You are on page 1of 12

Composite Structures 201 (2018) 664–675

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Bond strength model for near-surface mounted (NSM) FRP bonded joints: T
Effect of concrete edge distance
S.S. Zhang
School of Civil Engineering and Mechanics, Huazhong University of Science and Technology, Wuhan, China

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

Keywords: As one of the mainstream topics in the research field of infrastructure application of FRP composites, the near-
Fiber-reinforced polymer (FRP) surface mounted (NSM) fiber-reinforced polymer (FRP) strengthening technique has attracted an increasing
Concrete attention over the last decade and become an effective alternative to the externally bonded FRP strengthening
Near-surface mounted (NSM) method. The bond strength (i.e., maximum force that can be developed in the FRP reinforcement) of NSM FRP
Strengthening
bonded joints has been extensively studied by worldwide researchers, and several bond strength models have
Bond strength model
Edge distance
been established. All the existing bond strength models, however, are not able to consider the detrimental effect
of insufficient concrete edge distance (i.e., the distance between the groove and the nearer edge of the concrete).
To clarify such detrimental effect on the bond strength between NSM FRP strip and concrete, a numerical
parametric study, employing a three-dimensional meso-scale model developed by the author for NSM FRP
bonded joints, is conducted in the present study. Based on the results from the parametric study, reduction
factors are formulated to consider such detrimental effect and a bond strength model incorporating the reduction
factors are proposed. The accuracy of the proposed bond strength model is verified with a large test database
containing 86 specimens.

1. Introduction method e.g. [1,4], including: (1) higher bond strength between FRP and
concrete; (2) better protection of the FRP reinforcement; and (3) better
Most of the world’s civil structures have been constructed in re- durability and long-term performance e.g. [5,6]. Existing studies have
inforced concrete, and corrosion of steel reinforcement has been the shown that the NSM FRP strengthening method is very effective in
main cause of their deterioration. Simple demolition and reconstruction enhancing the flexural capacity and/or shear capacity of RC beams e.g.
will be time-consuming and will also lead to a great waste of resources. [7–9]. So far, square, round and rectangular FRP bars have been studied
Therefore, the development of effective and reliable methods for the as NSM FRP reinforcement in experimental tests, which indicated that
service life extension of existing structures has become extremely im- FRP strips (i.e., rectangular bars with a large aspect ratio) possesses a
portant. Against this background, extensive research has been con- much better bonding efficiency than FRP bars with other cross-sectional
ducted on the use of fiber-reinforced polymer (FRP) for the strength- shapes, as the former has a larger embedment depth as well as a larger
ening of existing deficient structures e.g. [1–3]. The near-surface perimeter-to-cross-sectional area ratio e.g. [10–13]. Carbon FRP (CFRP)
mounted (NSM) FRP strengthening technique, as a promising alter- is preferable in the application of NSM FRP strengthening method
native to the externally bonded (EB) FRP method, has attracted an in- compared with other types of FRP (e.g. glass FRP and aramid FRP). This
creasing worldwide attention over the last decade. In the NSM FRP is because CFRP has a higher tensile strength and stiffness, and thus a
strengthening method, grooves of desired dimensions are first cut in the smaller cross-sectional area of FRP is needed to achieve the same in-
concrete cover of a concrete member and then filled halfway with the crease demand in the load-carrying capacity. As a result, strips made of
adhesive after cleaning the groove surface. The cleaned FRP re- CFRP have been recommended for the application of NSM strength-
inforcement is then inserted into the groove gently until the desired ening technique e.g. [13–16].
position is reached. Finally, an additional amount of adhesive is applied So far, several bond strength (the maximum force that can be de-
to fill the groove completely, and then the surface of the adhesive is veloped in the FRP reinforcement in the test of bonded joints) models
levelled and polished. have been established through regression of the test results e.g. [17,18]
The NSM FRP strengthening method has been reported to own a or numerical simulation results e.g. [19,20]. Some of the existing bond
number of advantages compared with the EB FRP strengthening strength models e.g. [17,19,20] have identified the main parameters

E-mail address: shishun@hust.edu.cn.

https://doi.org/10.1016/j.compstruct.2018.06.089
Received 6 August 2017; Received in revised form 4 June 2018; Accepted 22 June 2018
Available online 25 June 2018
0263-8223/ © 2018 Elsevier Ltd. All rights reserved.
S.S. Zhang Composite Structures 201 (2018) 664–675

model previously proposed by the author [20] for bonded joints with
Stirrups
sufficient concrete edge distance, a new bond strength model is de-
veloped for bonded joints with insufficient concrete edge distance. The
accuracy of the proposed bond strength model is verified with the ex-
Longitudinal steel bar isting test results.

2. Finite element (FE) model


NSM FRP Groove
2.1. General

The 3-D meso-scale model, developed by the author [21] using the
Fig. 1. Flexural strengthening of RC beams using NSM FRP method. software package MSC.MARC [22], has been proved to be capable of
accurately predicting not only the failure process and bond strength of
NSM FRP bonded joints, but also the strain distribution in the FRP and
(e.g. concrete strength and the groove depth-to-width ratio) that can
the bond-slip curves between NSM FRP and concrete [21]. The failure
influence the bond strength of a bonded joint, and can well consider the
mechanism of the specimens studied in the present study is the same as
effect of bond length on the bond strength. However, all the existing
that modelled in Ref. [21], with the only difference being that concrete
models are based on the assumption that the concrete edge distance
edge distance will be varied to investigate its effect on the bond
(i.e., ae in Figs. 1 and 2, the distance between the groove and the nearer
strength in the present study.
edge of the concrete) is sufficiently large and thus there is no detri-
In practical applications of FRP strengthening of deficient struc-
mental effect of edge distance on the bond strength. In real applica-
tures, it is much desirable that the final failure is governed by the
tions, NSM CFRP strips usually need to be placed near the edge of the
failure in concrete (i.e., in the original structure) but not in the
member in order to accommodate more strips over the width of the
strengthening system itself (e.g. failure of the adhesive) to maximize the
member to meet a certain enhancement of the load-carrying capacity.
strengthening efficiency. Experimental studies have shown that if a
When the edge distance is relatively small, the bond behaviour between
relatively strong adhesive (i.e., the tensile strength of adhesive is much
the NSM CFRP strip and concrete will be detrimentally influenced and
higher than that of the concrete) is used and the groove surface is ap-
such detrimental effect needs to be considered for an accurate/safe
propriately cleaned, the final failure in the concrete can be ensured e.g.
design of the NSM FRP strengthening system. To clarify the detrimental
[20,23]. As a result, the accurate modelling of concrete behaviour is of
effect of edge distance on the bond strength of NSM FRP strips, the
critical important in the numerical simulation of NSM bonded joints. In
three-dimensional (3-D) meso-scale finite element (FE) model devel-
the present study, the modelling of concrete (in particular the tensile
oped by the author [21] for bonded joints with NSM FRP strips (re-
and shear behaviour of the cracked concrete) was carefully treated
ferred to as NSM bonded joints hereafter for simplicity), is adopted in
following Ref. [21].
the present paper to conduct a numerical parametric study. Based on
the results from the parametric study, reduction factors are proposed to
account for the effect of concrete edge distance on the bond strength. By 2.2. FE model
introducing the proposed reduction factors into the bond strength
The schematic of the NSM bonded joints modelled in the present

25mm 450mm 75mm


ae
P
50mm

50mm

Fig. 2. Schematic of the NSM bonded joint in the parametric study.

