeIcr 3.38 Ig / Icr 64.42
0.001 ␳FRP 0.038
where a⫽shear span.
0.3 ␳FRP / ␳bal 9.1
Eq. 共12兲 has limited use because it is not clear what assump-
tions for the application of the effective moment of inertia should
be used for other load cases. However, it worked quite accurately The statistical analysis was performed by applying a log trans-
for predicting the deflection of the beams tested by Faza and formation to the ratios of the experimental to calculated deflection
GangaRao 共1992兲. ratios. A log transformation was employed to give equal weight
The CSA S806-02 共CSA 2002兲 suggests that the moment– to those ratios which were below one and those which were
curvature method of calculating deflection is well suited for FRP above one. Without this transformation, more weight is given to
reinforced members because the moment–curvature diagram can the ratios above one and information is lost when statistics are
be approximated by two linear regions: one before the concrete applied.
cracks, and the second one after the concrete cracks 共Razaqpur When considering long-term deflection, perhaps only the
et al. 2000兲. Therefore, there is no need for calculating curvature accuracy of short-term deflection equation is required since this
at different sections along the length of the beam as for steel number will be further modified by other coefficients. However,
reinforced concrete. There are only three pairs of moments with since only short-term deflection has been considered here, the
corresponding curvature that define the entire moment–curvature predicted deflection should be also consistently conservative. To
diagram: at cracking, immediately after cracking, and at ultimate. determine this, the 95% confidence interval was checked and only
With this in mind, simple formulas were derived for deflection those formulas whose entire confidence interval was conservative
calculation of simply supported FRP reinforced beams and are were deemed to be acceptable. This will minimize the chance of a
used in CSA S806-02 共CSA 2002兲. The deflection due to four- code equation underpredicting the deflection of a member. After
point bending can be found using the following equation: completing the log transformation, the 95% confidence interval is
defined by the following equation:

␦max =
Pa
24EcIcr
冋 冉 冊冉 冊 册
3L2 − 4a2 − 8 1 −
Icr
Ig
M cr 3 2
Ma
a 共13兲
冑 冑

e␮y−1.96␴y/ n,e␮y+1.96␴y/ n
where ␮y⫽mean of the ratios in the natural log scale;
共14兲

␴y⫽standard deviation of the ratios in the natural log scale; and

n⫽number of members in the sample set.
Verification of Proposed Methods When two confidently conservative methods were compared,
the geometric mean and the variance were used to establish the
The nine methods of deflection calculation presented in this paper most accurate one. Here the geometric mean should be as close to
were used to analyze 197 simply supported beams and slabs 1.0 as possible and the variance should be as small as possible.
tested by other investigators. Material and geometric properties of When using a log transformation, the geometric mean and vari-
the beams used in this investigation could not be published due to ance, in terms of a real ratio, are as defined in the following
the extent of the statistical sample but can be found in Mota equations:
共2005兲. Table 1 shows the range of some of the important prop-
mean = e␮y 共15兲
erties of the members in the database. All information used in the
analysis, such as cracking moment and modulus of elasticity of
variance = e␴y − 1 共16兲
concrete, was calculated using CSA A23.3-94 共CSA 1998兲 based
on input given by researchers. It was deemed necessary that val-
ues are not taken from experimental results to simulate design
Parametric Studies
office conditions.
To check the accuracy of formulas developed by other inves- From preliminary data analysis, it became apparent that the
tigators, a statistical analysis has been performed on each of the deflection values were sensitive to the value of the cracking
equations comparing the calculated deflection to the experimental moment. Even though it is not the usual practice to account for
deflection at several given load levels. It must be noted that the the moment due to the self weight for reinforced concrete beams,
deflection is typically only checked at the service load level. in the case of FRP reinforced concrete slabs, the addition of
However, since the service load criteria is only explicitly stated in moment due to the self weight became essential. The equations
the ISIS M03-01 共Rizkalla and Mufti 2001兲, it is unclear at this used to predict the deflection at a given load were modified in
point what the service load level for each code is. Thus, a statis- order to account for their respective self weights. It is important
tical analysis was carried out at both low loads and at elevated to note that without accounting for the moment due to the self-
loads to encompass the entire load–deflection curve, as well as at weight, the calculated cracking moment of the section was larger
the service level given by ISIS M03-01 共Rizkalla and Mufti 2001兲. than the cracking moment measured during the test and caused
This will allow the designers to choose an accurate formula, incorrect estimates of deflection. After the adjustment of the
based on the results of the analysis, at the load level which most moment, the accuracy of the formulas for predicting the cracking
closely resembles their service load criteria. load improved by approximately 1% on average but improved

JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / MAY/JUNE 2006 / 185

Table 2. Accuracy of Design Methods for Calculating Modulus of Rupture
Experimental/Calculated modulus of rupture
Statistical
property CSA ACI OHBDC CEB-FIP Raphael Oluokun Légeron
Geometric mean 1.01 0.86 1.21 1.06 0.46 1.37 0.64
Variance 0.39 0.39 0.39 0.39 0.39 0.40 0.39
95% confidence 共⫺兲 0.96 0.83 1.16 1.02 0.44 1.31 0.61
95% confidence 共⫹兲 1.06 0.91 1.27 1.11 0.48 1.44 0.67

