Construction and Building Materials 17 (2003) 471–489

Design-oriented stress–strain model for FRP-confined concrete

L. Lam, J.G. Teng*
Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, Hung Hom, Hong Kong, PR China

Abstract

External confinement by the wrapping of FRP sheets (or FRP jacketing) provides a very effective method for the retrofit of
reinforced concrete (RC) columns subject to either static or seismic loads. For the reliable and cost-effective design of FRP
jackets, an accurate stress–strain model is required for FRP-confined concrete. In this paper, a new design-oriented stress–strain
model is proposed for concrete confined by FRP wraps with fibres only or predominantly in the hoop direction based on a careful
interpretation of existing test data and observations. This model is simple, so it is suitable for direct use in design, but in the
meantime, it captures all the main characteristics of the stress–strain behavior of concrete confined by different types of FRP. In
addition, for unconfined concrete, this model reduces directly to idealized stress–strain curves in existing design codes. In the
development of this model, a number of important issues including the actual hoop strains in FRP jackets at rupture, the
sufficiency of FRP confinement for a significant strength enhancement, and the effect of jacket stiffness on the ultimate axial
strain, were all carefully examined and appropriately resolved. The predictions of the model are shown to agree well with test
data.
䊚 2003 Elsevier Ltd. All rights reserved.

Keywords: Concrete; Fibre reinforced polymer; Confinement; Stress–strain model; Compressive strength; Ultimate strain; Design

1. Introduction confining pressure is assumed, which is the case for

steel-confined concrete when the steel is in plastic flow,
Fibre reinforced polymer (FRP) composites have but not the case for FRP-confined concrete. As FRP
found increasingly wide applications in civil engineering composites remain linear elastic until final rupture, the
due to their high strength-to-weight ratio and high lateral confining pressure in FRP-confined concrete
corrosion resistance. One important application of FRP- increases continuously with the applied load.
composites is as a confining material for concrete, in Many investigations have been conducted into the
both the retrofit of existing reinforced concrete (RC) behavior of FRP-confined concrete and as a result, a
columns by the provision of an FRP jacket, and in number of stress–strain models have been proposed.
concrete-filled FRP tubes in new construction. As a These models can be classified into two categories: (a)
result of FRP confinement, both the compressive design-oriented models w4–11x, and (b) analysis-orient-
strength and ultimate strain of concrete can be greatly ed models w12–14x. In the first category, the compressive
enhanced. In both types of applications, an accurate strength, ultimate axial strain (hereafter, often referred
axial stress–axial strain model (referred to simply as to as ultimate strain for brevity) and stress–strain
stress–strain model hereafter) is required for FRP- behavior of FRP-confined concrete are predicted using
confined concrete for reliable and cost-effective designs. closed-form equations based directly on the interpreta-
In early studies of FRP retrofit of RC columns, the tion of experimental results. In the second category,
stress–strain model of Mander et al. for steel-confined stress–strain curves of FRP-confined concrete are gen-
concrete w1x was directly used in the analysis of FRP- erated using an incremental numerical procedure. In this
confined concrete columns w2,3x. Subsequent studies, second approach, an active confinement model for con-
however, showed that this direct use is inappropriate. crete is used to evaluate the axial stress and strain of
This is because in Mander et al.’s model w1x, a constant passively confined concrete at a given confining pressure
*Corresponding author. Tel.: q852-2766-6012; fax: q852-2334-
and the interaction between the concrete and the confin-
6389. ing material is explicitly accounted for by equilibrium
E-mail address: cejgteng@[Link] (J.G. Teng). and radial displacement compatibility considerations. In

0950-0618/03/$ - see front matter 䊚 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0950-0618(03)00045-X
472 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489

the three studies cited above w12–14x, the model of

Mander et al. w1x was used as the active confinement
model.
Although analysis-oriented models have advantages
in accounting for the interaction between concrete and
confining materials including both steel and FRP com-
posites, the complexity of the incremental process pre-
vents analysis-oriented models from direct use in design.
They are, however, suitable for incorporation in com-
puter-based numerical analysis such as non-linear finite
element analysis. Compared to analysis-oriented models,
design-oriented models are particularly suitable for
direct application in design calculations. A simple and
accurate design-oriented stress–strain model offers an
approach that is familiar to engineers for determining
the strength and ductility of FRP-confined RC structural
members. It may be worth noting that in Eurocode 2
w15x, although a stress–strain model for concrete in
uniaxial compression is provided for structural analysis,
a simpler idealized model is recommended for design
use, which thus represents a similar differentiation
between analysis-oriented and design-oriented models.
This paper is concerned with the development of a
new design-oriented stress–strain model for concrete
confined by an FRP jacket in which the reinforcing
fibres are only or predominantly oriented in the hoop
direction, so that the jacket has little longitudinal stiff-
ness. That is, the jacket can be simplified as a unidirec-
tional material providing only hoop resistance to any
expansion of the concrete. The ultimate condition of Fig. 1. Confining action of (b) FRP jacket to (a) Concrete.
such FRP-confined concrete is reached when the FRP
ruptures due to hoop tensile stresses; failure of insuffi- ent from steel-confined concrete in which the lateral
cient vertical lap joints is excluded from consideration confining pressure is constant following the yielding of
here. Concrete-filled FRP tubes for new construction steel, the confining pressure provided by the FRP jacket
also make use of FRP confinement but they differ increases with the lateral strain of concrete because FRP
significantly in behavior as a result of the substantial does not yield. The confining action in FRP-confined
longitudinal stiffness possessed by the tube (e.g. part of concrete can be schematically illustrated in Fig. 1, where
the hoop strain of the tube comes from its own Poison’s all stresses are shown in their positive directions. In the
effect). Concrete confined by such FRP tubes is thus concrete, compressive stresses and strains are defined as
excluded from consideration here. positive but in the FRP, tensile stresses and strains are
The paper begins with a review of the fundamental positive. The lateral (radial) confining pressure acting
behavior of FRP-confined concrete as established by on the concrete core sr is given by
existing test results, followed by a discussion of the
deficiencies of existing design-oriented stress–strain sht 2sht
srs s (1)
models based on available test observations. A new R d
design-oriented stress–strain model, which overcomes
these deficiencies is then presented and compared with where shstensile stress in the FRP jacket in the hoop
test data. direction, tstotal thickness of the FRP jacket, and R
and dsradius and diameter of the confined concrete
2. Confining action of FRP jacket core, respectively. If the FRP is loaded in hoop tension
only, then the hoop stress in the FRP jacket sh is
In most applications, the lateral confinement provided proportional to the hoop strain ´h due to the linearity of
by an FRP jacket to concrete is passive in nature. When FRP and is given by
the concrete is subject to axial compression, it expands shsEfrp´h (2)
laterally. This expansion is confined by the FRP jacket,
which is loaded in tension in the hoop direction. Differ- where Efrpselastic modulus of FRP.
 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489 473