Table 1
Design of parametric study of concrete edge distance.
Specimens fc tf hf Ef hf / t f ae
(MPa) (mm) (mm) (GPa) (mm)

Case-f20-h10-ae0, 15, 30, 45, 60 20 2 10 150 5 0, 15, 30, 45 and 60


Case-f20-h20-ae0, 25, 50, 75, 100 20 2 20 150 10 0, 25, 50, 75, and 100
Case-f20-h30-ae0, 35, 70, 105, 140 20 2 30 150 15 0, 35, 70, 105 and 140
Case-f30-h10-ae0, 15, 30, 45, 60 30 2 10 150 5 0, 15, 30, 45 and 60
Case-f30-h20-ae0, 25, 50, 75, 100 30 2 20 150 10 0, 25, 50, 75, and 100
Case-f30-h30-ae0, 35, 70, 105, 140 30 2 30 150 15 0, 35, 70, 105 and 140
Case-f40-h10-ae0, 15, 30, 45, 60 40 2 10 150 5 0, 15, 30, 45 and 60
Case-f40-h20-ae0, 25, 50, 75, 100 40 2 20 150 10 0, 25, 50, 75, and 100
Case-f40-h30-ae0, 35, 70, 105, 140 40 2 30 150 15 0, 35, 70, 105 and 140

665
S.S. Zhang Composite Structures 201 (2018) 664–675

study is shown in Fig. 2. The height and length of the concrete block are
150 mm and 550 mm respectively. The bond length of the NSM FRP
strips is selected to be 450 mm, which is larger than the effective bond
length predicted according to Seracino et al. [23] and Zhang et al. [20],
in order to eliminate its effect on the bond strength. An un-bonded
length of 75 mm is left near the loaded end to avoid local shear failure
at loaded end, while an un-bonded length of 25 mm is left near the free
end of the FRP strip. The concrete edge distance in the specimen is
varied according to the height of the FRP strip, as shown in Table 1. By
taking advantage of symmetry, only half of the specimen is included in
the FE model (Fig. 2). In addition, the bottom layer of concrete block
with a height of 50 mm was not included in the FE model to save the
computing time. Such simplification has only marginal effect on the
accuracy of modelling but can largely save the computational effort
[21]. Following the work presented in [21], the meso-scale FE approach
was adopted in the present study. In the meso-scale FE modelling, in-
terfacial elements were not needed anymore, instead, all the materials (a) Case-f20-h30-ae35
(i.e., the concrete block, the FRP strip and the adhesive layer) were
modelled using first-order eight-node solid elements with full Gaussian
integration scheme. With very fine elements (in the order of 1 mm)
being employed and proper constitutive laws adopted in the FE model,
the failure process of the bonded joint can be well reflected by the
failure of the material [21].

2.3. Boundary conditions

Proper boundary conditions are applied, including (see Fig. 2): (1)
the displacement of bottom surface of the concrete block in the vertical
direction is restrained; (2) the displacement of the plane of symmetry in
the width direction of the specimen is prevented; (3) the lower portion
(with a height of 50 mm) of the vertical surface of the concrete block at
the loaded end is restrained in the length direction of the specimen; and
(4) the vertical displacement on the top surface of the concrete block at (b) Case-f20-h30-ae70
the far end (i.e., far from the loaded end) is restrained.

2.4. Material models

The orthogonal fixed smeared crack model, available in MSC.MARC


[22], is used to model the cracked concrete. At each integration point, a
maximum of three cracks could form, and their directions can only be
orthogonal to each other. To eliminate/mitigate the problem of mesh
sensitivity, the crack band model proposed by Bazant and Oh [24] is
adopted in the FE model, and the tensile fracture energy of cracked
concrete given by CEB-FIP [25] is adopted. For the modelling of
cracked concrete: (1) the initiation of cracking is detected by the
maximum tensile stress criterion (as expressed in Eq. (1)); (2) Teng
et al. [21] showed that the tension-softening curves of cracked concrete
had nearly no influence on the modelling results if the crack band
model was adopted. Therefore, the widely adopted exponential soft-
ening curve (as expressed in Eq. (2)) proposed by Hordijk [26] is used
(c) Case-f20-h30-ae105
in the present study to model the tension-softening behaviour of Fig. 3. Typical failure modes.
cracked concrete, following Teng et al. [21]; and (3) the model (as
expressed in Eq. (3)) proposed by Okamura and Maekawa [27], which
f (I1, J2 , θ) = 2 3 J2 cosθ + I1−3ft = 0 (1)
was found to be the best model for the modelling of shear behaviour of
cracked concrete [21], is used to describe the shear stress-slip beha- where I1 and J2 are the first invariant of stresses and the second in-
viour of cracked concrete. For the modelling of compression-dominated 2

behaviour of concrete: (1) the yield law (as expressed in Eq. (4)) pro- variant of deviatoric stresses respectively; ft = 1.4 ( )
fc − 8 3
10
(according to
posed by Buyukozturk [28] (with the associated flow rule) is adopted; [25]) is the tensile strength of concrete; fc is the compressive strength of
and (2) the uniaxial stress-strain strain behaviour is defined by Elwi and concrete; and θ is the angle of similarity.
Murray’s [29] model (as expressed in Eq. (5)). Both FRP and adhesive
3
are modelled as isotropic elastic materials. The simplification in the σt w ⎞ ⎤ (−6.93 ww ) w (−6.93)
=⎡ ⎛
⎢1 + 3.0 w0 ⎥ e
⎜ ⎟ 0 −10 e
modelling of FRP and adhesive was found to have nearly no effect in ft ⎣ ⎝ ⎠ ⎦ w 0 (2)
such simulation [21]. The above constitutive laws/models have been
proven effective for the meso-scale modelling of cracked concrete, and where w is the crack width (i.e. crack opening displacement), and w0 is
more details can be found in Ref. [21,30,31]. the crack width at the complete release of stress or fracture energy.

666
S.S. Zhang Composite Structures 201 (2018) 664–675

Table 2
Threshold values of concrete edge distance.
Specimens Bond strength with different ae (kN) aet (mm)

0 mm 15 mm 30 mm 45 mm 60 mm FE analysis Pre. of Eq. (7) Pre./FE analysis

Case-f20-h10-ae0, 15, 30, 45, 60 0 16.01 24.05 27.08 27.40 45.5 46.2 1.016
Case-f30-h10-ae0, 15, 30, 45, 60 0 18.88 27.85 30.59 31.15 50.8 51.2 1.008
Case-f40-h10-ae0, 15, 30, 45, 60 0 20.94 30.89 33.79 34.49 54.6 55.5 1.018

Specimens Bond strength with different ae (kN) aet (mm)

0 mm 25 mm 50 mm 75 mm 100 mm

Case-f20-h20-ae0, 25, 50, 75, 100 0 32.68 47.27 52.13 52.78 75.7 75.4 0.996
Case-f30-h20-ae0, 25, 50, 75, 100 0 37.88 54.34 60.37 61.95 84.9 82.6 0.974
Case-f40-h20-ae0, 25, 50, 75, 100 0 40.37 60.88 66.88 68.08 90.5 88.6 0.979

Specimens Bond strength with different ae (kN) aet (mm)

0 mm 35 mm 70 mm 105 mm 140 mm

Case-f20-h30-ae0, 35, 70, 105, 140 0 54.80 76.50 81.59 82.25 101.5 103.5 1.020
Case-f30-h30-ae0, 35, 70, 105, 140 0 59.91 85.88 92.49 93.65 110.3 112.5 1.020
Case-f40-h30-ae0, 35, 70, 105, 140 0 65.08 93.00 101.21 102.97 116.9 119.8 1.024

