Composites: Part B 38 (2007) 119–143

www.elsevier.com/locate/compositesb

Near-surface mounted FRP reinforcement: An emerging technique for strengthening structures

L. De Lorenzis a,*, J.G. Teng b
a Department of Innovation Engineering, University of Lecce, via per Monteroni, 73100 Lecce, Italy
b Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, Hong Kong, China

Received 27 January 2006; accepted 17 August 2006
Available online 18 October 2006

Abstract

Near-surface mounted (NSM) fiber-reinforced polymer (FRP) reinforcement is one of the latest and most promising strengthening
techniques for reinforced concrete (RC) structures. Research on this topic started only a few years ago but has by now attracted worldwide
 attention. Issues raised by the use of NSM FRP reinforcement include the optimization of construction details, models for the bond
behaviour between NSM FRP and concrete, reliable design methods for flexural and shear strengthening, and the maximization of the
advantages of this technique. This paper provides a critical review of existing research in this area, identifies gaps of knowledge, and
outlines directions for further research.

2006
Elsevier
Ltd.
All
rights
reserved.
Keywords: A. Polymer matrix composites (PMCs); B. Debonding; C. Analytical modeling; D. Mechanical testing; Near-surface mounted reinforcement

1.
Introduction
Over the past decade, extensive research has been
conducted on the strengthening of reinforced concrete
(RC) structures using externally bonded fiber-reinforced
polymer (FRP) laminates; the technology has also been
implemented in a large number of practical projects worldwide.
 More recently, near-surface mounted (NSM) FRP
reinforcement has attracted an increasing amount of
research as well as practical application. In the NSM
method, grooves are first cut into the concrete cover of
an RC element and the FRP reinforcement is bonded
therein with an appropriate groove filler (typically epoxy
paste or cement grout). What is herein called ‘‘NSM reinforcement’’
 was previously given other names such as
‘‘grouted reinforcement’’ [1], or ‘‘embedded reinforcement’’
[2,3].

*

Corresponding author. Tel.: +39 0832 297241; fax: +39 0832 297279.
E-mail address: laura.delorenzis@unile.it (L. De Lorenzis).

Examples of the use of NSM steel rebars in Europe for
the strengthening of RC structures date back to the early
1950s [1]. More recent applications of NSM stainless steel
bars for the strengthening of masonry buildings and arch
bridges have also been documented (e.g. [4]). The advantages
 of FRP versus steel as NSM reinforcement are better
resistance to corrosion, increased ease and speed of installation
 due to its lightweight, and a reduced groove size due
to the higher tensile strength and better corrosion resistance
 of FRP.

Compared to externally bondedFRP reinforcement, the
NSM system has a number of advantages: (a) the amount
of site installation work may be reduced, as surface preparation
 other than grooving is no longer required (e.g., plaster
 removal is not necessary; irregularities of the concrete
surface can be more easily accommodated; removal of
the weak laitance layer on the concrete surface is no longer
needed); (b) NSM reinforcement is less prone to debonding
from the concrete substrate; (c) NSM bars can be more easily
 anchored into adjacent members to prevent debonding
failures;this feature is particularly attractive in the flexural
strengthening of beams and columns in rigidly-jointed

1359-8368/$
-see
front
matter
2006
Elsevier
Ltd.
All
rights
reserved.
doi:10.1016/j.compositesb.2006.08.003

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
frames, where the maximum moments typically occur
at the ends of the member; (d) NSM reinforcement
can be more easily pre-stressed; (e) NSM bars are protected
 by the concrete cover and so are less exposed to accidental
 impact and mechanical damage, fire, and vandalism;
this aspect makes this technology particularly suitable for
the strengthening of negative moment regions of beams/
slabs; (f) the aesthetic of the strengthened structure is
virtually unchanged. Due to the above advantages, the
NSM FRP method is in many cases superior to the
externally bondedFRP method or can be used in combination
 with it, provided that the cover of the member is sufficiently
 thick for grooves of a desirable size to be
accommodated.

The existing knowledge on the NSM FRP method is
much more limited than that on the externally bonded
FRP method, as reflected by the absence of relevant provisions
 in the existing guidelines on the FRP strengthening of
concrete structures issuedby fib [5] and ACI-440 [6]. However,
 the international engineering community has become
increasingly aware of the practical advantages of this
method, which has led to accelerations of research and
practical applications worldwide. Both ACI-440 and fib
are currently considering revisions to their documents to
include NSM-related provisions. Against this background,
this paper provides a critical review of existing research in
this area, identifies gaps of knowledge, and outlines directions
 for further research.

This paper focuses on research work on the structural
aspects of NSM strengthening of concrete structures. Discussions
 on some significant practical applications of the
NSM method, on NSMstrengthening of masonryand timber
 structures, and on durability-related aspects are given
in Ref. [7]. Space limitation does not allow them to be discussed
 in this paper.

