Sei sulla pagina 1di 16

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/222419382

Experimental study on FRP-to-concrete bonded joints

Article  in  Composites Part B Engineering · September 2004


DOI: 10.1016/j.compositesb.2004.06.001

CITATIONS READS

418 2,140

3 authors, including:

J.G. Teng Jian-Fei Chen


The Hong Kong Polytechnic University Southern University of Science and Technology
316 PUBLICATIONS   17,986 CITATIONS    215 PUBLICATIONS   6,785 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Hot weather concreting: Early age behaviour and cracking risk of concretes with GGBS View project

SMArtPlate (Funding scheme: H2020-MSCA-IF) - A ductile, high energy absorptive and rapid post-tensioning system for extending life of concrete structures View project

All content following this page was uploaded by Jian-Fei Chen on 08 January 2018.

The user has requested enhancement of the downloaded file.


Composites: Part B 36 (2005) 99–113
www.elsevier.com/locate/compositesb

Experimental study on FRP-to-concrete bonded joints


J. Yaoa,b, J.G. Tengb, J.F. Chenc,*
a
Department of Civil Engineering, Zhejiang University, Hangzhou, 310027, P.R. China
b
Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, Hong Kong, China
c
Institute for Infrastructure and Environment, School of Engineering and Electronics, Edinburgh University, Alexander Graham Bell Building,
The King’s Buildings, Edinburgh EH9 3JN, UK
Received 16 January 2004; revised 2 June 2004; accepted 20 June 2004
Available online 7 August 2004

Abstract
The behaviour of bond between FRP and concrete is a key factor controlling the behaviour of concrete structures strengthened with FRP
composites. This article presents an experimental study on the bond shear strength between FRP and concrete using a near-end supported
(NES) single-shear pull test. The test results are found to be in close agreement with the predictions of Chen and Teng’s [J. Struct. Eng.
127(2001) 784] bond strength model, which mutually verifies the reliability of both the test method and the Chen and Teng model in general.
The NES single-shear pull test, given its simplicity and reliability, is therefore a good candidate as a standard bond test. The test results also
showed that Chen and Teng’s [J. Struct. Eng. 127(2001) 784] bond strength model is slightly conservative when the FRP-to-concrete width
ratios are at the two extremes, but this small weakness can be easily removed when more test results of good quality become available.
q 2004 Elsevier Ltd. All rights reserved.

Keywords: A. Polymer-matrix composites (PMCS); B. Debonding; B. Interface; Strength

1. Introduction provided a useful initial basis. The existing work has


included experimental studies conducted using single shear
External bonding of fibre reinforced polymer (FRP) tests, e.g. [11–15], double shear tests, e.g. [16–23] and
composites has become a popular technique for strength- modified beam tests, e.g. [23–25], theoretical studies using
ening concrete structures all over the world [2]. An fracture mechanics analysis [15,26–33] and finite element
important issue in the strengthening of concrete structures analysis [34,35], and the development of empirical
using FRP composites is to design against various models [1,23,36,37]. A review of these studies can be
debonding failure modes, some of which were first studied found in Ref. [1].
for concrete beams bonded with a steel plate, including: (a) Existing studies suggest that the main failure mode of
cover separation [3–5]; (b) plate end interfacial debonding FRP-to-concrete joints in shear tests is cracking of concrete
[3,4,6]; (c) intermediate (flexural or flexural-shear) crack under shear, occurring commonly at a few millimetres from
(IC) induced interfacial debonding [7] and (d) critical diagonal the adhesive-concrete interface [1]. The bond strength (i.e.
crack (CDC) induced interfacial debonding [8–10]. the maximum transferable load) of the joint therefore
The bond strength between FRP and concrete is a key depends significantly on concrete strength. In addition, the
factor controlling debonding failures of various forms in FRP-to-concrete member width ratio has a significant effect.
FRP-strengthened structures. As a result, extensive research A very important aspect of the behaviour of these bonded
on this topic has been carried out, in addition to earlier work joints is that there exists an effective bond length beyond
concerned with steel plates bonded to concrete which which an extension of the bond length cannot increase the
ultimate load. This is the fundamental difference between
externally bonded reinforcement and internal reinforcement
* Corresponding author. Tel.: C44-131-650-6768; fax: C44-131-650-
6789.
for which a sufficiently long anchorage length can always be
E-mail address: jchen@staffmail.ed.ac.uk (J.F. Chen). found that the full tensile strength of the reinforcement can

1359-8368/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compositesb.2004.06.001
100 J. Yao et al. / Composites: Part B 36 (2005) 99–113

be achieved. The majority of existing studies have been their simplicity [33]. Both numerical [33] and experimental
concerned with the prediction of the ultimate load and the [38] studies have shown that different test set-ups can lead to
effective bond length [1]. significantly different test results. Within each test method,
This article presents an experimental study on the bond small variations in the test set-up such as the height of the
shear strength between FRP and concrete using a near-end support block in a NES single- or double-shear test may also
supported (NES) single-shear pull test in which the concrete have significant effects based on a recent stress analysis
prism is supported at the end nearer the applied load. These [33].
tests have been conducted with the following purposes: (a) An FRP-to-concrete bond strength model is the key to the
to examine the reliability and robustness of the NES single- accurate prediction of debonding failures in FRP-strength-
shear pull test as a candidate standard bond test; and (b) to ened RC beams, including shear crack-induced debonding
verify the accuracy of the bond strength model recently failures [8,39] as well as intermediate flexural or flexural-
developed by Chen and Teng [1]. shear crack-induced debonding failures [7].
In debonding failures in FRP shear-strengthened RC
beams with transverse plates, the bond strength model
2. Test program developed from pull tests is directly applicable [39]. Such a
model is also important in understanding the mechanism of
2.1. Test methods debonding induced by a critical diagonal crack near the end
of a longitudinal tension face plate for flexural strengthening
A recent survey [33] showed that many different [8,10], where the longitudinal plate increases the concrete
experimental set-ups have been used for determining the component of the shear capacity and where the bond
FRP-to-concrete bond strength, but no consensus on a strength developed from pull tests is also directly
standard test procedure has been reached. Chen et al. [33] applicable.
classified the existing test set-ups into the following five Furthermore, in intermediate crack-induced debonding
types: (a) double-shear pull tests; (b) double-shear push failures, the stress state in the critical region of the beam is
tests; (c) single-shear pull tests; (d) single-shear push tests; also closely similar to that of the concrete prism in a NES
and (e) beam (or bending) tests. For better clarity, the first single-shear pull test. The NES single-shear pull test
four test methods are renamed here as: (a) far end supported therefore appears to be a promising candidate as a standard
(FES) double-shear tests; (b) near end supported (NES) set-up for determining the FRP-to-concrete bond strength
double-shear tests; (c) far end supported (FES) single-shear and was therefore adopted in the present study. One of the
tests; and (d) near end supported (NES) single-shear tests aims of the present experimental study is to examine the
(Fig. 1). Collectively, all these four tests may also be effect of a number of small variations in this test set-up on
referred to as pull tests, as the plate is always directly pulled the resulting bond strength to aid in fine-tuning this test
by a tensile force. method as a standard bond test method. Results from
FES double-shear pull tests and NES single-shear pull previous NES single-shear pull tests also formed part of the
tests have been the most popular test methods so far due to database on which Chen and Teng’s [1] recent bond strength