Statistical characteristics Average = 1.006


STD = 0.018
CoV = 0.018

bcr2 εntcr FRP strip height is varied to obtain different heights of the groove (thus
Gcr = 3.8fc1/3
w 2 + Δ2 (3) different aspect ratio of the groove). The elastic modulus of the CFRP
strip in the longitudinal direction is chosen to be 150 GPa, which is a
where Gcr is the shear modulus of cracked concrete; Δ = bcr εntcr is the
cr typical value for pultruded CFRP strips.
crack slip; w = bcr εnn is the crack width; bcr is the crack band width; εntcr
cr In the present study, the adhesive is assumed to be linear-elastic
and εnn are the shear strain and the normal strain of the crack respec-
material, so the slip caused by the deformation in the adhesive under a
tively.
certain local bond stress is only dependent on the thickness and the
3J2 + 3 χσ ∗I1 + ξI12 = (σ ∗)2 (4) shear modulus of the adhesive layer. It has been reported by Zhang
et al. [19] that for commonly used adhesives in NSM CFRP strength-
where ξ , χ and σ∗
are material constants and equal to 0.2, 3 and
f ening technique (with an elastic modulus not larger than 5 GPa), the
σ ∧ /3 respectively; σ ∧ is taken to be c according to Ref. [28].
3 adhesive thickness, which varies in the practical range (e.g. around
E0 ε 1–4 mm), has only marginal effect on the slip between the NSM FRP
σc =
ε 2 ε 3 and concrete, as most slip is contributed by the thin-layer of cracked
1+ R+( E0
Es
−2 )( )−(2R−1) ( )
ε
ε0 ε0
+R ()
ε0 (5) concrete near the concrete-to-adhesive interface. In the present study,
E0/ Es (σ0/ σu−1) 1 the elastic modulus and the thickness of adhesive are taken to be 3 GPa
R= − and 2 mm respectively (both are typical values in practice).It has been
(εu/ ε0−1)2 (εu/ ε0) (6)
shown in existing studies that the bond strength of NSM bonded joint
where σ0 and ε0 are the maximum stress and the corresponding strain increases with the bond length, until the bond length is larger than the
respectively; εu and σu are the ultimate strain and the corresponding so-called effective bond length (Le ) [20,23,32]. To eliminate the effect
stress respectively, E0 is the initial elastic modulus of concrete; and Es is of bond length on the bond strength, a sufficiently large bond length of
the secant modulus at the maximum stress: Es = σ0/ ε0 . 450 mm was selected according to Zhang et al. [20] and Seracino et al.
[23].
3. Parametric study
3.2. Numerical specimens
3.1. Analysis of parameters
Based on the above analysis, a total of 45 specimens are proposed
Preliminary analysis indicates that the strength of concrete ( fc ) is and modelled in the parametric study, as listed in Table 1. The studied
obviously one of the most important parameters in determining the parameters cover the concrete strength, the groove height-to-width
bond strength, because as mentioned earlier the final failure mode of ratio, and the concrete edge distance. Three values of the cylinder
interest in the present study is the failure in the thin-layer of concrete compressive strength of concrete are used respectively: 20 MPa, 30 MPa
near the epoxy-to-concrete interface. It can be expected that, the higher and 40 MPa; three groove height-to-width ratios are considered re-
the concrete strength, the higher the fracture energy of concrete and spectively: 2.33, 4.00 and 5.67. The groove height-to-width ratios are
thus larger bond strength. achieved by changing the height of the grooves with the same width
Usually speaking, when the groove is deeper, the embedment depth being used for all numerical specimens. The groove width is 6 mm,
of the FRP strip is larger and thus the confinement from the surrounding which is the summation of the thickness of CFRP strip (2 mm) and the
concrete to the FRP strip is higher. As a result, the aspect ratio of the thickness of the adhesive layer (2 mm on each side of the CFRP strip).
groove (i.e., the groove height hg to groove width wg ratio) should also The heights of the grooves are 14 mm, 24 mm and 34 mm respectively
have significantly effect on the bond behaviour and thus should be for CFRP strips with a height of 10 mm, 20 mm and 30 mm, as a 2 mm-
covered in the parametric study. In the present study, the FRP strip thick adhesive layer exists on the top as well as the bottom of the CFRP
thickness is 2 mm (a typical value for commercial CFRP strips), and the strip. The maximum value of 34 mm is chosen based on the

667
S.S. Zhang Composite Structures 201 (2018) 664–675

90 each of the nine combinations of concrete strength and groove height-


to-width ratio, five values of concrete edge distance are chosen based
80
on the height of the FRP strip: 0 mm, 15 mm, 30 mm, 45 mm and 60 mm
70 for bond joints with NSM FRP strips with a height of 10 mm; 0 mm,
25 mm, 50 mm, 75 mm and 100 mm for bond joints with NSM FRP
60 strips with a height of 20 mm; and 0 mm, 35 mm, 70 mm, 105 mm and
Load (kN)

50 140 mm for bond joints with NSM FRP strips with a height of 30 mm.
The bond strength of specimens with a 0 mm concrete edge distance is
40 assumed to be 0 kN. As shown in Table 1, for ease of reference, each
h_g=14 numerical case is given a name, which starts with a word “Case”, fol-
30 99% load (h_g=14)
h_g=24 lowed by a letter “f” and a Roman numeral to represent the concrete
20 99% load (h_g=24) strength ( fc ), a letter “h” and a Roman numeral to represent the height
h_g=34 of the CFRP strip (hf ), and two letters “ae” and a Roman numeral to
10 99% load (h_g=34) represent the concrete edge distance (ae ). For instance, Case-f20-h10-
0 ae30 refers to the specimen which has a concrete strength of 20 MPa, a
0 50 100 150 FRP strip height of 10 mm and a concrete edge distance of 20 mm
Edge distance (mm)
4. Results of parametric study
a) fc=20 MPa
4.1. Failure modes
100
90 Specimens Case-f20-h30-ae35, Case-f20-h30-ae70 and Case-f20-
h30-ae105 are selected as the examples to illustrate the typical failure
80 modes of the simulated specimens (see Fig. 3). In Fig. 3, the distribution
70 of maximum principal cracking strains in the concrete of the three
specimens is plotted. It can be seen from Fig. 3a (Case-f20-h30-ae35)
Load (kN)

60 that when the concrete edge distance is small (i.e., 35 mm), the con-
50 crete cracking region reaches the edge of the specimen and several large
inclined cracks appear on the side surface of the specimen, indicating
40 h_g=14
99% load (h_g=14) that part of the concrete is peeled off from the specimen at the final
30 h_g=24 stage of the loading process. When the concrete edge distance is in-
99% load (h_g=24) creased to 70 mm (i.e., Case-f20-h30-ae70, see Fig. 3b), the penetration
20
h_g=34 of the cracking strain in the concrete is mitigated, but the cracking
10 99% load (h_g=34) region can still reach the edge of the specimen and one large inclined
crack appears on the side surface of the specimen, indicating the edge
0
distance is still insufficient to fully develop the bond strength of the
0 50 100 150
bonded joints. When the concrete edge distance is further increased to
Edge distance (mm) 105 mm (i.e., Case-f20-h30-ae105, see Fig. 3c), the concrete cracking
b) fc =30 MPa region do not reach the edge of the specimen and no inclined crack
appears on the side surface of the specimen, indicating that the edge
120 distance of 105 mm is sufficiently large to eliminate the detrimental
effect of edge distance on the bond strength of the specimen. Based on
100 the above observation, it is not unreasonable to expect that when the
concrete edge distance is relatively small, the crack propagation in the
concrete can be significantly influenced by the concrete edge distance;
80 when the concrete edge distance is larger than a certain value, its effect
Load (kN)

on the crack propagation becomes insignificant.


60
h_g=14 4.2. Bond strength
99% load (h_g=14)
40
h_g=24 The bond strength of the whole specimen was obtained by multi-
99% load (h_g=24) plying the ultimate load directly obtained from the FE modelling by 2,
20 h_g=34 as only half of the specimen was included in the FE modelling by taking
99% load (h_g=34) advantage of symmetry. The bond strengths obtained from the para-
0 metric study are listed in Table 2, and the bond strength versus edge
0 50 100 150 distance curves are plotted in Fig. 4. From Fig. 4, it can be seen that: (1)
with other parameters are kept the same, a larger height of FRP strip or
Edge distance (mm)
a larger concrete strength leads to a larger bond strength of the spe-
c) fc =40 MPa cimen; (2) the bond strength increases with the value of edge distance
but the increasing rate gradually decreases with the value of edge dis-
Fig. 4. Load versus edge distance curves.
tance. When the value of edge distance is larger than a certain value,
further increase in the edge distance only gives marginal increase in the
consideration that the concrete cover thickness is not much larger than bond strength. It can be seen from Table 2 that, for all studied series of
35 mm in most practical cases, while the minimum value of 14 mm numerical specimens, the bond strength of the specimens with the
corresponds to FRP strips with a height-to-thickness ratio of 5, which is largest edge distance studied in that series is very close to that with the
the lower bound for CFRP strips suggested by Zhang et al. [19]. For second largest edge distance, indicating that the value of bond strength

668
S.S. Zhang Composite Structures 201 (2018) 664–675

140.00 6
Threshed value of edge distance y = 0.0723x + 2.7703
120.00 y = 5.620x0.866 5

Coefficient A in Eq. (7)


100.00
y = 5.024x0.881 4
80.00
y = 4.174x0.907 3
(mm)

60.00
2
40.00
fc=20 MPa
20.00 fc=30 MPa 1
fc=40 MPa
0.00 0
0 10 20 30 40 0 10 20 30 40 50
Groove height (mm) Concrete strength (MPa)

edge distance curves (b) Coefficient A in Eq. (7)