2.
Materials
and
systems
2.1. FRP reinforcement
In most existing studies, carbon FRP (CFRP) NSM
reinforcement has been used to strengthen concrete structures.
 Glass FRP (GFRP) has been used in most applications
 of the NSM method to masonry and timber
structures. The present authors are not aware of any study
or practical application in which aramid FRP (AFRP) was
used. The tensile strength and elastic modulus of CFRP are
much higher than those of GFRP, so for the same tensile
capacity, a CFRP bar has a smaller cross-sectional area
than a GFRP bar and requires a smaller groove. This in
turn leads to easier installation, with less risks of interfering
with the internal steel reinforcement, and with savings in
the groove-filling material.

FRP bars can be manufactured in a virtually endless
variety of shapes. Hence, the NSM FRP reinforcement
may be round, square, rectangular and oval bars, as well
as strips (Figs. 1 and 2). For brevity, the term ‘‘bars’’ is

Fig. 1. Types of FRP bars for NSM applications.

Fig. 2. Different NSM systems and nomenclature.

used herein as a generic term encompassing all cross-
sectional shapes, while the term ‘‘strips’’ is reserved for
thin narrow strips. Different cross-sectional shapes have
different advantages, and offer different choices for practical
 applications. For example, square bars maximize the
bar sectional area for a given size of square groove while
round bars are more readily available and can be more
easily anchored in pre-stressing operations. Narrow strips

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
maximize the surface area-to-sectional area ratio for a
given volume and thus minimize the risk of debonding,
but require a thicker cover for a given cross-sectional area.
In practical applications, the choice depends strongly on
the constraints of a specific situation, such as the depth
of the cover, and the availability and cost of a particular
type of FRP bar.

FRP bars are also manufactured with a variety of surface
 textures, which strongly affect their bond behaviour
as NSM reinforcement. Their surface can be smooth,
sand-blasted, sand-coated, or roughened with a peel-ply
surface treatment. Round bars can also be spirally wound
with a fiber tow, or ribbed [8].

2.2. Groove filler
The groove filler is the medium for the transfer of stresses
 between the FRP bar and the concrete. In terms of
structural behaviour, its most relevant mechanical properties
 are the tensile and shear strengths. The tensile strength
is especially important when the embedded bars have a
deformed surface, which produces high circumferential
tensile stresses in the cover formed by the groove filler
(simply referred to as ‘‘the cover’’ or ‘‘the epoxy cover’’
hereafter) as a result of the bond action. In addition, the
shear strength is important when the bond capacity of
the NSMreinforcement is controlled by cohesive shear failure
 of the groove filler. The effect of the modulus of elasticity
 of the groove filler has never been experimentally
investigated.

The most common and best performing groove filler
is a two-component epoxy. Low-viscosity epoxy can be
selected for strengthening in negative moment regions as
the epoxy can be ‘‘poured’’ into the grooves. For other
cases, a high-viscosity epoxy is needed to avoid dripping
or flowing-away. The addition of sand to epoxy can
increase the volume, control the viscosity, lower the coefficient
 of thermal expansion, and raise the glass transition
temperature. A drawback of this addition seems to be
reduced adhesion at the bar–epoxy interface for a smooth
bar surface [2].

The use of cement paste or mortar in place of epoxy as a
groove filler has recently been explored in an attempt to
lower the material cost, reduce the hazard to workers, minimize
 the environmental impact, allow effective bonding to
wet substrates, and achieve better resistance to high temperatures
 and improved thermal compatibility with the
concrete substrate. However, cement mortar has inferior
mechanical properties and durability, with a tensile
strength an order of magnitude smaller than that of common
 epoxies. Results of bond tests and flexural tests
[9,10] have identified some significant limitations of cement
mortar as a groove filler. Given these limitations and the
very limited data available, the rest of this paper is focused
on epoxy-bonded NSM FRP reinforcement only, except
when future research needs are discussed. Nevertheless,
tests on NSM FRP reinforcement using cement grout as

the groove filler, if available, are included in Tables 1–4
for completeness.

2.3. Groove dimensions
Fig. 2 shows several configurations of NSM FRP reinforcement,
 where db is the nominal diameter of a round
bar, and tf and hf are the thickness/width and the height
of an FRPstrip or rectangular bar respectively. The groove
width bg, the groove depth hg, the net distance between two
adjacent grooves ag, and the net distance between a groove
and the beam edge ae are all relevant construction parameters,
 which can influence the bond performance and hence
the structural behaviour.

For round bars, De Lorenzis [11], based on results of
bond tests with square grooves (bg= hg) and defining
k = bg/db, proposed a minimum value of 1.5 for k for
smooth or lightly sand-blasted bars and a minimum value
of 2.0 for k for deformed bars. Parretti and Nanni [12] suggested
 that both bg and hg should be no less than 1.5db. For
NSM strips, Blaschko [13] suggested that the depth and
width of the cut groove should be about 3 mm larger than
the height and thickness of the corresponding FRP strip
respectively, so to obtain an adhesive layer thickness of
about 1–2 mm. Also for NSM strips, Parretti and Nanni

[12] recommended that the minimum width of a groove
be no less than 3tf and the minimum depth be no less than
1.5hf. For a more detailed discussion, see the section on
bond behaviour.
In the existing studies, NSM strips were bonded using
epoxy either along all four sides of the strip surface
[14,15], or along three sides of the strip surface only
[16,17](Fig.2). Dueto the large width to thickness ratio
of the strips, the reduction in the bond surface in the latter
case is negligible.In existing tests on NSM square bars [18],
only three sides of the bar surface were bonded to the concrete
 member.

