You are on page 1of 11

Composites: Part B 41 (2010) 423–433

Contents lists available at ScienceDirect

Composites: Part B
journal homepage: www.elsevier.com/locate/compositesb

Analytical modeling of the bond–slip relationship at FRP-concrete interfaces


for adhesively-bonded joints
Ying-Wu Zhou, Yu-Fei Wu *, Yanchun Yun
Department of Building and Construction, City University of Hong Kong, Hong Kong

a r t i c l e i n f o a b s t r a c t

Article history: The bond–slip relationship is essential to the study of the macro mechanical properties of composite
Received 13 March 2010 materials and structures. An analytical model is developed in this work to derive the bond–slip relation-
Received in revised form 26 May 2010 ship at the reinforcement-substrate concrete interface (joint) for externally bonded reinforcement. The
Accepted 19 June 2010
model is applicable to both long joints (infinite bond length) and short joints. When the bond length
Available online 23 June 2010
approaches infinity, the model reverts to a well-known existing analytical model. A bond–slip relation-
ship for short joints with a limited bond length is derived for the first time. It is concluded from the mod-
Keywords:
eling that the existing model for long joints is not applicable to short joints that have a bond length that is
B. Debonding
B. Interface
less than the effective bond length, or at locations in long joints that are closer than the effective bond
C. Analytical modeling length to the free end of the reinforcement.
D. Mechanical testing Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction local bond stress and slip are directly calculated from the variation
in strain, which is measured by closely spaced strain gauges
An analytical procedure is proposed here to derive the bond– mounted on top of the FRP laminate, as shown in Fig. 1
slip relationship at the reinforcement-concrete interface (joint) [19,22,29,30]. However, extensive research has concluded that it
using conventional pull-off tests. In principle, this procedure is is practically impossible to derive a consistent, reliable, and accu-
applicable to any externally bonded (EB) joint, such as joints exter- rate distribution of the bond stresses at the bond interface with
nally bonded with steel mesh or plate and fiber reinforced polymer this method for several reasons. First, the process of debonding is
(FRP). In this work, the procedure is applied to derive the bond–slip difficult to capture because of the brittle and highly nonlinear nat-
relationship of EB–FRP joints. ure of local fractures [22]. Second, the random distribution of dis-
The external bonding of FRP is an effective and popular method crete cracks and aggregates in the concrete causes large and
for the rehabilitation of reinforced concrete (RC) structures [1]. irregular fluctuations in the strain measurement and hence incon-
The mechanical performance, including the strength of the EB– sistent bond stresses [23]. Finally, irregularity in the adhesive layer
FRP structural system, is often determined by the bond between and surface preparation produces irregular measurements [22]. As
the FRP and the concrete. The bond interface is usually the weak- a result, the bond–slip relationship as measured directly from pull-
est link, and debonding at the interface is usually the critical fail- off tests is always significantly varied, as illustrated in Fig. 2, and
ure mode. Thus, investigating the bond behavior of EB–FRP cannot be used for structural analyses.
systems is an important subject, and has attracted extensive re- To overcome this problem, Dai et al. [18–22] recently developed
search [1–40]. another method to derive the bond–slip relationship at the EB–FRP
In general, investigations into the bond characteristics of EB– joints through the indirect analysis of test results, referred to here-
FRP joints can be divided into three categories: (1) direct measure- after as the analytical method. The analytical method derives the
ments taken with closely spaced strain gauges from experimental bond–slip relationship from strain–slip responses that are mea-
tests such as pull-off tests [2–17]; (2) indirect analytical solutions sured at the loaded end from a pull-off test. As the measurements
derived from experimental test results [18–22]; and (3) finite ele- of the load and slip at the loaded end in a pull-off test are much
ment analyses [23–28]. Of the three methods, the first is the most more stable and reliable than the strain measurements along the
popular and widely adopted by researchers. With this method, the bond interface, this method delivers a consistent and reliable
bond–slip relationship for EB–FRP joints. This work is a comple-
ment and supplement to the newly developed and award winning
* Corresponding author. Tel.: +852 27844259; fax: +852 27887612. analytical method, and presents a more general bond–slip relation-
E-mail address: yfwu00@cityu.edu.hk (Y.-F. Wu). ship for the first time.

1359-8368/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compositesb.2010.06.004
424 Y.-W. Zhou et al. / Composites: Part B 41 (2010) 423–433

2. Analytical solution for EB–FRP joints with an infinite bond


length

2.1. Bond–slip model

Direct pull-off tests, as schematically shown in Fig. 1, are gener-


ally considered to be the simplest but most effective method to
investigate the bond behavior of EB–FRP joints [2–17]. For a certain
pull load P at the loaded end of the FRP strip, there is a correspond-
ing and unique distribution of the internal deformations and stres-
ses at the bond interface, such as the slip of the FRP strip relative to
the concrete substrate s(x), the FRP strain along the interface e(x),
and the shear (bond) stress at the bond interface s(x), as shown
in Fig. 1. From the equilibrium and deformation compatibility con-
ditions of a typical segment dx, the following two equations gov-
erning the problem can be obtained [29–32].

s0 ðxÞ
eðxÞ ¼ ; ð1Þ
1þq

and
Fig. 1. Analytical model of EB–FRP joint.
Ef t f 00
sðxÞ ¼ s ðxÞ; ð2Þ
1þq

9000 where Ef and tf are the elastic modulus and the thickness of the FRP
23.5 kN strip, respectively, q is the stiffness ratio of the FRP strip to the con-
8000 23.2 KN
24.0 KN crete block that is given by
24.8 kN
7000 23.5 kN Ef t f bf
22.0 kN q¼ ; ð3Þ
Ec t c bc
FRP strain (με)