3. Experimental behavior of FRP-confined concrete axial strains are average values that were obtained either
using strain gauges (up to three) at the mid height of
3.1. Test database the specimens or from relative displacement measure-
ments of the middle region or between the two ends
Many tests have been conducted on FRP-confined using linear-variable differential transformers (LVDTs)
concrete. In the present study, a database containing the (up to two). The FRP hoop strains are also average
test results of 76 FRP-wrapped plain concrete circular values from strain gauges (up to four), or are taken to
specimens was assembled from an extensive survey of be the same as lateral strains deduced from measure-
the open literature (Table 1). These 76 specimens were ments of LVDTs at the mid height of specimens except
reported by Picher et al. w16x, Watanable et al. w17x, those reported by Pessiki et al. w25x. In Pessiki et al.’s
Matthys et al. w18x, Purba and Mufti w19x, Kshirsagar et study, an array of strain gauges was used to measure
al. w20x, Rochette and Labossiere w21x, Xiao and Wu FRP hoop strains, and where possible, average values
w11x, Aire et al. w22x, Dias da Silva and Santos w23x, from a number of gauges of the critical regions (near
Micelli et al. w24x, Pessiki et al. w25x, Wang and Cheong locations of rupture) were reported. It should be noted
w26x, De Lorenzis et al. w27x, and Shehata et al. w28x. that assuming the deformation of the confined concrete
The specimens included in the database have diameters cylinder is truly axisymmetric, the lateral strain and the
d from 100 mm to 200 mm and unconfined concrete hoop strain of the FRP jacket are always equal in
strengths f9co from 26.2 to 55.2 MPa. Two specimens magnitude but opposite in sign according to the present
with a very small diameter (ds55 mm) tested by De sign convention. This is taken to be true in the present
Lorenzis et al. w27x have been excluded. As this study study, despite that a small amount of asymmetry is
is limited to normal strength concrete, 10 specimens unavoidable in the deformation due to factors such as
tested by Aire et al. w22x with f9cos69 MPa have also the inhomogeneity of concrete and eccentricity of
been excluded. It should be noted that the term ‘speci- loading.
men’ is used loosely here for convenience, as some of
the specimens represent the average performance of up 3.2. Failure mode and FRP hoop rupture strain
to three nominally identical physical specimens.
Different types of FRP were used in the specimens All specimens included in Table 1 failed by the
in the database, namely carbon FRP (CFRP), aramid rupture of the FRP jacket due to hoop tension. This is
FRP (AFRP), and glass FRP (GFRP). The carbon fibres the most common mode of failure for FRP-confined
used include high strength and high modulus carbon concrete, although premature failure due to the separa-
fibres. In the following discussion, the FRP prepared tion of the FRP at the vertical lap joint has also been
from high strength carbon fibres is referred to simply as reported for specimens with an insufficient lap length
CFRP, but that prepared from high modulus carbon w19x. Specimens failing by a mode other than FRP
fibres is referred to as HM CFRP. The fibres employed rupture have been excluded from the present database.
were supplied in the form of unidirectional tow sheets In existing models for FRP-confined concrete, it is
(carbon fibres), or woven fabrics with the fibres (aramid commonly assumed that the FRP ruptures when the
and glass fibres) mainly in one direction. They were hoop stress in the FRP jacket reaches its tensile strength
wrapped on concrete cylinders with the main fibres from either flat coupon tests w29x or ring splitting tests
running in the hoop direction, so the resulting FRP w30x which is herein referred to as the FRP material
jacket had an insignificant stiffness in the axial direction. tensile strength. This assumption is the basis for calcu-
A few specimens that were wrapped with FRP jackets lating the maximum confining pressure f l (the confining
with a significant stiffness in the axial direction have pressure reached when the FRP ruptures) using the
been excluded from the database. These include two following equation:
specimens tested by Pessiki et al. w25x with glass fibres
at 0 and "458 from the hoop direction, and three 2ffrpt
fls (3)
specimens tested by Dias da Silva and Santos w23x with d
glass fibre woven fabrics having a fibre thickness of
0.094 mm and 0.040 mm in the circumferential and where f frpsFRP material tensile strength in the hoop
axial directions, respectively. For most specimens (Table direction. The confinement ratio of an FRP-confined
1a), the FRP properties were determined from flat specimen is defined as the ratio of the maximum
coupon tensile tests w29x by researchers themselves. For confining pressure to the unconfined concrete strength
the rest (Table 1b), the FRP properties were supplied (fl yf9co).
by manufacturers. However, experimental results show that in most
Table 1 reports the compressive strength f9cc and cases, the FRP material tensile strength was not reached
ultimate axial strain ´cu of confined concrete, and the at the rupture of FRP in FRP-confined concrete. Table
FRP hoop strain at rupture ´h,rup for all specimens. The 2 provides the average ratios between the measured
 474
Table 1
Test results of FRP-wrapped concrete specimens

No. Source of data d L f9co ´co Fiber type t f frp Efrp ´cc ´cu ´h,rup f9cc f9cu fo
(mm) (mm) (MPa) (%) (mm) (MPa) (GPa) (%) (%) (%) (MPa) (MPa) (MPa)
(a) FRP properties from flat coupon tests by researchers
1 Watanable et al. w17x 100 200 30.2 0.23 Carbon 0.17 2716 224.6 1.51 0.94 46.6 32.0
2 Watanable et al. w17x 100 200 30.2 0.23 Carbon 0.50 2873 224.6 3.11 0.82 87.2 35.0
3 Watanable et al. w17x 100 200 30.2 0.23 Carbon 0.67 2658 224.6 4.15 0.76 104.6 35.0
4 Watanable et al. w17x 100 200 30.2 0.23 HM carbon 0.14 1579 628.6 0.57 0.23 41.7 30.0
5 Watanable et al. w17x 100 200 30.2 0.23 HM carbon 0.28 1824 629.6 0.88 0.22 56.0 36.0

L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489

6 Watanable et al. w17x 100 200 30.2 0.23 HM carbon 0.42 1285 576.6 1.30 0.22 63.3 40.0
7 Watanable et al. w17x 100 200 30.2 0.23 Aramid 0.15 2589 97.1 1.58 2.36 39.0 30.0
8 Watanable et al. w17x 100 200 30.2 0.23 Aramid 0.29 2707 87.3 4.75 3.09 68.5 30.5
9 Watanable et al. w17x 100 200 30.2 0.23 Aramid 0.43 2667 87.3 5.55 2.65 92.1 35.0
10 Matthys et al. w18x 150 300 34.9 0.21 Carbon 0.12 2600 200.0 0.85 1.15 44.3 32.5
11 Matthys et al. w18x 150 300 34.9 0.21 Carbon 0.12 2600 200.0 0.72 1.08 42.2 32.5
12 Matthys et al. w18x 150 300 34.9 0.21 HM carbon 0.24 1100 420.0 0.40 0.19 41.3 31.0
13 Matthys et al. w18x 150 300 34.9 0.21 HM carbon 0.24 1100 420.0 0.36 0.18 40.7
14 Kshirsagar et al. w20x 102 204 38.0 0.22 E-glass 1.42 363 19.9 1.73 1.74 57.0 38.1
15 Kshirsagar et al. w20x 102 204 39.4 E-glass 1.42 363 19.9 1.60 2.07 63.1 40.0
16 Kshirsagar et al. w20x 102 204 39.5 E-glass 1.42 363 19.9 1.79 1.89 60.4 40.0
17 Rochette and Labossiere w21x 100 200 42.0 Carbon 0.60 1265 82.7 1.65 0.89 73.5
18 Rochette and Labossiere w21x 100 200 42.0 Carbon 0.60 1265 82.7 1.57 0.95 73.5
19 Rochette and Labossiere w21x 100 200 42.0 Carbon 0.60 1265 82.7 1.35 0.80 67.6
20 Rochette and Labossiere w21x 150 300 43.0 Aramid 1.27 230 13.6 1.11 1.53 47.3
21 Rochette and Labossiere w21x 150 300 43.0 Aramid 2.56 230 13.6 1.47 1.39 58.9
22 Rochette and Labossiere w21x 150 300 43.0 Aramid 3.86 230 13.6 1.69 1.33 71.0
23 Rochette and Labossiere w21x 150 300 43.0 Aramid 5.21 230 13.6 1.74 1.18 74.4
24 Xiao and Wu w11x 152 305 33.7 Carbon 0.38 1577 105.0 1.20 0.84 47.9 31.2
25 Xiao and Wu w11x 152 305 33.7 Carbon 0.38 1577 105.0 1.40 1.15 49.7 31.2
26 Xiao and Wu w11x 152 305 33.7 Carbon 0.38 1577 105.0 1.24 0.87 49.4 31.2
27 Xiao and Wu w11x 152 305 33.7 Carbon 0.76 1577 105.0 1.65 0.91 64.6 36.0
28 Xiao and Wu w11x 152 305 33.7 Carbon 0.76 1577 105.0 2.25 1.00 75.2 36.0
29 Xiao and Wu w11x 152 305 33.7 Carbon 0.76 1577 105.0 2.16 1.00 71.8 36.0
30 Xiao and Wu w11x 152 305 33.7 Carbon 1.14 1577 105.0 2.45 0.82 82.9 38.4
31 Xiao and Wu w11x 152 305 33.7 Carbon 1.14 1577 105.0 3.03 0.90 95.4 38.4
32 Xiao and Wu w11x 152 305 43.8 Carbon 0.38 1577 105.0 0.60 0.98 0.81 54.8 54.4 50.4
33 Xiao and Wu w11x 152 305 43.8 Carbon 0.38 1577 105.0 0.39 0.47 0.76 52.1 51.4 50.4
34 Xiao and Wu w11x 152 305 43.8 Carbon 0.38 1577 105.0 0.28 0.37 0.28 48.7 39.2 50.4
35 Xiao and Wu w11x 152 305 43.8 Carbon 0.76 1577 105.0 1.57 0.92 84.0 50.4
36 Xiao and Wu w11x 152 305 43.8 Carbon 0.76 1577 105.0 1.37 1.00 79.2 50.4
37 Xiao and Wu w11x 152 305 43.8 Carbon 0.76 1577 105.0 1.66 1.01 85.0 50.4
38 Xiao and Wu w11x 152 305 43.8 Carbon 1.14 1577 105.0 1.74 0.79 96.5 50.4
39 Xiao and Wu w11x 152 305 43.8 Carbon 1.14 1577 105.0 1.68 0.71 92.6 50.4
40 Xiao and Wu w11x 152 305 43.8 Carbon 1.14 1577 105.0 1.75 0.84 94.0 50.4
41 Xiao and Wu w11x 152 305 55.2 Carbon 0.38 1577 105.0 0.28 0.69 0.70 57.9 50.3
42 Xiao and Wu w11x 152 305 55.2 Carbon 0.38 1577 105.0 0.41 0.48 0.62 62.9 53.1
43 Xiao and Wu w11x 152 305 55.2 Carbon 0.38 1577 105.0 0.31 0.49 0.19 58.1 52.0
44 Xiao and Wu w11x 152 305 55.2 Carbon 0.76 1577 105.0 0.39 1.21 0.74 74.6 73.1 68.6
45 Xiao and Wu w11x 152 305 55.2 Carbon 0.76 1577 105.0 0.81 0.83 77.6 68.6
46 Xiao and Wu w11x 152 305 55.2 Carbon 1.14 1577 105.0 1.43 0.76 106.5 61.2
47 Xiao and Wu w11x 152 305 55.2 Carbon 1.14 1577 105.0 1.45 0.85 108.0 61.2
48 Xiao and Wu w11x 152 305 55.2 Carbon 1.14 1577 105.0 1.18 0.70 103.3 61.2
49 De Lorenzis et al. w27x 120 240 43 Carbon 0.3 1028 91.1 0.87 1.16 0.70 58.5 55.6
50 De Lorenzis et al. w27x 120 240 43 Carbon 0.3 1028 91.1 0.82 0.95 0.80 65.6 63.5
51 De Lorenzis et al. w27x 150 300 38 Carbon 0.45 1028 91.1 0.71 0.95 0.80 62
52 De Lorenzis et al. w27x 150 300 38 Carbon 0.45 1028 91.1 1.24 1.35 0.80 67.3
(b) FRP properties from manufacturers
53 Picher et al. w16x 152 304 39.7 Carbon 0.36 1266 83.0 1.070 0.84 56.0 42.5

