Engineering Structures 174 (2018) 629–645

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Behavior of large-scale FRP-confined rectangular RC columns under axial T

compression
⁎
J.J. Zenga,b, G. Linb, J.G. Tengb, , L.J. Lia
a
School of Civil and Transportation Engineering, Guangdong University of Technology, Guangzhou 510006, China
b
Department of Civil and Environmental Engineering, The Hong Kong Polytechnic University, Hong Kong 999077, China

A R T I C LE I N FO A B S T R A C T

Keywords: Fiber-reinforced polymer (FRP) jacketing has become an attractive technique for strengthening/retrofitting re-
Fiber-reinforced polymer (FRP) inforced concrete (RC) columns. Extensive research has been conducted on FRP-confined rectangular columns
Confinement under axial compression, leading to a significant number of stress-strain models for FRP-confined concrete in
Rectangular column these columns. However, most of these models have been developed based on test results of small-scale columns,
Reinforced concrete (RC)
so their applicability to large FRP-confined rectangular RC columns has yet to be properly validated. To this end,
Stress-strain behavior
the present paper first presents the test results of an experimental study consisting of nine large-scale rectangular
RC columns, including eight FRP-confined RC columns and one RC column without FRP jacketing as the control
specimen, tested under axial compression. The experimental program examined the sectional corner radius and
the FRP jacket thickness as the key test variables. Five representative design-oriented stress-strain models for
FRP-confined concrete in rectangular columns, identified from critical reviews of the existing literature, are then
assessed using the test results to examine their validity for these large-scale columns.

1. Introduction an FRP jacket [23].

Fiber-reinforced polymer (FRP) jacketing has become a widely ac-
cepted technique for strengthening/retrofitting reinforced concrete
(RC) columns [51,6]. Extensive research has been conducted on FRP-
confined concrete columns aiming at gaining an in-depth understanding
of the confining mechanism in FRP-confined concrete. While existing
studies [51,22,52] have demonstrated that FRP confinement can sub-
stantially enhance both the compressive strength and ductility of con-
fined concrete in circular columns, the same method has been found to
be much less effective for rectangular columns (e.g., [35,47,23,13]).
Corner rounding is generally recommended to enhance the confinement
effectiveness in a rectangular column and to reduce the detrimental
effect of sharp corners on the rupture strain of the FRP jacket. The lower
FRP confinement effectiveness in a rectangular column is mainly at-
tributed to the non-uniform FRP confinement around the column,
whereas the concrete in an FRP-confined circular column is nominally
uniformly confined. In an FRP-confined rectangular section, the flat
sides of the FRP jacket are subjected to bending, to which the jacket has
little resistance due to its negligible flexural rigidity; as a result, the
concrete near the flat sides receives the lowest confinement, while that
in the four corner regions receives the highest confinement [29].
Therefore, only part of the rectangular section is effectively confined by
A significant number of experimental studies have been conducted
on FRP-confined rectangular concrete columns (e.g.,
[47,23,11,30,64,40,67,19,38,61,62]), leading to many stress-strain
models for FRP-confined concrete in such columns (e.g.,
[23,63,64,67,17,62,27]). However, most of those experimental studies
have been conducted on small-scale plain concrete columns; only a
limited number of those studies have been concerned with large-scale
RC columns (e.g., [59,44,53,13]). FRP jacketing has also been found to
be effective to enhance the seismic performance of large-scale rectan-
gular RC columns (e.g., [48,7,39,16,55,58]). The lack of a reliable
stress-strain model for FRP-confined concrete in large rectangular col-
umns also makes it difficult to accurately predict the seismic perfor-
mance of such columns. The existing experimental results indicate that
there exists a significant behavioral difference between small and large
rectangular columns, which has been referred to as the column size
effect (e.g., [42,34,9,45,46,60,13,57]). This size effect, however, has
been found to be negligible for FRP-confined circular RC columns
[35,33,72,50]. As a result, significant uncertainty exists with the ap-
plicability to large columns of existing stress-strain models for FRP-
confined concrete in rectangular columns developed on the basis of
studies on small-scale columns.
Against this background, the present paper presents the test results

⁎
Corresponding author.
E-mail address: cejgteng@polyu.edu.hk (J.G. Teng).

https://doi.org/10.1016/j.engstruct.2018.07.086
Received 13 November 2017; Received in revised form 26 July 2018; Accepted 29 July 2018
Available online 07 August 2018
0141-0296/ © 2018 Elsevier Ltd. All rights reserved.
J.J. Zeng et al. Engineering Structures 174 (2018) 629–645

of an experimental study that included nine large-scale rectangular RC
column specimens, including eight FRP-confined RC columns and one
un-confined RC column, tested under axial compression. The experi-
mental program included the sectional corner radius and the FRP jacket
thickness as the key test variables. The test results are then used to
assess five existing design-oriented stress-strain models for FRP-con-
fined concrete in rectangular columns. It should be noted that, in this
paper, “axial stress-axial strain” is simply referred to as “stress-strain”
unless otherwise specified.
2. Experimental program
2.1. Test specimens
Nine rectangular RC columns were cast in two batches and tested.
All the columns had a cross-section of 435 mm in depth and 290 mm in
width (h × b = 435 mm × 290 mm) and a column length of 1300 mm.
The sharp corners of columns were rounded into a corner radius (rc ) of
25 mm, 45 mm or 65 mm before FRP jacketing, leading to three corner
radius ratios (2rc / h ) of around 1/9, 1/5 and 2/7. Apart from the corner
radius ratio, the experimental program included the FRP jacket thick-
ness as another test variable. The detailed information of the test col-
umns is given in Table 1.
The columns were longitudinally reinforced with 20-mm-diameter
deformed steel bars and transversely reinforced with 8-mm-diameter
round bars in the form of steel ties. The columns in Batch 1 were re-
inforced with ten longitudinal steel bars, while those in Batch 2 were
reinforced with six longitudinal steel bars, corresponding to long-
itudinal steel reinforcement ratios of 2.49% and 1.49%, respectively
(Fig. 1). A smaller longitudinal steel reinforcement ratio was adopted
for Batch 2 to reduce the small additional confinement on the concrete
from longitudinal steel bars (e.g., [49,31,1]). Nevertheless, for both
longitudinal steel reinforcement ratios used in the present study, the
additional confinement as estimated using the model of Sheikh and
Uzumeri [49] is much smaller than the confinement provided by the
FRP jacket at the ultimate condition for the present tests. It is thus
believed that the use of two different longitudinal steel reinforcement
ratios had little implication for understanding the behavior of confined
concrete in these columns.
The longitudinal steel bars were welded onto a 30-mm-thick steel
plate at each end. Each column was designed with a test portion of
700 mm in length in the middle region where failure was expected to
occur, and a loading portion of 300 mm in length at each end (Fig. 1.
The spacing of steel ties was 100 mm in the loading portions but much
larger in the test portion to minimize the confinement effect from the
steel ties. However, the tie spacing should not exceed 16 times the
longitudinal bar diameter (320 mm) according to ACI 318 [1]. There-
fore, the tie spacing arrangement of 200 mm-300 mm-200 mm was
adopted in the test portion. The steel ties were bent to 135° hooks at the
ends (Fig. 1). The concrete cover thickness, measured from the centers
of the longitudinal steel bars to the concrete surface, was 35 mm. All the
columns were cast in wooden moulds, demoulded on Day 2 and cured
for 28 days or more in the laboratory environment before FRP jack-
eting.
Eight of the nine columns were wrapped with carbon FRP (CFRP)
jackets, with fibers oriented only in the hoop direction (Table 1). The
jacket was formed in a wet lay-up process, in which the carbon fiber
sheet was impregnated with epoxy resin and then wrapped around the
column. A 300-mm-long overlapping zone in the hoop direction was
adopted to avoid FRP debonding of the wrapping end. The overlapping
zone was horizontally centered at one of the longer sides of the column
section. The FRP-wrapped columns were further cured in the laboratory
environment for a period much longer than 7 days until testing.
Each column specimen was given a name in the following format:
RmLrn, where “R” denotes a rectangular column, m is the number of
FRP layers (plies), and n is the radius of the rounded corners. For ex-
ample, “R1Lr25” refers to a rectangular column with a corner radius of
25 mm wrapped with a one-layer CFRP jacket, and “R0Lr25” refers to
the control RC column without FRP wrapping. To avoid unexpected
failure outside the testing portion, an additional layer of CFRP of
200 mm in height was wrapped near each end of the column.
2.2. Material properties
Ready-mixed concrete from a local supplier was used in casting the
columns. Crushed granite with a maximum nominal size of 20 mm was
used as the coarse aggregate. The concrete slump was around 150 mm.
Several standard concrete cylinders (150 mm in diameter and 300 mm
in height) were cast and tested under axial compression at the time of
testing each column to determine the unconfined concrete properties
(e.g., compressive strength fc′ and axial strain at peak axial stress εco )
following ASTM C469 [4]. Tensile tests on three steel bar specimens
were conducted to determine properties of each type of steel bars (i.e.,
deformed bars and round bars) according to BS 18 [8]. The yield stress,
tensile strength, and elastic modulus of the 20-mm-diameter deformed
steel bars were 491.4 MPa, 602.8 MPa and 199.5 GPa, respectively. The
corresponding values of the 8-mm-diameter round bars were
380.8 MPa, 448.4 MPa and 187.3 GPa, respectively.
Unidirectional high tensile strength carbon fiber sheets, with a
nominal layer thickness of 0.334 mm, were used to form the FRP
jackets. The average modulus of elasticity, tensile strength and rupture
strain of single-layer CFRP were found to be 245.6 GPa, 3993.3 MPa
and 1.71%, respectively, based on the results of three FRP flat coupons
tested following ASTM D3039 [5].
It is well known that the hoop rupture strain of an FRP jacket in a
column (εh, rup ) is significantly lower than that from FRP coupon tensile
tests (εf ) [24]. Lam and Teng [22] found an average value of 0.586 for
the FRP strain efficiency factor (εh, rup/ εf ) for CFRP jackets in circular
concrete columns. For FRP-confined rectangular concrete columns, this
strain efficiency factor is still needed as the analysis of such columns
requires some information from an equivalent FRP-confined circular