6000 22.1 kN
20.0 kN in which bf is the width of the FRP strip and Ec, tc, and bc are the elas-
5000 12.5 kN
7.5 kN tic modulus, thickness, and width, respectively, of the concrete
4000 block. As the value of Ef tf bf is in most cases much smaller than that
3000 of Ectcbc, q is usually a very small value that approaches zero. Eqs.
(1) and (2) can be further simplified by neglecting the deformation
2000 of the concrete block or letting q = 0.
1000 EB–FRP joints of a length of more than 300 mm can be practi-
cally considered as joints of infinite length [19,22]. Careful obser-
0 vation of previous experimental results [15–19,36] reveals that
0 50 100 150 200 250 300
both the variation and distribution of the slip s(x) along the bond
Distance from loaded end (mm) interface at different loading levels are regular, and can be repre-
(a) FRP strain distribution measured sented with sufficient accuracy by the simple mathematical
function:
directly by strain gauges
 xx0 
sðxÞ ¼ a ln 1 þ e b ; ð4Þ
Distance from the loaded end
10 10mm 30mm 50mm
70mm 80mm 90mm where a, b, and x0 are coefficients that can be determined from
100mm 110mm 120mm tests. Eq. (4) is the basic and only assumption of this derivation,
8 130mm which is further discussed and validated in Section 2.2. Substituting
Bond stress (MPa)

Eq. (4) into Eqs. (1) and (2) yields


6 1 a 1
eðxÞ ¼   xx  ; ð5Þ
1 þ q b 1 þ e b 0
4
and
2 xx0
Ef t f a e b
sðxÞ ¼   2 : ð6Þ
1 þ q b2 xx0

0 1 þ e b

0 0.1 0.2 0.3 0.4 0.5


Slip (mm) Rewriting Eq. (4) leads to
xx0
(b) Bond-slip responses e b
sðxÞ
¼ e a  1: ð7Þ
Fig. 2. Pull-off test results of specimen EB300-1. Substituting Eq. (7) into Eqs. (5) and (6) gives
Y.-W. Zhou et al. / Composites: Part B 41 (2010) 423–433 425

1 a s model proposed in the previous section is in fact identical to that


eðsÞ ¼ 1  ea ; ð8Þ
1þq b developed by Dai et al. [19], although the starting point (Eq. (4))
is different.
and The advantage of the analytical procedure proposed in [19] is
Ef t f a as  s that a more stable slip response at the loaded end is used to derive
sðsÞ ¼ e 1  ea ; ð9Þ the bond–slip relationship, rather than the highly unstable strain
1 þ q b2
reading response, as clearly shown in Figs. 2 and 3. The test results
respectively, where Eq. (9) is the bond–slip relationship and Eq. (8) in these figures are from an experimental study recently completed
gives the FRP strain–slip relationship. by the authors [17]. Details of the test specimen are given in Table
1. Fig. 2 shows that the strain readings measured on top of the FRP
2.2. Comparison with the work of Dai et al. [19] strip are highly unstable and fluctuate significantly, which pro-
duces inconsistent and unreliable bond–slip responses at different
When q = 0, Eq. (9) is mathematically identical to the solution of positions, as shown in Fig. 2b. In Fig. 3a, the slip at the loaded end,
Dai et al. [19], except that the parameters a and b in Eq. (9) are re- which is calculated from the integration of the FRP strains along
placed by another two parameters A and B that have different the bonded length, is much more stable than the strain readings.
physical meanings. It is interesting to note that the derivation of Hence, Fig. 3a is suitable for use to determine the coefficients a
Dai et al. [19] is based on an assumed relationship between the and b in Eq. (8), where the load P in Fig. 3a is directly related to
FRP strain e and the slip s, that is, eðsÞ ¼ Að1  eBs Þ, which happens the strain e in Eq. (8) by P ¼ Ef  Af  e, in which Af is the area of
to have the same mathematical form as Eq. (8). Thus, the analytical the FRP strip. A regression analysis is used to calculate a and b such

Fig. 3. Load–slip response and bond–slip response of EB300-1.

Table 1
Details of specimens.

Reference Specimen ID fc0 (MPa) Block dimensions (mm  mm  mm) bf (mm) Lf (mm) Ef tf (GPa?mm) a (mm) b (mm) Correlative factor R2

Dai et al. [19] CRL1R2 35 400  200  400 100 350 25.3 0.1445 17.7828 0.9997
Yun et al. [17] EB300-1 28 150  150  400 50 300 82.5 0.1129 21.6326 0.9995
EB300-2 28 150  150  400 50 300 82.5 0.1083 20.9563 0.9995
M-EB 32.5 200  200  400 50 100 165 0.1263 24.1274 0.9997
Chajes et al. [6] S4inch 36.4 152.4  152.4  228.6 25.4 101.6 109.7 0.1334 30.0632 0.9999
Zhou [36] GI60-100-50-1 52.5 150  150  300 50 100 12.3 0.1457 13.5125 0.9999
GI50-60-50-1 46.8 150  150  300 50 60 12.3 0.0832 8.3792 0.9997
GI80-100-50-1 67.1 150  150  300 50 100 12.3 0.0483 4.7916 0.9995
GI80-60-50-1 67.1 150  150  300 50 60 12.3 0.0395 5.6611 0.9981
GII80-200-50-2 67.1 150  150  300 50 200 24.2 0.0474 7.3796 0.9989
GII80-100-50-1 67.1 150  150  300 50 100 24.2 0.0842 10.1039 0.9995
GII80-60-50-1 67.1 150  150  300 50 60 24.2 0.1015 12.2554 0.9988
CII60-100-60-1 52.5 150  150  300 60 100 24.5 0.0470 6.8065 0.9968
CII60-60-60-1 52.5 150  150  300 60 60 24.5 0.0613 8.4511 0.9983
CII80-100-60-1 67.1 150  150  300 60 100 24.5 0.0494 7.6517 0.9994
CII80-60-60-1 67.1 150  150  300 60 60 24.5 0.0573 8.4259 0.9997
CI80-100-60-1 67.1 150  150  300 60 100 39.4 0.0831 10.4891 0.9992
CI80-60-60-1 67.1 150  150  300 60 60 39.4 0.1452 16.6892 0.9992
CI80-100-30-1 67.1 150  150  300 30 100 39.4 0.0582 12.9648 0.9967
CI50-80-60-2 46.8 150  150  300 60 80 39.4 0.0520 9.4684 0.9986
CI50-100–60-1 46.8 150  150  300 60 100 39.4 0.0492 8.6265 0.9988
426 Y.-W. Zhou et al. / Composites: Part B 41 (2010) 423–433