2.4. Groove position
IfasingleNSMbaristobe providedtothe tensionside
of an RC member, it should naturally be centrally located
over the beam width. When two or more NSM bars need to
be provided, then the distance between two adjacent NSM
bars and the distance between the edge of the member and
the adjacent bar become important design parameters. The
effect of these parameters is discussed in the section on
bond behaviour.

2.5. Constructional aspects
Compared with the use of externally bonded FRP laminates,
 the need to cut grooves into the concrete member in
the construction process of the NSMmethod is the keydifference.
 A detailed discussion of the construction process
can be found in Ref. [7].

Table 1
Summary of existing experimental work on the bond behaviour of NSM FRP reinforcement bonded to concrete

Round bars Square bars Strips
[2] [38] [3] [21] [23] [26] [9] [13,16] [20] [14]
Test method Direct pull-out Beam pull-out Direct pull-out Beam pull-out Direct pull-out Beam
pull-out
Specimen Epoxy-filled Concrete Two concrete Inverted T beam, C-shaped C-shaped Beam with two Concrete block, Concrete block, RILEM
shape and plastic pipe, block, blocks, each total specimen, specimen, halves connected 300 · 300 · 150 · 150 · 350 beam pull-
dimensions diameter = 152 · 152 · 150 · 150 · height = 254 mm, external external by a steel hinge, 1100 mm, one NSM mm, one NSM out test
51 mm, 203 mm, two 300 mm, two flange width = dimensions dimensions net span = 1.73 m, strip strip per test
one bar NSM bars on NSM bars on 254 mm, net are 300 · are cross-sectional per test
embedded opposite opposite sides span = 1.07 m, 300 · 400 mm, 300 · 300 · dimensions N/Aa ,
concentrically sides one NSM bar one NSM bar 400 mm, one NSM bar
one NSM bar
f
0 c (MPa) – 34 20 28 22 22 N/Aa 32 and 46 29 cube 35, 45, 70
strength
Groove-filler Epoxy or Epoxy Epoxy Epoxy Epoxy Epoxy or Epoxy or Epoxy Epoxy Epoxy
epoxy + sand cement paste cement paste
Direct tensile 15–20 from 13.8 from N/Aa 13.8, from 27.4, from 27.4, from 31, from manufacturer 33.3 from 42.6 from 16–22, from
strength of manufacturer manufacturer manufacturer testing testing (epoxy) (epoxy)e 9, bending testing testing testing
groove filler (value with 6.3, bending tensile
(MPa) no sand) tensile strength from
strength manufacturer (cement)
from testing
(cement)
Type of FRP/ CFRP/ CFRP/ CFRP/round/ CFRP/round/ CFRP/round/ CFRP/round/ CFRP/square/SM CFRP/strip/ CFRP/strip/ CFRP/strip/
crossround/
SM round/SB SB SB SWS SWS roughened roughened N/Aa
sectional CFRP/ CFRP/round/ GFRP/round/ GFRP/round/ CFRP/square/SC
shape/ round/SB RB RB RB
surface GFRP/round/
configuration RB
Nominal db or 9.5 11 9.5 9.5 and 12.7 7.5 and 9.5 7.5 and 9.5 10 · 10 (1.2–2) · 20 5· 16 1.4 · 9.3
tf · hf of
FRP (mm)
bg · hg (mm) – 14.3 · 19.0 16 · 25 bg = hg, bg = hg, bg = hg, 16 · 16 (2.8–4.3) · (21–23) 9· 22 3.3 · 15
k= 1.12–2.67 k= 1.24–2.50 k= 1.24–2.50 average 3.3 · 22
Groove surface – Saw-cut Saw-cut Saw-cut Pre-formed Saw-cut Saw-cut Saw-cut Saw-cut Saw-cut
Test variables Bar surface, BL BL BL, type of bar, BL, type of BL, type of BL, bar surface BL, type of strip,
0 c, BL BL, concrete
type of epoxy bar diameter, k bar, k bar, groove-configuration, type of loading, ae strength
filler groove-filler
Observed bond BE-I Concrete Concrete BE-I, SP-E, EC-I, SP-C1 SP-C1, BE-C N/Aa (indicated as BE-C, SP-ED BE-I BEd
failure shear at shear failurec SP-C1, SP-C2 ‘‘concrete failure’’ and
modesb edgec, BE-I ‘‘delamination’’)

L. De Lorenzis, J.G. Teng / Composites: Part B 38 (2007) 119–143
Acronyms: BL = bond length; RB = ribbed; SB= sand-blasted; SC = sand-coated; SM = smooth; SWS = spirally wound and sand-coated.

a N/A = not available.

b For acronyms see Fig. 3.

c Failure mode influenced by specimen details.

d Unspecified whether BE-I or BE-C.

e Tensile test method unspecified.