Table 1
Key information of test columns and test results.
Specimen Corner radius r fc′ ′
fco Ec εco ρl tf ′
fcc ′ / fco
fcc ′ εcc εcc / εco εh, max Batch
(mm) (MPa) (MPa) (GPa) (%) (mm) (MPa)

R1Lr25 25 43.4 40.8 32.0 0.00249 2.50 0.334 46.5 1.14 0.0094 3.76 0.0164 1
R2Lr25 25 37.4 35.2 28.0 0.00251 2.50 0.668 42.1 1.20 0.0143 5.72 0.0120
R0Lr25 25 42.2 39.6 30.7 0.00252 2.50 0 39.6 1.00 0.0025 1.00 N.A.
R2Lr45 45 34.1 32.1 26.9 0.00250 1.52 0.668 42.2 1.32 0.0216 8.64 0.0130 2
R2Lr65 65 34.1 32.1 26.9 0.00250 1.54 0.668 44.9 1.40 0.0230 9.20 0.0133
R4Lr45 45 30.8 28.9 26.3 0.00250 1.52 1.336 45.2 1.56 0.0248 9.92 0.0137
R4Lr65 65 30.8 28.9 26.3 0.00250 1.54 1.336 51.1 1.77 0.0262 10.48 0.0107
R6Lr45 45 34.1 32.1 26.9 0.00250 1.52 2.004 63.9 1.99 0.0387 15.48 0.0162
R6Lr65 65 34.1 32.1 26.9 0.00250 1.54 2.004 68.4 2.13 0.0437 17.48 0.0108

Note: “N.A.” – Not applicable.

630
J.J. Zeng et al. Engineering Structures 174 (2018) 629–645

Fig. 1. Geometry and reinforcement details of test columns.

concrete column [23]. For the determination of the strain efficiency Code for Infrastructure Application of FRP Composites [15], three of
factor for this particular type of CFRP, six FRP-confined standard con- them had a one-layer CFRP jacket while the other three had a two-layer
crete cylinders (150 mm in diameter and 300 mm in height) were tested CFRP jacket. Existing studies have shown that the FRP thickness has a
under axial compression as recommended by the Chinese Technical limited effect on the hoop rupture strain of an FRP jacket in concrete

631
J.J. Zeng et al. Engineering Structures 174 (2018) 629–645

erl
ap
SG6
Ov

SG1,7
SG5

SG2
SG4,8

SG3
Fig. 2. Layout of strain gauges on FRP-confined standard circular concrete
cylinders. (a) One-layer FRP jacket

cylinders provided the concrete is sufficiently confined to exhibit a

bilinear axial stress-strain response [24]. Therefore, it is not necessary
for the thickness of the FRP jacket in an accompanying cylinder test to
be identical to that in the rectangular column tests. For each specimen,
eight strain gauges were installed on the FRP jacket at the mid-height
(Fig. 2): six were evenly distributed around the circumference to
measure hoop strains and the other two at 180° apart to measure axial
strains. An overlap length of 150 mm was used in all those cylinders. An
additional 20-mm-wide FRP strip was wrapped near each end of the
cylinder to avoid premature local failure. The unconfined concrete
strength and the axial strain at peak axial stress from three unconfined
concrete cylinder tests were 38.6 MPa and 0.0027, respectively. The
stress-strain curves of these CFRP-confined concrete cylinders are
shown in Fig. 3, and the FRP hoop strains recorded by the strain gauges
at jacket rupture are listed in Table 2. The average FRP hoop strain
(excluding the hoop strain gauges within the overlapping zone) at
jacket rupture is 1.13%, leading to an FRP strain efficiency factor
(εh, rup/ εf ) of 0.661. In Fig. 3, the stress-strain curves from these cylinder
tests are compared with the analysis-oriented stress-strain model of
Jiang and Teng [21] for FRP-confined concrete in circular columns, (b) Two-layer FRP jacket
showing close agreement between them.
Fig. 3. Stress-strain curves from tests on FRP-confined concrete cylinders.

2.3. Experimental set-up and instrumentation
Six linear variable displacement transducers (LVDTs) were used to
monitor the axial shortening of each column (Fig. 4); four of them were
centered on the four sides covering a mid-height length of 300 mm; the
other two were mounted on two opposite sides between the loading
platens at the two ends to monitor the full-height axial shortening.
For each FRP-confined RC column, a large number of strain gauges
were installed to monitor strains in the FRP jacket. The strain gauges
were installed at two levels of the column: (1) the mid-height level (i.e.,
Level 1 in Fig. 4), and (2) 300 mm below the mid-height level (i.e.,
Level 2 in Fig. 4). For each level, the strain gauges were installed at the
mid-width positions of the flat sides, the mid-arc positions of the
rounded corners, and the transition points between each rounded
corner and its adjacent flat sides (i.e., locations with a curvature
change) (Fig. 4). At the mid-width position of a flat side, two strain
gauges were installed to measure both hoop strains and axial strains
(Fig. 4). Totally 20 strain gauges were installed at each level. One strain
gauge was installed on each longitudinal steel bar at Level 1 where a
smooth surface had previously been provided by grinding. Waterproof
coating was subsequently applied before concrete casting to protect the
strain gauges on the steel bars. An automatic data acquisition system
was used to collect the test data, including axial loads, displacements,
and strains.
The columns were tested under axial compression on a testing
machine with a capacity of 10,000 kN (Fig. 5). The specimens were
carefully centered and preloaded with an axial load of around 20% of
the estimated load-carrying capacity of the corresponding unconfined
RC column. The columns were first loaded with a load-controlled rate of
5 kN/s until around 80% of the estimated load-carrying capacity of the
corresponding unconfined RC column; afterwards, a displacement-
controlled rate of 0.72 mm/min was adopted until column failure due
to FRP jacket rupture.
3. Test results and disscussions
3.1. Failure modes
The failure of the control specimen R0Lr25 started with the cracking
of concrete cover in the upper region of the specimen after the attain-
ment of the peak load, followed by the spalling of cover concrete and
buckling of longitudinal steel bars in the mid-height region (Fig. 6(a)).

632
J.J. Zeng et al. Engineering Structures 174 (2018) 629–645

Table 2
FRP hoop rupture strains obtained from FRP-confined concrete cylinders.
Specimen ′ (MPa)
fco ′ (MPa)
fcc SG1 SG2 SG3 SG4 SG5* SG6 Average

C1-1 38.6 71.6 0.0118 0.0120 0.0119 0.0102 0.0090 0.0113 0.0114
C1-2 70.4 0.0125 0.0094 / 0.0112 0.0081 0.0122 0.0113
C1-3 68.0 0.0124 0.0117 0.0111 0.0114 0.0088 / 0.0117
Average rupture strain 0.0115

Specimen SG1 SG2 SG3 SG4 SG5a SG6

C2-1 38.6 116.4 / 0.0118 0.0112 0.0117 0.0103 0.0124 0.0118

C2-2 107.5 0.0094 0.0095 0.0097 0.0096 0.0089 0.0112 0.0099
C2-3 106.0 0.0125 0.0110 0.0097 0.0110 0.0087 0.0127 0.0114
Average rupture strain 0.0110

Note: “/” – strain gauge failed before FRP rupture.

a
Rupture strains in the overlapping zone (SG5) were excluded in calculating the average rupture strains.

