Professional Documents
Culture Documents
Polymer-Confined Concrete
J. G. Teng1 and L. Lam2
Abstract: One important application of fiber reinforced polymer (FRP) composites in the retrofit of reinforced concrete structures is to
provide confinement to columns for enhanced strength and ductility. As a result, many theoretical and experimental studies have been
carried out on FRP-confined concrete. This paper provides a critical review of existing studies, with the emphasis being on the revelation
of the fundamental behavior of FRP-confined concrete and the modeling of this behavior. Aspects covered in this paper include stress–
strain behavior, dilation properties, ultimate condition, and stress–strain models. The paper concludes with a brief outline of issues which
require further research. Although the paper is explicitly limited to concrete confined by FRP jackets in which the fibers are oriented only
or predominantly in the hoop direction, many of the observations made in this paper are also applicable or relevant to concrete confined
by FRP jackets with a significant axial stiffness, as found in concrete-filled FRP tubes as new columns.
DOI: 10.1061/(ASCE)0733-9445(2004)130:11(1713)
CE Database subject headings: Fiber reinforced polymers; Concrete structures; Retrofitting; Ductility; Confinement; Stress strain
curves; Models.
Ultimate Condition
As eventual failure of FRP-confined concrete is by the rupture of
FRP jacket, the ultimate condition of the confined concrete, often
characterized by its compressive strength and ultimate axial
strain, is intimately related to the ultimate tensile strain or tensile
strength of the confining FRP jacket in the hoop direction. In most
existing theoretical models for FRP-confined concrete, it has been
assumed that tensile rupture of FRP occurs when the hoop stress
in the FRP reaches its tensile strength from material tests, either
flat coupon tests (e.g., ASTM 1995) or ring splitting tests (e.g.,
ASTM 1992). However, extensive experimental results have
shown that the material tensile strength of FRP cannot be reached
in FRP-confined concrete as the hoop rupture strains of FRP mea-
sured in FRP-confined cylinder tests have been found to be con-
siderably smaller than those obtained from material tensile tests
(e.g., Shahawy et al. 2000; Xiao and Wu 2000; Pessiki et al.
2001).
This uncertainty with FRP hoop rupture strains has led to dif-
Fig. 3. Dilation properties of fiber reinforced polymer-confined and
ficulties in predicting the ultimate condition of FRP-confined con-
actively confined concrete
crete, particularly the ultimate axial strain. This is because the
ratio between the FRP hoop rupture strain and the material tensile
strain varies with the type of FRP (Lam and Teng 2003b). De
and steel-confined concrete which behaves similarly to actively Lorenzis and Tepfers (2003) showed that of the models they re-
confined concrete after the yielding of the confining steel. They viewed and assessed (Fardis and Khalili 1981; Saadatmanesh et
demonstrated that for steel-confined concrete, unstable dilation al. 1994; Miyauchi et al. 1997; Kono et al. 1998; Samaan et al.
occurs when steel yields, but for FRP-confined concrete, the lin- 1998; Saafi et al. 1999; Spoelstra and Monti 1999; Toutanji 1999;
early increasing hoop stress of the FRP jacket can eventually Xiao and Wu 2000), none was able to predict the ultimate axial
curtail the dilation if the amount of FRP is large enough. strain with reasonable accuracy if the hoop strain in the FRP
In Fig. 3, the lateral-to-axial strain ratio and the volumetric jacket at rupture is taken to be equal to the ultimate material
response of FRP-confined concrete under axial compression are tensile strain. Xiao and Wu (2000), based on their own test ob-
compared with those of unconfined and actively confined con- servations, suggested that 50% of the flat coupon ultimate tensile
crete using the test data of the specimens of Fig. 2. In Fig. 3(a), strain be taken as the hoop rupture strain. Jin et al. (2003), how-
the lateral-to-axial strain ratio, given in absolute values, is plotted ever, suggested using 0.96 times the flat coupon ultimate tensile
against the normalized axial strain, while in Fig. 3(b), the normal- strain. Moran and Pantelides (2002) suggested using the hoop
ized axial stress is plotted against the volumetric strain V, which strain limits of 0.0085 for CFRP and 0.0125 for glass FRP
is given by (GFRP). Lam and Teng (2003b) recently suggested that in devel-
oping confinement models, the actual hoop rupture strain h,rup
V = c + 2r 共4兲
measured in the FRP jacket should be used to evaluate the stress
A positive volumetric strain indicates compaction while a nega- in the FRP rather than simply using the FRP material tensile
tive value corresponds to dilation. strength f frp.