Table
2
Local bond strength of NSM systems (saw-cut grooves)

Type of bar (material, Surface Nominal db or k or bg · hg Groove Direct tensile strength of
f
0 c (MPa) Local bond Mainb Reference
cross-sectional shape) configuration tf · hf (mm) (mm) filler groove filler (MPa) strength (MPa) failure mode
CFRP, round Sand-blasted 9.5 1.34–2.67 Epoxy 13.8 From manufacturer 28 8.6 BE-I [21]
12.7 1.25 9.7
CFRP, round Ribbed 9.5 1.33 27.4 From testing 22 11.2 SP-C1 [26]
1.59 15.4
2.12 16.6
GFRP, round Ribbed 9.5 1.36 9.1
1.64 10.0
2.18 12.5
CFRP, round Spirally wound and 7.5 1.50 18.0
sand-coated 2.00 20.8
2.50 21.9
CFRP, square Smooth 10 · 10 N/Aa 31 From manufacturer N/Aa 9.0c N/Aa [9]
CFRP, strip N/Aa 1.5 · 9.6 3.3 · 15 16–22 From testing 35, 45, 70 19.8 BE [14]
Roughened (1.2–2) · 20 Average 33.3 From testing 32 and 46 20.0d BE-C [16]
3.3 · 22
Roughened 5· 16 9· 22 42.6 From testing 29 Cube 10–12 BE-I [20]
strength
CFRP, round Ribbed 9.5 1.59 Cement 6.3, Bendingtensile strength 22 9.7 SP-C1, BE-C [26]
2.21 paste from testing 6.4
GFRP, round Ribbed 9.5 1.64 8.0
2.27 8.3
CFRP, round Spirally wound and 7.5 1.50 6.7
sand-coated
CFRP, square Sand-coated 10· 10 N/Aa 9, Bending tensile strength N/Aa 4.3c N/Aa [9]
from manufacturer
a N/A = not available.
b In some cases more than one failure mode were concurrent.
c Taken from the specimen with the shortest bonded length (100 mm).
d With no influence of edge distance (i.e. a0 e
P
150 mm).

L.DeLorenzis,J.G.Teng /Composites:PartB 38(2007)119–143

Table 3
Summary of existing experimental work on flexural strengthening with NSM FRP reinforcement – beams with limited bonded lengths, slabs, and columns

Beams with limited bonded lengths Slabs Columns
Reference [17] [19] [20] [2] [42] [43] [42]
Test method 3-Point bending 4-Point bending Patch loading at centre 3-Point bending Uniform line load Horizontal load
at mid-span or at 330 at top
from free end of
cantilever slab
Cross-sectional T, total height = 300, Rectangular, 5400 · 229 RC, simply 1800 · 229 PC, simply 7600 · 460 1180 · 406 PC 600 · 600
shape and web height = 250, 150 · 300 supported supportedc
dimensions flange width = 300,
(mm) web width = 150
Net span (m) 2.5 3 3 7.9 4.9 (mid-span), 1.8 Height
(cantilever) 1.8–3.4
Shear span (m) – – – – – –
f
0 c (MPa) 48 35 28 41 56 45–50 56
Groove filler Epoxy
Direct tensile N/Aa 42.6 from testing N/Aa N/Aa N/Aa N/Aa N/Aa
strength of
groove filler
(MPa)
Type of FRP/crossCFRP/
strip/ CFRP/ CFRP/ CFRP/round/SM CFRP/round/SM CFRP/round/SB CFRP/round/N/Aa CFRP/round/
sectional shape/ N/Aa round/RB strip/PP CFRP/strip/N/Aa CFRP/strip/N/Aa SB
surface CFRP/round/RB
configuration
FRP nominal db or 1.2 · 25 9.5 4· 16 3 9.5, 1.2 · 25 11 10; 1.4 · 25; 9.5 11
tf · hf (mm)
Number of FRP 1 1 1 4 Bars every 102 mm 1 Bar every 102 mm 20 bars (one every 10; 18; (mid-span) 3 on each face
bars o.c. (long.), 1bar every o.c. (long.), 1 strip 381 mm o.c.) 6; 6; 8 (cantilever) or 7 on each
102 mm o.c. (transv.) every 51 mm o.c. face
(long.)
Elastic modulus of 150 111 151 172 172 (bars), 119 147; 150; 111 119
FRP (GPa) 130 (strips)
Tensile strength of 2000 1918 2068 1550 1550 (bars), 1240 1970; 2000; 1918 1240
FRP (MPa) 1790 (strips)
bg · hg (mm) 5· 25 18 · 30 8· 22 9.5 · 13 16 · 25, 3.2 · 29 14 · 19 16 · 30; 5 · 25; 16 · 30 14 · 19
Steel tension Two 10-mm bars Two 12-mm bars 16-mm bars @102 mm None 25-mm bars @127 mm Five 16-mm mild bars Four 19-mm
reinforcement o.c. (long.) and o.c. (long.) and 13-mm (mid-span), four 16-mm bars
204 mm o.c. (transv.) bars @457 mm o.c. mild bars (cantilever) +
(transv.) twelve 15-mm 7-wire
pre-stressing strands
Bar anchorage 150–1200 150–1200 0–1150 Termination: 152 from N/Aa Bar length 6 m N/Aa 381 mm in the
lengthb (mm) support, 762 from free footing
edge
Test variables Bar anchorage length – – – Type of bar Number of bars
Observed failure Debonding by EC-Cb at Concrete CCS (cut-off), Punching CC + DBd (bars), BR CC BR, CC
modes cut-off and MMR, BR Splittingc CCS (MMR) shear + DBd CC (strips)
(MMR)
Increase in ultimate 0–54 0–41 0–106 15 300 27 36; 43; 39 102, 177
load (%)