All FRP-confined RC columns failed by the explosive rupture of CFRP
jacket due to hoop tension near the mid-height (Fig. 6(b)–(i)). Some
cracking sounds of epoxy rupture were first heard, followed by a sudden
explosive sound of FRP rupture. The explosive sound of FRP rupture
was much louder for columns with a thinner FRP jacket. FRP rupture
occurred at or near one of the rounded corners in all those columns.
Inspection of the columns after the removal of ruptured FRP jacket and
concrete cover revealed that the longitudinal steel bars had buckled.
The buckling deformation of steel bars near a longer side was often
more serious than that near a shorter side or a rounded corner as the
confinement provided by the FRP jacket is smaller at the longer sides
due to the small flexural rigidity of the FRP jacket.
3.2. Stress-strain behavior of confined concrete
The axial stress of concrete over the rectangular section was ob-
tained by dividing the applied axial load resisted by the concrete by the
cross-sectional area of concrete. Due to non-uniform confinement, the
axial stress in such a rectangular column varies over the section, so the
axial stress so obtained is the average axial stress of the rectangular
section. The axial load resisted by the concrete was obtained by de-
ducting the load taken by the longitudinal steel bars from the axial load
resisted by the entire column section. In calculating the axial load taken
by the longitudinal steel bars at a certain axial strain, an elastic-per-
fectly plastic stress-strain response was assumed for the steel bars. The
axial strains were averaged from the readings of the four LVDTs for the
mid-height region. The axial strains obtained from the readings of the
full-height LVDTs were not used as the axial shortenings recorded by
these LVDTs included the movements of components of the loading
system into tight contact with the column. The axial strains obtained
from strain gauges were found to be much smaller than those from
LVDTs due to the local wrinkling of the FRP jacket and thus they were
not used either [54,71].
The stress-strain responses of FRP-confined concrete in the test
columns are shown in Fig. 7. All the stress-strain curves terminate at
FRP rupture due to hoop tension. It can be seen that, except for spe-
cimens R1Lr25 and R2Lr25, all the stress-strain curves have a typical
bilinear shape with an ascending second portion. For specimens R1Lr25

LVDT 3

LVDT 4 LVDT 2

LVDT 1

Level 1
1 1

2 2

LVDT 5 LVDT 6

Level 2

Fig. 4. Locations of strain gauges and LVDTs on FRP-confined rectangular RC columns.

633
J.J. Zeng et al.
Fig. 5. Test set-up.
and R2Lr25, the axial stress fluctuates in the second portion, but the
overall stress-strain curves still feature a bilinear shape. For specimen
R1Lr25, the axial stress decreases slightly in the transition region and
then increases monotonically, but the overall stress-strain curve still has
a bilinear shape. The decrease in the axial stress in the transition region
can be attributed to the relatively low level of FRP confinement in this
column, whose behavior can be expected to be as follows. During the
early loading stage (i.e., before the transition region), the concrete has
not cracked, and the dilation of concrete is small. The flat sides of the
rectangular section remain almost flat during this stage. Due to the
small bending rigidity of the FRP jacket, the confining action of FRP to
the concrete near the flat sides is not significant during this stage.
However, as the dilation of concrete increases, the dilation of concrete
pushes the four flat sides into curved shapes, leading to substantial
curvatures in the FRP jacket on the flat sides and substantial tensile
stresses in the FRP jacket. As a result, the FRP jacket starts to offer
substantial confinement on the flat sides, leading to an ascending stress-
strain response after the transition region.
Fig. 8 shows a comparison of stress-strain curves found from test
columns with different corner radii. In order to eliminate the effect of
different concrete strengths, the axial stresses and strains are normal-
ized by the unconfined concrete strength fco ′ and the corresponding
′ refers to the compressive strength of
axial strain εco , respectively. fco
unconfined concrete in large-scale RC columns, which is discussed in
detail in the next section. Fig. 8(a) shows that the stress-strain curves of
columns R2Lr45 and R2Lr65 with a corner radius of 45 mm and 65 mm
respectively indicate higher strain capacities than that of column
R2Lr25 with a corner radius of 25 mm. Column R2Lr65 had a slightly
larger ultimate axial stress but a slightly lower ultimate axial strain than
those of column R2Lr45. Fig. 8(b) and (c) also show that the stress-
strain curves from the two columns with a corner radius of 45 mm, and
65 mm respectively, are similar, but the latter is slightly larger in both
the ultimate axial stress and the ultimate axial strain. These results
indicate that a corner radius of 25 mm may not be sufficient to effec-
tively activate the FRP confinement for columns of such sizes. Never-
theless, Column R2Lr25, which had a corner radius of 25 mm, still at-
tained an ultimate stress enhancement ratio ( fcc′ / fco′ ) of around 1.2,
634
Engineering Structures 174 (2018) 629–645
which is also valuable in practice. Note that existing design guidelines
all specify a minimum corner radius for a rectangular RC column to be
strengthened with FRP. For example, ACI 440.2R [2] specifies a
minimum corner radius of 0.5 in. (13 mm); Concrete Society [12] spe-
cifies a minimum corner radius of 15 mm; fib (2001) recommends a
minimum corner radius of 15–25 mm; the Chinese and Italian codes
[15]; CNR DT-200 2013) both specify a minimum corner radius of
20 mm; the Turkey seismic code (TSDC 2007) specifies a minimum
corner radius of 30 mm.
Fig. 9 shows a comparison of the normalized stress-strain curves of
FRP-confined concrete in columns confined with different numbers of
layers of FRP (i.e., different FRP confinement levels). It is obvious that
both the ultimate axial stress and the ultimate axial strain increase with
the number of FRP layers.
3.3. Ultimate axial stresses and ultimate axial strains
The ultimate axial stresses ( fcc′ ) and ultimate axial strains (εcu ) of
concrete in the test columns are listed in Table 1. The compressive
strength and the axial strain at peak axial stress of unconfined concrete
obtained from the large-scale control column R0Lr25 are 39.6 MPa and
0.00252, respectively. This compressive strength is lower than the un-
confined concrete strength determined from the standard concrete cy-
linders ( fc′ = 42.15 MPa). The issue of lower compressive strength of
concrete in large-scale columns has also been widely reported by other
researchers, although the difference between fco ′ and fc′ varies and is
somewhat uncertain (e.g., [38,13,69,71]. A ratio of 0.85 between fco ′
and fc′ is commonly specified for use in the design of RC columns (e.g.,
[1]), whereas in the present study a ratio of 0.94 was identified. This
relatively high ratio of fco′ / fc′ may be a result of the good quality control
of the tests (including both specimen preparation and testing) in the
laboratory. The ratio of 0.94 is adopted to evaluate fco ′ for all the FRP-
confined RC test columns in the present study unless otherwise speci-
fied (see Table 1).
As can be seen from Table 1, FRP confinement significantly en-
hances both the ultimate axial stress and ultimate axial strain of con-
crete in rectangular RC columns. The strength enhancement ratio
′ ) ranges from 1.14 to 2.13, and the ultimate strain enhancement
( fcc′ / fco
ratio (εcu /εco ) ranges from 3.76 to 17.48 for the FRP-confined RC test
columns. Both the strength enhancement ratio and the strain en-
hancement ratio increase with the number of FRP layers, and for col-
umns with the same number of FRP layers, an increase in the corner
radius ratio leads to an increase in both the strength enhancement ratio
and the strain enhancement ratio.
3.4. FRP hoop strains at ultimate condition
Fig. 10 shows distributions of FRP hoop strains around the peri-
meter of column section at the ultimate condition (i.e., FRP rupture).
All these strains were obtained from strain gauges on the FRP jacket at
the mid-height of the column. The crosses in the figure indicate only the
horizontal locations of FRP rupture around the column perimeter, and
they should not be mistaken to mean that FRP rupture occurred at the
mid-height of the column (see Fig. 6 for the FRP rupture locations). As
expected, these distributions of FRP hoop strains are highly non-uni-
form, with the maximum FRP hoop strain being generally from the
strain gauge either at one of the flat side centers or one of the transition
points between the rounded corner and the two adjacent flat sides (see
the numbers in boxes in Fig. 10). However, FRP rupture generally did
not occur at the location of maximum FRP hoop strain, but occurred at
or near one of the transition points (Fig. 10). The same observation has
also been reported by other researchers (e.g., [56,61]). FRP rupture at
or near one of the transition points is mainly caused by the non-uniform
dilation of confined concrete in rectangular columns: the dilation of
concrete near the flat sides is much larger due to the small flexural
rigidity of the FRP jacket, while that in the rounded corner regions is
J.J. Zeng et al.
(a) R0Lr25 (b) R1Lr25 (c) R2Lr25
(f) R4Lr45 (g) R4Lr65
Fig. 6. Failure of test columns.
much lower [29]. This non-uniform dilation of concrete causes large
bending strains in the FRP jacket, with the largest bending strains being
at the transition points where an abrupt change in curvature occurs
[29], which explains the location of FRP jacket rupture.
The average FRP hoop strains at different section locations of the
test columns at the ultimate condition are listed in Table 3. εh, l and εh, s
refer to the average FRP hoop strains at the centers of the longer sides
and the shorter sides of the rectangular section, respectively; εh, tl (εh, ts )
refers to the average FRP hoop strain at the transition point between
each rounded corner and its adjacent longer (shorter) side of the rec-
tangular section; εh, corner is the average FRP hoop strain at the centers of
four rounded corners; and εh, max is the maximum FRP hoop strain
captured by the strain gauges. The strain gauges in the overlapping
zone were excluded in calculating the above average strains. The last
two rows of Table 3 list the ratios between the average FRP hoop strains
and the FRP rupture strain obtained from coupon tests (εf = 1.71%),
and the ratios between the average FRP hoop strains and the FRP
rupture strain obtained from FRP-confined standard concrete cylinder
tests (εh, rup _ cyl = 1.13%). The following conclusions can subsequently
635
Engineering Structures 174 (2018) 629–645
(d) R2Lr45 (e) R2Lr65
(h) R6Lr45 (i) R6Lr65
be drawn: (1) the maximum hoop strains at FRP rupture in the test
columns are lower than the rupture strain obtained from coupon tests;
(2) the average FRP hoop strain on the flat sides [Column (3) in
Table 3] is the largest, while that at the corner centers [Column (7) in
Table 3] is the lowest, with that at the transition points in between
[Column (6) in Table 3]; (3) the average FRP hoop strain on the flat
sides is very close to the average rupture strain obtained from FRP-
confined concrete cylinder tests (εh/ εh, rup _ cyl = 0.946); (4) the maximum
FRP hoop strains, εh, max , are generally larger than the average rupture
strain obtained from FRP-confined concrete cylinder tests, εh, rup _ cyl
[Column (8) in Table 3].
4. Stress-strain models for FRP-confined concrete in rectangular
columns
4.1. General
As mentioned in the preceding sections, a significant number of
stress-strain models for FRP-confined concrete in rectangular columns
J.J. Zeng et al. Engineering Structures 174 (2018) 629–645