Fig. 1. Classification of bond tests (Chen et al. 2001).


J. Yao et al. / Composites: Part B 36 (2005) 99–113 101

Fig. 2. Test specimen.

model was based, so the present test results also provide an results of all specimens are listed in Table 1. Specimens
appropriate independent check of the validity of this bond II-1 and II-4 had a loading offset of dZ4 mm (equivalent
strength model. to an initial loading angle of 1.78) whilst Specimens II-3
and II-6 had a loading offset of dZK4 mm (equivalent to
2.2. Specimen design an initial loading angle of K1.78). All other specimens
had no loading offset.
The NES single-shear pull test specimens consisted of a Concrete prisms of two different sizes were used. Half of
concrete prism bonded with an FRP strip (Fig. 2). The the specimens in Series III and V used 100!150!350 mm
factors considered in the present test program include the concrete prisms so that a desired range of bfrp/bc ratios could
bond length Lfrp, the width ratio between the FRP strip and be achieved. All other specimens used 150!150!350 mm
the concrete prism bfrp/bc, the height of the concrete free concrete prisms. Concrete cubes and cylinders were tested
edge hc (Zheight of concrete prism hKheight of the according to BS 1881 [43] to determine the material
support block hb) (Fig. 2) and the offset in the load position properties at the time when the series of specimens made
d. The first two factors have been identified to have a from the same batch of concrete were tested.
significant effect on the bond strength but there have been GFRP was used in Specimens III-7 and III-8 while CFRP
insufficient test data to rigorously verify the proposed was used in all others. The nominal thicknesses for the
relationships [1]. The height of the concrete free edge hc CFRP and GFRP strips were 0.165 and 1.27 mm respect-
(Fig. 2b) has been shown to have a significant effect on the ively, the former being roughly the fibre sheet thickness
stress distribution in the specimen [33], but its effect on the before resin impregnation with the latter being similar to the
ultimate bond strength is yet unclear. In practical pull tests, thickness of the cured FRP strip. The FRP strips were
there may be a small unintended offset d in the position of bonded to the concrete prisms following the manufacturer’s
the load (Fig. 2b). This offset may alternatively be expressed instructions. The mechanical properties of the FRP
as the initial loading angle q. The effect of this loading angle composites are shown in Table 2. The tensile strengths of
needs to be understood if standardisation of the test set-up is FRPs were determined according to ASTM D3039/
to be considered in the future. Furthermore, in flexurally D3039M-95a [44] on the basis of the nominal thicknesses.
strengthened concrete structures, when debonding is The nominal thicknesses were also used in all other
induced by the opening up of a flexural-shear crack, there calculations of the present study. FRP composites were
exists a relative vertical displacement between the two sides
of the crack, e.g. [40–42], so the FRP strip (or plate or sheet)
is loaded at a small positive (peeling) inclination angle to
the longitudinal axis on one side and at the same but
negative angle on the other side of the crack (Fig. 3). This is
thus another reason why the effect of a small loading angle
is worthy of some attention.
A total of 72 specimens in seven series were prepared
to investigate the effects of the above factors on the bond
strength (Table 1). The variables considered in Series I
(Specimens I-1–16) include the bond length Lfrp and the
support height hb (or height of the free concrete edge on
the loading side hcZhKhb). Series II (Specimens II-1–6)
and III (Specimens III-1–8) were designed to investigate
the effects of the loading offset and the FRP-to-concrete
width ratio respectively. Series IV–VII (Specimens IV-1–
14, V-1–12, VI-1–8 and VII-1–8) were designed following
the completion of the first three series to further explore Fig. 3. Relative vertical displacement between two sides of a flexural-shear
the effects of Lfrp, bfrp/bc and hc. Key parameters and test crack.
102 J. Yao et al. / Composites: Part B 36 (2005) 99–113