L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489

54 Purba and Mufti w19x 191 788 27.1 Carbon 0.22 3483 230.5 0.576 0.67 53.9 43.8
55 Aire et al. w22x 150 300 42.0 Glass 0.149 3000 65.0 0.30 0.73 0.55 41.0 35.7
56 Aire et al. w22x 150 300 42.0 Glass 0.447 3000 65.0 1.74 1.30 61 35.7
57 Aire et al. w22x 150 300 42.0 Glass 0.894 3000 65.0 2.5 1.10 85 41.1
58 Aire et al. w22x 150 300 42.0 Carbon 0.117 3900 240.0 1.1 0.95 46 40.7
59 Aire et al. w22x 150 300 42.0 Carbon 0.351 3900 240.0 2.26 1.05 77 40.7
60 Aire et al. w22x 150 300 42.0 Carbon 0.702 3900 240.0 3.23 1.06 108 40.7
61 Dias da Silva and Santos w23x 150 600 28.2 Carbon 0.111 3700 240.0 0.39 0.26 31.4 28.2
62 Dias da Silva and Santos w23x 150 600 28.2 Carbon 0.222 3700 240.0 2.05 1.18 57.4 26.8
63 Dias da Silva and Santos w23x 150 600 28.2 Carbon 0.333 3700 240.0 2.59 1.14 69.5 29.6
64 Dias da Silva and Santos w23x 150 600 28.2 HM carbon 0.167 3000 390.0 0.75 0.37 41.5 24
65 Dias da Silva and Santos w23x 150 600 28.2 HM carbon 0.334 3000 390.0 1.81 0.69 65.6 28.2
66 Dias da Silva and Santos w23x 150 600 28.2 HM carbon 0.501 3000 390.0 1.69 0.64 79.4 28.2
67 Micelli et al. w24x 102 204 37 Carbon 0.16 3790 227.0 1.02 1.2 60 37
68 Micelli et al. w24x 102 204 32 Glass 0.35 1520 72.0 1.25 1.25 52 33
69 Pessiki et al. w25x 152 610 26.2 0.22 E-glass 1.00 383 21.6 1.30 1.15 38.4 31.5
70 Pessiki et al. w25x 152 610 26.2 0.22 E-glass 2.00 383 21.6 1.82 1.24 52.5 31.5
71 Pessiki et al. w25x 152 610 26.2 0.22 Carbon 1.00 580 38.1 1.44 0.81 50.6 33.9
72 Pessiki et al. w25x 152 610 26.2 0.22 Carbon 2.00 580 38.1 1.65 0.72 64.0 33.9
73 Wang and Cheong w26x 200 600 27.9 0.16 Carbon 0.36 4400 235.0 1.52 0.85 82.8 32.5
74 Wang and Cheong w26x 200 600 27.9 0.16 Carbon 0.36 4400 235.0 1.43 1.07 81.2 32.5
75 Shehata et al. w28x 150 300 29.8 0.21 Carbon 0.165 3550 235.0 1.23 1.23 57.0 33
76 Shehata et al. w28x 150 300 29.8 0.21 Carbon 0.33 3550 235.0 1.74 1.19 72.1 34

475
476 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489

Table 2 curvature of an FRP jacket on the tensile strength of

Average hoop rupture strain ratios FRP. Shahawy et al. w31x suggested that ring splitting
Type of fibre No. of FRP material Ratio of hoop tests could provide a closer estimation of the hoop
specimens ultimate rupture strain rupture strain of FRP. Thorough studies on these aspects
tensile to FRP are not yet available. The hoop rupture strains reported
strain material in the existing literature are in general average values
´frp from ultimate around the circumference, so strain distributions around
coupon tensile strain
tests ´h,rupy´frp (%) the circumference in FRP-confined specimens are
unclear at this stage.
Average S.D. Average S.D.
CFRP 52 0.0148 0.0015 58.6 15.3 3.3. Stress–strain response
High modulus CFRP 8 0.0045 0.0027 78.8 16.8
AFRP 7 0.0223 0.0068 85.1 9.5
GFRP 9 0.0280 0.0136 62.4 36.4 It has been well recognized that the stress–strain
Total 76 0.0160 0.0080 63.2 20.5 curve of FRP-confined concrete features a monotonically
ascending bi-linear shape as shown in Fig. 3a, if the
hoop strain at FRP rupture ´h,rup and the FRP material amount of FRP exceeds a certain threshold. Such FRP-
ultimate tensile strain ´frp for a number of categories. It confined concrete is said to be sufficiently confined.
is seen that the average ratio differs for specimens This type of stress–strain curves (the increasing type)
confined by a different type of FRP, and has a value of was observed in the vast majority of the tests covered
0.63 when all specimens of the present database are by the present database. With this type of stress–strain
considered together. Thus, the maximum confining pres- curves, both the compressive strength and the ultimate
sure given by Eq. (3) is only a nominal value. The strain are reached at the same point and are significantly
actual maximum confining pressure should be given by enhanced. However, existing tests have also shown that
in some cases such a bi-linear stress–strain behavior
2Efrpt´h,rup cannot be expected. Instead, the stress–strain curve
fl,as (4) features a post-peak descending branch and the com-
d pressive strength is reached before FRP rupture (the
decreasing type). This decreasing type of stress–strain
where f l,a is the actual maximum confining pressure. curves can be further differentiated in terms of the stress
The actual confinement ratio is then given by the ratio in concrete at the ultimate strain f9cu (Fig. 3b,c). If the
between f l,a and f9co. stress–strain curve terminates at a concrete stress f9cu
Table 2 indicates that the assumption that the FRP above the compressive strength of unconfined concrete
ruptures when the stress in the jacket reaches the FRP f9co as illustrated in Fig. 3b, the FRP confinement is still
material tensile strength is invalid for concrete confined sufficient to lead to strength enhancement. Such concrete
by FRP wraps. Fig. 2 shows the ratio of FRP hoop is also referred to as sufficiently confined concrete in
rupture strain from tests to the FRP material ultimate the present study. However, if the stress–strain curve
tensile strain against the actual confinement ratio terminates at a stress f9cu-f9co as illustrated in Fig. 3c,
fl,a yf9co, which illustrates the great scatter displayed by the specimen is said to be insufficiently confined, where
the data. In particular, this figure indicates that in
concrete confined by a small amount of FRP, premature
rupture of the FRP at a hoop strain much lower than
the material ultimate tensile strain is very likely. For
example, the four specimens with very low levels of
FRP confinement failed with hoop rupture strains below
20% of the material ultimate tensile strain. The maxi-
mum value of fl,a yf9co of these four specimens is only
0.034.
The difference between the FRP tensile strength or
ultimate strain from material tests and that reached in
tests of FRP-confined concrete specimens has been
discussed in a number of recent papers w13,25,31–33x.
Several causes have been suggested for this phenome-
non. The two main causes are believed to be (a)
deformation localization in cracked concrete leading to
a non-uniform stress distribution in the FRP jacket and
thus premature rupture of FRP, and (b) the effect of Fig. 2. Hoop strains at rupture of FRP jackets.
 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489 477

and were wrapped with one, two, and three layers of

CFRP, respectively, (specimens 32, 35 and 40 in Table
1a). Specimen E had an unconfined concrete strength
of 33.7 MPa and was wrapped with three layers of
CFRP (specimen 31 in Table 1a). The actual confine-
ment ratios of these five specimens are 0.067, 0.097,
0.221, 0.302 and 0.421, respectively. Further details of
these specimens can be found in Table 1a. Specimen A
exhibits insignificant strength enhancement and the
stress–strain curve terminates at a stress below the
unconfined concrete strength (Fig. 4). Specimen B also
has a decreasing stress–strain response after the peak
stress, but the stress at the ultimate strain is higher than
the unconfined concrete strength. The behavior of this
specimen thus belongs to the type illustrated in Fig. 3b.
Specimens C, D and E all has a stress–strain curve of
the increasing type. For these three specimens, the
enhancement in both the compressive strength and the
ultimate strain increases with the amount of
confinement.
The volumetric change of confined concrete under
axial compression can be represented by the volumetric
strain ´v, which is defined by