that the difference or error between the curve in Fig. 3a and Eq. (8) culated to be greater than 0.99, as shown in Table 1, which vali-
is minimized. From the bond–slip response curve shown in Fig. 3b, dates the accuracy of Eq. (4).
it is clear that the analytical method of Dai et al. provides a much
more accurate and stable way of determining the bond–slip rela- 2.3. Physical meanings of the parameters
tionship than the method of direct measurement in Fig. 2b. Fig. 3
s
shows predictions with both q = 0.6% and q = 0, and confirms that Differentiating Eq. (8) with respect to s gives e0 ðsÞ ¼ 1b ea , from
the deformation of the concrete block has little effect on the bond which the following equation is obtained by letting s = 0,
behavior of the EB–FRP joint.
1
In this work, the measured slip distribution s(x) is used to derive e0 ð0Þ ¼ : ð10Þ
the bond–slip relationship, or Eq. (9). In Fig. 4, the same example as b
that used for Figs. 2 and 3 is adopted to illustrate the difference be- Thus, 1/b is the initial slope of the FRP strain versus the slip
tween the method proposed in this work and that suggested in curve. As the FRP strain of the loaded end is directly related to
[19]. The slip distributions shown in Fig. 4a are obtained by the the pull load P, we have dP j
ds s¼0
¼ 1b Ef tf bf ¼ K 0 . It is clear that the re-
integration of the FRP strains along the bonded length. Each curve ciprocal of b determines the initial tangential rigidity K0 of the
in Fig. 4a corresponds to a particular load P at the loaded end. The load–slip response curve.
coefficients a, b, and x0 are calculated from the regression analyses Assuming that q = 0 in Eq. (8), the maximum strain of the FRP
by best fitting the curves in Fig. 4a to Eq. (4). In the next section strip is obtained as emax ¼ ab ¼ a  1b. Multiplying the equation by
(see Eq. (12a)), it is shown that x0 is in fact a function of a, b, and Eftfbf on both sides gives
P, and thus is not an additional parameter at a particular load P.
Pmax Pmax
In other words, there are only two parameters a and b for the par- a¼1 ¼ ¼ sm : ð11Þ
ticular slip distribution curve in Fig. 4a. Hence, similar to the b
Ef t f bf K0
regression analysis for Fig. 3, the two parameters a and b can be
Thus, a is the slip at the turning point of the equivalent bilinear
determined by regressing all of the curves together. Once a and b
load–slip curve, or the peak elastic slip sm, as shown in Fig. 5.
are determined, the bond–slip relationship can be calculated from
Differentiating Eq. (6) with respect to x and letting it equal zero
Eq. (9), which is plotted in Fig. 4b with a thick solid line together
yields x = x 0, which reveals that the peak bond stress occurs at the
with the same relationships from Fig. 2b and Fig. 3b, which are
shown by the thin lines and a thick dashed line, respectively.
Fig. 4b clearly shows that the proposed method produces a signif-
icantly different bond–slip relation from that produced using the
procedure introduced in [19]. A direct comparison of the solid line
and the dashed line in Fig. 4b against the directly measured re-
sponses does not determine which method produces a more accu-
rate result. However, it can be observed by comparing Fig. 3a with
Fig. 4a that the points in Fig. 4a fluctuate much less than those in
Fig. 3a. Furthermore, the bond–slip curve in Fig. 4b is derived by
regressing a series of curves (i.e., the slips at many locations under
different loads), whereas the result in Fig. 3b is derived from the
slip at only one location – the loaded end. From a statistical point
of view, the result derived by the proposed method should be more
reliable than that derived using the other method.
As mentioned in Section 2.1, the basic assumption of Eq. (4) de-
rives from observed test results. In the previous example, Eq. (4)
fits the experimental slip distributions so well that the correlative
factor R2 is calculated to be 0.9995. In all of the regression analyses
carried out by the authors so far, the correlative factors R2 are cal- Fig. 5. Equivalent bi-linear load–slip curve.

Fig. 4. Analytical results of EB300-1 from the proposed model.