1 150
0.9
y = 1.0211x - 1.1821
0.8 R² = 0.9962
Prediction of Eq. (7) (mm)
Coefficient B in Eq. (7)

0.7
y = -0.00205x + 0.94617 100
0.6
0.5
0.4
0.3 50
0.2
0.1
0 0
0 10 20 30 40 50
0 50 100 150
Concrete strength (MPa)
FE result (mm)
(c) Power B in Eq. (7) (d) Verification of Eq. (7)
Fig. 5. The threshold value of edge distance.

has been converged with respect to the value of edge distance. power functions are also given. It can be seen from Fig. 5a that the
relationship between the threshold value of edge distance and the
4.3. Threshold value of concrete edge distance groove height can be described by the following power function:

In the present study, the minimum required value of edge distance aet = A × hgB (7)
for the full development of bond strength of the NSM bonded joint is
termed as the threshold value of edge distance (i.e., aet in Table 2). In where the coefficient A and the power B are related to the concrete
the present study, the threshold values of edge distance are obtained strength fc ; hg is the groove height. The relationship between the
using the following steps: (1) for each of the nine series of numerical coefficient A and the concrete strength is shown in Fig. 5b, and the
specimens, the best-fit four-order polynomial function to describe the relationship between the power B and the concrete strength is shown in
relationship between the bond strength and the value of edge distance Fig. 5c. Based on the best-fit curves shown in Fig. 5b and c, two liner
is formulated first; (2) the obtained best-fit four-order polynomial functions are respectively proposed for the coefficient A and the power
function is then used to calculate the value of edge distance which B:
corresponds to 99% of the bond strength obtained with the largest edge
A = 0.072fc + 2.77 (8)
distance in that series, and this edge distance value is treated as the
threshold value of edge distance in the present study. The threshold
values of edge distance obtained using this method are listed in Table 2. B = −0.0021fc + 0.95 (9)
It can be seen from Table 2 that a larger groove height (i.e., a larger FRP
height) or a larger concrete strength leads to a larger threshold value of The predictions of the threshold value of edge distance from Eq. (7)
edge distance when other parameters are kept the same, which is not are compared with the results from the FE analysis (see Table 2) in
difficult to understand: a deeper groove or a higher strength of concrete Fig. 5d, from which good agreement can be observed. It can be seen
usually incurs a larger motivated stress zone around the groove. from Table 2 that the ratios between predictions of Eq. (7) and FE
To formulate the calculation equation for the threshold value of analysis have an average value of 1.006, a standard deviation (STD) of
edge distance, the relationship between the threshold value of edge 0.018, and a coefficient of variation (CoV) of 0.018.
distance and the groove height is plotted in Fig. 5a, in which the best-fit

669
S.S. Zhang Composite Structures 201 (2018) 664–675

1 1.66
Le =
η (14)
Normalized bond strength

2
0.8 τmax Cfailure
η2 =
2Gf Ef Af (15)
h_g=14 mm, fc=20 MPa τmax = 1.15γ 0.138f c0.613
0.6 (16)
h_g=24 mm, fc=20 MPa
h_g=34 mm, fc=20 MPa Lb ⎛ L
h_g=14 mm, fc=30 MPa βL = ⎜2.08−1.08 b ⎞ ⎟

0.4 Le ⎝ Le ⎠ (17)
h_g=24 mm, fc=30 MPa
2
h_g=34 mm, fc=30 MPa where Af (mm ) is the cross sectional area of a single FRP strip; Cfailure
0.2 h_g=14 mm, fc=40 MPa (mm) is the cross-sectional contour of the failure surface which is taken
h_g=24 mm, fc=40 MPa to be the sum of the three side lengths of the groove; Ef (MPa) is the
h_g=34 mm, fc=40 MPa elastic modulus of the FRP strip; fc (MPa) is the compressive strength of
0 concrete cylinder; Gf (N/mm) is the interfacial fracture energy; Lb (mm)
0 0.2 0.4 0.6 0.8 1 is the bond length; Le (mm) is the effective bond length; Pt (N) is the full
tensile capacity of a single FRP strip; τmax (MPa) is the maximum bond
Normalized edge distance stress, γ is the groove height-to-width ratio, βL the a reduction factor to
(a) Normalized bond strength versus normalized edge distance account for the effect of insufficient bond lengths.

1 5.2. Reduction factor

To formulate βe , for each studied series in the parametric study, the


0.8 threshold value of edge distance and the corresponding bond strength
are treated as the references, with respect to which other edge distances
Reduction factor

and bond strengths are normalized respectively. The normalized bond


0.6 strength (i.e., the bond strength corresponding to a certain concrete
edge distance divided by the bond strength corresponding to the
threshold value of concrete edge distance aet ) versus the normalized
0.4 edge distance (i.e., concrete edge distance divided by the threshold
value aet ) curves for each studied series just represent the reduction
factor βe , as shown in Fig. 6a. It can be seen from Fig. 6a that although
0.2 the curves do not coincide with each other, the scatter is very small. By
Accurate reduction factor (Eq. 18) regressing all the points on Fig. 6a, the following equation is proposed
Simplified reduction factor (Eq. 19) for the reduction factor βe :
0 a
3
a
2
a
0 0.2 0.4 0.6 0.8 1 βe = 0.83 ⎛ e ⎞ −2.57 ⎛ e ⎞ + 2.74 ⎛ e ⎞ ⩽ 1
⎜ ⎟ ⎜ ⎟ ⎜ ⎟

⎝ aet ⎠ ⎝ aet ⎠ ⎝ aet ⎠ (18)


ae/aet
(b) Comparison of the two reduction factor models Although Eq. (18) is able to give accurate predictions of the re-
duction factor as it was obtained by directly regressing the numerical
Fig. 6. Proposed reduction factor models: (a) Normalized bond strength versus results, its form is a little complex and thus not preferable for the use by
normalized edge distance; (b) Comparison of the two reduction factor models. engineers. Therefore, a simplified reduction factor, as expressed in Eq.
(19), is also proposed and assessed in the present study.
5. Bond strength model and reduction factor
ae ⎛ ae ⎞
βe = ⎜ 2− ⩽1 ⎟

aet ⎝ aet ⎠ (19)


5.1. Bond strength model
A general comparison of the reduction factors predicted by Eqs. (18)
When the edge distance is smaller than the threshold value aet , the and (19) is shown in Fig. 6b, from which it can be seen that the sim-
bond strength of the NSM bonded joint can be detrimentally influenced. plified reduction factor given by Eq. (19) is very close to and con-
To consider such detrimental effect, a reduction factor βe , which is sistently smaller than the accurate reduction factor given by Eq. (18).
defined as the ratio between the bond strength with insufficient edge
distance and that with sufficient bond strength, should be introduced 6. Verification of the proposed model
into the bond strength model as:
Pu∗ = βe Pu (10) 6.1. Test database

where Pu∗ is the bond strength of NSM bonded joint with insufficient To verify the accuracy of the proposed bond strength model (Eq.
edge distance; Pu is the bond strength of NSM bonded joint with suffi- (10)) with the reduction factors (Eqs. (18) and (19)), a large database
cient edge distance. In the present study, Pu can be obtained using the containing 86 tests on NSM FRP strip bonded joints collected from 11
equation proposed by Zhang et al.[20]: experimental studies [17,23,33–41] is established in the present study.
The selection criteria are as follows: (1) debonding failure of the
Pu = 2Gf Ef Af Cfailure ⩽ Pt when Lb ⩾ Le (11) bonded joints is caused the cohesive failure (i.e. cracking) in the con-
crete adjacent to the adhesive layer; specimens failed duo to adhesive
Pu = βL 2Gf Ef Af Cfailure ⩽ Pt when Lb < Le (12) failure (usually a weak adhesive is used) or interfacial debonding
failure (usually caused by insufficient clean of the groove surface or
Gf = 0.40γ 0.422f c0.619 (13) FRP strip) are not included; (2) the height-to-thickness ratio of the FRP

670
Table 3
Bond strength comparison between predictions and test results.
S.S. Zhang

Source Specimen Lb tf hf wg hg Ef fc ae aet βe by Eq. (18) βe by Eq. (19) Test Without βe With βe by Eq. (18) With βe by Eq. (19)
(mm) (mm) (mm) (mm) (mm) (MPa) (MPa) (mm) (mm) (mm) (mm) (kN)
Pre. (kN) Pre./Test Pre.(kN) Pre./Test Pre.(kN) Pre./Test