´vs´cq´rq´us´cq2´r (5)

where ´uscircumferential strain and s´rslateral (radi-

al) strain. It is commonly known that unconfined con-
crete in axial compression experiences a volumetric
reduction or compaction up to 90% of the peak stress,
but thereafter the concrete shows volumetric expansion
or dilation which becomes unstable after the peak stress
w34,35x. Unstable dilation has also been observed in
actively confined concrete in tri-axial compression tests
w36,37x. Recently, Mirmiran and his co-authors w6,35x
compared the volumetric responses of FRP-confined
Fig. 3. Classification of stress–strain curves of FRP-confined concrete. concrete with those of plain concrete and steel-confined
(a) Increasing type; (b) Decreasing type with f9cu)f9co; (c) Decreasing
type with f9cu-f9co.
concrete. They demonstrated that for steel-confined con-
crete, unstable dilation occurs when steel yields, but for

little strength enhancement can be expected. The behav-

ior of concrete with insufficient FRP confinement has
been observed in some tests conducted by Xiao and Wu
w11x and Aire et al. w22x. For those specimens with the
compressive strength reached before FRP rupture, Table
1 also provides the axial strain at the peak stress ´cc
and the stress of concrete at the ultimate strain f9cu where
available.
A set of stress–strain curves of concrete confined by
different amounts of FRP is shown in Fig. 4, using the
test data obtained by Xiao and Wu w11x, where the axial
stress is normalized by the unconfined concrete strength.
In Fig. 4, specimen A had an unconfined concrete
strength of 55.2 MPa and was wrapped with one layer
of CFRP (specimen 41 of Table 1a). Specimens B, C
and D had an unconfined concrete strength of 43.8 MPa Fig. 4. Typical stress–strain curves of FRP-confined concrete.
478 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489

that such weak confinement should not be allowed in

practical design. Consequently, the new stress–strain
model presented in Section 5 is intended for application
to sufficiently confined concrete only. For this limitation
to be observed in practical design, the minimum amount
of FRP deemed necessary to achieve sufficient confine-
ment needs to be defined.
The issue of insufficient confinement has been dis-
cussed in a number of previous papers. Mirmiran et al.
w38x suggested that for FRP-confined rectangular con-
crete specimens with rounded corners, enhancement in
the compressive strength of confined concrete should
not be expected if the following modified confinement
ratio (MCR) is less than 0.15:

Fig. 5. Volumetric responses of FRP-confined concrete.

B 2R E fl
MCRsC F
c
FRP-confined concrete, the linearly increasing hoop (6)
D D G f9co
stress of FRP can eventually curtail the dilation if the
amount of FRP is large enough.
Fig. 5 illustrates the volumetric changes of FRP- where Rcsradius of rounded corners and Dsside length
confined concrete using the test data of Xiao and Wu of confined square section, f9coscompressive strength of
w11x for the five specimens shown in Fig. 4. A positive
unconfined concrete and f lsequivalent maximum con-
volumetric strain indicates compaction while a negative fining pressure. For circular specimens, this criterion
value corresponds to dilation. It can be seen that for reduces to fl yf9co-0.15. However, Spoelstra and Monti
specimens A, B and C, the volumetric strain changes w13x showed that the stress of concrete at the ultimate
from compaction to dilation at an axial stress above the
strain f9cu falls below f9co if fl yf9co-0.07 using their
compressive strength of unconfined concrete, and this
analysis-oriented model. Besides, based on experimental
dilation continues to increase until failure. For specimen
data, Xiao and Wu w11x suggested that for FRP-confined
D, dilation is taken over by compaction at a normalized 2
axial stress of approximately 1.75. For specimen E, no
concrete with EfrptyRf9co -0.2 (MPay1), a post-peak
dilation is found during the entire loading history. The descending branch could be expected.
different volumetric responses of FRP-confined concrete In Table 1, a total of five specimens have a concrete
from steel-confined concrete are due to the linear elastic stress at the ultimate strain f9cu below the unconfined
behavior of FRP. This linearity of FRP provides a concrete strength f9co. The nominal confinement ratio
continuously increasing confining pressure until rupture, fl yf9co is less than 0.15 for thirteen of the specimens,
and limits the lateral strain of confined concrete to the including four with f9cu yf9co-1 (Fig. 6a). Another spec-
hoop rupture strain of FRP. As a result, concrete con- imen showing f9cu yf9co-1 has a nominal confinement
fined by a large amount of FRP may not show dilation ratio of fl yf9cos0.18. Fig. 6b shows that in terms of the
at all. actual confinement ratio, of the seven specimens with
an actual confinement ratio fl,a yf9co-0.07, five have
3.4. Minimum amount of FRP for sufficient confinement f9cu yf9co-1 (Fig. 6b). The maximum value of fl,a yf9co for
these five specimens is 0.069. Fig. 7 shows that the
As explained above, FRP-confined concrete with a criterion of Xiao and Wu w11x for insufficient confine-
2
stress–strain curve of the decreasing type and with a ment (EfrptyRf9co -0.2) is met by five specimens,
concrete stress at the ultimate strain below the compres- including four with f9cu yf9co-1 and one with f9cu yf9cos
sive strength of unconfined concrete is taken to be 1.01. For the other specimen with f9cu yf9co-1.01,
2
insufficiently confined, as in such cases little strength EfrptyRf9co s0.275.
enhancement can be expected from the FRP confinement Based on the above discussion, it can be concluded
and the FRP is likely to rupture at a small hoop strain. that in judging whether sufficient confinement is avail-
The latter phenomenon, believed to be due to the able, Spoelstra and Monti’s w13x criterion for insufficient
sensitivity of a weak jacket to the non-uniform defor- confinement can be used, with the maximum confining
mation of concrete, is particularly important as it means pressure being the actual rather than the nominal value.
that the use of such a weak jacket leads to little strength That is, FRP-confined concrete with an actual confine-
or strain enhancement, and any enhancement cannot be ment ratio fl,a yf9coG0.07 can be taken to be sufficiently
reliably predicted. It is, therefore, recommended here confined.
 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489 479

shape. A smooth transition between the two portions is

also provided. Based on the same general equations,
two models were proposed separately for FRP-wrapped
concrete and concrete-filled FRP tubes by calibrating
the model parameters with corresponding test data. The
models of Samaan et al. w6x, Toutanji w9x and Saafi et
al. w8x can predict the shape of a bilinear stress–strain
curve reasonably closely, provided their predictions of
the compressive strength and ultimate strain are accurate.
However, the relative complexity of these three models
in form means inconvenience or difficulty in section
analysis for the determination of section capacity or
ductility, where integration of the stress distribution over
the section is required.
Miyauchi et al. w7x used Hognestad’s parabola w39x
followed by a straight line to describe both the increas-
ing and decreasing types of stress–strain curves of FRP-
confined concrete. This parabola, given by the following
equation, is commonly adopted in codes of practice
such as BS 8110 w40x and Eurocode 2 w15x to describe
the ascending part of the stress–strain curve of uncon-
fined concrete for design use:

w
2´c B ´c E2z
scsf9cox yC F | (7)
y ´co D ´co G ~

Fig. 6. Stress of concrete at ultimate strain vs. confinement ratio. (a)

where sc and ´c are the axial stress and strain, respec-
Variation with nominal confinement ratio; (b) Variation with actual tively, and ´co is the axial strain at the peak stress of
confinement ratio. concrete. However, the direct use of Hognestad’s parab-
ola as adopted in Miyauchi et al.’s model w7x cannot
4. Deficiencies of existing design-oriented stress– reflect the process of gradual development of confine-
strain models ment. In fact, the FRP confinement is activated once
micro-cracks in concrete are initiated under loading.
4.1. Shape of stress–strain curve Lillistone and Jolly w10x attempted to account for this
effect in their stress–strain model for concrete-filled
FRP tubes in which the first portion of the stress–strain
Existing design-oriented stress–strain models for
curve is described using Hognestad’s parabola plus an
FRP-confined concrete w4–11x have adopted different
additional term related to the hoop stiffness of the FRP
approximations to a typical bilinear stress–strain curve.
In the models of Karbhari and Gao w5x and Xiao and
Wu w11x, the two portions of a bilinear curve are
approximated using two straight lines. This approach is
simple but not realistic. Samaan et al. w6x proposed a
model for FRP-confined concrete in which the nearly
linear second portion of the stress–strain curve is char-
acterized by its slope E2 and its intercept with the stress
axis. Another salient feature of this model is that the
stress–strain curve is represented by a single equation,
with the transition from the first portion to the second
portion being controlled by a shape parameter n. The
use of a single equation, however, necessarily leads to
an equation of a more complex form. Toutanji w9x and
Saafi et al. w8x proposed an alternative form for the
stress–strain curve, in which the two portions of a
bilinear curve are approximated using two separate Fig. 7. Stress of concrete at ultimate strain vs. Xiao and Wu’s con-
equations, with both equations producing a curved finement stiffness parameter.
480 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489