Y.-W. Zhou et al. / Composites: Part B 41 (2010) 423–433 427

location of x = x0, as shown by the typical bond stress distribution face does not equal the pull force P, or the balance of force is not
in Fig. 1. Substituting e(Lf) = P/Eftfbf into Eq. (5) gives satisfied for joints with a finite bond length.
  To meet the boundary condition of e(0) = 0, the strain distribu-
1 a Ef t f bf tion function Eq. (5) can be slightly adjusted to
x0 ¼ Lf þ b ln  1 ; ð12aÞ
1þq b P
1 a x=Lf
where Lf is the total length of the bonded zone (see Fig. 1). As a and eðxÞ ¼ xx ; ð14Þ
1 þ q b x=L þ e b 0
b are constants related to the stiffness and strength of the FRP sys- f
tem, x0 is a variable that depends on the pull load P. Substituting
x = x0 into Eq. (6) gives which always satisfies the boundary condition of e(0) = 0. Substitut-
ing Eq. (14) into Eq. (1) yields the slip distribution
Ef t f a Z Z
smax ¼  : ð12bÞ x
a x
t=Lf
4ð1 þ qÞ b2
sðxÞ ¼ s0 þ eðtÞdt ¼ s0 þ tx0 dt; ð15Þ
0 b 0 t=Lf þ e b
From Eq. (12), it is clear that the peak bond stress has a constant
value and moves toward the free end when the load P increases.
where s0 is the FRP slip at the free end and can be directly measured
Eq. (5) shows that, mathematically, e(x) is an asymptotic curve
from tests. Eq. (14) is the basic assumption in this part of derivation.
and its asymptotic value corresponds to the peak strength of the
As slip measurements are more stable than strain measurements,
bond system Pmax. Eq. (9) further indicates that once a and b are
and Eqs. (14) and (15) are mathematically equivalent, Eq. (15) is
known, the bond–slip relationship s(s) can be uniquely deter-
used to verify the basic assumption in Section 3.2.
mined. Meanwhile, it is clear from Eqs. (4)–(8) that all of the distri-
Substituting Eq. (14) into P = Ef  Af  e(Lf) gives
bution functions of s(x), e(x), s(x), and e(s) are fixed once the value
of a, b, and x0 are fixed. If any of these distribution functions is E f Af a 1
known from tests, then the parameters can be calculated by fitting P¼ : ð16Þ
1þq b Lf x0

the mathematical equations to the test results to determine the 1 þ e b

bond–slip relationship. From Eqs. (14) and (15), the strain–slip relationship at the
Numerous investigations have pointed out [33–36] that there is loaded end is obtained as
an effective bond length Le for an EB–FRP joint beyond which the
8 1 a 1
pull load P does not increase as the bond length Lf further increases. < eðLf Þ ¼ 1þq
> b

Lf x0

This effective bond length Le can be identified from the typical 1þe b
: ð17Þ
: sðLf Þ ¼ s0 þ a R Lf
> tx0
bond stress distribution shown in Fig. 1. As the enclosed area of t=Lf
b 0 t=Lf
þ e b dt
s equals the pull load P, the total bond resistance is essentially pro-
vided by the area near x0. As the mathematical asymptotic value Differentiating e(x) in Eq. (14) with respect to x gives the shear
Pmax can never be achieved in reality, we can take the close value stress distribution
dPmax as the maximum bond strength, where d is a value close to
xx0
1, such as 0.96, 0.98, or 0.99 [22]. The area of the bond stress that sðxÞ 1 a xþb e b
corresponds to the value of dPmax is located in a region in which the ¼  xx0 2
: ð18Þ
Ef tf 1 þ q b Lf
2
FRP strains are (1 + d)emax/2 and (1  d)emax/2, respectively, on the x=Lf þ e b
two sides of x0 such that the difference in FRP strain is demax.
Combining Eqs. (18) and (15) gives the bond–slip relationship
Substituting the two strain values of (1 + d)emax/2 and
as
(1  d)emax/2 into Eq. (5) allows the two locations xL and xR to be
determined, which in turn leads to Le = xL  xR, or 8 xx0
>
> E t a xþb

b
  > sðxÞ ¼ 1þf qf
< b2 Lf  e
xx0
2
1þd x=Lf þe

b
Le ¼ 2b ln : ð13Þ : ð19Þ
1d >
>
: sðxÞ ¼ s0 þ a R x
> tx0
t=Lf
b 0 t=Lf
þ e b dt
Regardless of the differences between the above model and that
proposed by Dai et al. [19], in principle, these two models are gen- Mathematically, the parameter x0 can be eliminated from the
erally the same. Therefore, the originality of the solution with infi- foregoing equation set to give a single equation relating the bond
nite bond length belongs to Dai et al. [19]. The significance and stress to the slip. The bond–slip relationship is then not only a
originality of this work is the extension of the above work to joints function of a and b, but also of x. In other words, the bond–slip
with finite bond length, as given in the following section. relation is different at different locations x, which is distinct from
the case for an infinite bond length.
3. Analytical solution for EB–FRP joints with a finite bond
length 3.1. Comparison with the case of an infinite bond length

When (x  x0) is a positive large value, that is, for points on the It can be mathematically shown that Eqs. (14)–(19) all approach
left-hand side of and far away from x0 (see Fig. 1), the strain value the respective cases for an infinite bond length derived in Section 2
given by Eq. (5) is very small, and approaches zero asymptotically when Lf approaches infinity. This means that the analytical solu-
and quickly toward the free end. If the equations derived in the tions for an infinite bond length are a special case of the solutions
previous section are applied to EB–FRP joints with a limited bond for a finite bond length. For example, Eq. (14) can be re-written as
length, say x = 0, then Eq. (5) gives eð0Þ ¼ 1þ1 q  ab  1 x0 , which does
1þe b
1 a x=Lf 1 a 1
not equal zero unless x0 approaches infinity. The strain at the free
eðxÞ ¼ xx ¼ xx : ð20Þ
1 þ q b x=L þ e b 0 1 þ q b 1 þ Lf e b 0
f x
end is then not zero, which is unreasonable. Thus, the analytical
solution for EB–FRP joints with an infinite bond length is not appli- When Lf ? 1, the bonded zone is so long that the bond stress
cable to joints with a limited bond length. Furthermore, it can be block only occurs near the loaded end where x is very large and
easily shown that the integration of bond stress at the bond inter- in the same order of Lf, or mathematically, where Lf/x ? 1. This
428 Y.-W. Zhou et al. / Composites: Part B 41 (2010) 423–433