Li et al. [33] CS2-30 30 4 16 8 20 131,000 23.2 71.0 66.1 1.00 1.00 14.82 15.0 1.015 15.0 1.015 15.0 1.015
CS2-100 100 4 16 8 20 131,000 23.2 71.0 66.1 1.00 1.00 36.28 41.3 1.139 41.3 1.139 41.3 1.139
CS2-150 150 4 16 8 20 131,000 23.2 71.0 66.1 1.00 1.00 46.07 52.5 1.140 52.5 1.140 52.5 1.140
Shield et al. [34] DP460NS 152 2 16 6.4 19 131,000 64.5 98.3 81.6 1.00 1.00 73.40 55.7 0.759 55.7 0.759 55.7 0.759
3M-90-1 305 2 16 6.4 19 131,000 60 225.8 80.2 1.00 1.00 60.50 54.5 0.901 54.5 0.901 54.5 0.901
3M-90-2 305 2 16 6.4 19 131,000 60 225.8 80.2 1.00 1.00 62.50 54.5 0.872 54.5 0.872 54.5 0.872
3M-90-3V 305 2 16 6.4 19 131,000 60 225.8 80.2 1.00 1.00 60.50 54.5 0.901 54.5 0.901 54.5 0.901
Seracino et al. [17] 30-MPa-100-10 100 1.2 10 3.2 11 161,800 30 148.4 41.4 1.00 1.00 22.60 21.8 0.963 21.8 0.963 21.8 0.963
30-MPa-100-10 100 1.22 10.02 3.22 11 161,800 30 148.4 41.4 1.00 1.00 20.40 21.9 1.074 21.9 1.074 21.9 1.074
30-MPa-150-10 150 1.23 10.33 3.23 11 161,800 30 148.4 42.5 1.00 1.00 23.20 24.4 1.050 24.4 1.050 24.4 1.050
30-MPa-200-10 200 1.22 10.48 3.22 11 161,800 30 148.4 43.0 1.00 1.00 27.90 24.7 0.884 24.7 0.884 24.7 0.884
30-MPa-250-10 250 1.22 10.29 3.22 11 161,800 30 148.4 42.3 1.00 1.00 26.60 24.2 0.909 24.2 0.909 24.2 0.909
30-MPa-300-10 300 1.22 10.38 3.22 11 161,800 30 148.4 42.6 1.00 1.00 26.00 24.4 0.939 24.4 0.939 24.4 0.939
30-MPa-350-10 350 1.22 10.35 3.22 11 161,800 30 148.4 42.5 1.00 1.00 23.00 24.3 1.058 24.3 1.058 24.3 1.058
42-MPa-200-10 200 1.27 10.29 3.27 11 161,800 41.8 148.4 46.7 1.00 1.00 30.60 27.3 0.891 27.3 0.891 27.3 0.891
30-MPa-100-20 100 1.2 20 3.2 21 162,300 30 148.4 73.4 1.00 1.00 51.40 45.2 0.878 45.2 0.878 45.2 0.878
30-MPa-200-20 200 1.2 20 3.2 21 162,300 30 148.4 73.4 1.00 1.00 57.80 50.6 0.875 50.6 0.875 50.6 0.875
30-MPa-300-20 300 1.2 20 3.2 21 162,300 30 148.4 73.4 1.00 1.00 66.70 50.6 0.758 50.6 0.758 50.6 0.758
65-MPa-200-10 200 2.88 10.08 4.88 11 144,600 64.8 147.6 52.7 1.00 1.00 45.00 41.2 0.916 41.2 0.916 41.2 0.916
65-MPa-200-20 200 2.97 19.77 4.97 21 162,300 64.8 147.5 87.8 1.00 1.00 108.80 92.6 0.851 92.6 0.851 92.6 0.851
53-MPa-200-10 200 1.24 10.23 3.24 11 161,800 52.8 148.4 50.0 1.00 1.00 31.90 28.8 0.903 28.8 0.903 28.8 0.903
53-MPa-200-10 200 1.3 10.43 3.3 11 161,800 53 148.4 50.8 1.00 1.00 34.00 30.1 0.885 30.1 0.885 30.1 0.885
53-MPa-100-20 100 1.25 20.23 3.25 21 162,300 53 148.4 85.4 1.00 1.00 63.80 59.3 0.930 59.3 0.930 59.3 0.930

671
33-MPa-200-15 200 1.26 15.65 3.26 17 162,050 33.4 148.4 61.5 1.00 1.00 47.50 40.4 0.850 40.4 0.850 40.4 0.850
33-MPa-300-15 300 1.26 15.31 3.26 16 162,050 33.4 148.4 60.4 1.00 1.00 51.60 39.4 0.764 39.4 0.764 39.4 0.764
65-MPa-200-10 200 2.9 9.95 4.9 11 144,600 64.8 147.6 52.2 1.00 1.00 45.10 40.8 0.904 40.8 0.904 40.8 0.904
33-MPa-200-20 200 1.2 20 3.2 21 162,300 33.4 148.4 75.4 1.00 1.00 60.70 52.3 0.861 52.3 0.861 52.3 0.861
Seracino et al. [23] 12x3 350 2.76 12.37 4.76 13 146,348 36.7 147.6 52.1 1.00 1.00 59.20 42.6 0.719 42.6 0.719 42.6 0.719
12x4 350 4.24 12.47 6.24 13 134,467 36.7 146.9 52.4 1.00 1.00 54.10 49.3 0.911 49.3 0.911 49.3 0.911
12x6 350 5.73 12.35 7.73 13 130,489 36.7 146.1 52.0 1.00 1.00 47.60 54.6 1.147 54.6 1.147 54.6 1.147
24x4 350 4.33 24.06 6.33 25 141,434 36.7 146.8 90.1 1.00 1.00 130.00 105.2 0.809 105.2 0.809 105.2 0.809
12x12 350 12 12 14 13 131,566 36.7 143.0 50.8 1.00 1.00 85.90 73.9 0.861 73.9 0.861 73.9 0.861
30x7 350 7.3 30.6 9.3 32 134,562 36.7 145.4 110.3 1.00 1.00 165.30 164.8 0.997 164.8 0.997 164.8 0.997
26x20 350 20.6 25.26 22.6 26 129,837 36.7 138.7 93.8 1.00 1.00 199.40 197.8 0.992 197.8 0.992 197.8 0.992
Rashid et al. [35] C150NSMa 350 1.2 20 3 21 161,000 35.5 150.0 76.5 1.00 1.00 63.00 53.7 0.852 53.7 0.852 53.7 0.852
C10NSMa 350 1.2 20 3 21 161,000 35.5 10.0 76.5 0.32 0.24 22.20 53.7 2.418 17.0 0.764 13.1 0.590
C20NSMa 350 1.2 20 3 21 161,000 35.5 20.0 76.5 0.56 0.45 33.80 53.7 1.588 29.8 0.882 24.4 0.721
C30NSMa 350 1.2 20 3 21 161,000 35.5 30.0 76.5 0.73 0.63 40.60 53.7 1.322 39.1 0.964 33.8 0.833
C40NSMa 350 1.2 20 3 21 161,000 35.5 40.0 76.5 0.85 0.77 50.20 53.7 1.069 45.5 0.907 41.4 0.825
C60NSMa 350 1.2 20 3 21 161,000 35.5 60.0 76.5 0.97 0.95 59.20 53.7 0.907 52.0 0.878 51.2 0.864
C85NSMa 350 1.2 20 3 21 161,000 35.5 85.0 76.5 1.00 1.00 75.70 53.7 0.709 53.7 0.709 53.7 0.709
C150NSMb 350 2.4 40 3 41 173,000 35.5 150.0 137.5 1.00 1.00 205.10 176.1 0.859 176.1 0.859 176.1 0.859
C50NSMb 350 2.4 40 3 41 173,000 35.5 50.0 137.5 0.70 0.60 119.90 176.1 1.469 122.7 1.023 104.8 0.874
C75NSMb 350 2.4 40 3 41 173,000 35.5 75.0 137.5 0.86 0.79 170.90 176.1 1.031 152.3 0.891 139.7 0.818
C100NSMb 350 2.4 40 3 41 173,000 35.5 100.0 137.5 0.95 0.93 194.90 176.1 0.904 167.8 0.861 163.0 0.836
(continued on next page)
Composite Structures 201 (2018) 664–675
Table 3 (continued)
S.S. Zhang