by up to 20% for slabs. Given that there were 65 slabs present in members that had service moments less than the cracking moment
the database, this was deemed to be necessary. The self-weight were not included in the service load analysis. This greatly
was accounted for in the analysis by including the moment due reduced the number of members in the analysis and led to under-
to the self weight, as well as the moment caused by the four-point reinforced members not being analyzed at service load, as they all
loading. Assuming that the density of the concrete, ␥sw, was encountered this phenomenon.
23.5 kN/m3, the applied moment M a was defined according to the Given that almost half of the beams in the database had,
following equation: according to ISIS M03-01 共Rizkalla and Mufti 2001兲, service
moments smaller than the cracking moment, it was desired to
␥swbhl2 perform a statistical analysis at a load level which encompassed
M a = Pa + 共17兲
8 all of the beams in the database to improve the reliability of
the statistical equations. Upon viewing a plot of the log of the
where ␥sw⫽density of concrete; b⫽width of the member; and deflection ratios versus the ratio of the applied moment to the
h⫽height of the member. cracking moment as shown in Fig. 2, a funnel effect could be
The accuracy of the calculated cracking moment, M cr, is a key seen. The proximity of the points to the horizontal axis signifies
aspect to the accuracy of the deflection calculations. The control- higher accuracy of calculated deflections. Inspection of Fig. 2
ling variable for predicting the cracking moment is the modulus indicates that the accuracy of the equations is improved drama-
of rupture of concrete, f r. Thus, an analysis was done on the 197 tically at elevated loads for M a / M cr greater than 9, where the
members in the database to determine the most accurate formula points are gathered around the horizontal axis, and that the great-
for calculating f r. The formulas were taken from Reda Taha and est area of concern is slightly after cracking, where the points are
Hassanain 共2003兲 and the accuracy of the formulas compared scattered in the plot. Similar behavior was also observed earlier in
against the experimental modulus of rupture in the database can steel RC 共Choi et al. 2004兲. Since it was found in Fig. 2 that all
be seen in Table 2. The most accurate methods for calculating the equations are accurate at higher loads, it was desired to perform
cracking moment are given in the CSA A23.3-94 共CSA 1998兲 and an analysis of the behavior of the equations at a lower load where
CEB-FIP MC-90 共CEB-FIP 1990兲 codes. Thus, the CSA A23.3-94 there is a greater spread between the outputs of each equation.
共CSA 1998兲 equation, given by the following equation, has been This load was chosen to be 1.1Pcr. A statistical analysis was also
used in the database to calculate the modulus of rupture: performed on each of the code equations at 50 and 80% of the
maximum experimental loads to validate the equations for the
f r = 0.6␭冑 f ⬘c 共18兲 entire loading range.
where f r⫽modulus of rupture; f ⬘c ⫽concrete compressive strength;
and ␭⫽factor accounting for concrete density. For normal density Analysis at Service Loads
concrete ␭⫽1.00.
The trend lines for the log of the deflection ratios versus the
modulus of elasticity for glass FRP-RC 共GFRP-RC兲 beams at the
service load level can be seen in Fig. 3. In order for the formulas
Data Analysis
to be conservative, the log of the deflection ratio has to be smaller
than zero. Fig. 3 shows that the formula proposed by Yost et al.
In the following sections, the service load of FRP-RC beams will
共2003兲 is quite accurate at predicting the service deflection of
be referred to as the load at which the strain in the outermost FRP
layer in tension reaches a strain of 2,000⫻ 10−6 共ISIS M03-01
Rizkalla and Mufti 2001兲.
The service conditions proposed by ISIS M03-01 共Rizkalla and
Mufti 2001兲 often led to instances where the service moment was
smaller than the cracking moment, implying that the beam was
not cracked. However, when the moment–curvature diagrams of
these members were examined, all of these beams had curvature
greater than the cracking curvature and therefore were cracked.
A typical moment–curvature diagram for these members can be
seen in Fig. 1. The reason for this is the sudden decrease in
stiffness immediately after cracking, when concrete in tension is
not considered in calculating the capacity of the section. Since
the cracking moment is greater than the service moment, all the
formulas recommend the use of the gross moment of inertia to Fig. 1. Moment–curvature relationship for beam BC2H 共adapted
calculate the service deflection. This is incorrect and therefore from Theriault and Benmokrane 1997兲