Table 1
Details of specimens and test results

Test Concrete Width of FRP width FRP bond Height of Test failure Test failure Predicted Ptest/Ppred
specimen cylinder concrete bfrp (mm) length Lfrp free concrete load Ptest mode failure load
strength f 0 c prism bc (mm) edge hc (kN) Ppred (kN)
(MPa) (mm) (mm)
I-1 23.0 150 25 75 5 4.75 DB-C 5.72 0.83
I-2 23.0 150 25 85 5 5.69 DB-C 5.96 0.96
I-3 23.0 150 25 95 5 5.76 DB-C 6.02 0.96
I-4 23.0 150 25 95 5 5.76 DB-C 6.02 0.96
I-5 23.0 150 25 95 5 6.17 DB-C 6.02 1.02
I-6 23.0 150 25 115 5 5.96 DB-C 6.02 0.99
I-7 23.0 150 25 145 5 5.95 DB-C 6.02 0.99
I-8 23.0 150 25 190 5 6.68 DB-C 6.02 1.10
I-9 23.0 150 25 190 5 6.35 DB-C 6.02 1.05
I-10 23.0 150 25 95 75 6.17 DB-C 6.02 1.02
I-11 23.0 150 25 75 120 5.72 DB-C 5.72 1.00
I-12 23.0 150 25 85 120 6 DB-C 5.96 1.01
I-13 23.0 150 25 95 120 6.14 DB-C 6.02 1.02
I-14 23.0 150 25 115 120 6.19 DB-C 6.02 1.03
I-15 23.0 150 25 145 120 6.27 DB-C 6.02 1.04
I-16 23.0 150 25 190 120 7.03 DB-C 6.02 1.17
II-1 22.9 150 25 95 120 5.2 DB-C 6.02 0.86
II-2 22.9 150 25 95 120 6.75 DB-C 6.02 1.12
II-3 22.9 150 25 95 120 5.51 DB-C 6.02 0.92
II-4 22.9 150 25 190 120 7.02 DB-C 6.02 1.17
II-5 22.9 150 25 190 120 7.07 DB-C 6.02 1.17
II-6 22.9 150 25 190 120 6.98 DB-C 6.02 1.16
III-1 27.1 150 25 100 120 5.94 DB-C 6.27 0.95
III-2 27.1 150 50 100 120 11.66 DB-C 11.19 1.04
III-3 27.1 150 75 100 120 14.63 DB-C 15.02 0.97
III-4 27.1 150 100 100 120 19.07 DB-C 17.91 1.06
III-5 27.1 100 85 100 120 15.08 CPF 13.42 1.12
III-6 27.1 100 100 100 120 15.75 CPF 14.16 1.11
III-7 27.1 100 25.3 100 120 4.78 DB-C 4.92 0.97
III-8 27.1 100 50.6 100 120 8.02 DB-C 8.30 0.97
IV-1 18.9 150 25 95 5 5.86 DB-C 5.72 1.02
IV-2 18.9 150 25 95 5 5.9 DB-C 5.72 1.03
IV-3 19.8 150 25 95 5 5.43 DB-C 5.80 0.94
IV-4 19.8 150 25 95 5 5.76 DB-C 5.80 0.99
IV-5 18.9 150 25 95 15 5 DB-C 5.72 0.87
IV-6 19.8 150 25 95 15 7.08 DB-C 5.80 1.22
IV-7 18.9 150 25 95 30 5.5 DB-C 5.72 0.96
IV-8 19.8 150 25 95 30 5.93 DB-C 5.80 1.02
IV-9 18.9 150 25 95 45 5.38 DB-C 5.72 0.94
IV-10 19.8 150 25 95 45 6.6 DB-C 5.80 1.14
IV-11 18.9 150 25 95 60 5.51 DB-C 5.72 0.96
IV-12 19.8 150 25 95 60 5.67 DB-C 5.80 0.98
IV-13 18.9 150 25 95 90 6.31 DB-C 5.72 1.10
IV-14 19.8 150 25 95 90 6.19 DB-C 5.80 1.07
V-1 21.1 150 15 95 60 3.81 DB-C 3.71 1.03
V-2 21.1 150 15 95 60 4.41 DB-C 3.71 1.19
V-3 21.1 150 25 95 60 6.26 DB-C 5.89 1.06
V-4 21.1 150 50 95 60 12.22 DB-C 10.51 1.16
V-5 21.1 150 75 95 60 14.29 DB-C 14.10 1.01
V-6 21.1 150 100 95 60 15.58 DB-C 16.82 0.93
V-7 21.1 100 80 95 60 14.27 CPF 12.28 1.16
V-8 21.1 100 80 95 60 13.78 CPF 12.28 1.12
V-9 21.1 100 90 95 30 13.56 CPF 12.88 1.05
V-10 21.1 100 90 95 5 15.66 CPF 12.88 1.22
V-11 21.1 100 100 95 30 15.57 CPF 13.30 1.17
V-12 21.1 100 100 95 5 17.43 CPF 13.30 1.31
VI-1 21.9 150 25 95 60 6.01 DB-I 5.95 1.01
VI-2 21.9 150 25 95 60 5.85 DB-I 5.95 0.98
VI-3 21.9 150 25 145 60 5.76 DB-I 5.95 0.97
VI-4 21.9 150 25 145 60 5.73 DB-I 5.95 0.96
(continued on next page)
J. Yao et al. / Composites: Part B 36 (2005) 99–113 103

Table 1 (continued)

Test Concrete Width of FRP width FRP bond Height of Test failure Test failure Predicted Ptest/Ppred
specimen cylinder concrete bfrp (mm) length Lfrp free concrete load Ptest mode failure load
strength f 0 c prism bc (mm) edge hc (kN) Ppred (kN)
(MPa) (mm) (mm)
VI-5 21.9 150 25 190 60 5.56 DB-I 5.95 0.93
VI-6 21.9 150 25 190 60 5.58 DB-I 5.95 0.94
VI-7 21.9 150 25 240 60 5.91 DB-I 5.95 0.99
VI-8 21.9 150 25 240 60 5.05 DB-I 5.95 0.85
VII-1 24.9 150 25 95 60 6.8 DB-C 6.14 1.11
VII-2 24.9 150 25 95 60 6.62 DB-C 6.14 1.08
VII-3 24.9 150 25 145 60 7.33 DB-C 6.14 1.19
VII-4 24.9 150 25 145 60 6.49 DB-C 6.14 1.06
VII-5 24.9 150 25 190 60 7.07 DB-C 6.14 1.15
VII-6 24.9 150 25 190 60 7.44 DB-C 6.14 1.21
VII-7 24.9 150 25 240 60 7.16 DB-C 6.14 1.17
VII-8 24.9 150 25 240 60 6.24 DB-C 6.14 1.02
Average 1.04
CoV 9.6%

Note: (a) CFRP was used in all specimens except III-7 and III-8 in which GFRP was used; (b) all concrete prisms had a height of 150 mm; (c) concrete cylinder
strength determined from cube strength according to fcLZ0.79 fcuL (d) DB-C, debonding in concrete; DB-I, debonding at adhesive-concrete interface; CPF,
Concrete prism failure.

bonded to the concrete prisms with epoxy resins. More Table 2


details of the material properties and specimen preparation Properties of FRPs
procedures are available in Ref. [45]. Type Thickness Tensile Elastic mod- Ultimate
(mm) strength ffrp ulus Efrp tensile
2.3. Test set-up (MPa) (GPa) strain 3frp
(%)
A steel rig for NES single-shear pull tests (Fig. 4a) was CFRP 0.165 4114 256 1.61
carefully fabricated to carry out all the tests reported in this GFRP 1.27 351 22.5 1.56
article. In this rig, the load could be accurately positioned
vertically by adjusting the height of the bearing plate.
Different support blocks could be used to achieve the
required support heights on the loaded end (i.e. the end
nearer the applied load or the near end) of the concrete
prism. A positioning frame was used to prevent the far end
of the concrete prism from uplifting. The concrete prism
was separated from the positioning frame by a thin layer of
rubber to allow horizontal sliding of the concrete prism.