L. De Lorenzis, J.G. Teng / Composites: Part B 38 (2007) 119–143
Acronyms: BR = bar rupture, CC = concrete crushing, CCS = concrete cover separation; DB = debonding; MMR = maximum moment region; RC = reinforced concrete; PC = pre-stressed concrete; PP = roughened with peel-ply
surface treatment; RB = ribbed; SB = sand-blasted; SM = smooth.

a N/A = not available.

b See Fig. 3.

c Not clear whether it was CCS.

d Debonding mechanism not clear.

Table 4
Summary of existing experimental work on flexural strengthening with NSM FRP reinforcement –beams

Reference

[61] [44] [11] [10] [16] [41] [62] [63] [15]
Test method 4-Point bending 3-Point bending 4-Point bending
Cross-sectional Rectangular, T, total height = 406, Rectangular, Rectangular, Rectangular, T, total height = 300, Rectangular, Rectangular, Rectangular,
shape and 457 · 152, web height = 305, 200 · 400 200 · 300 200 · 500, web height = 250, 152 304 · 188 200 · 120 100 · 170 180
dimensions 914 · 356 flange width = 381, 600 · 500 flange width=300,
(mm) web width = 152 web width=150
Net span (m) 2.4, 7.5 3.9 4 3.6 2.8, 7.5 2.5 1.3 1.1 1.5
Shear span (m) 0.8, 2.5 1.83 1.75 1.3 1.15, 3.25 – 0.5 0.5 0.5
f
0 c (MPa) N/Aa 36 15 61 44 48 37 20–63 46
Groove filler Epoxy Epoxy Epoxy Epoxy and Epoxy Epoxy Epoxy Epoxy Epoxy
cement grout
Direct tensile 30 13.8 27.4 N/Aa 33.3 48 (bars) 70 (strips) N/Aa N/Aa 16–22 from testing
strength of
groove filler
(MPa)
Type of FRP/ CFRP/round/ GFRP/round/RB CFRP/round/ CFRP/square/ CFRP/strip/ CFRP/round/SW CFRP/strip/ CFRP/round/ CFRP/strip/N/A
cross-sectional SM CFRP/round/SB SWS SM roughened CFRP/strip/N/A roughened N/Aa
shape/surface CFRP/square/ CFRP/strip/ N/A
configuration SC GFRP/strip/N/A
FRP nominal db 4.75, 6.35 12.7 (GFRP), 9.5 and 7.5 10 · 10 2· 20 9.5, 2 · 16, 1.2 · 25, 0.25 · 15.5 7 (net); 8 1.5 · 9.6
or tf · hf (mm) 12.7 (CFRP) 2· 20 (external)
Number of FRP 4, 11 2 1, 2 2 3, 11 1, 2 (CFRP), 5 1, 2 1, 3 1–3
bars (GFRP)
Elastic modulus 122 38.6 (GFRP) 175 230 156 122.5, 140, 150, 45 136.6 201 159
of FRP (GPa)
Tensile strength 1326 773 (GFRP) 2214 4140 1813 1408, 1525, 2000, 1000 1656 1940 2740
of FRP (MPa)
bg · hg (mm) bg = 10.2, 19 · 19 (db = 9.5) 16 · 16 N/Aa 3.3 · 23 18 · 30, 6.4 · 19, 6.4 · 19 N/Aa 4· 12
varying hg 25 · 25 (db = 12.7) 6.4 · 25, 6.4 · 25
Steel tension 1.14, 1.19 0.89 0.38–0.64 0.67 0.63, 0.84 0.48 0.83–1.74 0.28, 0.57 0.33–0.84
reinforcement
ratio (%)
Cut-off distance N/Aa Extended over Extended over 300 or extended 150, 300 50 Extended over 50 50
from support supports supports over supports supports
(mm)
Test variables Beam size, steel Type of FRP bar, Steel ratio, NSM bar length, End anchorage, Type of FRP bar Section width, Number of FRP Steel ratio,
ratio, groove size bar diameter number of FRP groove filler type of loading steel ratio, number bars, concrete number of FRP
bars of FRP bars strength bars
Observed failure CC, secondary DBb of NSM SB bar, CC, CCS DBc of NSM CCS from CCS (MMR) (CFRP CC, BR CC, DB CCSd
modes debonding and CCS (MMR) (MMR), edge bars, BR (when cut-off, BR round bars and (described as
partial BR failure extended over GFRP strips), BR ‘‘rod slippage’’)
supports) (CFRP strips)
Increase in 20–50 26–44 21–61 56–92 67–82 69–99 15–55 140–430 78–98
capacity (%)

Acronyms: BR = bar rupture, CC = concrete crushing, CCS = concrete cover separation; DB = debonding; MMR = maximum moment region; RB = ribbed; SB = sand-blasted; SC = sand-coated; SM = smooth;
SW = spirally wound; SWS = spirally wound and sand-coated.

a

N/A = not available.
b Debonding mechanism BE-I(Fig.3), see Fig.7a.
Debonding mechanism not clear.