tube (EfrptyR), while the second portion is a straight

line. This additional term used to account for the effect
of confinement (not the contribution of the longitudinal
stiffness), being equal to 1.282 Efrpt´c yR, means that
the initial slope of the predicted stress–strain curve can
be significantly greater than that of unconfined concrete,
which is obviously not supported by the test results.
In summary, the present authors believe the best
approach for describing a typical bilinear stress–strain
curve of FRP-confined concrete is to use a modified
parabola for the first portion and a straight line for the
second portion, with the various parameters being
dependent on the FRP properties. The modified parabola
should be able to reflect the gradual development of
confinement as axial stress increases. This same view
has also recently been expressed by Monti w41x. Such a
Fig. 8. Proposed stress–strain model for FRP-confined concrete.
parabola leads to ease in section analysis where integra-
tion of the stress–strain curve is necessary and offers
Thirdly, as a result of the above two deficiencies and
an approach that is familiar to engineers. For the
due to the use of a limited database, there is room for
definition of the linear second portion, the use of its
improvement to the accuracy of the predictive equations
slope and its intercept with the stress axis as done by
for the compressive strength and ultimate strain of FRP-
Samaan et al. w6x offers a rational and simple approach.
confined concrete in existing design-oriented models
using a larger test database.
4.2. Definition of ultimate condition
In summary, improvements to existing stress–strain
models in defining the ultimate condition should be
Central to any stress–strain model for FRP-confined
made in three aspects: (a) the actual hoop rupture strain
concrete is the determination of the ultimate condition
should be used instead of the ultimate material tensile
of FRP-confined concrete, which is reached when the
strain; (b) the effect of jacket stiffness should be
FRP ruptures. This ultimate condition is characterized
explicitly and properly reflected; and (c) the ultimate
by two parameters: the ultimate axial strain and the
strain and compressive strength equations should be
corresponding stress level, which is generally but not
based on the largest test database which can be assem-
always the compressive strength of FRP-confined con-
bled from the open literature.
crete. There are three major deficiencies in existing
design-oriented stress–strain models in predicting the
ultimate condition of confined concrete. 5. Assumptions and general equations of new model
Firstly, it is commonly assumed that rupture of FRP
occurs when the hoop stress in the FRP jacket reaches 5.1. Assumptions
the tensile strength determined from material tests, with
the only exception being Xiao and Wu’s model w11x. A new stress–strain model is proposed here for FRP-
This assumption is, however, not valid as shown in the confined concrete based on the various observations
preceding section, and leads to difficulty in producing a discussed in the preceding sections. The basic assump-
unified stress–strain model for FRP-confined concrete tions of this simple model are: (i) the stress–strain
as the ratio of hoop rupture strain to FRP material curve consists of a parabolic first portion and a straight-
tensile strain varies with the type of FRP (Table 2). line second portion, as given in Fig. 8; (ii) the slope of
Existing analysis-oriented models also suffer from this the parabola at ´cs0 (initial slope) is the same as the
deficiency, so they are also incapable of accurate predic- elastic modulus of unconfined concrete Ec; (iii) the non-
tions of the ultimate condition. linear part of the first portion is affected to some degree
Secondly, the effect of the stiffness of the FRP jacket by the presence of an FRP jacket; (iv) the parabolic
on the ultimate condition has not been well established first portion meets the linear second portion smoothly
and explicitly accounted for, although it is implied to (i.e. there is no change in slope between the two portions
some degree in the ultimate strain equations of Samaan where they meet); (v) the linear second portion ends at
et al. w6x, Toutanji w9x and Saafi et al. w8x. The stiffness a point where both the compressive strength and the
of the FRP jacket in fact has an important effect on the ultimate axial strain of confined concrete are reached.
stress–strain response of FRP-confined concrete, partic- These basic assumptions of the proposed model are
ularly the ultimate axial strain as shown later in the in accordance with the test observations of FRP-confined
paper. concrete with a monotonically increasing stress–strain
 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489 481

curve as illustrated in Fig. 3a. The first assumption leads

to a stress–strain curve which is similar to those adopted
by existing design codes for unconfined concrete and
thus familiar to engineers. The second assumption is to
account for the fact that the initial stiffness of FRP-
confined concrete is little affected by the FRP due to
the passive nature of confinement. The third assumption
is to reflect the fact that the FRP confinement is
activated when the behavior of the concrete becomes
non-linear. This third assumption makes the new model
different from Miyauchi et al.’s model w7x in which the
shape of the parabola remains the same as that for
unconfined concrete and is not affected by the FRP
confinement at all. The fourth assumption ensures a
smooth stress–strain curve, while the last assumption is
obviously valid for FRP-confined concrete with a mon- Fig. 9. Stress–strain curves predicted by Spoelstra and Monti’s anal-
otonically increasing stress–strain curve. For FRP-con- ysis-oriented model for concrete confined by different materials.
fined concrete whose behavior is as illustrated in Fig.
3b, the last assumption is not reflective of reality, but defined are: the ultimate strain ´cu, the compressive
the present model provides a good approximation for strength f9cc, and the intercept of the stress axis by the
design use. It may be noted that even for unconfined second linear portion f o.
concrete, the design stress–strain curve is represented
by a parabola followed by a horizontal straight line in 6. Ultimate strain of FRP-confined concrete
both BS 8110 w40x and Eurocode 2 w15x, despite that
test stress–strain curves display a descending post-peak 6.1. Theoretical basis
branch.
Based on the assumptions listed above, the proposed Studies on actively confined concrete and steel-con-
stress–strain model for FRP-confined concrete is given fined concrete showed that the axial strain at the
by the following expressions: compressive strength of confined concrete ´cc can be
linearly related to the maximum confining pressure
ŽEcyE2.2 w42,43x. This approach has been adopted in some design-
scsEc´cy ´c2 for 0F´cF´t (8a) oriented models w5,7x for predicting the axial strain at
4fo
the compressive strength of FRP-confined concrete ´cc,
which in most cases is also the ultimate strain of FRP-
and
confined concrete ´cu. A deficiency of this approach is
that the stiffness of the confining jacket is not properly
scsfoqE2´c for ´tF´cF´cu (8b)
accounted for. This seems not important for steel-
confined concrete because the elastic moduli of all types
where f osintercept of the stress axis by the linear
of steel are similar, but is important for FRP-confined
second portion. The parabolic first portion meets the
concrete as the elastic modulus of FRP varies over a
linear second portion with a smooth transition at ´t,
wide range. A number of studies w6,13,44x have noticed
which is given by
that similar levels of lateral pressure do not result in
similar ultimate strains of confined concrete. Although
2fo
´ts (9) the effect of jacket stiffness has not been properly
E
Ž cyE2. accounted for in design-oriented models, it is always
accurately represented in analysis-oriented models
where E2 is the slope of the linear second portion, given through equilibrium and compatibility considerations of
by the concrete and the jacket. This issue can thus be
explained by making use of predictions from an analysis-
f9ccyfo oriented model.
E 2s (10)
´cu Fig. 9 shows four stress–strain curves predicted by
the analysis-oriented model of Spoelstra and Monti w13x
where f9ccscompressive strength of confined concrete. for concrete cylinders confined by three different con-
This proposed model allows the use of test values or fining materials: steel, CFRP and GFRP with their
values suggested by design codes for the elastic modulus properties given in Table 3. For confinement by CFRP,
of unconfined concrete. The three parameters yet to be predictions are provided for two scenarios: the FRP
482 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489

Table 3
Properties of confining materials used for predictions shown in Fig. 9

Confining Elastic modulus Rupture or yield Rupture or yield Thickness of confining

material (MPa) stress (MPa) strain (%) jacket (mm)
Steel 2=105 300 0.15 4
CFRP 2.35=105 3530 1.5 0.34
CFRP 2.35=105 2115 0.9 0.567
(actual rupture strain)
GFRP 23 100 462 2.0 2.6