2.4. Instrumentation and loading procedure

Strain gauges and LVDTs were used to measure strains


in the FRP and displacements at various positions. Details of
these measurements are not given here, but are available
elsewhere [45], as the main concern of the present paper is
with the bond strength. Loading was applied through a
hydraulic jack at increments of about 5% of the ultimate
load predicted by Chen and Teng’s model [1]. Fig. 4b shows
a specimen during the test.

3. Chen and Teng’s bond strength model

As the specimens were designed based on Chen and


Teng’s bond strength model [1] and the results are Fig. 4. Test rig.
104 J. Yao et al. / Composites: Part B 36 (2005) 99–113

Fig. 5. Debonding in concrete.

compared with its predictions later in the article, it is It may be noted that this model was developed based on
necessary to introduce this model before the test results are a fracture mechanics solution [31] with rational simplifica-
presented. The bond strength expressed as per unit width of tion, with its coefficient regressed from a set of either
the FRP strip, qu, is single-shear or double-shear pull tests on FRP and steel plate-
pffiffiffiffi to-concrete bonded joints. The model is thus semi-empirical
Pu
qZ Z abw bl Le fc0 (1) and generic, being applicable to both FRP (wet lay-up or
bfrp prefabricated) and steel plates, subject to the condition that
failure is not due to yielding of steel or rupture of FRP. The
where Pu is the ultimate load in N, bfrp is the width of the
model was developed for debonding failure in the concrete,
FRP strip in mm, bw and bl are dimensionless coefficients
but may also be applicable to debonding failure at the
reflecting the effects of the FRP-to-concrete width ratio bfrp/
adhesive–concrete interface as shown later.
bc and the bond length Lfrp respectively, Le is the effective
bond length in mm and f 0 c is the cylinder compressive
strength of concrete in MPa. Based on the regression of test
data collected from the literature, Chen and Teng [1] 4. Test results and discussions
obtained the best fit value of aZ0.427. It was proposed to
use the 95th percentile of aZ0.315 as the lower bound for
4.1. Failure modes
design. bw, bl and Le are given by
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 K bfrp =bc Fifty-six out of the 72 specimens failed due to debonding
bw Z (2) in concrete adjacent to the adhesive-concrete interface in
1 C bfrp =bc
which a thin layer of concrete is attached to the FRP strip
8 after failure (Fig. 5). It may be noted that this is not strictly
>
< 1:0 if Lfrp R Le ‘debonding’ because the failure actually occurs in concrete.
b1 Z p Lfrp (3) Nevertheless, the term is still adopted here because it has
>
: sin if Lfrp ! Le been widely used by the research community as discussed in
2 Le
Ref. [1]. Eight specimens failed by debonding at the
sffiffiffiffiffiffiffiffiffiffiffiffiffiffi adhesive-concrete interface where much less concrete is
Efrp tfrp attached to the FRP strip after failure (Fig. 6). The
Le Z pffiffiffiffi (4)
fc0 remaining eight specimens failed in the concrete prism by
the formation of a fracture plane that starts at the far end
in which Efrpand f 0 c are in MPa while tfrp and Le are in mm. of the FRP strip and extends to the top of the support

Fig. 6. Debonding at the adhesive–concrete interface.


J. Yao et al. / Composites: Part B 36 (2005) 99–113 105

the mixing of the primer [45]. In light of this, the


concrete prisms used in Series VI were reused in a
following series of tests (series VII) with the FRP strip
bonded to the opposite side of the prism, while all other
parameters remained unchanged. All Series VII speci-
mens failed by debonding in concrete, confirming that
the results of Series VI had been influenced to some
extent by interfacial weakness introduced during
preparation.

4.1.3. Concrete prism failure


Fig. 7. Failure in the concrete prism.
Specimens III-5 and III-6, and V-7–12 failed in the
concrete prism. The failure started by the initiation of a
block (Fig. 7). The failure mode of each specimen is
crack in the concrete prism near the far end of the FRP strip.
indicated in Table 1.
Once the crack appeared, it propagated almost immediately
towards the upper end of the support block and the specimen
4.1.1. Debonding in concrete failed (Fig. 7). The FRP–concrete interface was intact after
For those specimens which failed by debonding in failure and the failure process was catastrophic.
concrete, the failure process started with visible concrete All specimens failed in the concrete prism had a concrete
cracking near the loaded end of the concrete prism. The width of 100 mm (compared with 150 mm for most of other
surface cracks observed on the sides of the FRP strip were specimens) and an FRP strip which was quite wide (bfrp/bc
at about 458 to the longitudinal axis of the FRP strip. As R0.8). This failure is obviously more likely when the FRP
the load increased, visible cracking in the concrete initiated strip and the concrete prism have similar widths, and is more
debonding of the FRP from the concrete at the loaded end. a consequence of the test set-up than any other factors. The
Debonding then propagated towards the far end of the FRP use of the positioning frame to prevent the concrete prism
strip and eventually led to the complete detachment of the from uplifting (Fig. 4a) introduces tensile bending stresses at
FRP strip from the concrete. The duration of this the upper surface of the concrete prism at the far end of the
debonding process depended on the bond length of the FRP strip, while the use of a low support block allows
FRP strip. It was very short or could not be noticed at all formation of a fracture plane at a relatively low load.
for a small bond length but could be easily seen for a long
one. Debonding was due to failure in the concrete at a 4.2. Load–displacement behaviour
small distance beneath the adhesive-concrete interface. A
lump of concrete from the loaded end was generally Fig. 8 shows the load–displacement curves of Series VI
attached to the debonded FRP strip, while a smaller (which failed due to debonding at the adhesive–concrete
concrete lump was sometimes found at the far end of the interface) and VII specimens (which failed due to debond-
FRP strip (Fig. 5). The thickness of the concrete layer on ing in the concrete). The displacement was measured at the
the debonded FRP strip elsewhere varied approximately right end of the grip (Fig. 4a) so it includes not only the
between 1 and 5 mm. The surface of the failure zone of the displacement due to interfacial slip, but also a number of
concrete prism was uneven, with the aggregate being other components such as the elastic deformation of the un-
clearly visible (Fig. 5). The phenomenon that more bonded part of the FRP strip and possible slip of the FRP
concrete was usually attached to the FRP strip at both strip in the grip. Therefore, these curves shall only be treated
ends of the interface than elsewhere may be related to the as qualitative information reflecting the global load–
stress concentration at the ends [34]. displacement response.
Initially, the displacement increases almost linearly with
4.1.2. Debonding at the adhesive–concrete interface the load and the slopes of different curves are similar. Faster
The failure process was almost same as the above, but the increases in the displacement indicate the initiation of
failure was mostly along the adhesive–concrete interface. micro-cracking at the loaded end. Substantial differences
The FRP strip generally also pulled off a lump of concrete at between the curves are observed for the later stage of
the loaded end. Much less or little concrete was attached to loading as failure was approached and these differences are
the FRP strip elsewhere (Fig. 6). attributable to different bond lengths and different failure
This failure occurred only in eight Series VI speci- modes. For Series VI specimens with debonding at the
mens. It may be noted that Series IV, V and VI adhesive-concrete interface, all curves feature a plateau
specimens were prepared by an assistant with limited before ultimate failure, and the length of the plateau
experience under the supervision of an experienced increases with the bond length. For Series VII specimens
researcher. There was some uncertainty with regard with debonding in the concrete, only those for a long bond
to the surface preparation of the concrete prisms and length feature such a plateau and it is much short than that of
106 J. Yao et al. / Composites: Part B 36 (2005) 99–113