(a) Specimen R1Lr25 (b) Specimen R2Lr25

(c) Specimen R2Lr45 (d) Specimen R2Lr65

(e) Specimen R4Lr45 (f) Specimen R4Lr65

(g) Specimen R6Lr45 (h) Specimen R6Lr65

Fig. 7. Axial stress-axial strain curves of FRP-confined concrete in test columns.

have been proposed. Similar to the case of circular columns, these [23,64,67,17,62,27,10]); and (2) analysis-oriented stress-strain models
stress-strain models can be classified into two categories: (1) design- that predict stress-stain responses via an incremental process (e.g.,
oriented stress-strain models in closed-form expressions (e.g., [59,32,40,25,37]). Only design-oriented stress-strain models are

636
J.J. Zeng et al. Engineering Structures 174 (2018) 629–645

(a) R0Lr25, R2Lr25, R2Lr45 and R2Lr65 (a) R2Lr45, R4Lr45 and R6Lr45

(b) R4Lr45 and R4Lr65

(b) R2Lr65, R4Lr65 and R6Lr65

Fig. 9. Normalized axial stress-strain curves of FRP-confined concrete in col-
umns with different FRP thicknesses.

rectangular columns can generally be classified into two categories: (1)

single-segment models which use a single expression to describe the
entire stress-strain curve; and (2) two-segment model which consists of
two segments defined by two separate expressions. Most of these stress-
strain models predict only stress-strain curves with an ascending second
branch. Such stress-strain curves are referred to as strain hardening
stress-strain curves while those with a descending second branch are
referred to as strain softening stress-strain curves.

4.2.1. Single-segment models

The single-segment models (e.g., [40,18,14]) generally employ the
following four-parameter curve of Richard and Abbott [68]:
(c) R6Lr45 and R6Lr65
(Ec−E2) εc
Fig. 8. Normalized axial stress-axial strain curves of FRP-confined concrete in σc = E2 εc +
n 1/ n
columns with different corner radii. ⎡1 +
⎣
( Ec − E2
fi
εc ) ⎤⎦ (1)

examined in this paper as for rectangular columns analysis-oriented
models are rather involved due to the non-uniformity of axial stress
distribution.
4.2. Stress-strain models
The existing stress-strain models for FRP-confined concrete in
where σc and εc are the axial stress and axial strain of confined concrete,
respectively; Ec is the elastic modulus of unconfined concrete; E2 is the
slope of the linear second portion; fi is the intercept of the stress axis by
the asymptote of the second portion of the stress-strain curve; and n is a
shape parameter controlling the transition region between the two
portions of the stress-strain curve. Ilki et al. [18] assumed that fi = fco′ ,
whereas in Pantelides and Yan [40], fi is determined by the ultimate

637
J.J. Zeng et al. Engineering Structures 174 (2018) 629–645

(a) R1Lr25 (b) R2Lr25

(c) R2Lr45 (d) R2Lr65

(e) R4Lr45
(f) R4Lr65

(g) R6Lr45 (h) R6Lr65

Fig. 10. FRP hoop strain distributions around section perimeter at ultimate condition (in microstrains).

point and the transition point of the stress-strain curve. where n is a parameter controlling the transition region between the
Cao et al. [10] employed the following equation proposed by Zhou two portions of the stress-strain curve; and fo is the stress limit of the
and Wu [70] for their single-segment stress-strain model: linear first portion of the stress-strain curve, which is taken as fco
′.
εc εc
(
σc = [(Ec εn−fo ) e− εn + fo + E2 εc ] 1−e− εn ) (2a) 4.2.2. Two-segment models
A typical two-segment stress-strain model is that of Lam and Teng
εn = n × fo / Ec (2b) [23], which consists of a parabolic first segment and a linear second

638
J.J. Zeng et al. Engineering Structures 174 (2018) 629–645

Table 3
FRP hoop strains in test columns at ultimate condition.
Specimen εh, l εh, s Average εh, tl εh, ts Average εh, corner εh, max
(1) (2) (1) + (2) (4) (5) (4) + (5) (7) (8)
(3) = (6) =
2 2

R1Lr25 0.0119 0.0154 0.0137 0.0096 0.0094 0.0095 0.0025 0.0164

R2Lr25 0.0042 0.0086 0.0064 0.0070 0.0074 0.0072 0.0025 0.0120
R2Lr45 0.0129 0.0096 0.0113 0.0098 0.0072 0.0085 0.0055 0.0130
R2Lr65 0.0133 0.0106 0.0120 0.0080 0.0116 0.0098 0.0082 0.0133
R4Lr45 0.0112 0.0098 0.0105 0.0107 0.0097 0.0102 0.0066 0.0137
R4Lr65 0.0107 0.0075 0.0091 0.0072 0.0065 0.0069 0.0017 0.0107
R6Lr45 0.0123 0.014 0.0132 0.0092 0.0106 0.0099 0.0033 0.0162
R6Lr65 0.0088 0.0097 0.0093 0.0073 0.0074 0.0074 0.0031 0.0108