Fig. 3(a) shows that the lateral-to-axial strain ratio of FRP- Several causes have been suggested for the difference in the
confined concrete eventually stabilizes at values that depend on ultimate tensile strain between FRP tensile test specimens and
the amount of FRP provided. This is in contrast to unconfined and FRP jackets confining concrete (Spoelstra and Monti 1999; Sha-
actively confined concrete for which this ratio continuously in- hawy et al. 2000; Xiao and Wu 2000; Pessiki et al. 2001; Moran
creases with the axial strain. Fig. 3(b) shows that for unconfined and Pantelides 2002; De Lorenzis and Tepfers 2003; Lam and
and actively confined concrete, the change from compaction to Teng 2003b). These suggestions are generally speculative without
dilation occurs at different stress levels depending on the confin- sound experimental evidence. Lam and Teng (2003a, 2004) re-
ing pressure, and thereafter the dilation tendency remains until cently conducted the first carefully planned study involving com-
failure. This unchanged dilation tendency is only observed for one parative experiments in an attempt to clarify the causes for the
of the three CFRP-confined concrete specimens under consider- reduced strain capacity of FRP when used to confine concrete.
ation which is confined by a one-ply CFRP jacket. For the speci- The experimental program covered flat coupon tensile tests
men confined by two plies of CFRP, dilation is taken over by (ASTM 1995) and ring splitting tests (ASTM 1992) on CFRP and
compaction at a normalized axial stress of about 1.5. For the GFRP specimens, and compression tests on concrete cylinders
specimen with three plies of CFRP, no dilation is found during the wrapped with one to three plies of CFRP and GFRP. Based on the
冉 冊
(1998) was based on the test results of rectangular specimens = 共5兲
⬘
f co c c 2
共100⫻ 200⫻ 400 mm兲 confined by FRP sheets and spirals. Their 1 + 共A − 2兲 +D
co co
model predicts only the descending type of stress–strain curves.
All these models are suitable for direct application in design cal- where A and D = constants controlling the initial slope and the
culations by hand or spreadsheets and are thus referred to as descending path of the stress–strain curve, respectively. Ahmad
design-oriented models, although they may also be used in non- and Shah (1982a, b) used Eq. (5) to model the stress–strain curve
linear computer analysis of structures with FRP confinement. In of concrete confined by steel spirals. This equation has also been
addition, Moran and Pantelides (2002) proposed a stress–strain used in Eurocode 2 (CEN 1991) to represent the stress–strain
model which may be classified as a design-oriented model as its curve of unconfined concrete for structural analysis. However, the
many parameters are directly based on test results, but the model stress-strain curve for design use in the same document takes the
is far more complicated than other design-oriented models and following form which is a special case of Eq. (5) with A = 2 and
共E1 − E2兲
D = 0, which is usually referred to as Hognestad’s (1951) parabola
c
f ⬘co
冋 冉 冊册
=
2c
co
−
c
co
2
共6兲
=
再 冏
1+
共E1 − E2兲
fo
冏冎n 1/n + E 2 共7兲
Analysis-Oriented Models 冉 冑
f ⬘cc = f ⬘co 2.254 1 + 7.94
fl
⬘
f co
fl
− 2 − 1.254
f ⬘co
冊 共8兲
A number of stress–strain models have been developed based on
an incremental iterative numerical approach in which the interac- where f ⬘cc = peak compressive stress of concrete under a constant
tion between the concrete core and the confining FRP is explicitly confining pressure f l. Harries and Kharel (2002) also used
accounted for. The use of an incremental approach makes it in- Mander’s model, except that Eq. (8) was replaced by an equation
convenient for these models to be adopted in hand or spreadsheet proposed by Mirmiran and Shahawy (1997a), which is given by
calculations in design, but they are suitable for use in computer
f ⬘cc = f co
⬘ + 4.269f 0.587
l 共MPa兲 共9兲
analysis such as nonlinear finite element analysis, particularly if
the lateral strain of the confined concrete is also required in the The lateral strain is determined by different means in the five