ruptures at its material ultimate tensile strain from axial strain of concrete, ´2s´rslateral (radial) strain
coupon tests and the FRP ruptures at an assumed hoop of concretes´3s´uscircumferential strain of concrete,
rupture strain of 60% of the material ultimate tensile s1sscscompressive stress of concrete, and s2ss3s
strain. The compressive strength of unconfined concrete srslateral confining pressure. The following equation
is 35 MPa, while the diameter of the cylinders is 150 can then be obtained from Eqs. (11a), (11b) and (11c):
mm. For all four cylinders, the FRP jackets are assumed
to supply the same ultimate tensile capacity in the hoop 2
´r Ž[Link]
direction and thus the same maximum confining pres- ´csy q (12)
nsec nsecEsec
sure, but they have different stiffnesses. The substantial
differences between the predicted responses including
the ultimate strain are due to the differences in the Further, as sr can be expressed as a function of the
stiffness of the four jackets only. While one may argue hoop strain in the FRP according to Eq. (1) and Eq.
that analysis-oriented models can be inaccurate, the (2), there is
differences as shown in Fig. 9 are obviously too large 2
to be attributed to the inaccuracy of the active confine- ´r Ž1ynsecy2nsec. Efrpt´h
´csy q (13)
ment model since the same concrete is modeled and the nsec nsecEsec R
stress–strain curve predicted by Spoelstra and Monti’s
model w13x for a similar CFRP-wrapped specimen has Under the ultimate condition of FRP rupture, ´rs
previously been shown to match the test curve closely y´h,rup. The secant modulus of elasticity of concrete
w33x. under the ultimate condition Esecu is given by w45x.
The dependence of the ultimate strain of FRP-con-
fined concrete on the stiffness of the confining jacket Eseco
can also be shown by examining the constitutive model Esecus (14)
1q4ŽAy1.x
for concrete under a triaxial state of stress proposed by
Ottosen w45x. This model is based on non-linear elastic-
ity, with the properties of concrete being represented by where Esecossecant modulus of elasticity at the com-
the secant values of elastic modulus and Poisson’s ratio. pressive strength of unconfined concrete and sf9co y ´co;
This is the constitutive model recommended by the As2 if the Hognestad’s w39x parabola is assumed for
CEB-FIP Model Code 1990 w46x for concrete under the ascending part of the stress–strain curve of uncon-
fined concrete; and x is given by
multi-axial stresses. The basic equations of the model
are:
xsŽyJ2 yf9co.fy1y y3 (15)

1 w where J2 is the second invariant of the deviatoric stresses

´ 1s x

s1ynsecŽs2qs3.z~ |

(11a)
Esec y 1
ands =Žscysr.2 for confined concrete. The term
3
´ 2s
1
ws2ynsec(s1qs3)x (11b)
ŽyJ2 yf9co.f denotes the value of the invariant under the
Esec ultimate condition of FRP-confined concrete, so
scsf9cc and srsf l,a. The compressive strength of con-
fined concrete can be expressed in the following com-
1 w mon form w47x:
´ 3s x

s3ynsecŽs1qs2.z~ |

(11c)
Esec y
f9cc fl,a
s1qk1 (16)
where Esecssecant modulus of elasticity and nsecs f9co f9co
secant Poisson’s ratio. For confined concrete, ´1s´cs
 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489 483

6.2. Proposed equation

Eq. (18) shows clearly the dependence of the ultimate

strain of FRP-confined concrete ´cu on the confinement
stiffness ratio and the strain capacity of FRP at hoop
rupture. Existing test data on FRP-confined concrete
also show that the secant Poisson’s ratio of FRP-
confined concrete under the ultimate condition nsecu
depends strongly and predominantly on the confinement
stiffness ratio (Fig. 11). Thus, the ultimate strain of
FRP-confined concrete ´cu can be taken to be a function
only of the confinement stiffness ratio Efrpty ŽEsecoR. and
the strain ratio ´h,rup y ´co. The following general equation
is, therefore, proposed to predict the ultimate strain:
Fig. 10. Definition of secant modulus of elasticity.
B Efrpt EaB ´h,rup Eb
´cu y ´coscqk2C F C F (20)
where k1sconfinement effectiveness coefficient. Eq. D EsecoR G D ´co G

(15) then becomes

where csnormalized ultimate strain of unconfined con-
crete, k2sstrain enhancement coefficient, and a and b
Žk1y1. B fl,a E Žk1y1. B Efrpt EB ´ E
are exponents to be determined. The effect of the secant
xs C Fs C FC h,rup
F (17)
y3 D f9co G y3 D EsecoR GD ´co G Poisson’s ratio nsecu is reflected by the choice of appro-
priate values for a, b and k2.
Eq. (20) explicitly accounts for the stiffness and the
Substituting Eqs. (14) and (17) into Eq. (13), the actual ultimate condition of the jacket. If both a and b
following equation can be obtained for the normalized are taken as unity, Eq. (20) reduces to
ultimate strain of confined concrete:
fl,a
2
Ž1ynsec uy2nsecu. B Efrpt E
´cu y ´coscqk2 (21)
´cu ´h,rup f9co
s q C F
´co nsecu´co nsecu D EsecoR G

B´ E
2
4Žk1y1. Ž1ynsecuy2nsecu . which then relates the normalized ultimate strain to the
=C Fq
h,rup
actual confinement ratio only.
D ´co G y3 nsecu
B Efrpt E2B ´ E2 6.3. Determination of a and b
=C F C h,rup F (18)
D EsecoR G D ´co G
The values of the two exponents, a and b, are
determined here using the test data of the present
where nsecussecant Poisson’s ratio of the confined
concrete under the ultimate condition and the term Efrpty
EsecoR is the confinement stiffness ratio, representing
the stiffness ratio between the FRP jacket and the
concrete core.
The definitions of the secant moduli Esec and Esecu
are given in Fig. 10, which can be found by subtracting
Eq. (11b) from Eq. (11a). The secant Poisson’s ratio
can be found from the constitutive equations (Eqs. (11a),
(11b) and (11c)) and related to the lateral-to-axial strain
ratio ´r y ´c through the following equation:

sr ´r
y
sc ´c
nsecs (19)
´r sr sr
1y2 q
´c sc sc
Fig. 11. Dependence of ultimate secant Poisson’s ratio on confinement
stiffness ratio.
484 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489

Fig. 13a,b show that the trends of the test data are
similar for both CFRP and AFRP wraps. For HM CFRP-
wrapped specimens (Fig. 13c), if one of the three
specimens tested by Dias da Silva and Santos (2001) is
excluded (shaded) as a statistical outlier, the remaining
specimens also show a trend similar to those observed
for CFRP or AFRP wraps. A large scatter is observed
for GFRP-wrapped specimens (Fig. 13d), the cause of
which is difficult to pinpoint at the present, but the
predictions of Eq. (22) are well covered by the scatter
of the test data. When all test data are plotted together
(Fig. 13e), a close overall agreement between the test
data and Eq. (22) is observed. Eq. (22), therefore,
provides a unified expression for the ultimate strain of
FRP-confined concrete that is applicable to different
types of FRP. Obviously, more test data for GFRP wraps
should be obtained in the future for further verification
of Eq. (22).

7. Compressive strength of confined concrete

The compressive strength of FRP-confined concrete

has been discussed in detail in Lam and Teng (2002b).
In that paper, the compressive strength of FRP-confined
concrete f9cc is related to the nominal confinement ratio
through
f9cc fl
s1q2 (23)
Fig. 12. Strain enhancement ratio vs. actual confinement ratio. (a) f9co f9co
CFRP wraps; (b) AFRP wraps.
Fig. 14 is a plot of the strengthening ratio f9cc yf9co
database. In determining these exponents, the strain at against the actual confinement ratio fl,a yf9co of the
the compressive strength of unconfined concrete ´co was present test data. The trend line of these test data can
taken as 0.002, a value, which was assumed in all data be closely approximated using the following equation:
interpretation in the present study. f9cc fl,a
Fig. 12 shows two plots of the strain enhancement s1q3.3 (24)
ratio against the actual confinement ratio from the tests f9co f9co
for CFRP wraps and AFRP wraps, respectively. A linear This equation implies that in terms of the actual
relationship clearly exists in both cases but the two confinement ratio, the confinement effectiveness coeffi-
trend lines are very different. These diagrams indicate cient k1 becomes 3.3 instead of 2. It should be noted
that the ultimate strain of FRP-confined concrete can be that the stiffness of the confining jacket also has an
related linearly to the actual confinement ratio for a effect on the compressive strength of concrete with the
given type of FRP, but separate expressions are needed same confinement ratio, as demonstrated in Fig. 9. As
for different types of FRP due to differences in stiffness. this effect on the compressive strength is far less than
To achieve a unified expression for the ultimate strain that on the ultimate strain, it is ignored in Eq. (24).
of FRP-confined concrete, the confinement stiffness ratio It should be reminded that a significant strength
needs to be included, which means that the exponents, enhancement can only be expected with an actual
a and b, cannot both be unity. Thus, the following confinement ratio fl,a yf9coG0.07. The use of Eq. (24) is
expression is suggested for FRP-wrapped concrete based recommended to be subjected to this condition. For the
the trends of the test data: case of FRP-confined concrete with fl,a yf9co-0.07, no
strength enhancement is assumed.
B Efrpt EB ´h,rup E1.45
´cu y ´cos1.75q12C FC F 8. Intercept of the stress axis by the linear second
D EsecoR GD ´co G
portion
Bf EB ´ E0.45
s1.75q12C FC F
l,a h,rup
(22)
D f9co GD ´co G The intercept of the stress axis by the linear second
portion, f o, is an important parameter in the proposed
 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489 485

Fig. 13. Performance of proposed equation for the ultimate strain of FRP-confined concrete. (a) CFRP wraps; (b) AFRP wraps; (c) HM CFRP
wraps; (d) GFRP wraps; (e) all specimens.

stress–strain model as it, together with the ultimate 1.0 and 1.2 in most cases, which appear to be independ-
point, determines the slope of the second portion. ent of the confinement ratio. For the 63 specimens, for
Samaan et al. w6x proposed the following expression of which the values of the intercept are available, the
f o based on experimental data available to him: average ratio of fo yf9co is 1.09 with a standard deviation
of 0.13. It is, therefore, suggested for simplicity that in
fos0.872f9coq0.371flq6.258 (MPa) (25) the proposed model

In the present database, values of f o obtained by the fosf9co (26)

present authors from test stress–strain curves have been
included. These values are normalized by the compres- It should be noted that with fosf9co, the second linear
sive strength of unconfined concrete f9co and are plotted portion of the proposed stress–strain model reduces to
against the actual confinement ratio fl,a yf9co as shown in a horizontal straight line (i.e. E2s0) as assumed in both
Fig. 15. The values of fo yf9co are seen to fall between BS 8110 w40x and Eurocode 2 w15x, and at the same
486 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489

plain cylinder tests, be conducted to determine the

efficiency factor of a given FRP product in confinement
applications. For this purpose, a standard confined cyl-
inder test method should be formulated in the future.
FRP manufacturers can then supply the results of such
tests as part of their product information. If this infor-
mation is not available from the manufacturer, the user
should conduct these tests instead.