Average FRP strain ratio ε h / εf 0.625 0.625 0.625 0.503 0.511 0.508 0.244 0.776
εh/ εh, rup _ cyl 0.946 0.947 0.946 0.761 0.772 0.764 0.368 1.174

segment. The model was originally proposed by Lam and Teng [22] for [23] model, which also consists of a parabolic first segment and a linear
FRP-confined concrete in circular columns. Lam and Teng’s [22,23] second segment:
model has been widely adopted by national standards/guidelines due to ft − Ec εt 2
its simplicity in form and high accuracy (e.g., [15,2,3,12]). The stress- ⎧ Ec εc + εc (0 ⩽ εc ⩽ εt )
εt2
strain curve is described by the following expression: σc =
⎨ ft + E2 (εc−εt ) (εt < εc ⩽ εcu )
⎩ (6a)
2
⎧ Ec εc− (Ec − E2) εc2 (0 ⩽ εc ⩽ εt )
σc = ′
4f co
′ + Ec εcu )2−8ft Ec εcu
′ + Ec εcu− (ft + fcu
ft + fcu
⎨ f′ + E ε (εt ⩽ εc ⩽ εcu ) εt =
⎩ co 2 c (3a) 2Ec (6b)
fcc′ −fco′ where E2 is calculated by Eq. (5b). They proposed a predictive equation
E2 =
εcu (3b) for the transition stress which is assumed to be a function of the section
shape and FRP confinement (Table 4).
where fcc′ is the ultimate axial stress of FRP-confined concrete (equal to A number of two-segment stress-strain models employ the following
′
the axial stress at ultimate strain, f cu , for a hardening stress-strain second-degree parabola for the first segment (e.g., [17,64]):
curve); εcu is the ultimate axial strain; and εt is the transition axial strain
2
determined by the condition that the two segments are connected 2εc ⎛ εc ⎞ ⎤
σc = ft ⎡
⎢ εt − εt ⎥ (0 ⩽ εc ⩽ εt )
⎜ ⎟

smoothly: ⎣ ⎝ ⎠⎦ (7)
2fcc′ This expression, however, implies that the slope at the transition
εt =
Ec−E2 (4) point is always equal to zero; therefore, the two segments are not
connected smoothly if the slope of the linear second segment is not
Unlike Lam and Teng’s [23] model where the transition strain is
zero.
determined based on the smooth connection condition between the two
It should be mentioned that apart from the above models for pre-
segments, other two-segment stress-strain models generally have ex-
dicting the entire stress-strain curve of FRP-confined concrete, a sig-
plicit predictive equations for the transition axial stress and/or transi-
nificant number of studies have been concerned with only the ultimate
tion axial strain. Typical examples are the models of Youssef et al. [67],
condition (i.e., ultimate axial stress fcc′ and ultimate axial strain εcu )
Wu et al. [64], and Wei and Wu [62].
(e.g., [27,43,14,26]).
Youssef et al.’s [67] model consists of a four-parameter polynomial
first segment and a linear second segment:
4.2.3. Ultimate condition
εc n − 1 The equations for the ultimate axial stress and the ultimate axial
⎧ Ec εc ⎡1− 1
⎪ ⎣ n ( )( )
E
1− E2
c εt
⎤
⎦
(E2 > 0;0 ⩽ εc < εt ) strain in most of the existing stress-strain models for FRP-confined
⎪ concrete in rectangular columns were modified from the corresponding
σc = E ε ⎡1− 1 εc n − 1⎤
⎨ c c
⎪ ⎣ n εt
()
⎦
(E2 ⩽ 0;0 ⩽ εc < εt ) equations for circular columns by introducing appropriate “shape fac-
ft + E2 (εc−εt ) (εt ⩽ εc ⩽ εcu ) tors” to reflect the effectiveness of FRP confinement (e.g.,
⎪
⎩ (5a) [23,18,64,67,53,27]). The concept of “shape factor” was originally
proposed by Sheikh and Uzumeri [49] for rectangular RC columns. For
′ −ft
fcu
E2 = FRP-confined rectangular concrete columns, the shape factor has often
εcu−εt (5b) been defined as a function of the effective-confinement area ratio ke
(i.e., the ratio of the effective-confinement area to the total area of
(Ec − E2) εt
⎧ (E2 > 0) concrete, ke = Ae / Ac ) (e.g., [23,18,17,40,67,53]). The effective-con-
⎪ Ec εt − ft
n= Ec εt finement area in Lam and Teng’s [23] model is taken as the area en-
⎨ (E2 ⩽ 0)
⎪ Ec εt − ft closed by the four parabolas that intersect the flat sides at an angle
⎩ (5c)
equal to that of the adjacent diagonal of the rectangular section. The
where εt and ft = axial strain and axial stress at the transition point detailed equations for the shape factors and the effective-confinement
between the two segments; and n is a constant determined by the area ratio of Lam and Teng’s [23] model are listed in Table 4. Some
condition that the two segments are connected smoothly. In Youssef other stress-strain models employ complex exponential equations for
et al.’s [67] model, predictive equations are given for both εt and ft . This the ultimate axial stress and the ultimate axial strain, which are not
model predicts both the hardening and the softening types of stress- directly related to the corresponding equations for circular columns
strain curve. (e.g., [62,10,20]). The equations in the models of Wei and Wu [62] and
Wei and Wu’s [62] stress-strain model is similar to Lam and Teng’s Cao et al. [10] are of this type, which can be found in Table 4.

639
 J.J. Zeng et al.

Table 4
Summary of existing stress–strain models for concrete in FRP-confined rectangular columns.
Model Ultimate condition Transition point Shape factor Diameter of FRP strain
equivalent circular efficiency factor
column

Lam and ′
fcc f Determined by the smooth connection condition (b / h)(h − 2r )2 + (h / b)(b − 2r )2 D= h2 + b2 Obtained from
= 1 + 3.3ks1 ⎛ lu ⎞ k e = 1−
Teng [23] ′
fco ′ 3Ag (1 − ρsc ) accompanying
⎝ fco ⎠
εcu flu εh, rup 0.45 b 2 h 0.5 compression
= 1.75 + 12ks2 ks1 = e s2 = ke
εco ′
fco ( εco ) h
( )k,k b
() tests of FRP-
confined circular
columns
Wu et al. [64] f ′cu f fo (Ef ⩽ 250 GPa) 2bh εh, rup = εf
= ksσ + 2.4ksσ ⎛ lu ⎞ = 1.0 + 0.0008αk1 γ1 ⎧ (2−α ) r / h + 0.5α D=
′
fco ′ ′
fco ksσ = b+h
⎝ fco ⎠ ⎨ (2−α Ef /250 ) r /h + 0.5α Ef /250 (Ef > 250 GPa)
fo ⎩
ε ′cu f = 1.0 + 0.0034αk2 γ1
= 1.3ksε + 6.3ksε ⎛ lu ⎞ ′
fco
εco ′ ⎧ (2−1.6α ) r /h + 0.8α (Ef ⩽ 250 GPa)
⎝ fco ⎠ ρf Ef ksε =
γ1 = , ⎨ (2−1.6α Ef /250 ) r /h + 0.8α Ef /250 (Ef > 250 GPa)
′
fco ⎩
′,
α = 30/ fco

⎧1 Ef ⩽ 250GPa
k1 = ,
⎨ Ef /250 Ef > 250GPa
⎩

640
⎧1 Ef ⩽ 250GPa
k2 =
⎨
⎩
250/ Ef Ef > 250GPa
Youssef et al. 0.6 (b − 2r )2 + (h − 2r )2 2Ag εh, rup = εf
f ′cu f′ fo ρf Ef ε ft 1.25
D=
= 0.5 + 1.225 ⎛ lu ⎞ = 1.0 + 1.135 ⎛ ⎞ ks = 1−
3bh [(b + h) − (4 − π ) r ]
[67] ′
fco ′ ′
fco ′
fco
⎝ fco ⎠ ⎝ ⎠
E ε 0.5 6
f ′lu ρf Ef ε ft 7 Ef ε f 0.5
εcu = 0.004325 + 0.2625 ⎛ ⎞⎛ f f ⎞ ⎞ ⎛ ⎞
′
fco ′
fco εo = 0.002 + 0.0775 ⎛
⎝ ⎠⎝ ⎠ ′
fco ′
fco
⎝ ⎠ ⎝ ⎠
εft = 0.002
2t f [(b + h) − (4 − π ) r ]
ρf =
bh − (4 − π ) r 2
Wei and 0.73 ks = 2r / b D=b εh, rup = εf
f ′cu f h −1 fo h −1 ⎛ flu ⎞
= 0.5 + 2.7ks0.4 ⎛ lu ⎞ = 1.0 + 0.43ks0.68
Wu [62] ′
fco ′ () b ′
fco ()b ′
fco
⎝ fco ⎠ ⎝ ⎠
0.75 0.62
εcu f h −0.3 (fo + f ′cu + Ec εcu) − (fo + f ′cu + Ec εcu)2 − 8fo Ec εcu
= 1.75 + 12(0.36ks + 0.64)0.4 ⎛ lu ⎞ ⎛ 30 ⎞ ε0 =
εco ′ ′ b
() 2Ec
⎝ fco ⎠ ⎝ fco ⎠
Cao et al. [10] f ′cu 2Ef t f 1.03 2rc 0.81 30 0.54 h −1.9 ε f 0.82 fo ks = 2r / b D=b εh, rup = εf
= 1 + 8.34 ⎛ ⎞ ⎛ ⎞ = 1.0
′
fco
′
fco ⎝ bEc ⎠ b
( ) ′ b εco() ( )
⎝ fco ⎠
0.79 ′ b/2Ef t f εf
εo = fco
εcu 2Ef t f 0.68 2r 30 h −0.64 ε f 1.14
= 1.75 + 9.45 ⎛ ⎞ 0.54 + 0.46 ⎛ ⎞
εco ⎝ bEc ⎠
(b ′
fco b
)⎝ ⎠ ( ) ( ) εco

Note: Ag : gross area of the column section with rounded corners.