leads to eðxÞ ! 1þ1 q a 1


, which is identical to Eq. (5). For points Eq. (24) can then be re-written as
xx0
b 
1þe b

near the free end or an x that has a limited value, Eq. (20) becomes 2  
d s 1 ds
e(x) ? 0. Ef tf 2
¼ 2Gf B 1  ð1  eBs Þ: ð25Þ
dx A dx
Differentiating Eq. (18) with respect to x0 and setting it to zero
leads to Solving Eq. (25) leads to the solution for the slip distribution,
 
x which can then be used to derive all of the other solutions, such as
x0 ¼ x þ b ln : ð21Þ the bond–slip relationship at different locations. A typical bond–slip
Lf
relationship for specimen M-EB produced with this method is
As x0 is directly related to the load P through Eq. (16), when P shown in Fig. 6a. For comparison, a typical bond–slip relationship
increases, x0 moves from the loaded end to the free end, which derived from Eq. (19) and the test results is shown in Fig. 6b. Clearly,
causes variation in the bond stress s at location x. Eq. (21) thus the two solutions are fundamentally different. The bond–slip re-
gives the particular x0 or load P at which the bond stress s reaches sponse derived from [22] approaches a plateau near the free end,
the maximum value at location x. Substituting Eq. (21) into Eq. (18) which indicates a constant bond resistance at an ever-increasing
gives the maximum bond stress at a particular position x slip. This result is unreasonable and incorrect. Moreover, according
 
Ef t f a b to Fig. 6a, the fracture energy rates Gf (the area enclosed by the
smax ¼ 1 þ : ð22Þ
1 þ q 4b2 x bond–slip curve) at most locations, particularly those near the free
end, approach infinity, which is also incorrect. In fact, in [22] Gf is
It can be seen from Eq. (22) that the maximum bond stress is
assumed to be a material property and has a certain value for a
different at different positions. This phenomenon is distinct from
particular EB–FRP joint, which contradicts the results in Fig. 6a.
the case of an infinite bond length, where a constant peak bond
Furthermore, Eq. (24) does not automatically degenerate into its
stress moves from the loaded end towards the free end when the
own solution for an infinite bond length, s = 2GfBeBs(1  eBs),
load P increases. However, when Lf approaches infinity, the value
when the bond length approaches infinity.
of x near the loaded end approaches infinity and Eq. (22) becomes
Ef t f a
smax ¼ 4ð1þ qÞ b2 , which is identical to the case for an infinite bond
length given by Eq. (12b). 3.4. Analysis of the results for long joints
The physical meaning of the parameters a and b also changes
when the bond length is limited. Similar to the derivations for This section analyses long joints using an example with a bond
Eqs. (10) and (11), we have length of 300 mm (specimen EB300-2 in Table 1). The slip distribu-
   
de de ds  1 1
K 0 ¼ Ef bf tf ¼ E b t ¼ E b t þ ð23aÞ 35
dse¼0 dx dxe¼0
f f f f f f
b Lf
s¼0 s¼0
30 x=0 mm
Pmax 1 x=10 mm
sm ¼ ¼a : ð23bÞ 25 x=20 mm
Bond stress (MPa)

K0 1 þ b=Lf
x=30 mm
When the bond length approaches infinity, Eq. (23) is identical 20 x=40 mm
to Eqs. (10) and (11). x=50 mm
15 x=60 mm
3.2. Verification of the basic assumption x=70 mm
10 x=80 mm
The slip distributions shown in Fig. 4a are typical of those for x=90 mm
5
EB–FRP joints in a pull-off test. Each of the slip distributions corre-
sponds to a particular pull load P. As x0 is related to P through Eq. 0
(16), the two parameters a and b can be uniquely determined using 0 0.5 1 1.5 2
regression analysis to fit any of the slip distribution curves to Eq. Slip (mm)
(15). To eliminate the incidental error of the test measurements
and increase the accuracy of the analytical results as much as pos-
(a) Bond-slip response based on Eq. (25)
sible, all of the slip distribution curves are used to calculate a and b
in a single regression analysis. Twenty-one test specimens from
both the authors’ own tests and those reported in the literature
are used to validate the accuracy of Eq. (15). Most of the specimens
listed in Table 1 have a short bond length of 100 mm or less. The
correlative factors R2 between Eq. (15) and the test measurements
for the specimens are calculated and provided in the last column of
Table 1, and range from 0.9968 to 0.9999. This verification vali-
dates the basic assumption of Eq. (14) and substantiates the ana-
lytical model developed in Section 3.