Source Specimen Lb tf hf wg hg Ef fc ae aet βe by Eq. (18) βe by Eq. (19) Test Without βe With βe by Eq. (18) With βe by Eq. (19)
(mm) (mm) (mm) (mm) (mm) (MPa) (MPa) (mm) (mm) (mm) (mm) (kN)
Pre. (kN) Pre./Test Pre.(kN) Pre./Test Pre.(kN) Pre./Test

Oehlers et al. [36] TS1-3.6-C0 350 3.6 10 5.6 11 150,000 38.8 147.2 44.7 1.00 1.00 40.00 39.1 0.977 39.1 0.977 39.1 0.977
TS1-3.6-C0R 350 3.6 10 5.6 11 160,000 38.8 147.2 44.7 1.00 1.00 39.20 40.3 1.029 40.3 1.029 40.3 1.029
TS1-3.6-C10 350 3.6 10 5.6 21 165,000 38.8 147.2 78.3 1.00 1.00 61.80 61.7 0.998 61.7 0.998 61.7 0.998
TS2-6.0-C0 350 6 10 8 11 166,000 38.8 146.0 44.7 1.00 1.00 54.80 51.3 0.936 51.3 0.936 51.3 0.936
TS2-6.0-C10 350 6 10 8 21 165,000 38.8 146.0 78.3 1.00 1.00 86.10 75.7 0.879 75.7 0.879 75.7 0.879
TS2-6.0-C20 350 6 10 8 31 169,000 38.8 146.0 109.8 1.00 1.00 136.00 98.4 0.724 98.4 0.724 98.4 0.724
TS3-6.0-C15 350 6 10 8 26 160,000 38.8 146.0 94.3 1.00 1.00 89.80 85.4 0.951 85.4 0.951 85.4 0.951
TS3-6.0-C25 350 6 10 8 36 161,000 38.8 146.0 125.0 1.00 1.00 117.00 106.0 0.906 106.0 0.906 106.0 0.906
TS3-6.0-C30 350 6 10 8 41 160,000 38.8 146.0 140.0 1.00 1.00 129.90 115.2 0.887 115.2 0.887 115.2 0.887
TS3-6.0-C40 350 6 10 8 51 154,000 38.8 146.0 169.2 0.98 0.98 130.60 130.8 1.001 128.7 0.985 128.3 0.983
Perera et al. [37] 7_R_60_S_6.4 230 2 16 6 20 151,000 71.1 52.0 86.8 0.90 0.84 50.79 64.3 1.267 57.7 1.137 54.0 1.063
Ceroni et al. [38] C-2.5 x 15-S-1 300 2.5 15 8 25 182,000 19 76.0 77.5 1.00 1.00 53.00 56.3 1.062 56.2 1.060 56.3 1.062
C-2.5 x 15-S-2 300 2.5 15 8 25 182,000 19 76.0 77.5 1.00 1.00 56.00 56.3 1.005 56.2 1.004 56.3 1.005
C-2.5 x 15-S-3 300 2.5 15 8 25 182,000 19 76.0 77.5 1.00 1.00 46.30 56.3 1.216 56.2 1.214 56.3 1.216
Seo et al. [39] N150-1 150 3.6 16 7.1 20 160,000 24 96.5 66.6 1.00 1.00 44.13 55.1 1.248 55.1 1.248 55.1 1.248
N200-1 200 3.6 16 7.1 20 160,000 24 96.5 66.6 1.00 1.00 45.11 61.3 1.358 61.3 1.358 61.3 1.358
N300-1 300 3.6 16 7.1 20 160,000 24 96.5 66.6 1.00 1.00 62.76 62.0 0.988 62.0 0.988 62.0 0.988
N150-1-1S 150 3.6 16 7.1 20 160,000 24 96.5 66.6 1.00 1.00 45.11 55.1 1.221 55.1 1.221 55.1 1.221
N200-1-1S 200 3.6 16 7.1 20 160,000 24 96.5 66.6 1.00 1.00 50.01 61.3 1.225 61.3 1.225 61.3 1.225
N300-1-1S 300 3.6 16 7.1 20 160,000 24 96.5 66.6 1.00 1.00 50.02 62.0 1.240 62.0 1.240 62.0 1.240
N150-1-2S 150 3.6 16 7.1 20 160,000 24 96.5 66.6 1.00 1.00 45.11 55.1 1.221 55.1 1.221 55.1 1.221
N200-1-2S 200 3.6 16 7.1 20 160,000 24 96.5 66.6 1.00 1.00 51.00 61.3 1.201 61.3 1.201 61.3 1.201
N300-1-2S 300 3.6 16 7.1 20 160,000 24 96.5 66.6 1.00 1.00 57.86 62.0 1.072 62.0 1.072 62.0 1.072
Khshain et al. [40] MS20E-1 180 1.42 19.72 5 30 210,000 39.5 35.0 107.2 0.65 0.55 37.98 79.6 2.096 51.7 1.362 43.5 1.146

672
MS20E-2 180 1.42 19.72 5 30 210,000 39.5 35.0 107.2 0.65 0.55 43.37 79.6 1.836 51.7 1.193 43.5 1.003
MS20E-3 180 1.42 19.72 5 30 210,000 39.5 35.0 107.2 0.65 0.55 38.23 79.6 2.083 51.7 1.353 43.5 1.138
MR10E-1 180 1.43 9.78 5 15 212,000 39.5 35.0 58.7 0.90 0.84 34.86 35.8 1.028 32.1 0.921 30.0 0.860
MR10E-2 180 1.43 9.78 5 15 212,000 39.5 35.0 58.7 0.90 0.84 31.49 35.8 1.138 32.1 1.020 30.0 0.952
MR10E-3 180 1.43 9.78 5 15 212,000 39.5 35.0 58.7 0.90 0.84 30.17 35.8 1.188 32.1 1.064 30.0 0.994
MR20E-1 180 1.42 19.72 5 30 212,000 39.5 35.0 107.2 0.65 0.55 39.65 80.0 2.018 52.0 1.311 43.7 1.103
MR20E-2 180 1.42 19.72 5 30 212,000 39.5 35.0 107.2 0.65 0.55 42.57 80.0 1.879 52.0 1.221 43.7 1.027
MR20E-3 180 1.42 19.72 5 30 212,000 39.5 35.0 107.2 0.65 0.55 41.02 80.0 1.950 52.0 1.267 43.7 1.066
Peng et al. [41] T4.5-C40 250 4.5 16 10 30 131,000 40.96 70.0 108.0 0.92 0.88 74.00 91.4 1.236 84.3 1.139 80.1 1.082
T4.5-C60 250 4.5 16 10 30 131,000 61.36 70.0 117.4 0.90 0.84 88.00 103.6 1.177 92.8 1.055 86.7 0.986
T4.5-BL250 250 4.5 16 10 30 131,000 40.96 70.0 108.0 0.92 0.88 76.00 91.4 1.203 84.3 1.109 80.1 1.054
T2.0-GE20 250 2 16 10 30 131,000 40.96 20.0 108.0 0.42 0.34 40.00 61.0 1.524 25.9 0.647 20.5 0.512
T2.0-GE40 250 2 16 10 30 131,000 40.96 40.0 108.0 0.70 0.60 54.00 61.0 1.129 42.9 0.795 36.8 0.681
T2.0-GE60 250 2 16 10 30 131,000 40.96 60.0 108.0 0.87 0.80 62.00 61.0 0.983 53.1 0.857 48.9 0.789
T4.5-GE20 250 4.5 16 10 30 131,000 40.96 20.0 108.0 0.42 0.34 48.00 91.4 1.905 38.8 0.809 30.7 0.640
T4.5-GE40 250 4.5 16 10 30 131,000 40.96 40.0 108.0 0.70 0.60 68.00 91.4 1.345 64.4 0.947 55.2 0.811
T4.5-GE60 250 4.5 16 10 30 131,000 40.96 60.0 108.0 0.87 0.80 73.00 91.4 1.252 79.7 1.091 73.4 1.005