ρsc : cross sectional area ratio of the longitudinal steel reinforcement.
ρf : FRP volumetric ratio.
Engineering Structures 174 (2018) 629–645
J.J. Zeng et al.
(a) Ultimate axial stresses
(b) Ultimate axial strains
Fig. 11. Performance of Lam and Teng’s model in predicting the ultimate
condition of test columns.
The effective-confinement area ratio ke is often employed in calcu-
lating the effective confining pressure in an FRP-confined rectangular
section as follows (e.g., [23,18,17,40,67,53,20]):
2Ef t f εh, rup
flu′ = ke flu = ke
D (8)
where flu′ the effective confining pressure; Ef and t f are the elastic
modulus and thickness of the FRP jacket, respectively; εh, rup is the
nominal FRP hoop rupture strain in the equivalent circular column; and
D is the diameter of the equivalent circular column. Both εh, rup and D are
defined rather differently in different stress-strain models. The most
frequently used method to define D is to assume that the equivalent
circular column and the original rectangular column have approxi-
mately the same FRP volumetric ratio (e.g., [18,17,53]), leading to the
following expression:
2bh
D=
b+h (9)
In Lam and Teng’s [23] model, a circular section circumscribing the
641
Engineering Structures 174 (2018) 629–645
rectangular section is used as the equivalent circular section (i.e.,
D = h2 + b2 ). Some other researchers used an equivalent circular
section with the diameter equal to the sectional width or depth of the
rectangular section (e.g., [66,61,65,62,14]).
For the nominal FRP hoop rupture strain in the equivalent circular
column (εh, rup ), some models include a predictive equation (e.g.,
[41,61,43,36]), while others directly use the FRP rupture strain from
coupon tests (εf ). As mentioned earlier, Lam and Teng [22] suggested
that the average FRP hoop rupture strain measured from accompanying
compression tests on FRP-confined standard circular concrete cylinders
(i.e., εh, rup = εh, rup _ cyl ) or deduced from test results of circular concrete
columns confined with the same FRP jacket be used. With the absence
of accompanying tests on concrete cylinders, an FRP strain efficiency
factor (εh, rup/ εf ) of 0.586, which was obtained based on a large test
database of CFRP-confined circular concrete columns [22], can be used
for CFRP-confined rectangular concrete columns. Wang et al. [61] and
Pham and Hadi [43] derived their predictive equations for εh, rup based
on FRP rupture strains measured at the centers of rounded corners of
test rectangular columns, as FRP rupture generally occurs at or near one
of the corner centers in rectangular columns.
4.3. Assessment of stress-strain models
Lim and Ozbakkaloglu [27] carried out a review and an assessment
of 25 existing stress-strain models for FRP-confined concrete in rec-
tangular columns (including square columns as a special case) and
found that the models of Lam and Teng [23,22] and Wei and Wu [62]
performed better than the other models. Lin [28] also reviewed and
assessed ten existing stress-strain models for FRP-confined concrete in
rectangular columns (i.e., [23,18,17,40,64,67,53,62,27,43]) using a
large test database. The models of Lam and Teng [23], Youssef et al.
[67], Wu et al. [64], and Wei and Wu [62] were found to predict the
ultimate axial stress and the ultimate axial strain with higher accuracy
than the other models. In addition to the models mentioned above, two
other design-oriented stress-strain models have appeared recently
[10,20]. The stress-strain model of Isleem et al. [20] however does not
reproduce the basic bilinear feature of stress-strain curves for FRP-
confined concrete and was thus excluded from the assessment. There-
fore, five models in total (i.e., [23,67,64,62,10]) were used to predict
the stress-strain behavior of FRP-confined concrete in the test columns.
The detailed equations for the five models are listed in Table 4. In the
assessment, the confinement effect from the transverse steel reinforce-
ment is ignored because of the low transverse steel ratio within the test
portion of the columns.
The stress-strain curves predicted with the above five models are
compared with the test results in Fig. 7. It can be seen that for columns
with a relatively low FRP confinement level (i.e., R1Lr25, R2Lr25,
R2Lr45, and R2Lr65), the models of Youssef et al. [67] and Wu et al.
[64] predict a descending second branch, while the test curves have an
ascending second branch. Wei and Wu’s [62] model also predicts a
descending second branch for R1Lr25 and R2Lr25 but predicts rea-
sonably well the stress-strain curves of R2Lr45 and R2Lr65. Lam and
Teng’s [23] model predicts well the four columns in terms of the shape
of second branch; however, the ultimate axial stresses and strains are
generally underestimated. Cao et al.’s [10] model predicts the four
columns very well in terms of the shape of the second branch. For
columns with a relatively high FRP confinement level (i.e., R4Lr45,
R4Lr65, R6Lr45, and R6Lr65), Lam and Teng’s [23] model under-
estimates the slopes of the second branch as well as the ultimate axial
stresses. This also applies to Youssef et al.’s [67] model. Cao et al.’s [10]
model predicts the responses of columns R4Lr45 and R4Lr65 very well
but underestimates the ultimate axial strains of columns R6Lr45 and
R6Lr65. Wu et al.’s [64] model and Wei and Wu’s [62] model, however,
both overestimate the slopes of the second branch for all the four col-
umns.
J.J. Zeng et al. Engineering Structures 174 (2018) 629–645

(a) Ultimate axial stresses (a) Ultimate axial stresses

(b) Ultimate axial strains (b) Ultimate axial strains

Fig. 12. Performance of Wu et al.’s model in predicting the ultimate condition Fig. 13. Performance of Youssef et al.’s model in predicting the ultimate con-
of test columns. dition of test columns.

Fig. 11 shows comparisons of ultimate axial stress and ultimate axial value (M), average absolute error (AAE), and standard deviation (SD) of
strain between the predictions of Lam and Teng’s [23] model and the ratios between the theoretical and the experimental values (theoi / expi )
test results. As mentioned earlier, in Lam and Teng’s [23] model, the for each model are also given in the figures; these parameters are de-
FRP rupture strain from the accompanying compression tests on FRP- fined by the following equations:
confined circular concrete cylinders (εh, rup = εh, rup _ cyl ) should be used. n
theoi
When such accompanying tests are not available, an FRP strain effi- ∑ expi
ciency factor of 0.586 (εh, rup = 0.586εf ) can be used. The predictions M= i=1
obtained using these two different definitions of ultimate hoop strains n (10)
are compared with the test results in Fig. 11. Obviously, as the ac- n
companying compression tests in the present study lead to an FRP strain expi − theoi
∑ expi
efficiency factor of 0.661 which is larger than 0.586, the ultimate axial i=1
AAE =
stresses and the ultimate axial strains predicted with εh, rup _ cyl are larger n (11)
than those with 0.586εf as shown in Fig. 11. The results predicted with
n 2
εh, rup _ cyl agree more closely with the test results, although the ultimate
axial stresses are still significantly underestimated. ∑
i=1
⎡ theoi − theo
⎣ expi exp ( ) ⎤
aver ⎦
The comparisons of ultimate axial stresses and ultimate axial strains SD =
n−1 (12)
for the models of Wu et al. [64], Youssef et al. [67], Wei and Wu [62],
and Cao et al. [10] are shown in Figs. 12–15 respectively. The mean where n is the number of data points; expi and theoi are the ith experi-
mental value and theoretical value, respectively. The AAE measures the