analysis. These models are referred to as analysis-oriented stress– models mentioned above. In the models of Mirmiran and Sha-
strain models. hawy (1996) and Chun and Park (2002), a cubic polynomial equa-
tion developed by Elwi and Murray (1979) based on the results of
Models Based on Active Confinement Model Kurfer et al. (1969) from biaxial tests of concrete is used to de-
The models of Mirmiran and Shahawy (1996), Spoelstra and termine the lateral strain. In the model of Spoelstra and Monti
Monti (1999), Fam and Rizkalla (2001), Chun and Park (2002), (1999), the lateral strain is determined using a simple constitutive
and Harries and Kharel (2002) are all based on the assumption model proposed by Pantazopoulou and Mills (1995), which de-
that the axial stress and axial strain of concrete confined by FRP scribes the decrease of secant modulus of concrete with an in-
at a given lateral strain are the same as those of the same concrete creasing area strain (or lateral strain). In the model of Fam and
actively confined with a constant confining pressure equal to that Rizkalla (2001), an equation representing the change of secant
supplied by the FRP jacket. This assumption is equivalent to as- Poisson’s ratio with confining pressure was developed based on
suming that the stress path of the confined concrete does not the results of Gardener (1969) from triaxial tests of concrete. In
affect its stress–strain behavior. This assumption has not been these four models, the lateral strain equations are based on test
rigorously validated, but appears to be widely accepted as any results of unconfined or actively confined concrete. Harries and
error caused by this assumption is believed to be small. Based on Kharel (2002) instead based the lateral strains of their model on
this assumption, an active confinement model for concrete can be the test results of FRP-confined concrete. In their model, separate
used to evaluate the axial stress and axial strain of FRP-confined equations are used for predicting the lateral strains of CFRP and
concrete at a given confining pressure and the interaction between GFRP confined concrete.
the concrete and the FRP jacket can be explicitly accounted for by
equilibrium and radial displacement compatibility considerations. Models Based on Other Approaches
As a result, the stress–strain curve of FRP-confined concrete Apart from the above five models based on an active confinement
crosses a series of stress–strain curves for the same concrete with model, other approaches have also been used for modeling FRP-
different confining pressures, as demonstrated in Fig. 7 using confined concrete. Harmon et al. (1998) developed a model for
Spoelstra and Monti’s (1999) model. This approach has previ- FRP-confined concrete based on the concept of crack slip and
ously been used in studies on steel-confined concrete (Ahmad and separation in the concrete. An earlier version of this model was
Shah 1982b; Madas and Elnashai 1992). presented by Harmon et al. (1995) and the application of this
The four studies by Mirmiran and Shahawy (1996), Spoelstra model in column analysis was described by Harmon et al. (2002)
and Monti (1999), Fam and Rizkalla (2001), and Chun and Park and Gould and Harmon (2002). In this model, it is assumed that
Analysis-Oriented Models
The models of Harmon et al. (1998), Spoelstra and Monti (1999),
Fam and Rizkalla (2001), Chun and Park (2002), Harries and
Kharel (2002), and Becque et al. (2003) are considered here.
Mirmiran and Shahawy’s (1996) model is not included in this
comparison as some of its parameters are not clearly defined. The
predicted stress–strain curves terminate at the point when the av-
erage FRP hoop rupture strain from tests is reached. It can be seen
from Figs 10(a and b) that these models lead to very different
predictions. For the CFRP-confined concrete cylinders [Fig.
10(a)], the models of Spoestra and Monti (1999), Fam and Riz-
kalla (2001), and Chun and Park (2002) perform better than other
models. However, for the GFRP-confined concrete cylinders [Fig.