3.3. Comparison with reference [22]

It is noted that an analytical model for EB–FRP joints with a lim-


ited bond length has also proposed by Dai et al. [22]. This model in-
volves the modification of the bond–slip relationship for EB–FRP
joints with an infinite bond length with a strain related factor, or
 e (b) Bond-slip response based on Eq. (19)
s ¼ 2Gf B 1  ð1  eBs Þ: ð24Þ
A Fig. 6. Comparison of the bond–slip relations for specimen M-EB.
Y.-W. Zhou et al. / Composites: Part B 41 (2010) 423–433 429

tions obtained from the test, as shown in Fig. 7a, are used to deter- the region that runs from 140 mm from the free end to the loaded
mine the parameters a and b using Eq. (15). From Eq. (13), the end. Ten response points are highlighted in Fig. 7b, including the
effective bond length Le is calculated to be 140 mm when the value point near the peak strength (P5) and the failure point (P10).
d = 0.96, which was adopted in [22], is selected. Comparisons be- Three processes can be identified from Fig. 7c: the loading pro-
tween the model predictions and the directly measured test results cess, the progressive debonding process, and the failure process. In
for the load–slip response, FRP strain distribution, bond stress dis- the loading process, the bond resistance increases gradually and
tribution, and bond–slip response are shown in Fig. 7b–e, respec- approaches the peak strength P5 while the FRP strip remains
tively, with solid lines. The analytical results from the model for soundly bonded to the substrate concrete. When the load in-
an infinite bond length are also plotted for comparison with creases, the bond stress at the loaded end of the FRP strip
dash-dot lines. The results from the two models are very close in (x = 300 mm) experiences a complete cycle of ascent and descent

Fig. 7. Analytical solutions for specimen EB300-2.


430 Y.-W. Zhou et al. / Composites: Part B 41 (2010) 423–433

(see Fig. 7d). As the bond stress drops to zero and a full stress block resistant stress can no longer balance the pull force, which causes
first appears at the loaded end in a region that has a length equal to the sudden separation of the FRP strip from the concrete substrate
the effective bond length Le (see the distribution for P5 in Fig. 7d), and hence final failure. In this short failure process, the bond resis-
the stress block starts to move from the loaded end toward the free tance near the free end still manages to balance the pull force,
end, signifying the onset of the debonding process at the loaded leading to different FRP strain and bond stress distributions that
end. In the progressive debonding process, the pull force approxi- have higher peak stresses but shorter bonded zones, as shown in
mately maintains in its peak value (it increases just a little) as Fig. 7d and e. Thus, the bond behavior near the free end is different
the slip increases from P5 to P10 (see Fig. 7b). When the remaining from that near the loaded end. Although this conclusion is derived
bonded zone becomes shorter than the effective bond length Le, the from the analytical model, it is physically logical due for the fol-

Fig. 8. Analytical solutions for specimen CII80-60-60-1.


Y.-W. Zhou et al. / Composites: Part B 41 (2010) 423–433 431

lowing reason: the slip is the displacement of the FRP strip relative Based on Eqs. (13) and (22), the ratio of the peak stress Rs be-
to the fixed base, which is largely contributed by the movement of tween the two locations at x = Le and x = Lf is calculated to be
the surface concrete relative to the base concrete. When the stress
1 þ b=Le
block moves to the free end, the bonded area diminishes. At the Rs ¼ : ð26Þ
1 þ b=Lf
same time, the movement of the surface concrete in the smaller
bonded zone near the free end is restrained by the concrete behind Substituting the values of b, Le, and Lf into Eq. (26) gives Rs the
the free end, which is not under a direct pull force at the surface value of 1.075. Thus, the variation in the peak stress in the stable
from the FRP strip. In other words, the movement of concrete in zone, which is defined as the area from the loaded end to the loca-
the smaller bonded zone carries with it the concrete behind the tion that is a distance Le away from the free end, is only 7.5%. As Le
free end. Thus, the resistance from the base against a certain dis- is always smaller than Lf in long joints, it can be clearly seen from
placement increases when the contact area is reduced. This phe- Eq. (26) that Rs is always greater than 1 but is close to 1 in the prac-
nomenon can be seen as an effect of the boundary restraint. tical range of variables. Hence, it is concluded that the peak bond

Fig. 9. Analytical solutions for specimen S4inch.