Statistical characteristics for all specimens Average= 1.116 0.983 0.946


STD= 0.345 0.162 0.157
CoV= 0.309 0.164 0.166

Statistical characteristics for the 31 specimens with insufficient concrete edge distance (i.e., βe < 1) Average= 1.395 1.024 0.921
STD= 0.412 0.177 0.170
CoV= 0.295 0.173 0.185

Note: Lb = bond length of CFRP strip; t f = thickness of CFRP strip; hf = height of CFRP strip; wg = width of groove; hg = height of groove; Ef = elastic modulus of CFRP; fc = cylinder compressive strength of concrete;
ae = edge distance; aet = threshold value of concrete edge distance; βe = reduction factor for concrete edge distance.
Composite Structures 201 (2018) 664–675
S.S. Zhang Composite Structures 201 (2018) 664–675

45% prediction-to-test load ratio of 0.983, while the bond strength model
40.7% with the simplified reduction factor (Eq. (19)) gives a slightly smaller
40% value of 0.946. The standard deviation (STD) of the prediction-to-test
35% load ratio, obtained by using the two reduction factors given by Eqs.
Percentage of test data

(18) and (19), are 0.162 and 0.157 respectively, while the coefficient of
30% variation (CoV) of the prediction-to-test load ratio are 0.164 and 0.166
24.4% respectively. If the detrimental effect of edge distance is not considered
25% (i.e., the reduction factor is set to 1.0), the performance of the bond
20% strength model (Eq. (10)) is significantly influenced, with the average
15.12% value, STD and CoV of prediction-to-test load ratio being 1.116,
15% 0.345and 0.309 respectively. The above comparison indicates that both
10.5% the accurate reduction factor (Eq. (18)) and the simplified reduction
10% factor (Eq. (19)) can lead to reasonably accurate predictions of the test
4.7% 4.7% results, with the latter one giving relatively more conservative predic-
5%
tions. The good performance of the proposed bond strength model with
0% either of the two reduction factors can be also evidenced by the small
<1.5 1.5-3 3-4.5 4.5-6 6-7.5 >7.5 scatter of the points shown in Fig. 8.
Groove height-to-width ratio
6.3. Specimens with insufficient edge distance
(a) Groove height-to-width ratio
It can be seen from Table 3 that the reduction factors of most spe-
cimens (55 out of 86) predicted by either Eqs. (18) or (19) are equal to
50% 46.5% 1, indicating that the concrete edge distance is equal or larger than the
45% threshold value of edge distance. This is because that most of the ex-
isting tests on the NSM FRP bonded joints intentionally adopted a
40%
sufficiently large edge distance to eliminate its detrimental influence on
Percentage of test data

35% the bond strength, while very limited studies [35,41] intentionally
adopted insufficient/small edge distance to investigate its detrimental
30%
25.6% effect. In the present study, the 31 specimens with insufficient edge
25% distance (i.e., with a reduction factor of less than 1) are separated from
the database to further assess the performance of the proposed bond
20% strength. The bond strengths predicted by the bond strength model with
15% the two proposed reduction factors are compared with the 31 tests in
10.47% Fig. 9, in which the predictions without applying the reduction factors
10% 7.0% 5.81% are also plotted for comparison. The corresponding statistical char-
5% 3.49% acteristics are shown in Fig. 9 and listed in Table 3. It can be clearly
1.16%
seen from Table 3 that, without taking into account the effect of edge
0% distance (i.e., the reduction factor is set to 1.0), the bond strength
<20 20-30 30-40 40-50 50-60 60-70 >70 model (Eq. (10)) significantly overestimates the test results, with the
Concrete cylinder compressive strength fc (MPa) average prediction-to-test load ratio being 1.395 (i.e., 39.5% over-
estimation). The poor performance of the bond strength mode without
(b) Concrete cylinder compressive strength taking into account the effect of edge distance can also be seen from the
large values of STD (0.412) and CoV (0.295) of prediction-to-test load
Fig. 7. Distributions of parameters in test database: (a) Groove height-to-width
ratio. If the reduction factor (given by either Eqs. (18) or (19)) is ap-
ratio; and (b) Concrete cylinder compressive strength.
plied, the accuracy of bond strength model (Eq. (10)) is largely im-
proved: the accurate reduction factor (Eq. (18)) leads to an average
prediction-to-test load ratio of 1.024, a STD of 0.177 and a CoV of 0.173
strip is larger than 1. Although a narrow rectangular FRP bar with a respectively, while the simplified reduction factor (Eq. (19)) also gives
sectional height-to-thickness ratio not less than 5 is preferred in the accurate predictions with the average, STD and CoV of prediction-to-
application of NSM FRP strengthening method, specimens with this test load ratio being 0.921, 0.170 and 0.185 respectively. The good
ratio being between 1 and 5 are also included (36 out of the 86 col- performance of the bond strength model with either of the two reduc-
lected tests) as additional data to explore the applicability of the pro- tion factors can be also seen from the small scatter of the points shown
posed model outside its original parametric space. The details of the 86 in Fig. 9.
collected specimens are given in Table 3. The test database covers a
range of groove height-to-width ratios from 0.93 to 13.6 and a range of 7. Concluding remarks
concrete strengths from 19 MPa to 71.1 MPa. The distributions of the
groove height-to-width ratio and concrete strength are shown in Fig. 7. In this paper, the detrimental effect of concrete edge distance (i.e.,
the distance between the groove and the nearer edge of the concrete) on
6.2. Overall performance the bond strength of NSM bonded joints is clarified through a numerical
parametric study, which employs the 3-D meso-scale FE model devel-
The comparison of bond strength between the prediction of Eq. (10) oped by the author for NSM FRP bonded joints. The numerical para-
and test results for all collected specimens is given in Table 3 and metric study indicates that the bond strength of NSM FRP bonded joints
plotted in Fig. 8. As can be seen from Table 3, the bond strength model can be significantly influenced by the concrete edge distance, when the
with the accurate reduction factor (Eq. (18)) gives an average concrete edge distance is smaller than a threshold value (i.e., the

673
S.S. Zhang Composite Structures 201 (2018) 664–675

250 250
With reduction factor
Average=0.983
STD=0.162 Without reduction factor
200 CoV=0.164 200
Prediction (kN)

Prediction (kN)
150 150

100 100 With reduction factor:


Average=1.024
STD=0.177
CoV=0.173
50 50 Without reduction factor:
Average=1.395
STD=0.412
CoV=0.295
0
0
0 50 100 150 200 250 0 50 100 150 200 250
Test (kN) Test (kN)
(a) Prediction using the accurate reduction factor (Eq. 18) (a) Prediction using the accurate reduction factor (Eq. 18)

250 250
With reduction factor
Average=0.946 Without reduction factor
STD=0.157
200 200
CoV=0.166
Prediction (kN)

Prediction (kN)

150 150

100 100 With reduction factor:


Average=0.921
STD=0.170
CoV=0.185
50 50 Without reduction factor:
Average=1.395
STD=0.412
CoV=0.295
0 0
0 50 100 150 200 250 0 50 100 150 200 250
Test (kN) Test (kN)

(b) Prediction using the simplified reduction factor (Eq. 19) (b) Prediction using the simplified reduction factor (Eq. 19)
Fig. 8. Comparison of bond strength between prediction and test for all col- Fig. 9. Comparison of bond strength between prediction and test for specimens
lected specimens: (a) Prediction using the accurate reduction factor (Eq. (18)); with insufficient concrete edge distance: (a) Prediction using the accurate re-
(b) Prediction using the simplified reduction factor (Eq. (19)). duction factor (Eq. (18)); (b) Prediction using the simplified reduction factor
(Eq. (19)).