10(b)], none of the models provides reasonably close predictions
for the stress and strain at the ultimate point. The model of Har-
ries and Kharel (2002) predicts a nearly horizontal line for the
second portion of the stress–strain curve for both the CFRP- and
the GFRP-confined concrete, which greatly deviates from the test
curves. This is believed to be at least partly due to the use of an
incorrect failure surface. If the failure surface is redefined using
Eq. (10) which represents an improvement to Eq. (9) from the
same research group, the predicted results are in much better
Fig. 10. Performance of analysis-oriented models using test fiber
agreement with test data. In addition, the accuracy of the lateral
reinforced polymer hoop rupture strain
strain equations in Harries and Kharel’s (2002) model are also in
doubt because according to this model, a three-ply CFRP jacket
Cheng et al. (2002), Moran and Pantelides (2002), Jin et al. does not lead to a higher compressive strength than a two-ply
(2003), and Lam and Teng (2003b) are selected for comparison jacket for the same concrete, which is clearly incorrect.
with the test data. These models were identified in a preliminary The key to analysis-oriented models is the lateral-to-axial
comparison as being the more accurate models. When more than strain relationship for FRP-confined concrete, which does not ap-
one model has been proposed by the same researchers (Miyauchi pear to have been well understood and accurately captured in
et al. 1997, 1999; Xiao and Wu 2000, 2003), only their latest existing analysis-oriented models. For models employing an ac-
model is included in the comparison. In Fig. 8, comparisons are tive confinement model, a properly defined failure surface is also
made for the CFRP-confined cylinders where the hoop strain or of importance, as illustrated by the dependence of the predictions
stress in the FRP at rupture as defined in the original model is of Hurries and Kharel’s (2002) model on the defined failure sur-
used. For Xiao and Wu’s model, 50% of the FRP material tensile face. Further research is required to develop a more accurate
strength is used. For Jin et al.’s (2003) model, a reduction factor analysis-oriented stress–strain model.
of 0.96 is used for the ultimate FRP strain. For Moran and Pan-
telides’ (2002) model, the CFRP hoop rupture strain is taken as
0.0085. For Lam and Teng’s model, the CFRP is assumed to reach Concluding Remarks
58.6% of the material tensile strength, based on the average value
of a large test database (Lam and Teng 2003b). For other models, The large number of studies on FRP-confined concrete have led to
the FRP tensile strength from flat coupon tests is used. In Figs 9(a a good understanding of its behavior. The main difference be-
and b), comparisons are made for both CFRP and GFRP-confined tween FRP-confined concrete and actively confined concrete
concrete and the actual average hoop rupture strains obtained (which approximates the behavior of steel-confined concrete)
from the confined concrete tests are used for all models. arises from the linear elastic material behavior of an FRP jacket
It can be observed from Fig. 8 that if the hoop strain at the which supplies a continuously increasing confining pressure as
rupture of the FRP jacket is assumed the value specified in the the axial strain increases. As a result, the stress–strain curve of
original model, the three models using a substantially reduced FRP-confined concrete features a ascending bilinear shape, unless
value for this rupture strain (Moran and Pantelides 2002; Lam and the amount of FRP provided is rather small. The fact that the FRP
Teng 2003b; Xiao and Wu 2003) perform better than other mod- jacket ruptures at a hoop rupture strain considerably lower than
els. However, if the actual hoop rupture strain is used instead, the ultimate tensile strain from material tests is another important
then the performance of the other models is improved. Among observation.
these models, the models of Saafi et al. (1999), Moran and Pan- Based on existing test observations, many stress–strain models
telides (2002), Jin et al. (2003), Xiao and Wu (2003), and Lam have been developed for FRP-confined concrete. These stress–
and Teng (2003b) give closer predictions of the shape of the strain models can be classified into two categories, design-
stress–strain curve and the ultimate condition (Fig. 9). This ob- oriented models and analysis-oriented models. In the first cat-
servation suggests that, central to a design-oriented model is an egory, stress–strain models are presented in closed-form
accurate definition of the ultimate condition. Provided that the expressions, while in the second category, stress–strain curves of