Fig. 8. Load–displacement curves: Series VI and VII specimens.


J. Yao et al. / Composites: Part B 36 (2005) 99–113 107

Fig. 9. Strain distribution along the FRP strip for Specimen I-1: LfrpZ75 mm.

the corresponding specimen failed in FRP debonding at For Specimen I-1 with a small bond length (LfrpZ
concrete/adhesive interface. 75 mm), the increase of FRP strain is gradual until P reaches
0.89Pu (PuZ4.75 kN) (Fig. 9b). Cracking at the loaded end
4.3. Strain distributions in FRP was first observed by naked eyes (i.e. visible cracking) at
PZ4.5 kN (P/PuZ0.95). This cracking led to an obvious
Figs. 9 and 10 show typical distributions of strains in change of the strain distribution in the FRP strip indicating
the FRP strip. These strains were found from strain the propagation of debonding, and the specimen failed soon
gauges mounted on the upper surface of the FRP strip, thereafter. The strain in the debonded part of the FRP strip is
except the strains at xZ0 which were deduced directly seen to be almost constant.
from the applied load and the geometric and material For Specimen I-16 with a large bond length (LfrpZ
properties of the FRP strip, as readings from the strain 190 mm), visible cracking occurred at a similar load (i.e.
gauge at this location were found to be significantly PZ4.75 kN) but ultimate failure occurred at a higher load
affected by local bending of the strip. When the applied (PuZ7.03 kN). The propagation of debonding is more
load P is smaller than about 60% of the ultimate load Pu, clearly reflected by the strain distribution as shown in
the FRP strain is minimal beyond a small distance of Fig. 10b. It may be noted that a large part of the FRP strip
about 0.5Le from the loaded end (Figs. 9a and 10a), near the far end still had minimal strain when the ultimate
indicating that almost all the applied load is resisted load was reached, confirming the concept of effective bond
within this small area. Here Le is the effective bond length implying that increasing the bond length beyond a
length according to Chen and Teng’s model [1]. certain value does not further increase the bond strength.
108 J. Yao et al. / Composites: Part B 36 (2005) 99–113

Fig. 10. Strain distribution along the FRP strip for Specimen I-16: LfrpZ190 mm.

However, a larger bond length can be expected to lead to a the FRP to the concrete by a small distance (less than
longer deformation process as debonding propagates along 0.1LeZ9.4 mm in this case) towards the free end of the FRP
the interface. strip. This phenomenon has also been noted by Yuan et al.
Careful inspection of Figs. 9 and 10 reveals that local [33] and may be attributed to local stress concentration near
debonding near the loaded end occurred much earlier than the loaded end [33,34]. The same phenomenon is evident
was observed by naked eyes. Fig. 9a shows that there is a from the strain distributions shown in Fig. 10 for Specimen
significant change in the local strain distribution near the I-16, where local debonding appears to have occurred at a
loaded end (xZ0) when the applied load increases from load P less than 0.31Pu (Fig. 10a).
0.21Pu to 0.38Pu. When PZ0.21Pu, the deduced axial strain
at xZ0 is significantly larger than that measured on the 4.4. Effect of height of free concrete edge
upper surface of FRP at xZ0.1Le. The strain decreases fast
away from the loaded end. When the load increases to over Test results from Series I and II for various heights of the
0.38Pu, the deduced strain at xZ0 becomes slightly smaller free zone at the near end of the concrete prism (i.e. height of
than that measured at xZ0.1Le and this pattern remains the free concrete edge hcZhKhb) are shown in Fig. 11a
unchanged until failure. This phenomenon is believed to be and b. It is seen that the bond strengths of specimens with
due to very local debonding (not visible to naked eyes) that hcZ120 mm are consistently larger than those with
occurred before the applied load reached 0.38Pu. This local hcZ5 mm, with the difference being of the order of 10%.
debonding moves the effective point of stress transfer from This indicates that the height of free concrete edge does
J. Yao et al. / Composites: Part B 36 (2005) 99–113 109

for hc, as they only cover three values of hc. Series IV was
thus designed to further explore this issue, as this
information is useful for the development of a standard
bond test method. However, no definite conclusion can be
drawn from the results of Series IV (Fig. 11c) because they
show a relatively large scatter which may be attributed to
the less stringent specimen preparation procedure of these
specimens as discussed earlier in the article.