642
J.J. Zeng et al.
(a) Ultimate axial stresses
(b) Ultimate axial strains
Fig. 14. Performance of Wei and Wu’s model in predicting the ultimate con-
dition of test columns.
average error magnitude of model prediction; and the SD shows the
degree of variation or the magnitude of the associated scatter of pre-
diction.
Figs. 11–15 show that Cao et al.’s [10] model performs the best in
predicting the ultimate axial stresses in terms of the M value and the
AEE value, followed by Lam and Teng’s [23] model (predicted with
εh, rup = εh, rup _ cyl ). Both models, however, underestimate the ultimate
axial strains, with Lam and Teng’s [23] model performing slightly
better. Wu et al.’s [64] model provides satisfactory predictions for the
ultimate axial stress but significantly underestimates the ultimate axial
strains (Fig. 12b). Predictions from Youssef et al.’s [67] model for both
the ultimate axial stresses and ultimate axial strains are significantly
lower than the test results. The ultimate axial strains are also sig-
nificantly underestimated by Wei and Wu’s [62] model, while the ul-
timate axial stresses are overestimated for most cases.
5. Conclusions
This paper has presented the results of axial compression tests on
nine large-scale rectangular RC columns (including eight FRP-confined
643
Engineering Structures 174 (2018) 629–645
(a) Ultimate axial stresses
(b) Ultimate axial strains
Fig. 15. Performance of Cao et al.’s model in predicting the ultimate condition
of test columns.
RC columns and one control RC column without FRP jacketing). Five
representative stress-strain models for FRP-confined concrete in rec-
tangular columns were used to predict the stress-strain curves of FRP-
confined concrete in the test columns and their performance was as-
sessed. The results and discussions presented in this paper allow the
following conclusions to be drawn:
1. The stress-strain curves of FRP-confined concrete in the large-scale
FRP-confined rectangular RC columns tested in the study generally
have a typical bilinear shape;
2. The compressive strength of concrete in a large-scale unconfined
concrete column was found to be lower than that of a standard
concrete cylinder; the ratio between the two was found to be 0.94
for the columns tested in the present study;
3. The FRP hoop strains at ultimate condition at the mid-width posi-
tions of the flat sides are generally larger than those at the mid-arc
positions of corners, with those at the transition points in between;
the average FRP hoop strain of the mid-width positions of flat sides
is very close to the rupture strain obtained from FRP-confined
standard concrete cylinder tests;
J.J. Zeng et al.
4. The corner radius ratio and the FRP jacket thickness both have a
significant effect on the stress-strain response of FRP-confined con-
crete in rectangular columns; both the compressive strength and the
ultimate axial strain increase with the increase of corner radius ratio
or the FRP jacket thickness;
5. Cao et al.’s [10] model performs the best in predicting the ultimate
axial stresses of the test columns, followed by Lam and Teng’s [23]
model. Both models, however, underestimate the ultimate axial
strains. Lam and Teng’s [23] model predicts very well the shape of
the linear second branch of the stress-strain curve for columns with
a relatively low level of FRP confinement but significantly under-
estimates the slope of the second branch for columns with a rela-
tively high level of FRP confinement. Cao et al.’s [10] model pre-
dicts very well the shape of the linear second branch for most of the
columns. The complex mathematical expressions for the stress-strain
curve (Eq. (2)) and the ultimate axial stress/strain (Table 4) in Cao
et al.’s [10] model, however, make it inconvenient for use in design.
It is therefore desirable to formulate a simpler but similarly accurate
model for design purposes.
Acknowledgements
The authors are grateful for the financial support received from the
Research Grants Council of the Hong Kong Special Administrative
Region (Project No. 152217/15E) and the National Basic Research
Programme of the Ministry of Science and Technology of China (i.e.,
973 Program) (Project No.: 2012CB026200).
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in the
online version, at https://doi.org/10.1016/j.engstruct.2018.07.086.
References
[1] ACI 318-08. Building code requirements for structural concrete and commentary,
Farmington Hills, Michigan, USA; 2008.
[2] ACI-440.2R. Guide for the design and construction of externally bonded FRP sys-
tems for strengthening concrete structures. Farmington Hills, Michigan, USA:
American Concrete Institute; 2008.
[3] AS 5100. Bridge design. Sydney, Australia: Council of Standards Australia; 2017.
[4] ASTM C469. Standard test method for static modulus of elasticity and poisson ratio
of concrete in compression, West Conshohocken, USA; 2002.
[5] ASTM D3039. Standard test method for tensile properties of plymer matrix com-
posite materials. West Conshohocken, USA: American Society for Testing and
Materials; 2014.
[6] Hollaway LC, Teng JG. Strengthening and Rehabilitation of Civil Infrastructures
using Fiber-reinforced Polymer (FRP) Composites. Cambridge, UK: Woodhead
Publishing; 2008.
[7] Bousias SN, Triantafillou TC, Fardis MN, Spathis L, O’Regan BA. Fiber-reinforced
polymer retrofitting of rectangular reinforced concrete columns with or without
corrosion. ACI Struct J 2004;101(4):512–20.
[8] BS 18. Method for tensile testing of metals. London, U.K.: British Standards
Institution; 1987.
[9] Campione G. Influence of FRP wrapping techniques on the compressive behaviour
of concrete prisms. Cem Concr Compos 2006;28(5):505.
[10] Cao YG, Jiang C, Wu YF. Cross-section unification on the stress-strain model of
concrete subjected to high passive confinement by fiber-reinforced polymer.
Polymers 2016;2016(8). polym9050186.
[11] Chaallal O, Hassan M, Shahawy M. Performance of axially loaded short columns
strengthened with CFRP wrapping. J Compos Constr, ASCE 2003;7(3):200–8.
[12] Concrete Society. Design guidance for strengthening concrete structures with fibre
composite materials. 3rd ed. Concrete Society Technical Report No. 55,
Crowthorne, Berkshire, U.K.; 2012.
[13] De Luca A, Nardone F, Matta F, Nanni A, Lignola G, Prota A. Structural evaluation
of full-scale FRP-confined reinforced concrete columns. J Compos Constr, ASCE
2011;15(1):112–23.
[14] Faustino P, Chastre C, Paula R. Design model for square RC columns under com-
pression confined with CFRP. Compos B Eng 2014;57:187–98.
[15] GB 50608. Technical code for infrastructure application of FRP composites. China:
China Architecture and Building Press; 2010.
[16] Ghatte HF, Comert M, Demir C, Ilki A. Evaluation of FRP confinement models for
substandard rectangular RC columns based on full-scale reversed cyclic lateral
loading tests in strong and weak directions. Polymers 2016;8:323.
[17] Harajli MH. Axial stress-strain relationship for FRP-confined circular and
644
Engineering Structures 174 (2018) 629–645
rectangular concrete columns. Cem Concr Compos 2006;28(10):938–48.
[18] Ilki A, Kumbasar N, Koc V. Low strength concrete members externally confined with
FRP sheets. Struct Eng Mech 2004;18(2):167–94.
[19] Ilki A, Peker O, Karamuk E, Demir C, Kumbasar N. FRP retrofit of low and medium
strength circular and rectangular reinforced concrete columns. J Mater Civ Eng
ASCE 2008;20(2):169–88.
[20] Isleem HF, Wang DY, Wang ZY. Modeling the axial compressive stress-strain be-
havior of CFRP-confined rectangular RC columns under monotonic and cyclic
loading. Compos Struct 2018;185:229–40.
[21] Jiang T, Teng JG. Analysis-oriented models for FRP-confined concrete: a com-
parative assessment. Eng Struct 2007;29(11):2968–86.
[22] Lam L, Teng JG. Design-oriented stress-strain model for FRP-confined concrete.
Constr Build Mater 2003;17:471–89.
[23] Lam L, Teng JG. Design-oriented stress-strain model for FRP-confined concrete in
rectangular columns. J Reinf Plast Compos 2003;22(13):1149–86.