L.DeLorenzis,J.G.Teng /Composites:PartB 38(2007)119–143

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
3.
Bond
behaviour
3.1. Summary of existing work
The bond between an NSM bar and the substrate material
 plays a key role in ensuring the effectiveness of the
NSM strengthening method. The performance of the bond
depends on a number of parameters: the groove and the
bar dimensions, the tensile and shear strengths of the concrete
 and the groove filler, the bar cross-sectional shape
and surface configuration, and the degree of roughness of
the groove surface. This large number of parameters calls
for extensive laboratory characterization, as well as analytical
 and numerical modelling. Table1 providesa summary
of the existing experimental data on bond behaviour. Note
that the tests in some existing studies [17,19,20], although
addressing bond-related aspects, were conducted on beams
under bending and are therefore discussed in a later
section.

3.2. Failure modes and mechanisms
The bond tests summarized in Table 1 identified different
 possible bond failure modes of NSM systems(Fig.3).
These modes are described in some detail below and the
underlying mechanisms examined.

3.2.1. Bond failure at the bar–epoxy interface
This mode may occur as either pure interfacial failure
(BE-I), or as cohesive shear failure in the groove
filler (BE-C). The BE-I mode is critical for bars with a
smooth or lightly sand-blasted surface, i.e. when the
degree of surface deformation is insufficient to provide
mechanical interlocking between the bar and the groove
filler and the bond resistance relies primarily on adhesion
between the bar and the filler. For round bars, this mode
becomes critical if the groove size is sufficiently large to
avoid splitting failure of the groove filler. For epoxy and
concrete of moderate strengths, De Lorenzis and Nanni

[21] estimated that for lightly sand-blasted round bars, a
k value of 1.5 was enough to prevent splitting failure of
the epoxy cover. For round bars, cracking of the epoxy
cover (Fig. 3) produced by the radial components of the
bond stresses can accelerate the occurrence of a BE-I
failure.
The BE-C failure mode was observed for NSM strips
with a roughened surface [13,16]. This mode occurs when
the shear strength of the epoxy is exceeded.

Inter-laminar shear failure within the bar, although theoreticallypossible,
 has never been observed. Shearing-off of
ribs in ribbed bars has never been reported as a failure
mode itself, unlike in the case of FRP ribbed bars as internal
 reinforcement for concrete [22]. However, in some tests
[21], the surfaceof the ribbedbars was found tohave been
damaged after bond failure, indicating that this could be an
upper-bound failure mode.

3.2.2. Bond failure at the epoxy–concrete interface
Bond failure at the epoxy–concrete interface may
occur as pure interfacial failure (EC-I), or as cohesive
shear failure in the concrete (EC-C). The EC-I failure
mode was found to be critical for pre-cast grooves
[23]. For spirally wound bars or ribbed bars with low
rib protrusions, this was found to be the critical failure
 mode whenever the groove was preformed, independent
 of the value of k. For ribbed bars with high
rib protrusions, this mode was found to be critical only
for k values larger than a minimum value (equal to
approximately 2.00 for ribbed bars in epoxy), and for
lower k values, splitting failure of the epoxy cover
dominated.

The EC-C failure mode has never been observed in bond
tests, but it has been observed in bending tests on beams
(Fig. 3, see also Fig. 7f) within the strengthened region
[11,20] or at the bar cut-off point [17]. The latter authors
considered this failure mode in their theoretical model for
debonding of NSM strips.

3.2.3. Splitting of the epoxy cover
Longitudinal cracking of the groove filler and/or fracture
 of the surrounding concrete along inclined planes is
herein referred to as cover splitting. This was observed to
be the critical failure mode for deformed (i.e. ribbed and
spirally wound) round bars.

The mechanics of cover splitting bond failure in an
NSM system is similar to that of splitting bond failure of
steel deformed bars in concrete, on which a good understanding
 has been developed from decades of research
[24]. For an NSM FRP bar, the radial component of the
bond stresses is balanced by circumferential tensile stresses
in the epoxy cover which may lead to the formation of longitudinal
 splitting cracks of the cover. The concrete surrounding
 the groove is also subjected to tensile stresses
and may eventually fail when its tensile strength is reached,
causing fracture along inclined planes. Whether fracture in
the concrete occurs before or after the appearance of splitting
 cracks in the cover or even after the complete fracture
of the cover, depends on the groove size and the tensile
strengths of the two materials.