432 Y.-W. Zhou et al. / Composites: Part B 41 (2010) 423–433

stress is approximately constant inside the stable zone, which It should be noted that the free-end slip s0 is taken as zero in the
moves towards the free end when the load increases, as shown foregoing examples. For long EB–FRP joints, this is certainly cor-
in Fig. 7d. The dash-dot line in Fig. 7d, which shows the bond stress rect. However, for short joints, a small degree of free-end slip has
distribution at point P5 for an infinite bond length (given by Eq. been observed in our tests just before complete debonding. Never-
(6)), further illustrates the insignificant difference of the two ana- theless, our observations [17,36] indicate that the slip at the free
lytical models within the stable zone. end is more than two orders of magnitude smaller than the maxi-
In the unstable zone, which is defined as the area from the free mum slip at the interface, and thus for simplicity it is neglected in
end to the point that is a distance of Le from the free end, as shown the examples with no effect on the accuracy of the calculations.
in Fig. 7d, the bond stress distribution changes rapidly. The length
of this area is equal to the effective bond length Le, which is just 4. Summary and conclusions
sufficient to accommodate the last full stress distribution. The
envelope curve of the peak stress in this area grows rapidly, as An analytical model is developed to study the bond–slip rela-
shown. However, the variation in bond stress in this zone occurs tionship of EB–FRP joints. The model is applicable to EB–FRP joints
in an instant and the complete debonding of the FRP strip occurs with any bond length, and can be reduced to the same mathemat-
abruptly, which cannot be captured in an experiment. ical form to that developed by Dai et al. [19] when the bond length
Fig. 7e shows that the bond–slip responses differ at different approaches infinity. Although the proposed model has the same
locations. However, the bond–slip response for any point inside mathematical form as that in [19] for infinitely long EB–FRP joints,
the stable zone is approximately the same, and close to that pre- the bond–slip relationship is derived using different experimental
dicted by the model for an infinite bond length, as shown by the measurements that are statistically more stable and less affected
dash-dot lines in the figure. It can thus be concluded that the mod- by incidental measurement error. For EB–FRP joints with a limited
el for an infinite bond length is practically applicable for the anal- bond length, and particularly for those that are shorter than the
ysis of bond behavior in the stable zone. However, in the unstable effective bond length Le, the existing models are shown to be inap-
zone, the bond–slip responses differ greatly. The peak stress in- propriate, whereas the proposed model is demonstrated to be
creases quickly near the free end in conjunction with a reduction mathematically reasonable and matches the test results very well.
in the horizontal span of the curve, yet at the free end the peak The following conclusions can be drawn from the proposed model.
stress approaches infinity.
1. The bond–slip response at EB–FRP joints with an infinite bond
3.5. Analysis of the results for short joints length is unique and does not vary along the bond face.
2. Conclusion 1 is not applicable when the bond length is limited,
This section discusses another two examples with shorter particularly when the bond length is shorter than the effective
bonds. The first example is specimen CII80-60-60-1 in Table 1, bond length Le.
which has a bond length of 60 mm and was tested by the authors 3. For long EB–FRP joints with a bond length of greater than Le,
[36]. The details of the test specimen are given in Table 1. The test Conclusion 1 and the model for an infinite bond length is
results are shown in Fig. 8 with discrete points. The parameters a approximately applicable with a reasonable accuracy in bond
and b are evaluated by regression analysis of the slip distribution areas that are a distance greater than Le from the free end.
curves and the results given in Table 1, from which the analytical 4. For short EB–FRP joints that have a bond length of less than Le or
results are calculated and plotted in Fig. 8 with solid lines. for long joints in an area that is a distance of less than Le from
Nine response points are marked in the load–slip curve in the free end, the bond–slip relation varies greatly and the model
Fig. 8b. The strain and stress distribution curves corresponding to for an infinite bond length is not applicable. The peak stress
these points are further investigated in Fig. 8c–e. In Fig. 8c, there increases quickly towards the free end, in conjunction with a
are only two rather than three processes in the development of reduction in the horizontal span of the bond–slip response
the FRP strains. As the bond length is close to the effective bond curve.
length Le = 50 mm, no clear progressive debonding process is ob-
served after the loading process, and failure occurs quickly after
the onset of zero bond stress at the loaded end.
Acknowledgement
Fig. 8d shows the stress developments during loading. The dis-
crete points show reasonable agreement between the directly
The work described in this paper was fully supported by a Grant
measured test results and the analytical results. Distinct from the
from the City University of Hong Kong (Project No. 7002328).
case for long joints, there is no clear stable zone, because the total
bond length is close to Le. When the load increases, the centroid of
References
the bond stress block moves toward the free end. The bond–slip re-
sponses along the whole interface are shown in Fig. 8e, along with [1] Teng JG, Chen JF, Smith ST, Lam L. FRP-strengthened RC structures. UK: John
the bond–slip responses directly measured from the test at three Wiley & Sons; 2002.
locations. The analytical results are similar to those derived from [2] Maeda T, Asano Y, Sato Y, Ueda T, Kakuta Y. A study on bond mechanism of
carbon fiber sheet. In: Proceedings of 3rd international symposium on non-
the direct strain measurements in Fig. 8e, and the bond–slip re- metallic (FRP) reinforcement for concrete structures, Sapporo, Japan Concrete
sponse differs significantly at different locations. Institute, vol. 1; 1997. p. 279–85.
The second example is taken from a pull-off test reported in the [3] Brosens K, Gemert D. Anchoring stresses between concrete and carbon fiber
reinforced laminates. In: Proceedings of 3rd international symposium on non-
literature [6] on specimen S4inch in Table 1, which has a bond metallic (FRP) reinforcement for concrete structures, Sapporo, Japan Concrete
length Lf = 102 mm. Similar results to those in Fig. 8 are produced Institute, vol. 1; 1997. p. 271–8.
and shown in Fig. 9. In this example, the effective bond length is [4] De Lorezis L, Miller B, Nanni A. Bond of fiber-reinforced polymer laminates to
concrete. ACI Mater J 2001;98(3):256–64.
Le = 123 mm, which means that the total bond length Lf is smaller
[5] Nakaba K, Toshiyuki K, Tomoki F, Hiroyuki Y. Bond behavior between fiber-
than the effective bond length. Thus, the pull load P should not reinforced polymer laminates and concrete. ACI Struct J 2001;98(3):359–67.
have peaked at final failure, as can clearly be seen in Fig. 9b. The [6] Chajes MJ, Finch WWJ, Januszka TF, Thonson TAJ. Bond and force transfer of
directly measured results are also shown in the figure with discrete composite material plates bonded to concrete. ACI Struct J 1996;93(2):
208–17.
points for comparison, and show a reasonably similar trend to the [7] Monti G, Renzelli M, Luciani P. FRP adhesion in uncracked and cracked
model predictions. concrete zones. In: Proceedings of 6th international symposium on FRP
Y.-W. Zhou et al. / Composites: Part B 41 (2010) 423–433 433