10. Comparison with test data

The proposed stress–strain model for FRP-confined

concrete is compared with the test data obtained by
Fig. 14. Strengthening ratio vs. actual confinement ratio.
Xiao and Wu w11x on CFRP-wrapped concrete cylinders,
as shown in Fig. 16. Details of these specimens can be
time, Eq. (8a) reduces to Hognestad’s parabola for found in Table 1a. Fig. 16a and b are for specimens
unconfined concrete. That is, the present model for FRP- having an unconfined concrete strength f9co of 55.2 MPa
confined concrete reduces directly to the design stress– and wrapped with one (specimens 41–43) and two
strain models in BS 8110 w40x and Eurocode 2 w15x for (specimens 44 and 45) layers of CFRP, respectively.
unconfined concrete, provided the same initial elastic Fig. 16c is for specimens having f9cos43.8 MPa and
modulus is used. wrapped with three layers of CFRP (specimens 38–40),
and Fig. 16d is for specimens having f9cos33.7 MPa
9. FRP efficiency factor and wrapped with three layers of CFRP (specimens 30
and 31). Each figure shows test stress–strain curves
In the preceding sections, the definitions of the ulti- from nominally identical specimens and predictions of
mate strain, compression strength and minimum amount the present model. For each case, two predicted curves
of FRP for sufficient confinement for the proposed are shown: one with the FRP hoop rupture strain ´h,rup
stress–strain model are all in terms of the actual con- estimated on the basis of the average efficiency factor
finement ratio, so the actual hoop rupture strain of the for CFRP of 0.586 wrapped specimens (Table 2) and
FRP is required. To facilitate the application of the the other with the hoop rupture strain being the average
proposed stress–strain model, an FRP efficiency factor of the actual values recorded during the tests. The elastic
is, therefore, proposed here, which is defined as the modulus of unconfined concrete was taken as Ecs
ratio of the actual FRP hoop rupture strain (´h,rup) in 4730yf9co w48x and the strain at the compressive strength
FRP-confined concrete to the FRP rupture strain from of unconfined concrete was taken as ´cos0.002 for
flat coupon tests (´frp). For the 52 CFRP-wrapped calculating the ultimate strain from Eq. (22). The
specimens, out of the total of 76 specimens in the specimens used in Fig. 16a had an average actual
present database (Table 1), this efficiency factor is 0.586 confinement ratio of fl,a yf9cos0.048, so these specimens
on average. Making use of this efficiency factor, the
ultimate strain of CFRP-confined concrete can be
expressed as

Bf EB ´ E0.45
´cu y ´cos1.75q5.53C FC F
l,a frp
(27)
D f9co GD ´co G

With this equation, the user only needs to know the

tensile strain ´frp from flat coupon tests. This average
efficiency factor of 0.586 for CFRP-wrapped concrete
also unifies Eq. (23) and Eq. (24).
For other types of FRP, insufficient information exists
to define this efficiency factor with confidence. Even
for CFRP-confined concrete, there is a considerable
scatter in the efficiency factor deduced from test results.
The present authors, therefore, recommend that for cost-
effective and safe applications, a small number (say 3) Fig. 15. Intercept of stress axis by the second linear portion vs. actual
of FRP-confined concrete cylinder tests, analogous to confinement ratio.
 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489 487

Fig. 16. Comparison between proposed model and test stress–strain curves. (a) fl,ayf9cos0.048; (b) fl,ayf9cos0.150 ; (c) fl,ayf9cos0.281; (d)
fl,af9cos0.403.

are insufficiently confined ones. Consequently, k1s0 ple and accurate stress–strain model for FRP-confined
was used in predicting the compressive strength of concrete has been presented for design use. The results
confined concrete. The average actual confinement ratios and discussions presented in this paper also allow the
for the specimens included in Fig. 16b,c and d are following conclusions to be drawn:
0.150, 0.281 and 0.403, respectively. For these cases,
k1s3.3 was used. It can be seen from Fig. 16 that the 1. The average hoop strain in FRP at rupture in FRP-
predictions compare well with the test results. It should wrapped concrete can be much lower than the FRP
be noted that the ultimate strains of specimens 41–43 material ultimate tensile strain from flat coupon tests,
are all considerably overestimated using the estimated indicating the assumption that FRP ruptures when the
´h,rup (Fig. 16a), which further justifies the exclusion of FRP material tensile strength reached is not valid in
such insufficiently confined concrete from practical con- the case of concrete confined by wrapped FRP. Based
siderations and from the range of applicability of the on this observation, a unified stress–strain model for
proposed stress–strain model. concrete confined by different types of FRP must be
based on the actual hoop rupture strain of FRP rather
11. Conclusions than the ultimate material tensile strain.
2. The stress–strain curve of FRP-confined concrete can
This paper has been concerned with the development be in one of several forms, but in the vast majority
of a stress–strain model for concrete confined by of cases, this curve is or can be approximated as a
wrapped FRP with fibres only or predominantly in the monotonically ascending bi-linear curve. Such FRP-
hoop direction. Existing experimental data have been confined concrete is said to be sufficiently confined.
thoroughly reviewed and discussed, and the deficiencies Any FRP-confined concrete with an actual confine-
of existing stress–strain models are highlighted. A sim- ment ratio less than 0.07 is said to be insufficiently-
488 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489