minimum required edge distance for the full development of bond


strength of the bonded joint); the threshold value of edge distance is Acknowledgements
mainly dependent on the groove height and the concrete strength: a
larger groove height or concrete strength generally leads to a larger The work presented in this paper was undertaken under the su-
threshold value of edge distance. pervision of Prof. Jin-Guang Teng from The Hong Kong Polytechnic
Based on the results from the numerical parametric study, the University. The author is grateful to Prof. Teng for his contributions to
equation for the threshold value of edge distance is formulated, and two this work.
reduction factors (i.e., the accurate reduction factor based on the direct
regression of the numerical results and the simplified reduction factor References
which is easy to use by engineers) to consider the detrimental effect of
[1] De Lorenzis L, Teng JG. Near-surface mounted FRP reinforcement: an emerging
insufficient edge distance on the bond strength are proposed. With in-
technique for strengthening structures. Compos Part B Eng 2007;38(2):119–43.
troducing the proposed reduction factors, a bond strength model is [2] American Concrete Institute (ACI). Guide for the design and construction of ex-
developed for NSM FRP bonded joints with insufficient edge distances. ternally bonded frp systems for strengthening concrete structures. MI, USA:
Farmington Hills; 2017. ACI 440.2R-17.
The comparison of the bond strength between the predictions and test
[3] Zhang SS, Yu T, Chen GM. Reinforced concrete beams strengthened in flexure with
results shows that the bond strength model with either of the two near-surface mounted (NSM) CFRP strips: current status and research needs.
proposed reduction factors can provide similarly accurate predictions of Compos B Eng 2017;131:30–42.
the test results. [4] Barros JAO, Dias SJE, Lima JLT. Efficacy of CFRP-based techniques for the flexural
and shear strengthening of concrete beams. Cem Concr Compos 2007;29(3):203–17.
[5] Garzon-Roca J, Sena-Cruz JM, Fernandes P, Xavier J. Effect of wet-dry cycles on the

674
S.S. Zhang Composite Structures 201 (2018) 664–675

bond behaviour of concrete elements strengthened with NSM CFRP laminate strips. [23] Seracino R, Jones NM, Ali MSM, Page MW, Oehlers DJ. Bond strength of near-
Compos Struct 2015;132:331–40. surface mounted FRP strip-to-concrete joints. J Compos Constr ASCE
[6] Fernandes P, Sena-Cruz J, Xavier J, Silva P, Pereira E, Cruz J. Durability of bond in 2007;11(4):401–9.
NSM CFRP-concrete systems under different environmental conditions. Compos [24] Bazant ZP, Oh BH. Crack band theory for fracture of concrete. Mater Struct
Part B Eng 2018;138:19–34. 1983;16(93):155–77.
[7] Bilotta A, Ceroni F, Nigro E, Pecce M. Efficiency of CFRP NSM strips and EBR plates [25] CEB-FIP. Model Code 90, Lausanne, Switzerland, 1993. 1990.
for flexural strengthening of RC beams and loading pattern influence. Compos [26] Hordijk DA. Local approach to fatigue of concrete. Delft University of Technology;
Struct 2015;124:163–75. 1991. PhD thesis.
[8] Daghash SM, Ozbulut OE. Flexural performance evaluation of NSM basalt FRP- [27] Okamura H, Maekawa K. Nonlinear analysis and constitutive models of reinforced
strengthened concrete beams using digital image correlation system. Compos Struct concrete. Tokyo: Gihodo-Shuppan; 1991.
2017;176:748–56. [28] Buyukozturk O. Nonlinear analysis of reinforced concrete structures. Comput Struct
[9] Kuntal VS, Chellapandian M, Prakash SS. Efficient near surface mounted CFRP shear 1977;7:149–56.
strengthening of high strength prestressed concrete beams – An experimental study. [29] Elwi AA, Murray DW. A 3D hypoelastic concrete constitutive relationship. J Eng
Compos Struct 2017;180:16–28. Mech Div ASCE 1979;105(4):623–41.
[10] Perera WKKG, Ibell TJ, Darby AP. Bond characteristics of near surface mounted [30] Zhang SS, Teng JG. Finite element analysis of end cover separation in RC beams
CFRP bars. Constr Build Mater 2013;43:58–68. strengthened in flexure with FRP. Eng Struct 2014;75:550–60.
[11] Lee D, Cheng LJ, Hui JYG. Bond characteristics of various NSM FRP reinforcements [31] Zhang SS, Teng JG. End cover separation in RC beams strengthened in flexure with
in concrete. J Compos Constr ASCE 2013;17(1):117–29. bonded FRP reinforcement: simplified finite element approach. Mater Struct
[12] Bilotta A, Ceroni F, Nigro E, Pecce M. Strain assessment for the design of NSM FRP 2015;49(6):2223–36.
systems for the strengthening of RC members. Constr Build Mater 2014;69:143–58. [32] Chen JF, Teng JG. Anchorage strength models for FRP and steel plates bonded to
[13] Coelho MRF, Sena-Cruz JM, Neves LAC. A review on the bond behavior of FRP NSM concrete. J Struct Eng ASCE 2001;127(7):784–91.
systems in concrete. Constr Build Mater 2015;93:1157–69. [33] Li R, Teng JG, Yue QR. Experimental study on bond behavior of NSM CFRP strips-
[14] Barros JAO, Costa IG, Ventura-Gouveia A. CFRP flexural and shear strengthening concrete Interface. Ind Constr 2005;35(8):31–5. (in Chinese).
technique for rc beams: experimental and numerical research. Adv Struct Eng [34] Shield CK, French CW, Milde E. The effect of adhesive type on the bond of NSM tape
2011;14(3):551–73. to concrete. In: ACI SP230: 7th International symposium on fiber-reinforced
[15] Ceroni F, Barros JAO, Pecce M, Ianniciello M. Assessment of nonlinear bond laws polymer (FRP) reinforcement for concrete structures. November. American
for near-surface-mounted systems in concrete elements. Compos Part B Eng Concrete Institute; 2005. p. 355–72.
2013;45(1):666–81. [35] Rashid R, Oehlers DJ, Seracino R. IC debonding of FRP NSM and EB retrofitted
[16] Emara M, Torres L, Baena M, Barris C, Cahis X. Bond response of NSM CFRP strips in concrete: plate and cover interaction tests. J Compos Constr ASCE
concrete under sustained loading and different temperature and humidity condi- 2008;12(2):160–7.
tions. Compos Struct 2018;192:1–7. [36] Oehlers DJ, Haskett M, Wu CQ, Seracino R. Embedding NSM FRP plates for im-
[17] Seracino R, Saifulnaz MRR, Oehlers DJ. Generic debonding resistance of EB and proved IC debonding resistance. J Compos Constr ASCE 2008;12(6):635–42.
NSM plate-to-concrete joints. J Compos Constr ASCE 2007;11(1):62–70. [37] Perera WKKG, Ibell TJ, Darby AP. Bond behaviour and effectiveness of various
[18] Sena Cruz JM, Barros JAO. Bond between near-surface mounted carbon-fiber-re- shapes of NSM CFRP bars. In: Proceedings, 9th international symposium on fiber-
inforced polymer laminate strips and concrete. J Compos Const ASCE reinforced polymers reinforcement for concrete structures (FRPRCS-9), 13–15 July,
2004;8(6):519–27. Sydney, Australia (CD-ROM). 2009.
[19] Zhang SS, Teng JG, Yu T. Bond-slip model for CFRP strips near-surface mounted to [38] Ceroni F, Pecce M, Bilotta A, Nigro E. Bond behavior of FRP NSM systems in con-
concrete. Eng Struct 2013;56:945–53. crete elements. Compos Part B Eng 2012;43(2):99–109.
[20] Zhang SS, Teng JG, Yu T. Bond strength model for CFRP strips near-surface [39] Seo SY, Feo L, Hui D. Bond strength of near surface-mounted FRP plate for retrofit
mounted to concrete. J Compos Constr ASCE 2014;18(3). http://dx.doi.org/10. of concrete structures. Compos Struct 2013;95:719–27.
1061/(asce)cc.1943-5614.0000402. Article Number: A4014003. [40] Khshain NT, Al-Mahaidi R, Abdouka K. Bond behaviour between NSM CFRP strips
[21] Teng JG, Zhang SS, Dai JG, Chen JF. Three-dimensional meso-scale finite element and concrete substrate using single-lap shear testing with epoxy adhesive. Compos
modeling of bonded joints between a near-surface mounted FRP strip and concrete. Struct 2015;132:205–14.
Comput Struct 2013;117:105–17. [41] Peng H, Hao HX, Zhang JR, Liu Y, Cai CS. Experimental investigation of the bond
[22] MSC.MAR. User's Manual. MSC. Software Corporation, 2 MacArthur Place, Santa behavior of the interface between near-surface-mounted CFRP strips and concrete.
Ana, California 92707, USA. 2005. Constr Build Mater 2015;96:11–9.

675

You might also like