4.5. Effect of bond length

Fig. 12 shows the relationship between the FRP bond


strength and the bond length for all the specimens with the
same CFRP width of 25 mm and no loading offset from
Series I, II, VI and VII. The predictions of Chen and Teng’s
model [1] are also shown for comparison. These test results
clearly support the concept of an effective bond length and
the accuracy of the effective bond length formula of Chen
and Teng’s model.
It is seen that the test results from Series I with hcZ5 mm
and those from Series VI are slightly below Chen and
Teng’s predictions, whilst those from Series VII are above
the predictions. Overall, the experimental results are nicely
scatted around the predictions of Chen and Teng’s model.
This observation agrees with expectation because Chen and
Teng’s model was developed to provide best-fit predictions
of test data collected from the literature which are expected
to have been obtained from slightly different test set-ups and
by different researchers.
It should be noted that Series VI specimens failed due to
debonding at the adhesive–concrete interface as a result of a
less stringent specimen preparation procedure, while the
more carefully prepared specimens of Series VII failed due
to debonding in concrete. Clearly, Series VII results are
significantly higher than those of Series VI, further
confirming the importance of careful specimen preparation.

4.6. Effect of loading offset

Fig. 13 shows that both a positive and a negative loading


offset have a significant effect on the bond strength when the
bond length is small (LfrpZ95 mm). The loading offsets of
Fig. 11. Effect of height of free concrete edge. G4 mm (i.e. initial loading angles of G1.78) reduced the
bond strength significantly. This may reflect the effect of the
have some effect on the bond strength, which is in loading angle on the local stresses near the loaded end. It is
agreement with previous numerical observations [34]. shown in Ref. [46] that both small positive and negative
This is because the local stiffness near the loaded end is angles increase the principal tensile stress locally and thus
increased when the top of the support block is closer to the may have a detrimental effect on the bond strength.
FRP plate (smaller hc values). This increased local stiffness The effect of a small loading angle is insignificant for a
attracts an increased local stress transfer from the FRP to the relatively long bond length of 190 mm (Fig. 13). A possible
concrete there, leading to early debonding and hence a explanation for this phenomenon may be as follows. For a
reduced bond strength. positive loading angle, as debonding propagates, the loading
Numerical results from linear elastic analysis [34] have angle and thus its effect reduces. When the loading angle is
shown that there is a range in which the interfacial stress negative, two factors may contribute to this phenomenon.
distribution is insensitive to hc. The test data shown in First, assuming that the bond length is sufficiently large, the
Fig. 11a and b do not allow the identification of such a range debonding crack appears first at the loaded end and then
110 J. Yao et al. / Composites: Part B 36 (2005) 99–113

Fig. 12. Effect of bond length.

progresses towards the far end of the FRP strip, in contrast of bfrp/bc (i.e. close to 0 and 1), such a modification is not
to specimens with a short bond length in which complete attempted here.
failure is reached immediately when debonding starts. Once
the debonding crack has progressed by a small distance, the
effect of a negative loading angle disappears as the
debonded portion of the FRP strip has to remain in contact 5. Comparison with Chen and Teng’s predictions
with the concrete. Second, a negative loading angle results
in compressive normal stresses on the debonded area, which A comparison between the present test data and the
produce frictional forces to help resist the applied load. predictions of Chen and Teng’s model [1] is shown in
Therefore, a small negative loading angle is expected to Fig. 15. Statistics of the test-to-predicted bond strength ratio
have no detrimental effect on the bond strength if the bond are given in Table 3. Here the effects of the height of free
length is sufficiently large. concrete edge at the loaded end and the loading offset are
These test results illustrate the importance of a reliable treated as factors contributing to the experimental scatter. It
set-up for the determination of bond strength in a pull test. is seen that Chen and Teng’s model [1] underestimates the
Since a small loading offset is hard to avoid, the bond length bond strength by 4% on average for failure by debonding in
of the FRP strip in a bond test specimen should be the concrete, but overestimates the bond strength by the
sufficiently long to minimise the effect of a loading offset. same percentage for failure by debonding at the adhesive-
They also imply that in the flexural strengthening of beams concrete interface, with the coefficient of variation being
and slabs, it is important to provide a sufficient bond less than 10% in both cases (Table 3). This comparison
(anchorage) length so that the effect of relative vertical confirms that Chen and Teng’s model [1] represents very
displacements between the two sides of a flexural-shear closely the bond strength overall.
crack can be minimised. The average test-to-predicted bond strength ratio and its
standard deviation for the complete data set containing both

4.7. Effect of FRP-to-concrete width ratio

Fig. 14 shows the effect of the FRP-to-concrete width


ratio bfrp/bc on the bond strength. It is seen that Chen and
Teng’s model [1] underestimates slightly the bond strength
both when bfrp/bc is close to 0 or 1. It may be noted that the
specimens with bfrp/bc close to 1 failed in concrete prism so
their actual bond strength should be even higher than the test
results. Whilst Eq. (2) may be modified to slightly better fit
the data points shown in Fig. 14, such a modifications is
statistically insignificant for a combined database contain-
ing test data presented in this article and those in Ref. [1]. As
there is still a lack of high quality test data at both extremes Fig. 13. Effect of loading offset displacement.
J. Yao et al. / Composites: Part B 36 (2005) 99–113 111

Fig. 14. Effect of FRP-to-concrete width ratio.