confined. Such concrete is not expected to possess a w7x Miyauchi K, Inoue S, Kuroda T, Kobayashi A. Strengthening
compressive strength significantly above that of effects of concrete columns with carbon fiber sheet. Trans Jpn
Concr Institute 1999;21:143 –50.
unconfined concrete and the FRP may rupture at a w8x Saafi M, Toutanji HA, Li Z. Behavior of concrete columns
low hoop strain. Such insufficiently-confined concrete confined with fiber reinforced polymer tubes. ACI Mater J
should not be allowed in design. 1999;96(4):500 –9.
3. The new design-oriented stress–strain model pro- w9x Toutanji HA. Stress–strain characteristics of concrete columns
posed in this paper is in a form that is familiar to externally confined with advanced fiber composite sheets. ACI
engineers. This model is simple, so it is suitable for Mater J 1999;96(3):397 –404.
w10x Lillistone D, Jolly CK. An innovative form of reinforcement
direct use in design, but in the meantime, it captures
for concrete columns using advanced composites. Struct Eng
all the main characteristics of the stress–strain behav- 2000;78(23–24):20 –8.
ior of FRP-confined concrete. The model reduces w11x Xiao Y, Wu H. Compressive behavior of concrete confined by
directly to idealized stress–strain curves adopted by carbon fiber composite jackets. J Mater Civ Eng ASCE
existing design codes for unconfined concrete. 2000;12(2):139 –46.
4. The stiffness of the FRP jacket and the actual ultimate w12x Mirmiran A, Shahawy M. A new concrete-filled hollow FRP
composite column. Compos Part B: Eng 1996;27B(3–4):263 –
condition of the jacket are explicitly accounted for in
8.
the proposed model. As a result, the proposed model w13x Spoelstra MR, Monti G. FRP-confined concrete model. J
is applicable to concrete confined by different types Compos Constr, ASCE 1999;3(3):143 –50.
of FRP. The predictions of the model have been w14x Fam AZ, Rizkalla SH. Confinement model for axially loaded
shown to agree well with a set of test data. concrete confined by circular fiber-reinforced polymer tubes.
5. For the application of the proposed model in design, ACI Struct J 2001;98(4):451 –61.
w15x ENV 1992-1-1. Eurocode 2: Design of Concrete Structures—
the FRP efficiency factor (ratio between the actual
Part 1: General Rules and Rules for Buildings, European
hoop rupture strain of FRP in FRP-confined concrete Committee for Standardization, Brussels, 1991.
and the ultimate tensile strain from material tests) w16x Picher F, Rochette P, Labossiere P. Confinement of concrete
needs to be established. For this purpose, confined cylinders with CFRP. In: Saadatmanesh H, and Ehsani MR.
cylinder tests have been suggested as a standard type editors. Proceedings of the First International Conference on
of test to supply this information. Composites for Infrastructures, Tucson, Arizona, USA: Uni-
versity of Arizona, 1996:829–841.
w17x Watanable K, Nakamura H, Honda T, Toyoshima M, Iso M,
Acknowledgments Fujimaki T,et al., Confinement effect of FRP sheet on strength
and ductility of concrete cylinders under uniaxial compression.
In: Non-Metallic (FRP) Reinforcement for Concrete Structures,
The work presented in this paper forms part of a Proceedings of the Third International Symposium, vol. 1,
research project (Project No: PolyU 5064y01E) funded Sapporo, Japan: Japan Concrete Institute, 1997:233–240.
by the Research Grants Council of Hong Kong SAR. w18x Matthys S, Taerwe L, Audenaert K. Tests on axially loaded
The first author has been financially supported by The concrete columns confined by fiber reinforced polymer sheet
Hong Kong Polytechnic University through a postdoc- wrapping. In: Dolan CW, Rizkalla SH, and Nanni SH, editors,
Proceedings of the Fourth International Symposium on Fiber
toral fellowship and through the Area of Strategic
Reinforced Polymer Reinforcement for Reinforced Concrete
Development (ASD) Scheme. The authors are grateful Structures, SP-188, Farmington, Michigan, USA: American
to both organizations for their financial support. Concrete Institute, 1999:217–229.
w19x Purba BK, Mufti AA. Investigation of the behavior of circular
References concrete columns reinforced with carbon fiber reinforced poly-
mer (CFRP) jackets. Can J Civ Eng 1999;26:590 –6.
w1x Mander JB, Priestley MJN, Park R. Theoretical stress–strain w20x Kshirsagar S, Lopez-Anido RA, Gupta RK. Environmental
model for confined concrete. J Struct Eng, ASCE aging of fiber-reinforced polymer-wrapped concrete cylinders.
1988;114(8):1804 –26. ACI Mater J 2000;97(6):703 –12.
w2x Saadatmanesh H, Ehsani MR, Li MW. Strength and ductility w21x Rochette P, Labossiere P. Axial testing of rectangular column
of concrete columns externally reinforced with fiber composite models confined with composites. J Compos Constr, ASCE
straps. ACI Struct J 1994;91(4):434 –47. 2000;4(3):129 –36.
w3x Seible F, Burgueno A, Abdallah MG, Nuismer R. Advanced w22x Aire C, Gettu R, Casas JR. Study of the compressive behavior
composites carbon shell system for bridge columns under of concrete confined by fiber reinforced composites. In:
seismic loads. In: Proceedings, National Seismic Conference Figueiras J, Juvandes L, Faria R, Marques AT, Ferreira A,
on Bridges and Highways: San Diago, USA, 1995. Barros J, Appleton J, editors, Composites in Constructions,
w4x Fardis MN, Khalili H. FRP-encased concrete as a structural Proceedings of the International Conference, Lisse, The Neth-
material. Mag Concr Res 1982;34(122):191 –202. erlands: A.A. Balkema Publishers, 2001:239–243.
w5x Karbhari VM, Gao Y. Composite jacketed concrete under w23x Dias da Silva V, Santos JMC. Strengthening of axially loaded
uniaxial compression–verification of simple design equations. concrete cylinders by surface composites. In: Figueiras J,
J Mater Civ Eng, ASCE 1997;9(4):185 –93. Juvandes L, Faria R, Marques AT, Ferreira A, Barros J,
w6x Samaan M, Mirmiran A, Shahawy M. Model of concrete Appleton J, editors, Composites in Constructions, Proceedings
confined by fiber composite. J Struct Eng, ASCE of the International Conference, Lisse, The Netherlands: A.A.
1998;124(9):1025 –31. Balkema Publishers, 2001:257–262.
 L. Lam, J.G. Teng / Construction and Building Materials 17 (2003) 471–489 489

w24x Micelli F, Myers JJ, Murthy S. Effect of environmental cycles ence, Lisse, The Netherlands: A.A. Balkema Publishers
on concrete cylinders confined with FRP. In: Figueiras J, 2002:39–44.
Juvandes L, Faria R, Marques AT, Ferreira A, Barros J, w34x Chen WF. Plasticity in reinforced concrete. New York, USA:
Appleton J, editors, Composites in Constructions, Proceedings McGraw-Hill Inc, 1982.
of the International Conference, Lisse, The Netherlands: A.A. w35x Mirmiran A, Shahawy M. Behavior of concrete columns
Balkema Publishers, 2001:317–322. confined by fiber composites. J Struct Eng, ASCE
w25x Pessiki S, Harries KA, Kestner JT, Sause R, Ricles JM. Axial 1997;123(5):583 –90.
behavior of reinforced concrete columns confined with FRP w36x Gardner NJ. Triaxial behavior of concrete. ACI J
jackets. J Compos Constr ASCE 2001;5(4):237 –45. 1969;66(February):136 –46.
w26x Wang P, Cheong KK. RC columns strengthened by FRP under w37x Imran I, Pantazopoulou SJ. Experimental study of plain con-
uniaxial compression. In: Teng, JG, editor, FRP Composites in crete under triaxial stress. ACI Mater J 1996;93(6):589 –601.
Civil Engineering, Proceedings of the International Conference, w38x Mirmiran A, Shahawy M, Samaan M, El Echary H. Effect of
edited by Teng JG, Oxford, UK: Elsevier Science Ltd, column parameters on FRP-confined concrete. J Compos
2001:327–334. Constr ASCE 1998;2(4):175 –85.
w27x De Lorenzis L, Micelli F, La Tegola A. Influence of specimen w39x Hognestad E. A Study of Combined Bending and Axial Load
size and resin type on the behavior of FRP-confined concrete in Reinforced Concrete Members. Bulletin Series No. 399,
cylinders. In: Shenoi RA, Moy SSJ, Hollaway LC, editors, Engineering Experiment Station, Urbana, USA: University of
Advanced Polymer Composites for Structural Applications in Illinois, 1951.
w40x BS 8110. Structural Use of Concrete, Part 1, Code of Practice
Construction, Proceedings of the First International Conference,
London, UK: Thomas Telford, 2002:231–239. for Design and Construction, London, UK: British Standards
w28x Shehata IAEM, Carneiro LAV, Shehata LCD. Strength of short Institution, 1997.
w41x Monti G. Confining reinforced concrete with FRP: behavior
concrete columns confined with CFRP sheets. Mater Struct
and modelling. In: Cosenza E, Manfredi G, Nanni A, editors,
2002;35(1):50 –8.
Composites in Construction: A reality, Proceedings of the
w29x ASTM D3039yD3039M–95. Standard test method for tensile
International Workshop, Virginia, USA: ASCE, 2002:213–222.
properties of polymer matrix composite materials. Annual Book w42x Richart FE, Brandtzaeg A, Brown RL. The Failure of Plain
of ASTM Standards, 1995:14.02.
and Spirally Reinforced Concrete in Compression, Bulletin No.
w30x ASTMD 2290–92. Standard test method for apparent tensile
190, Engineering Experiment Station, Urbana, USA: University
strength of ring or tubular plastics and reinforced plastics by of Illinois 1929.
split disk method. Annual Book of ASTM Standards, w43x Candappa DC, Sanjayan JG, Setunge S. Complete triaxial
1992:15.03. stress–strain curves of high-strength concrete. J Mater Civ
w31x Shahawy M, Mirmiran A, Beitelman A. Test and modeling of Eng, ASCE 2001;13(3):209 –15.
carbon-wrapped concrete columns. Compos: Part B w44x Pantazopoulou SJ. Role of expansion on mechanical behavior
2000;31:471 –80. of concrete. J Struct Eng, ASCE 1995;121(12):1795 –805.
w32x De Lorenzis L, Tepfers R. Performance assessment of FRP- w45x Ottosen NS. Constitutive model for short-time loading of
confinement models–Part II: comparison of experiments and concrete. J Eng Mech Div, ASCE 1979;105(1):127 –41.
predictions. In: Shenoi RA, Moy SSJ, Hollaway LC, editors, w46x CEB-FIP. CEB-FIP model code 1990: design code. London,
Advanced Polymer Composites for Structural Applications in UK: Thomas Telford, 1993.
Construction, Proceedings of the First International Conference, w47x Lam L, Teng JG. Strength models for FRP-confined concrete.
London, UK: Thomas Telford, 2002:261–269. J Struct Eng, ASCE 2002;128(5):612 –23.
w33x Lam L, Teng JG. Stress–strain models for concrete confined w48x ACI 318-95. Building code requirements for structural concrete
by fiber-reinforced polymer. In: Loo YC, Chowdhury SH, (318-95) and commentary (318R-95). Farmington Hills, Mich-
Fragomeni S, editors, Advances in Mechanics of Structures igan, USA: American Concrete Institute (ACI), 1999. (Fifth
and Materials, Proceedings of the 17th Australasian Confer- Printing).