The tensile strength of epoxy is one order of magnitude
larger than that of concrete, but the epoxy cover thickness
for NSM FRP reinforcement is one order of magnitude
smaller than the thickness of concrete cover to internal
steel reinforcement in an ordinary RC member. Moreover,
longitudinal steel reinforcement in RC beams benefits from
the restraint of shear links, but this restraint is not available
for NSM longitudinal reinforcement, unless external
restraint of some form (e.g. FRP U jackets as shear reinforcement)
 is provided. These factors explain why cover
splitting is a likely bond failure mode for an NSM system.
Figs. 4b–c illustrate how the bond mechanism of an NSM
system can be modelled in the plane perpendicular to the
bar axis, as further explained later. These figures also clarify
 the difference in bond mechanism between round bars

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
Fig. 3. Bond failure modes of NSM systems observed in bond tests.

and strips. In the latter case, the normal component of the
bond stresses is transverse to the thick lateral sides of
the groove [13,25] so that splitting failure is less likely to
occur.

The different patterns of cover splitting failure of NSM
systems are shown in Fig. 3. When the k ratio is very low

(e.g. specimens in Ref. [21] with k= 1.12–1.18), failure is
limited to the epoxy cover and involves little damage in
the surrounding concrete (SP-E failure). For higher values
of k, failure involves a combination of longitudinal cracking
 in the epoxy cover andfracture of the surrounding concrete
 along inclined planes (SP-C-1 failure); concrete
fracture starts as soon as the epoxy cover cracks and the
tensile stresses are redistributed [26]. The inclined fracture
planes in the concrete have been observed to form an angle
(b
in Fig. 3)of approximately 30with the horizontal. For

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
Fig. 4. Schematics of the bond behaviour of NSM FRP reinforcement: (a) bond stresses in the longitudinal plane; (b) normal stresses in the transverse
plane generated by a round bar; (c) normal stresses in the transverse plane generated by a rectangular bar.

large groove depths and/or when the tensile strength ratio
between concrete and epoxy is small, fracture of concrete
may occur before the epoxy crack has reached the external
surface (SP-C-2 failure).

The bond failure modes discussed above are for an
NSM bar located centrally in a wide member, where
edge effects are unimportant. When an NSM bar is close
to the edge of a concrete member, failure involves the
splitting of the edge concrete (SP-ED failure) [16].In tests,
this failure mode was found to occur when a0 <
20 mm

e

[16], with the angle b0 defined in Fig. 3 ranging from 45
to 70.

The bond strength associated with the SP-E mode is
expected to depend strongly on the tensile strength of the
epoxy, whereas those associated with the SP-C-1 and SPC-
2 modes are expected todepend strongly on the concrete
tensile strength. In all cases, the bond strength is expected
to increase with the cover thickness of the NSM bar (i.e.
the groove depth). However, the rate of strength increase
with the groove depth has been observed to reduce after
the groove depth exceeds about 2 times the diameter,
around which the SP-C-2 mode replaces the SP-C-1 mode
as the critical bond failure mode. The bond strength of
the SP-E mode also increases as the surface deformations
become less pronounced [26].

3.3. Effect of groove detailing on bond performance
The effect of the groove width-to-depth ratio on the
bond performance has not yet been investigated in detail.
Through finite element modelling, the debonding load of
an NSM strip was found to increase with the groove width

[17]. As the groove depth was kept constant, an increased
groove width implied a larger interfacial area between
epoxy and concrete. This in turn implied a larger debonding
 load since debonding was assumed to occur by cohesive
shear failure in the concrete at the epoxy–concrete interface
[17].

By simplified analytical modelling for deformed round
bars [25], the cracking load of the epoxy cover was found
to decrease with an increase in the groove width-to-depth
ratio (for a given depth) but the failure load was found
to remain substantially unchanged due to failure being in
the concrete along fracture surfaces nearly independent of
the groove width-to-depth ratio. The first result was confirmed
 by Hassan and Rizkalla [19] through finite element
modelling. It was also found that the tensile stresses in the
concrete decrease with an increase in the groove width,
which implies a larger concrete cracking load (but not necessarily
 a larger failure load). However, experimental evidence
 on the effect of the groove width-to-depth ratio on
bond performance is still lacking.

From bond tests on NSM strips, Blaschko [13,16]

indicated that a minimum a0ea0.tbg =2T of about

ee ae

20 mm was required to avoid a splitting failure of the con

crete corner, and for a0 e values larger than 30 mm, no cracks
were observed in the concrete at bond failure. He suggested
that a0 e be no less than 30 mm or the maximum aggregate
size, whichever is greater. The maximum aggregate size
was suggested as a limit to avoid damaging the concrete
during the cutting of the groove. In his bond tests, ae 0 did
still influence the bond behaviour until the maximum investigated
 value of 150 mm, beyond which no further influence
 was assumed.