specimens which failed by debonding in the concrete and 6. Conclusions


those which failed by debonding at the adhesive-concrete
interface are 1.03 and 0.093, respectively. The a value in This article has presented an experimental study on the
Eq. (1) for the 95th percentile can be easily found to be 0.37 bond shear strength between FRP and concrete using a near-
(Z0.427!(1.03K1.64!0.093)) which is about 20% larger end supported (NES) single-shear pull test in which the
than the value of 0.315 obtained from the database presented concrete prism is supported at the end nearer the applied
in Ref. [1]. This is understandable because the standard load. These tests have been conducted with the following
deviation of the test data presented here is (and should be) purposes: (a) to examine the reliability and robustness of the
smaller than that of the data presented in Ref. [1] which NES single-shear pull test as a candidate standard bond test;
were obtained by different researchers. As the actual quality and (b) to verify the accuracy of the bond strength model
variations at practical construction sites with different recently developed by Chen and Teng [1]. The results and
application personnel may be larger than those experienced discussions presented in the present article allow the
in laboratory tests, aZ0.315 is still recommended here as a following conclusions to be made.
conservative value for design use. However, a more precise (1) Since the NES single-shear pull test as presented in this
value may be proposed when sufficient confidence in such a article produced results which are in close agreement
value has been gained with more extensive research. with the predictions of Chen and Teng’s model [1], the
It may be noted that the bond strength model was not reliability of both the test method and the Chen and Teng
developed for the concrete prism failure mode. This failure model are mutually verified in general. The NES single-
mode can be prevented through the use of a higher support shear pull test, given its simplicity and reliability, is
block and a longer bond length. It is desirably to avoid this therefore a good candidate as a standard bond test. This
and other failure modes which do not have a direct bearing test method is also robust provided a sufficiently long
on the interfacial behaviour of FRP-to-concrete bonded bond length is employed to minimise the effect of
joints in a standard bond test procedure. unintended loading offsets and a sufficiently high support
block is used to avoid non-interfacial failures. Based on
the present test results, it may be recommended that the
bond length in a standard test should be around two times
Table 3
Statistics of test-to-predicted bond strength ratios

Test failure mode Average Standard CoV


deviation (%)
(1) Debonding in concrete 1.04 0.093 8.9
(56 specimens)
(2) Debonding at the adhesive– 0.96 0.050 5.2
concrete interface
(eight specimens)
(1)C(2) 1.03 0.093 9.0
(3) Concrete prism failure (eight 1.16 0.078 6.7
specimens)
All 1.04 0.100 9.6
Fig. 15. Comparison with Chen and Teng’s (2001) model.
112 J. Yao et al. / Composites: Part B 36 (2005) 99–113

the effective bond length specified by Chen and Teng’s Engineering, Vol. 6: Structures—Composites Materials, Structural
model and the height of the free concrete edge should be Systems, Telecommunications Towers, Sherbrooke, Québec, Canada
1997 p. 51–60.
around 50 mm for a concrete prism of 150 mm in height.
[12] Bizindavyi L, Neale KW. Transfer lengths and bond strengths for
In addition, the distance between the positioning frame composites bonded to concrete. J Compos Construct, ASCE 1999;
preventing the uplifting of the concrete prism and the far 3(4):153–60.
end of the FRP strip should be appropriate to avoid high [13] Chajes MJ, Finch Jr WW, Januszka TF, Thomson Jr TA. Bond and
flexural tensile stresses near the far end of the FRP strip force transfer of composites material plates bonded to concrete. ACI
Struct J 1996;93(2):209–17.
as well as interference with interfacial behaviour.
[14] Chajes MJ, Januszka TF, Mertz DR, Thomson Jr TA, Finch Jr WW.
(2) The test results showed that Chen and Teng’s bond Shear strengthening of reinforced concrete beams using externally
strength model [1] is slightly conservative when the applied composite fabrics. ACI Struct J 1995;92(3):295–303.
FRP-to-concrete width ratios are at the two extremes of [15] Täljsten B. Defining anchor lengths of steel and CFRP plates bonded
0 and 1. When more reliable test results become to concrete. Int J Adhes Adhes 1997;17(4):319–27.
available, this small weakness can be easily removed. [16] Brosens K, van Gemert D. Anchoring stresses between concrete and
carbon fibre reinforced laminates. Non-metallic (FRP) reinforcement
(3) The test results highlighted the importance of careful
for concrete structures, Proceedings of the Third International Sympo-
specimen preparation as the results can be significantly sium, vol. 1. Sapporo, Japan: Japan Concrete Institute; 1997 p. 271–8.
affected. In the development of a standard bond test [17] Fukuzawa K, Numao T, Wu Z, Yoshizawa H, Mitsui M. Critical strain
procedure, measures should be included to minimise energy release rate of interface debonding between carbon fibre sheet
this effect, while in the development of design methods, and mortar. Non-metallic (FRP) reinforcement for concrete structures,
due allowance should be made for the expected quality Proceedings of the Third International Symposium, vol. 1. Sapporo,
Japan: Japan Concrete Institute; 1997 p. 295–302.
variations at sites. [18] Hiroyuki Y, Wu Z. Analysis of debonding fracture properties of CFS
strengthened member subject to tension. Non-metallic (FRP)
reinforcement for concrete structures, Proceedings of the Third
International Symposium, vol. 1. Sapporo, Japan: Japan Concrete
Acknowledgements
Institute; 1997 p. 287–94.
[19] Kobatake Y, Kimura K, Ktsumata H. A retrofitting method for
The authors are grateful for the financial support from reinforced concrete structures using carbon fibre. In: Nanni A, editor.
The Hong Kong Polytechnic University (G-V784) and from Fibre-reinforced-plastic (FRP) reinforcement for concrete structures:
the Research Grants Council of the Hong Kong SAR (PolyU properties and applications. The Netherlands: Elsevier Science; 1993,
5151/03E). p. 435–50.
[20] Maeda T, Asano Y, Ueda T, Kakuta Y. A study on bond mechanism of
carbon fiber sheet. Non-metallic (FRP) reinforcement for concrete
structures, Proceedings of Third International Symposium, Sapporo,
References Japan 1997 p. 287–95.
[21] Neubauer U, Rostásy FS. Design aspects of concrete structures
strengthened with externally bonded CFRP plates. Proceedings of
[1] Chen JF, Teng JG. Anchorage strength models for FRP and steel
Seventh International Conference on Structural Faults and Repairs,
plates bonded to concrete. J Struct Eng, ASCE 2001;127(7):784–91.
vol. 2. Edinburgh: ECS Publications; 1997 p. 109–18.
[2] Teng JG, Chen JF, Smith ST, Lam L. FRP-strengthened RC
[22] Swamy RN, Jones R, Charif A. Shear adhesion properties of epoxy
structures. Chichester: Wiley; 2002.
resin adhesives. Proceedings of International Symposium on
[3] Smith ST, Teng JG. FRP-strengthened RC beams-I: review of
Adhesion between Polymers and Concrete. London: Chapman &
debonding strength models. Eng Struct 2002;24(4):385–95.
[4] Smith ST, Teng JG. FRP-strengthened RC beams-II: assessment of Hall; 1986 p. 741–55.
debonding strength models. Eng Struct 2002;24(4):397–417. [23] van Gemert D. Force transfer in epoxy-bonded steel-concrete joints.
[5] Smith ST, Teng JG. Shear-bending interaction in debonding failures Int J Adhes Adhes 1980;1:67–72.
of FRP-plated RC beams. Adv Struct Eng 2003;6(3):183–99. [24] De Lorenzis L, Miller B, Nanni A. Bond of fiber-reinforced polymer
[6] Oehlers DJ, Moran JP. Premature failure of externally plated laminates to concrete. ACI Mater J 2001;98(3):256–64.
reinforced concrete beams. J Struct Div Am Soc Civil Engr 1990; [25] Ziraba YN, Baluch MH, Basunbul AM, Azad AK, Al-Sulaimani GJ,
116(4):978–95. Sharif IA. Combined experimental—numerical approach to charac-
[7] Teng JG, Smith ST, Yao J, Chen JF. Intermediate crack-induced terization of steel-glue-concrete interface. Mater Struct 1995;28:
debonding in RC beams and slabs. Construct Build Mater 2003;17(6– 518–25.
7):447–62. [26] Brosens K, van Gemert D. Plate end shear design for external CFRP
[8] Mohamed Ali MS, Oehlers DJ, Bradford MA. Shear peeling of steel laminates. Proceedings of FRAMCOS-3, Freiburg, Germany.:
plates bonded to the tension faces of RC beams. ASCE J Struct Eng Aedificatio Publishers; 1998 p. 1793–804.
2001;127(12):1453–60. [27] Holzenkämpfer O. Ingenieurmodelle des Verbundes geklebter
[9] Mohamed Ali MS, Oehlers DJ, Bradford MA. Interaction between Bewehrung für Betonbauteile, Dissertation, TU Braunschweig; 1994.
flexure and shear on the debonding of RC beams retrofitted with [28] Täljsten B. Plate bonding. Strengthening of existing concrete
compression face plates. Adv Struct Eng 2002;5(4):223–30. structures with epoxy bonded plates of steel or fibre reinforced
[10] Oehler DJ, Park SM, Mohamed Ali MS. A structural engineering plastics. Doctoral thesis. Sweden: Luleå university of Technology;
approach to adhesive bonding longitudinal plates to RC beams and 1994.
slabs. Compos: Part A 2003;34(12):887–97. [29] Täljsten B. Strengthening of concrete prisms using the plate-bonding
[11] Bizindavyi L, Neale KW. Experimental and theoretical investigation technique. Int J Fracture 1996;82:253–66.
of transfer lengths for composite laminates bonded to concrete. [30] Triantafillou TC, Plevris N. Strengthening of RC beams with expoxy-
Proceedings, Annual Conference of Canadian Society for Civil bonded fibre-composite materials. Mater Struct 1992;25:201–11.
J. Yao et al. / Composites: Part B 36 (2005) 99–113 113