186 / JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / MAY/JUNE 2006

 Fig. 2. Effect of applied moment on deflection prediction

GFRP-RC members and its accuracy is not affected by changes in found to satisfy this criteria the best. The area of concern is that it
the modulus of elasticity. This is important since the modulus of always underpredicted the deflection, which is undesired. Thus,
elasticity changes rapidly among different types of FRP. Table 3 the use of the Yost et al. 共2003兲 formula for calculating the service
shows the statistical analysis of each of the formulas at the service load deflection of GFRP-RC beams and either the CSA S806-02
load level. The formulas are conservative if the 95% confidence 共CSA 2002兲 or ISIS M03-01 共Rizkalla and Mufti 2001兲 formulas
interval is less than one. The formula given by Yost et al. 共2003兲 for calculating the service load deflection of CFRP-RC beams are
is the most accurate for GFRP-RC beams with mean value of 0.95 recommended.
while also providing conservative results. Other areas of concern in predicting the deflection at service
The following observations were made from the trend lines for load are revealed by Figs. 5 and 6, which show the trend lines
the deflection ratio versus the modulus of elasticity for CFRP-RC of each of the equations plotted against the slenderness ratio,
beams at the service load level plotted in Fig. 4. The accuracy of a / d, and relative reinforcement ratio, ␳ / ␳bal. It can be seen that
the methods by Yost et al. 共2003兲, the Proposed ACI 440.1R 共ACI the accuracy of all of the formulas presented varies as either
2004兲, Faza and GangaRao 共1992兲, and Toutanji and Saafi 共2000兲 the slenderness ratio or relative reinforcement ratio changes. The
do not vary as greatly with the modulus of elasticity of the CFRP, latter is of less concern as FRP-RC tends to usually be only
as do the methods proposed by ISIS M03-01 共Rizkalla and Mufti slightly over-reinforced for economic purposes. However, the
2001兲 or CSA S806-02 共CSA 2002兲. Although the accuracy of the effect of the slenderness ratio on deflection prediction needs
latter formulas does seem to vary with the modulus of the CFRP, further investigation. It should also be noticed that the formula
they do have the advantage of giving conservative deflection es- proposed by Faza and GangaRao 共1992兲 is the least affected by
timates. Table 3 verifies the above and shows that the formula either of these relationships.
given by Faza and GangaRao 共1992兲 is the most accurate formula
for CFRP-RC beams. Table 3 shows that the equations of the CSA
S806-02 共CSA 2002兲 and ISIS M03-01 共Rizkalla and Mufti 2001兲
are, on average, quite accurate and are conservative for their
entire 95% confidence intervals.
Table 3 also presents the accuracy of each of the formulas for
the AFRP-RC beams present in the database. It shows that the
CSA S806-02 共CSA 2002兲 equation is the most accurate equation
for calculating the deflection of AFRP-RC beams at the service
load level. However, given that there were only two AFRP-RC
members present in the database, this information should be taken
lightly and was not analyzed at any other load level.
Given that the method proposed by Yost et al. 共2003兲 over-
predicts the deflection of GFRP-RC and underpredicts the
deflection of CFRP-RC at service load, it is no coincidence that
when the entire dataset is considered at the service load level,
this method performs the most accurately as shown in Table 3.
This trend can also be seen in the formula given by Faza and
GangaRao 共1992兲, which also gives accurate results when looking
at the entire dataset. However, if consistency of a formula is the Fig. 3. Effect of modulus of elasticity on service load deflection
key, the formula proposed by the ACI 440.1R 共ACI 2004兲 was for GFRP-RC

JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / MAY/JUNE 2006 / 187

Table 3. Statistical Analysis of Predicted Deflections at Service
Experimental/calculated deflection

Statistical GFRP-RC CFRP-RC AFRP-RC All members

Method property 共sample size 67兲 共sample size 41兲 共sample size 2兲 共sample size 110兲
Faza and GangaRao 共1992兲 Geometric mean 0.81 1.10 1.92 0.92
Variation 0.94 0.36 1.24 0.79
95% confidence 共⫺兲 0.69 1.00 0.63 0.83
95% confidence 共⫹兲 0.95 1.21 5.87 1.03
Benmokrane et al. 共1996兲 Geometric mean 0.49 0.72 0.52 0.57
Variation 0.74 0.46 0.05 0.69
95% confidence 共⫺兲 0.43 0.64 0.49 0.52
95% confidence 共⫹兲 0.56 0.81 0.56 0.63
Brown and Bartholomew 共1996兲 Geometric mean 1.45 1.27 1.49 1.38
Variation 0.95 0.67 1.71 0.85
95% confidence 共⫺兲 1.24 1.09 0.37 1.23
95% confidence 共⫹兲 1.70 1.49 5.93 1.55
Toutanji and Saafi 共2000兲 Geometric mean 1.28 1.17 1.39 1.24
Variation 0.96 0.67 1.77 0.85
95% confidence 共⫺兲 1.09 1.00 0.34 1.11
95% confidence 共⫹兲 1.50 1.37 5.71 1.39
ISIS M03-01 共2001兲 Geometric mean 0.60 0.90 0.69 0.70
Variation 0.73 0.50 0.02 0.70
95% confidence 共⫺兲 0.53 0.79 0.67 0.63
95% confidence 共⫹兲 0.68 1.02 0.71 0.77
CSA S806-02 共2002兲 Geometric mean 0.59 0.90 1.05 0.70
Variation 0.78 0.41 0.81 0.71
95% confidence 共⫺兲 0.51 0.81 0.46 0.63
95% confidence 共⫹兲 0.68 1.00 2.39 0.77
ACI 440.1R-03 共2003兲 Geometric mean 1.35 1.53 1.35 1.42
Variation 0.80 0.59 0.89 0.72
95% confidence 共⫺兲 1.17 1.33 0.56 1.28
95% confidence 共⫹兲 1.55 1.76 3.26 1.57
Yost et al. 共2003兲 Geometric mean 0.95 1.19 1.37 1.04
Variation 0.77 0.41 0.53 0.66
95% confidence 共⫺兲 0.83 1.07 0.76 0.95
95% confidence 共⫹兲 1.09 1.32 2.47 1.14
Proposed ACI 440.1R 共2004兲 Geometric mean 1.24 1.19 1.87 1.23
Variation 0.84 0.41 1.16 0.69
95% confidence 共⫺兲 1.07 1.07 0.64 1.12
95% confidence 共⫹兲 1.43 1.32 5.44 1.36

Analysis at Load Levels Close to Cracking

The log of experimental over calculated deflection was plotted in
Fig. 7 with respect to the modulus of elasticity of GFRP. Fig. 7
shows that the Proposed ACI 440.1R 共ACI 2004兲 and Yost et al.
共2003兲 formulas are both quite accurate at predicting the deflec-
tion of GFRP-RC beams at a load of 1.1Pcr, with the accuracy of
the formulas having little variance with the value of the modulus
of elasticity of the GFRP. Table 4 shows the statistical analysis
of each of the formulas at this load. It can be seen in Table 4 that
the formula proposed by Yost et al. 共2003兲 is the most accurate
formula for calculating the service deflection of GFRP-RC
members, with an average experimental to predicted deflection Fig. 4. Effect of modulus of elasticity on service load deflection
ratio of 1.01. for CFRP-RC