[24] Lam L, Teng JG. Ultimate condition of fiber reinforced polymer-confined concrete. J
Compos Constr, ASCE 2004;8(6):539–48.
[25] Lee CS, Hegemier GA, Phillippi DJ. Analytical model for fiber-reinforced polymer-
jacketed square concrete columns in axial compression. ACI Struct J
2010;107(2):208–17.
[26] Li X, Lu J, Ding DD, Wang W. Axial strength of FRP-confined rectangular RC col-
umns with different cross-sectional aspect ratios. Mag Concr Res
2017;69(19):1011–26.
[27] Lim JC, Ozbakkaloglu T. Design model for FRP-confined normal- and high-strength
concrete square and rectangular columns. Mag Concr Res 2014;66(20):1020–35.
[28] Lin G. Seismic performance of FRP-confined RC columns: stress-strain models and
numerical simulation [PhD. Thesis]. Hong Kong, China: The Hong Kong Polytechnic
University; 2016.
[29] Lin G, Teng JG. FRP-confined concrete in square columns: an advanced stress-strain
model based on a new approach. In: Proceedings, 4th Asia Pacific Conference on
FRP Structures (APFIS 2013), 11–13 December, Swinburne University of
Technology, Melbourne, Australia; 2013.
[30] Maalej M, Tanwongsval S, Paramasivam P. Modelling of rectangular RC columns
strengthened with FRP. Cem Concr Compos 2003;25:263–76.
[31] Mander JB, Priestley MJ, Park R. Theoretical stress-strain model for confined
concrete. J Struct Eng ASCE 1988;114(8):1804–26.
[32] Marques SPC, Marques DC, Silva JL, Cavalcante MA. Model for analysis of short
columns of concrete confined by fiber-reinforced polymer. J Compos Constr ASCE
2004;8(4):332–40.
[33] Matthys S, Toutanji H, Audenaert K, Taerwe L. Axial behavior of large-scale col-
umns confined with fiber-reinforced polymer composites. ACI Struct J
2005;102(2):258–67.
[34] Masia MJ, Gale TN, Shrive NG. Size effects in axially loaded square-section concrete
prisms strengthened using carbon fibre reinforced polymer wrapping. Can J Civ Eng
2004;31:1–13.
[35] Mirmiran A, Shahawy M, Samaan M, Echary H, Mastrapa J, Pico. Effect of column
parameters on FRP-confined concrete. J Compos Constr, ASCE 1998;2(4):175–85.
[36] Nisticò N, Monti G. RC square sections confined by FRP: analytical prediction of
peak strength. Compos B Eng 2013;45(1):127–37.
[37] Nisticò N. RC square sections confined by FRP: a numerical procedure for predicting
stress-strain relationship. Compos B Eng 2014;59:238–47.
[38] Ozbakkaloglu T, Oehlers DJ. Concrete-filled square and rectangular FRP tubes
under axial compression. J Compos Constr, ASCE 2008;12(4):469–77.
[39] Ozcan O, Binici B, Ozcebe G. Seismic strengthening of rectangular reinforced con-
crete columns using fiber reinforced polymers. Eng Struct 2010;32:964–73.
[40] Pantelides CP, Yan ZH. Confinement model of concrete with externally bonded FRP
jackets or posttensioned FRP shells. J Struct Eng, ASCE 2007;133(9):1288–96.
[41] Pellegrino C, Modena C. Analytical model for FRP confinement of concrete columns
with and without internal steel reinforcement. J Compos Constr ASCE
2010;14(6):693–705.
[42] Pessiki S, Harries KA, Kestner J, Sause R, Ricles JM. The axial behavior of concrete
confined with fiber reinforced composite jackets. J Compos Constr, ASCE
2001;5(4):237–45.
[43] Pham TM, Hadi MNS. Stress prediction model for FRP confined rectangular con-
crete columns with rounded corners. J Compos Constr ASCE 2014;18(1):04013019.
[44] Rocca S, Galati N, Nanni A. Large-size reinforced concrete columns strengthened
with carbon FRP: experimental evaluation. In: Proceedings, Third International
Conference on FRP Composites in Civil Engineering, December 13–15 2006, Miami,
Florida, USA; 2006.
[45] Rocca S. Experimental and analytical evaluation of FRP-confined large size re-
inforced concrete columns [PhD. Thesis]. Rolla, MO, USA: University of Missouri-
Rolla; 2007.
[46] Rocca S, Galati N, Nanni A. Review of design guidelines for FRP confinement of
reinforced concrete columns of noncircular cross sections. J Compos Constr ASCE
2008;12(1):80–92.
[47] Rochette P, Labossiere P. Axial testing of rectangular column models confined with
composites. J Compos Constr, ASCE 2000;4(3):129–36.
[48] Sheikh SA, Bayrak O. Seismic behavior of FRP retrofitted concrete columns. In:
Proceedings of the Structural Congress and Exposition, Washington, DC, USA,
21–23 May; 2001. p. 1–8.
[49] Sheikh SA, Uzumeri SM. Strength and ductility of tied concrete columns. J Struct
Div, ASCE 1980;106(ST5):1079–102.
[50] Silva M, Rodrigues C. Size and relative stiffness effects on compressive failure of
concrete columns wrapped with glass FRP. J Mater Civ Eng ASCE
2006;18(3):334–42.
[51] Teng JG, Chen JF, Smith ST, Lam L. FRP-strengthened RC structures. U.K.: John
J.J. Zeng et al.
Wiley and Sons Inc; 2002.
[52] Teng JG, Lam L. Behavior and modeling of fiber reinforced polymer-confined
concrete. J Struct Eng, ASCE 2004;130(11):1713–23.
[53] Toutanji H, Han M, Gilbert J, Matthys S. Behavior of large-scale rectangular col-
umns confined with FRP composites. J Compos Constr, ASCE 2010;14(1):62–71.
[54] Vincent T, Ozbakkaloglu T. Influence of concrete strength and confinement method
on axial compressive behavior of FRP-confined high- and ultra-high-strength con-
crete. Compos B Eng 2013;50:413–28.
[55] Vuggumudi S, Alagusundaramoorthy P. FRP strengthened RC rectangular columns
under combined axial and lateral loading: analytical study. Structures
2018;14:88–94.
[56] Wang LM, Wu YF. Effect of corner radius on the performance of CFRP-confined
square concrete columns: Test. Eng Struct 2008;30(2):493–505.
[57] Wang DY, Wang ZY, Smith ST, Yu T. Size effect on axial stress-strain behavior of
CFRP-confined square concrete columns. Constr Build Mater 2016;118:116–26.
[58] Wang DY, Wang ZY, Yu T, Li H. Seismic performance of CFRP-retrofitted large-scale
rectangular RC columns under lateral loading in different directions. Compos Struct
2018;192:475–88.
[59] Wang YC, Restrepo JI. Investigation of concentrically loaded reinforced concrete
columns confined with glass fiber-reinforced polymer jackets. ACI Struct J
2001;98(3):377–85.
[60] Wang YF, Wu HL. Size effect of concrete short columns confined with aramid FRP
jackets. J Compos Constr, ASCE 2011;15(4):535–44.
[61] Wang ZY, Wang DY, Smith S, Lu DG. CFRP-confined square RC columns. II: cyclic
axial compression stress-strain model. J Compos Constr, ASCE 2012;16(2):161–70.
645
Engineering Structures 174 (2018) 629–645
[62] Wei YY, Wu YF. Unified stress-strain model of concrete for FRP-confined columns.
Constr Build Mater 2012;26(1):381–92.
[63] Wu G, Lu ZG, Wu ZS. Strength and ductility of concrete cylinders confined with FRP
composites. Constr Build Mater 2006;20(3):134–48.
[64] Wu G, Wu ZS, Lu ZG. Design-oriented stress–strain model for concrete prisms
confined with FRP composites. Constr Build Mater 2007;21(5):1107–21.
[65] Wu YF, Wang LM. Unified strength model for square and circular concrete columns
confined by external jacket. J Struct Eng, ASCE 2009;135(3):253–61.
[66] Wu YF, Zhou YW. Unified strength model based on Hoek-Brown failure criterion for
circular and square concrete columns confined by FRP. J Compos Constr ASCE
2010;14(2):175–84.
[67] Youssef MN, Feng MQ, Mosallam AS. Stress-strain model for concrete confiend by
FRP composites. Compos B Eng 2007;38(5–6):614–28.
[68] Richard RM, Abbott BJ. Versatile elastic-plastic stress-strain formular. J Eng Mech,
ASCE 2003;101(4):511–5.
[69] Zhang YX. Behavior of large-size FRP-jacketed circular and rectangular reinforced
concrete columns [MPhil. Thesis]. Hong Kong, China: The Hong Kong Polytechnic
University; 2013.
[70] Zhou YW, Wu YF. General model for constitutive relationships of concrete and its
composite structures. Compos Struct 2012;94:580–92.
[71] Zhu JY. FRP-confined curvilinearized square concrete columns under axial com-
pression [MPhil. Thesis]. Hong Kong, China: The Hong Kong Polytechnic
University; 2014.
[72] Zhu Z, Ahmad I, Mirmiran A. Effect of column parameters on axial compression
behavior of concrete-filled FRP tubes. Adv Struct Eng 2005;8(4):443–9.