reinforcement for concrete structures. Singapore: World Scientific [23] Lu XZ, Teng JG, Ye LP, Jiang JJ. Bond–slip models for FRP sheets/plates bonded
Publications; 2003. p. 183–92. to concrete. Eng Struct 2005;27(6):920–37.
[8] Laura DL, Brian M, Antonio N. Bond of fiber-reinforced polymer laminates to [24] Lu XZ, Jiang JJ, Teng JG, Ye LP. Finite element simulation of debonding in FRP-
concrete. ACI Mater J 2001;98(3):256–64. to-concrete bonded joints. Constr Build Mater 2006;20(6):412–24.
[9] Ferracuti B, Savoia M, Mazzotti C. A numerical model for FRP-concrete [25] Lu XZ, Ye LP, Teng JG, Jiang JJ. Meso-scale finite element model for FRP sheets/
delamination. Composites: Part B 2006;37(4–5):356–64. plates bonded to concrete. Eng Struct 2005;27(4):564–75.
[10] Ferracuti B, Savoia M, Mazzotti C. Interface law for FRP-concrete delamination. [26] Chen JF, Pan WK. Three dimensional stress distribution in FRP-to-concrete
Compos Struct 2007;80(4):523–31. bond test specimens. Constr Build Mater 2006;20(1–2):46–58.
[11] Mazzotti C, Savoia M, Ferracuti B. An experimental study on delamination of [27] Wu YF, Wang ZY, Liu K, He W. Numerical analyses of hybrid-bonded FRP
FRP plates bonded to concrete. Constr Build Mater 2008;22(7):1409–21. strengthened concrete beams. Comput-Aided Civil Infrastruct Eng (CACAIE)
[12] Wu ZS, Yuan H, Hiroyuki Y, Toshiyuki K. Experimental/analytical study on 2009;24(5):371–84.
interfacial fracture energy and fracture propagation along FRP-concrete [28] Ayman MSK. Experimental and numerical analysis of FRP sheets bonded to
interface. ACI international SP-201-8; 2000. p. 133–52. concrete. PhD thesis, University of Alberta, Canada; 2003.
[13] Kasumassa N, Toshiyuki K, Tomoki F, Hiroyuki Y. Bond behavior between [29] Wu ZS, Yuan H, Niu HD. Stress transfer and fracture propagation in different
fiber-reinforced polymer laminates and concrete. ACI Struct J 2001;98(3): kinds of adhesive joints. J Eng Mech 2002;128(5):562–73.
359–67. [30] Yuan H, Teng JG, Seracino R, Wu ZS, Yao J. Full-range behavior of FRP-to-
[14] Mohamad AA, Kolluru S, Michel G. Experimental investigation and fracture concrete bonded joints. Eng Struct 2004;26(5):553–65.
analysis of debonding between concrete and FRP sheets. J Eng Mech ASCE [31] Yuan H, Chen JF, Teng JG, Lu XZ. Interfacial stress analysis of a thin plate
2006;132(9):914–23. bonded to a rigid substrate and subjected to inclined loading. Int J Solids Struct
[15] Kolluru VS, Christian C, Lucio N. Width effect in the interface fracture during 2007;44(16):5247–71.
shear debonding of FRP sheets from concrete. Eng Fract Mech [32] Teng JG, Yuan H, Chen JF. FRP-to-concrete interfaces between two adjacent
2007;74(4):578–94. cracks: theoretical model for debonding failure. Int J Solids Struct 2006;43(18–
[16] Sharma SK, Mohamed AMS, Goldar D, Sikdar PK. Plate-concrete interfacial 19):5750–78.
bond strength of FRP and metallic plated concrete specimens. Composites: [33] Bizindavyi L, Neale KW. Transfer lengths and bond strengths for composites
Part B 2006;37(1):54–63. bonded to concrete. J Compos Construct ASCE 1999;3(4):153–60.
[17] Yun YC, Wu YF, Tang WC. Performance of FRP bonding systems under fatigue [34] Chen JF, Teng JG. Anchorage strength models for FRP and steel plates bonded to
loading. Eng Struct 2008;30(11):3129–40. concrete. J Struct Eng 2001;127(7):784–91.
[18] Dai JG, Ueda T. Local bond stress slip relations for FRP sheets concrete interfaces. [35] Taljsten Björn. Defining anchor lengths of steel and CFRP plates bonded to
In: Proceedings of 6th international symposium on FRP reinforcement for concrete. Int J Adhes Adhes 1997;17(4):319–27.
concrete structures. Singapore: World Scientific Publications; 2003. p. 143–52. [36] Zhou YW. Analytical and experimental study on the strength and ductility of
[19] Dai JG, Ueda T, Sato Y. Development of the nonlinear bond stress-slip model of FRP-reinforced high strength concrete beam, PhD thesis, Dalian University of
fiber reinforced plastics sheet-concrete interfaces with a simple method. J Technology, Dalian, China; 2009.
Compos Construct ASCE 2005;9(1):52–62. [37] Ascione F, Berardi VP, Feo L, Giordano A. An experimental study on the long-
[20] Ueda T, Dai JG. Interface bond between FRP sheets and concrete substrates: term behavior of CFRP pultruded laminates suitable to concrete structures
properties, numerical modeling and roles in member behaviour. Prog Struct rehabilitation. Compos Part B – Eng 2008;39(7–8):1147–50.
Eng Mater 2005;7(1):27–43. [38] Aprile A, Feo L. Concrete cover rip-off of R/C beams strengthened with FRP
[21] Ueda T, Dai JG. Interface of fiber reinforced polymer laminates externally composites. Compos Part B – Eng 2007;38(5–6):759–71.
bonded to concrete substrate: form test methods to bond modeling. In: [39] Ascione L, Berardi VP, Feo L, Mancusi G. A numerical evaluation of the
Proceedings of the international symposium on bond behaviour of FRP in interlaminar stress state in externally FRP plated RC beams. Compos Part B –
structures; 2005. p. 23–34. Eng 2005;36(1):83–90.
[22] Dai JG, Ueda T, Sato Y. Unified analytical approaches for determining shear [40] Ascione L, Feo L. Modeling of composite/concrete interface of RC beams
bond characteristics of FRP-concrete interfaces through pullout tests. J Adv strengthened with composite laminates. Compos Part B – Eng 2000;31(6–
Concrete Technol 2006;4:133–45. 7):535–40.

You might also like