188 / JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / MAY/JUNE 2006

Table 4. Statistical Analysis of Predicted Deflections at 1.1Pcr
Experimental/calculated deflection

Statistical GFRP-RC CFRP-RC ␳ ⬎ ␳bal ␳ ⬍ ␳bal All members

Method property 共sample size 139兲 共sample size 56兲 共sample size 165兲 共sample size 32兲 共sample size 197兲
Faza and GangaRao 共1992兲 Geometric mean 0.64 0.83 0.70 0.68 0.70
Variation 1.21 0.53 1.02 1.29 1.06
95% confidence 共⫺兲 0.56 0.74 0.63 0.51 0.63
95% confidence 共⫹兲 0.73 0.93 0.78 0.91 0.77
Benmokrane et al. 共1996兲 Geometric mean 0.50 0.51 0.43 1.19 0.50
Variation 1.24 1.02 0.82 1.65 1.17
95% confidence 共⫺兲 0.44 0.42 0.39 0.85 0.45
95% confidence 共⫹兲 0.57 0.61 0.47 1.67 0.56
Brown and Bartholomew 共1996兲 Geometric mean 2.75 2.67 2.28 6.83 2.72
Variation 1.28 1.14 0.86 1.67 1.23
95% confidence 共-兲 2.40 2.19 2.07 4.86 2.43
95% confidence 共⫹兲 3.15 3.26 2.51 9.60 3.04
Toutanji and Saafi 共2000兲 Geometric mean 2.56 2.51 2.13 6.34 2.54
Variation 1.27 1.14 0.86 1.68 1.22
95% confidence 共⫺兲 2.23 2.06 1.94 4.51 2.27
95% confidence 共⫹兲 2.93 3.06 2.34 8.92 2.84
ISIS M03-01 共2001兲 Geometric mean 0.46 0.69 0.50 0.60 0.51
Variation 1.01 0.70 0.84 1.60 0.97
95% confidence 共⫺兲 0.41 0.60 0.46 0.43 0.46
95% confidence 共⫹兲 0.52 0.79 0.55 0.84 0.56
CSA S806-02 共2002兲 Geometric mean 0.50 0.69 0.55 0.56 0.55
Variation 1.09 0.51 0.90 1.34 0.97
95% confidence 共⫺兲 0.44 0.62 0.50 0.42 0.50
95% confidence 共⫹兲 0.57 0.77 0.61 0.75 0.60
ACI 440.1R-03 共2003兲 Geometric mean 1.94 2.53 1.75 5.23 2.09
Variation 1.28 1.14 0.86 1.79 1.25
95% confidence 共⫺兲 1.69 2.07 1.59 3.67 1.87
95% confidence 共⫹兲 2.22 3.09 1.92 7.46 2.34
Yost et al. 共2003兲 Geometric mean 1.01 1.38 1.01 1.80 1.11
Variation 1.16 0.81 0.89 1.71 1.09
95% confidence 共⫺兲 0.89 1.18 0.92 1.27 1.00
95% confidence 共⫹兲 1.15 1.61 1.11 2.54 1.23
Proposed ACI 440.1R 共2004兲 Geometric mean 1.11 1.22 1.20 0.93 1.15
Variation 1.26 0.65 1.00 1.52 1.09
95% confidence 共⫺兲 0.97 1.07 1.08 0.68 1.04
95% confidence 共⫹兲 1.27 1.39 1.33 1.28 1.27

Trend lines for the log of the deflection ratio versus the It was previously found by the ACI 440.1R 共ACI 2004兲
modulus of elasticity for CFRP-RC were plotted at a load of that the accuracy of the predicted deflection of FRP-RC depends
1.1Pcr and are shown in Fig. 8. This figure shows that the on the relative reinforcement ratio. Fig. 9 plots the log of experi-
trendlines for the Proposed ACI 440.1R 共ACI 2004兲, Faza and mental over calculated deflection for over-reinforced members.
GangaRao 共1992兲, ISIS M03-01 共Rizkalla and Mufti 2001兲, and It clearly shows that all of the methods have varying accuracy
CSA S806-02 共CSA 2002兲 are all parallel and become more depending on the relative reinforcement ratio. Given that mem-
conservative with increasing modulus of elasticity of FRP. bers typically have reinforcement ratios that are less than
However, only the latter three are conservative over the entire 2.5 times the balanced reinforcement ratio, the method given by
range of the modulus of elasticity for the CFRP. Table 4 shows Yost et al. 共2003兲 is performing most accurately for beams which
that the formula given by Faza and GangaRao 共1992兲 is the most are over-reinforced. These results are verified in Table 4. The
accurate for predicting the deflection of CFRP-RC beams at method proposed by Yost et al. 共2003兲 is the most accurate
this load level. The formula given by the Proposed ACI 440.1R and consistently conservative at reinforcement ratios of up to
共ACI 2004兲 is less accurate at this load level and gives deflection 2.5␳bal. The CSA S806-02 共CSA 2002兲, ISIS M03-01 共Rizkalla and
predictions that are not conservative. The formula presented by Mufti 2001兲, and Faza and Ganga Rao 共1992兲 formulas, are con-
Faza and Ganga Rao 共1992兲 on the other hand is consistently servative for nearly the entire range of reinforcement ratios, and
conservative. become increasingly accurate as the reinforcement ratio increases.

JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / MAY/JUNE 2006 / 189

 Fig. 5. Effect of slenderness on service load deflection Fig. 8. Effect of modulus of elasticity on deflection for CFRP-RC
at 1.1Pcr

Fig. 6. Effect of relative reinforcement ratio on service load Fig. 9. Effect of relative reinforcement ratio on deflection for
deflection overreinforced members at 1.1Pcr

Fig. 7. Effect of modulus of elasticity on deflection for GFRP-RC Fig. 10. Effect of relative reinforcement ratio on deflection for
at 1.1Pcr underreinforced members at 1.1Pcr

190 / JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / MAY/JUNE 2006

Table 5. Further Analysis of Predicted Deflections at 1.1Pcr
Experimental/calculated deflection

GFRP-RC CFRP-RC
Statistical ␳ ⬎ ␳bal ␳ ⬍ ␳bal ␳ ⬎ ␳bal ␳ ⬍ ␳bal
Method property 共sample size 115兲 共sample size 24兲 共sample size 48兲 共sample size 8兲
Faza and GangaRao 共1992兲 Geometric mean 0.65 0.61 0.81 0.94
Variation 1.15 1.52 0.55 0.35
95% confidence 共⫺兲 0.57 0.42 0.72 0.76
95% confidence 共⫹兲 0.75 0.88 0.92 1.16
Benmokrane et al. 共1996兲 Geometric mean 0.43 1.06 0.42 1.69
Variation 0.90 1.93 0.67 0.60
95% confidence 共⫺兲 0.38 0.69 0.36 1.22
95% confidence 共⫹兲 0.48 1.63 0.49 2.34
Brown and Bartholomew 共1996兲 Geometric mean 2.33 6.07 2.16 9.69
Variation 0.92 1.95 0.74 0.67
95% confidence 共⫺兲 2.07 3.94 1.85 6.79
95% confidence 共⫹兲 2.63 9.36 2.53 13.82
Toutanji and Saafi 共2000兲 Geometric mean 2.17 5.62 2.03 9.06
Variation 0.92 1.94 0.74 0.69
95% confidence 共⫺兲 1.93 3.65 1.74 6.30
95% confidence 共⫹兲 2.44 8.65 2.37 13.03
ISIS M03-01 共2001兲 Geometric mean 0.45 0.48 0.63 1.18
Variation 0.87 1.69 0.66 0.43
95% confidence 共⫺兲 0.40 0.32 0.55 0.92
95% confidence 共⫹兲 0.50 0.71 0.73 1.51
CSA S806-02 共2002兲 Geometric mean 0.50 0.48 0.66 0.88
Variation 1.00 1.52 0.51 0.37
95% confidence 共⫺兲 0.44 0.33 0.59 0.71
95% confidence 共⫹兲 0.57 0.69 0.74 1.09
ACI 440.1R-03 共2003兲 Geometric mean 1.64 4.28 2.03 9.56
Variation 0.92 1.96 0.72 0.62
95% confidence 共⫺兲 1.46 2.77 1.74 6.84
95% confidence 共⫹兲 1.85 6.61 2.37 13.36
Yost et al. 共2003兲 Geometric mean 0.93 1.47 1.19 3.28
Variation 0.97 1.87 0.62 0.54
95% confidence 共⫺兲 0.82 0.96 1.04 2.43
95% confidence 共⫹兲 1.05 2.24 1.36 4.42
Proposed ACI 440.1R 共2004兲 Geometric mean 1.18 0.81 1.19 1.36
Variation 1.13 1.76 0.68 0.48
95% confidence 共⫺兲 1.03 0.54 1.03 1.04
95% confidence 共⫹兲 1.35 1.22 1.38 1.78

When a similar plot for under-reinforced members is examined the over-reinforced from the under-reinforced members, rein-
in Fig. 10, it appears that the formula of the Proposed ACI 440.1R forced with either GFRP or CFRP. In Table 4, the formulas
共ACI 2004兲 is quite precise at calculating the deflection at 1.1Pcr presented by Yost et al. 共2003兲 and the Proposed ACI 440.1R
for under-reinforced members, regardless of the relative rein- 共ACI 2004兲 give satisfactory results for calculating the deflection
forcement ratio. This is verified by Table 4, which shows that of GFRP-RC beams. The performance of each of these equations
the formula is indeed accurate, with an average experimental with respect to the reinforcement ratio was investigated, as shown
over predicted deflection ratio of 0.93. However, the variance for in Table 5. The Yost et al. 共2003兲 equation is more precise
this equation is high so the 95% confidence interval becomes for over-reinforced beams, while the formula of the Proposed ACI
unconservative. 440.1R 共ACI 2004兲 is more accurate for under-reinforced beams.
It has been shown that the accuracies of the deflection It can also be seen that the equation by Benmokrane et al. 共1996兲
formulas are dependent on both the relative reinforcement ratio is, on average, very accurate for predicting the deflection of over-
as well as on the type of reinforcement used. For this reason, reinforced GFRP-RC members at 1.1Pcr. The benefit of the equa-
a further statistical analysis has been completed separating tion of the Proposed ACI 440.1R is that it is conservative

JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / MAY/JUNE 2006 / 191

Table 6. Analysis of Predicted Deflections at Elevated Loads
Experimental/calculated deflection