[31] Yuan H, Wu ZS, Yoshizawa H. Theoretical solutions on interfacial [39] Chen JF, Teng JG. Shear capacity of FRP strengthened RC beams:
stress transfer of externally bonded steel/composite laminates. J Struct FRP debonding. Construct Build Mater 2003;17(1):27–41.
Mech Earthquake Eng, JSCE 2001;18(1):27–39. [40] Arduini M, Di Tommaso A, Nanni A. Brittle failure in FRP and sheet
[32] Wu ZS, Yuan H, Niu H. Stress transfer and fracture propagation in bonded beams. ACI Struct J 1997;94(4):363–70.
different kinds of adhesive joints. J Eng Mech, ASCE 2002;128(5): [41] Meier U, Kaiser H. Strengthening of structures with CFRP laminates.
562–73. Advanced Composite Materials in Civil Engineering Stuctures,
[33] Yuan H, Teng JG, Seracino R, Wu ZS, Yao J. Full-range behavior of Proceedings of the speciality Conference, ASCE, Las Vegas 1991 p.
FRP-to-concrete bonded joints. Eng Struct 2004;26(5):553–64. 224–32.
[34] Chen JF, Yang ZJ, Holt GD. FRP or steel plate-to-concrete bonded [42] Swamy RN, Mukhopadhyaya P. Debonding of carbon-fiber-
joints: effect of test methods on experimental bond strength. Steel reinforced polymer plate from concrete beams. Proceedings of the
Compos Struct 2001;1(2):231–44. Institution of Civil Engineers: Structures and Buildings 1999;134:
[35] Niu H, Wu Z. Interfacial debonding mechanism influenced by flexural 301–17.
cracks in FRP-strengthened beams. J Struct Eng, JSCE 2001;47A:
[43] BS 1881. Testing concrete—Part 116: Method for determination of
1277–88.
compressive strength of concrete cubes; Part 117: Method for
[36] Chaallal O, Nollet MJ, Perraton D. Strengthening of reinforced
determination of tensile splitting strength; and Part 121: Method for
concrete beams with externally bonded fibre-reinforced-plastic plates:
determination of static modulus of elasticity in compression. London:
design guidelines for shear and flexure. Canad J Civil Eng 1998;25(4):
British Standards Institute; 1983.
692–704.
[44] ASTM D3039/D3039M-95a. Standard test method for tensile proper-
[37] Khalifa A, Gold WJ, Nanni A, Aziz A. Contribution of externally
bonded FRP to shear capacity of RC flexural members. J Compos ties of polymer matrix composite materials; 1995.
Construct, ASCE 1998;2(4):195–203. [45] Yao J. Debonding in FP-strengthened RC Structure. Doctorial thesis.
[38] Horiguchi T, Saeki N. Effect of test methods and quality of concrete The Hong Kong Polytechnic University; 2004.
on bond strength of CFRP sheet. Non-metallic (FRP) reinforcement [46] Yuan H, Chen JF, Teng JG. Interfacial stresses between FRP plate and
for concrete structures, Proceedings of International Symposium, concrete in a peel test: an analytical solution. Proceedings (CD-ROM),
vol. 1. Sapporo, Japan: Japan Concrete Institute; 1997 p. 265–70. Structural faultsCrepair—2003,1–3 July, London 2003.

View publication stats

Potrebbero piacerti anche