GFRP-RC CFRP-RC
␳ ⬎ ␳bal 共sample size 115兲 ␳ ⬍ ␳bal 共sample size 24兲 ␳ ⬎ ␳bal 共sample size 48兲 ␳ ⬍ ␳bal 共sample size 8兲
Statistical LL 50 LL 80 LL 50 LL 80 LL 50 LL 80 LL 50 LL 80
Method property 共%PmaxExp兲 共%PmaxExp兲 共%PmaxExp兲 共%PmaxExp兲 共%PmaxExp兲 共%PmaxExp兲 共%PmaxExp兲 共%PmaxExp兲
Faza and GM 1.08 1.17 1.40 1.24 1.13 1.19 2.66 1.14
GangaRao Variation 0.30 0.20 1.52 0.57 0.31 0.26 3.23 0.20
共1992兲 95% conf. 共⫺兲 1.03 1.13 0.97 1.04 1.05 1.11 0.98 1.00
95% conf. 共⫹兲 1.13 1.21 2.03 1.49 1.22 1.27 7.23 1.29
Benmokrane GM 0.71 0.87 1.44 1.13 0.80 0.94 2.88 1.14
et al. Variation 0.29 0.15 1.51 0.77 0.36 0.28 3.01 0.48
共1996兲 95% conf. 共⫺兲 0.68 0.85 1.00 0.90 0.73 0.88 1.10 0.87
95% conf. 共⫹兲 0.74 0.89 2.08 1.42 0.87 1.01 7.54 1.50
Brown and GM 1.02 1.05 2.64 1.95 1.02 1.12 5.29 2.22
Bartholomew Variation 0.37 0.17 1.53 1.74 0.31 0.28 2.15 1.26
共1996兲 95% conf. 共⫺兲 0.96 1.02 1.82 1.30 0.94 1.04 2.39 1.26
95% conf. 共⫹兲 1.08 1.08 3.83 2.92 1.10 1.20 11.72 3.91
Toutanji GM 0.93 1.02 2.28 1.62 0.97 1.11 4.89 1.88
and Saafi Variation 0.34 0.16 1.61 1.56 0.33 0.28 2.40 1.14
共2000兲 95% conf. 共⫺兲 0.88 0.99 1.55 1.11 0.89 1.04 2.09 1.11
95% conf. 共⫹兲 0.98 1.05 3.35 2.36 1.05 1.19 11.42 3.19
ISIS M03-01 GM 0.82 1.03 1.16 0.85 0.98 1.13 2.73 1.10
共2001兲 Variation 0.31 0.16 1.76 0.38 0.35 0.28 3.14 0.26
95% conf. 共⫺兲 0.78 1.00 0.77 0.75 0.90 1.05 1.02 0.94
95% conf. 共⫹兲 0.86 1.06 1.74 0.97 1.07 1.21 7.31 1.29
CSA S806-02 GM 0.80 1.01 1.08 0.82 0.95 1.11 2.46 0.98
共2002兲 Variation 0.30 0.16 1.86 0.37 0.36 0.28 3.52 0.22
95% conf. 共⫺兲 0.76 0.98 0.71 0.72 0.87 1.04 0.86 0.85
95% conf. 共⫹兲 0.84 1.04 1.64 0.93 1.04 1.19 7.00 1.12
ACI 440.1R-03 GM 1.21 1.18 3.01 2.59 1.28 1.23 6.41 3.39
共2003兲 Variation 0.36 0.22 1.16 1.55 0.33 0.26 1.44 1.29
95% conf. 共⫺兲 1.14 1.14 2.21 1.78 1.18 1.15 3.45 1.91
95% conf. 共⫹兲 1.28 1.22 4.10 3.77 1.39 1.31 11.89 6.02
Yost et al. GM 0.96 1.08 1.76 1.42 1.09 1.16 3.94 1.73
共2003兲 Variation 0.29 0.17 1.32 0.81 0.31 0.26 2.18 0.67
95% conf. 共⫺兲 0.92 1.05 1.26 1.12 1.01 1.09 1.77 1.21
95% conf. 共⫹兲 1.01 1.11 2.46 1.80 1.18 1.24 8.78 2.47
Proposed GM 1.03 1.10 1.42 1.04 1.08 1.16 2.84 1.11
ACI 440.1R Variation 0.34 0.18 1.52 0.48 0.31 0.26 3.00 0.24
共2004兲 95% conf. 共⫺兲 0.98 1.07 0.98 0.89 1.00 1.09 1.09 0.96
95% conf. 共⫹兲 1.09 1.13 2.06 1.22 1.17 1.24 7.42 1.29
Note: LL⫽load level; GM⫽geometric mean; and conf.⫽confidence.

at this load level. Table 5 shows that the Faza and GangaRao load levels, where tension stiffening does not play an important
共1992兲 formula is quite accurate at forecasting the deflection of role, and the level of conservatism in these methods decreases.
CFRP-RC members, while the CSA S806-02 共CSA 2002兲 gives
satisfactory performance for under-reinforced CFRP-RC. This
Analysis at Elevated Load Levels
indicates that moment–curvature based methods perform well
for CFRP-RC, mainly due to the larger stiffness of the beams Table 6 shows the statistical evaluation of the accuracy of each of
after cracking compared to GFRP-RC. It has to be pointed out the formulas at both 50 and 80% of the ultimate experimental
that the method of the CSA S806-02 共CSA 2002兲 does not loads for both over-reinforced and under-reinforced members,
incorporate tension stiffening into its model. It is assumed that reinforced with either GFRP or CFRP. A quick look at this table
if tension stiffening was added to the CSA S806-02 共CSA 2002兲, shows that almost all of the formulas are fairly precise at these
the results would improve dramatically for GFRP-RC beams. load levels. Fig. 2 supports this conclusion and shows that there is
This will become clearer as these beams are analyzed at higher less spread between the estimates of each method at elevated

192 / JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / MAY/JUNE 2006

Table 7. Suggested Methods for Deflection Calculation for Load Level of 1.1Pcr
GFRP-RC CFRP-RC
␳ ⬍ ␳bal ␳ ⬎ ␳bal ␳ ⬍ ␳bal ␳ ⬎ ␳bal
Proposed ACI 共2004兲 Yost 共2003兲 Faza and GangaRao 共1992兲 Faza and GangaRao 共1992兲

loads, and moreover each method converges to a ratio of 1.0 at worked well for all members and therefore results were
these elevated loads. However, in Table 6 it can be seen that none tabulated in Table 7.
of the methods were able to accurately calculate the deflection of 6. At loads of 50%PMaxExp and 80%PMaxExp, almost all of
the under-reinforced CFRP-RC members at a load of 50% of the the formulas performed accurately. However, none of the
experimental failure load. This may be due to the small number of methods was able to accurately calculate the deflection of
members present in this category 共eight兲 and should thus not be of the under-reinforced CFRP-RC members at a load of 50%
major concern at this instance. However, future testing of these of the experimental failure load. Further research is needed
members is recommended to improve the statistical evaluation in this area as the database had a very limited number of
of the formulas. these members.
This paper tried to point out that there is a critical need for
reliability analysis of FRP code equations. It became clear during
Conclusions the study that with an increasing number of samples in the data-
base, results change dramatically. If the number of samples is
A database of 197 beams and slabs was analyzed in order to find small, this can potentially lead to incorrect recommendations for
an accurate, and at the same time, conservative equation to future codes. It also became apparent that it is a strenuous task to
estimate deflections of FRP-RC. Beams reinforced with carbon try to characterize FRP-RC beams with only one code formula
FRP 共CFRP兲, GFRP, and aramid FRP 共AFRP兲 were found in this and assume that the results are conservative for all possible cases.
database, however the majority was GFRP-RC beams. The CSA An approach of using separate formulas for under- and over-
A23.3-94 共CSA 1998兲 was used for consistency to calculate reinforced beams, or beams reinforced with GFRP and CFRP may
the modulus of elasticity and cracking moment for all beams need to be considered in the future to provide designers with
and slabs, since they were not always provided by the original conservative estimates of such a critical value as service load
investigators. The work presented in this paper supports the deflection. Future formulas need to be developed based on the
following conclusions. mechanical principles that apply to FRP-RC rather than on the
1. The accuracy of the deflection equations improved when limited number of experimental data.
the moment due to the self weight of the beam was included
in the analysis. This lowered the predicted cracking moment,
when compared to an analysis not including the self-weight.
2. The accuracy of the deflection ratios was highly dependent Notation
on the accuracy of the calculated cracking moment. The
primary variable here was the modulus of rupture. Upon The following symbols are used in this paper:
performing a statistical analysis on various equations for a ⫽ shear span 共mm兲;
predicting the modulus of rupture, it was found that both the b ⫽ member width 共mm兲;
CSA A23.3-94 共CSA 1998兲 and CEB-FIP MC-90 共CEB-FIP d ⫽ effective depth 共mm兲;
1990兲 codes provided accurate equations. EC ⫽ modulus of elasticity of concrete 共MPa兲;
3. The accuracy of the deflection equations varies at different EF ⫽ modulus of elasticity of FRP 共MPa兲;
loads. Given that the service conditions are only explicitly EFRP ⫽ modulus of elasticity of FRP 共MPa兲;
stated in ISIS M03-01 共Rizkalla and Mufti 2001兲, an analysis Es ⫽ modulus of elasticity of steel 共MPa兲;
was performed at various loads to encompass the entire load f ⬘c ⫽ concrete compressive strength 共MPa兲;
range, including the service load levels based on ISIS f fu ⫽ ultimate strength of FRP reinforcement 共MPa兲;
M03-01 共Rizkalla and Mufti 2001兲. This will allow the f r ⫽ modulus of rupture of concrete 共MPa兲;
designer to choose the deflection formula which works best h ⫽ member height 共mm兲;
at their chosen service conditions. Results show that at Icr ⫽ moment of inertia of cracked section transformed to
higher loads, most of the methods perform similarly well. concrete 共mm4兲;
However, at lower loads, major discrepancies were found Ie ⫽ effective moment of inertia 共mm4兲;
between individual methods. Ig ⫽ moment of inertia of gross section 共mm4兲;
4. At the service load level, the formula proposed by Yost et al. IT ⫽ moment of inertia of uncracked section transformed
共2003兲 was sufficiently accurate and conservative for predict- to concrete 共mm4兲;
ing the deflection of GFRP-RC members, while the formulas L ⫽ span of beam 共mm兲;
of the CSA S806-02 共CSA 2002兲 and ISIS M03-01 共Rizkalla Lg ⫽ uncracked length of beam 共mm兲;
and Mufti 2001兲 were the most accurate and conservative for M a ⫽ applied moment 共N mm兲;
predicting the deflection of CFRP-RC members. The slender- M cr ⫽ cracking moment 共N mm兲;
ness ratio was also found to affect most of the deflection M / EI ⫽ curvature 共mm−1兲;
ratios. m ⫽ moment due to unit load applied at point where
5. At a load of 1.1Pcr, it was found that both the relative rein- deflection is to be calculated;
forcement ratio and the type of reinforcement affect the n ⫽ number of members in sample set;
accuracy of each of the deflection equations. No one formula Pcr ⫽ cracking load 共N兲;

JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / MAY/JUNE 2006 / 193

PMaxExp⫽ experimental failure load 共N兲; Canadian Standards Association 共CSA兲. 共2002兲. “Design and construction
␣ ⫽ bond dependent coefficient; of building components with fibre-reinforced polymers.” Concrete
␤ ⫽ reduction coefficient; design handbook: Canadian Standard S806-02, CSA, Toronto, 177.
␥sw ⫽ density of concrete; Choi, B. S., Scanlon, A., and Johnson, P. A. 共2004兲. “Monte Carlo
␦ ⫽ immediate deflection under given load 共mm兲; simulation of immediate and time-dependent deflections of reinforced
␦max ⫽ maximum deflection along length of member under concrete beams and slabs.” ACI Struct. J., 101共5兲, 633–641.
given load 共mm兲; Comite Euro-International du Beton—Federation International de la
␭ ⫽ factor accounting for concrete density. For normal Precontrainte 共CEB-FIP兲. 共1990兲. Model code for concrete structures
density concrete ␭⫽1.00; 共MC-90兲, Thomas Telford, London.
␮y ⫽ mean of ratios in natural log scale; Faza, S. S., and GangaRao, H. V. S. 共1992兲. “Pre- and post-cracking
␳bal ⫽ balanced reinforcement ratio; deflection behavior of concrete beams reinforced with fibre-reinforced
␳FRP ⫽ reinforcement ratio; and plastic rebars.” Proc., Advanced Composite Materials in Bridges and
␴y ⫽ standard deviation of ratios in natural log scale. Structures, Canadian Society for Civil Engineering, Sherbrooke, Que.,
Canada, 151–160.
Ghali, A., Hall, T., and Bobey, W. 共2001兲. “Minimum thickness of
concrete members reinforced with fibre reinforced polymer bars.”
References Can. J. Civ. Eng., 28, 583–592.
Mota, C. 共2005兲. “Flexural and Shear Behavior of FRP-RC Members.”
American Concrete Institute 共ACI兲. 共1999兲. “Building code requirements MSc thesis, Univ. of Manitoba, Winnipeg, Man., Canada.
for structural concrete.” Standard 318–99, ACI, Farmington Hills, Razaqpur, A. G., Svecova, D., and Cheung, M. S. 共2000兲. “Rational
Mich. method for calculating deflection of fiber-reinforced polymer
American Concrete Institute 共ACI兲. 共2003兲. Guide for the design and reinforced beams.” ACI Struct. J., 97共1兲, 175–184.
construction of concrete reinforced with FRP bars, 440.1R-03, Reda Taha, M. M., and Hassanain, M. A. 共2003兲. “Estimating the error
ACI Committee 440, Farmington Hills, Mich., 42. in calculating deflections of HPC slabs: A parametric study using
American Concrete Institute 共ACI兲. 共2004兲. Guide for the design and the theory of error propagation.” Deflection control for the future:
construction of concrete reinforced with FRP bars, Proposed ACI Special Publication SP-210, J. Gardner, ed., American Concrete
revisions, ACI Committee 440, Farmington Hills, Mich, 35. Institute, Detroit, 65–92.
Benmokrane, B., Chaallal, O., and Masmoudi, R. 共1996兲. “Flexural Rizkalla, S., and Mufti, A. 共2001兲. Reinforcing concrete structures
with fibre reinforced polymers—Design manual No. 3, ISIS Canada,
response of concrete beams reinforced with FRP reinforcing bars.”
Winnipeg, Man., Canada.
ACI Struct. J., 91共2兲, 46–55.
Theriault, M., and Benmokrane, B. 共1997兲. “Theoretical and experimental
Branson, D. E. 共1965兲. “Instantaneous and time-dependent deflections
investigation on crack width, deflection and deformability of concrete
of simple and continuous reinforced beams.” HPR Rep. No. 7,
beams reinforced with FRP rebars.” Proc., Annual Conf. of the
Part 1, Alabama Highway Department, Bureau of Public Roads,
Canadian Society for Civil Engineering, Canadian Society for Civil
Montgomery, Ala.
Engineering, Sherbrooke, Quebec, Canada, 151–160.
Brown, V. L., and Bartholomew, C. L. 共1996兲. “Long-term deflections
Toutanji, H. A., and Saafi, M. 共2000兲. “Flexural behavior of concrete
of GFRP-reinforced concrete beams.” Proc., ICCI, Tucson, Ariz., beams reinforced with glass fiber-reinforced polymer 共GFRP兲 bars.”
389–400. ACI Struct. J., 97共5兲, 712–719.
Canadian Standards Association 共CSA兲. 共1998兲. Concrete design hand- Yost, J. R., Gross, S. P., and Dinehart, D. W. 共2003兲. “Effective moment
book: Canadian Standard A23.3-94, 2nd Ed., CSA, Rexdale, Ont., of inertia for glass fiber-reinforced polymer-reinforced concrete
Canada. beams.” ACI Struct. J., 100共6兲, 732–739.

194 / JOURNAL OF COMPOSITES FOR CONSTRUCTION © ASCE / MAY/JUNE 2006