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
Based on finite element modelling for round deformed
bars, Hassan and Rizkalla [19] suggested a minimum ag
and a minimum ae of two and four bar diameters respectively,
 regardless of the groove width. However, one of the
beams tested by De Lorenzis [11], which was strengthened
with NSM spirally wound round bars with ag =30mm

(i.e. about 1.8 times the groove size and 3.6 times the bar
diameter) and ae =69mm (i.e. about 4.3 times the groove
size and 8.6 times the bar diameter), failed by debonding
of the NSM bars involving the spalling of the concrete cover
of the longitudinal steel reinforcement along the edges. This
test thus suggested that the minimum values for ag and ae as
specifiedinRef. [19] are insufficientto eliminate interactions
between an NSM bar and the edge of a beam.
3.4. Local bond strength
3.4.1. Experimental results
In anytype of bond test, the average bond strength usually
 decreases with increases in the bond length, as a result
of the non-uniform distribution of bond stresses (Fig.4a).
The local bond strength refers to the maximum value of
bond stress that the interface can resist, in contrast to the
overall bond strength (or simply bond strength as used in
this paper) which refers to the maximum transferable load
of the joint. The local bond strength must then be obtained
either from very short specimens or from a long specimen
by elaborative strain (and/or slip) measurements.

Several authors studied the local bond strengths of
NSM systems [13,27,16,20,21,25]. A summary of their values,
 reported in Table 2, allows the following observations
to be made:

 the local bond strength of the cover splitting mode
(deformed bars), as expected, is higher if the groove
depth is larger and the bar surface deformation less
pronounced;
 the local bond strength of the bar–epoxy interfacial (BEI)
 failure mode, which was observed for sand-blasted
bars, is not influenced by the groove size and is lower
than that for deformed bars;
 the local bond strengths of NSM strips from two test
series by different authors [14,16] are very close to each
other and are comparableto thatof spirally woundbars
[26]; by contrast, the local bond strength obtained from
a third test series [20] is notably lower.
Available information on the bond behaviour of square
bars is still very limited. Nordin and Taljsten [9] reported
average bond strengths from specimens whose lengths were
at least 10 times the width of the square cross-section but
no local bond strengths were reported. An unusual aspect
of the results reported by these authors is that the average
bond strength increased with the bond length.

De Lorenzis [25] reported the local bond strengths at the
epoxy–concrete interface for specimens with different types
of bars tested by De Lorenzis et al. [23]. As EC-I failure

was the critical mode, the surface configuration of the
bar did not have a significant effect on the bond behaviour
and the difference in local bond strength between ribbed
and spirally wound bars was basically due to the different
diameters, and hence to the different groove size corresponding
 to a given k value. The local bond strength was
found to decrease with an increasing groove size.

3.4.2. Theoretical models for NSM strips
It is interesting to compare the experimental local bond
strengths of NSM strips reported by Sena Cruz and Barros

[27] with the predictions by the formula proposed by
Blaschko [13], and with those given by Hassan and Rizkalla’s
 theoretical model [17]. Blaschko’s formula [13] is given
by:
pffiffiffiffiffi

4a00

smax . 0:2e saf eae 6
150 mmTe1T
where saf is the shear strength of the epoxy. Hassan and

Rizkalla’s formula [17] is given by:

f
0

fct smax

. c e2Tf
c 0tfct

where f
c 0 and fct are the (cylinder) compressive and tensile
strengths of concrete, respectively. The two formulae relate
the local bond strength to different parameters, consistent
with their own experimental observations: Blaschko observed
 cohesive shear failure in the epoxy and studied the
effect of ae0, whereas Hassan and Rizkalla observed cohesive
 shear failure in the concrete (hence, their value of smax
is the shear strength of concrete). The followingdifferences
between the two formulae should also be noted: (a) Blaschko
 performed pull-out bond tests to provide the experimental
 basis, while Hassan and Rizkalla conducted
flexural tests on RC beams embedded with bars of varying
lengths; (b) Blaschko’s formula was calibrated with bond
test results, while Hassan and Rizkalla’s formula was derived
 from Mohr’s circle for the pure shear stress state,
which, when used in finite element modelling, yielded predictions
 of the debonding load in good agreement with test
results.

The 95 percentile characteristic value of saf was indicated
 by Blaschko [16] to vary between 20 and 25 MPa
for common highly filled, two-component epoxies. According
 to the tests by Blaschko, the ratio between the characteristic
 and the average values of saf is about 0.89, hence the
average value of saf of common epoxies can be assumed to
vary between 22.5 and 28.1 MPa. For a0 e . 150mm (i.e.
with no edge effect), Eq. (1) thus yields a local bond
strength ranging between 15.8 and 19.8 MPa, whose upper
bound practicallycoincides with the values in Table2 from
tests in [14,16]. This is as expected as Eq. (1) was calibrated
using test results with an average value of 28.7MPa for saf,
very close to the upper bound of the average value range
mentioned above.

For f
c 0 ranging between 20 and 40MPa and taking fct

pffiffiffi ffi

as 0:53 f
c 0 [28], Eq. (2) predicts local bond strengths
between 2.1 and 3.1 MPa. The large difference between

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
the predictions of Eqs. (1) and (2) is a result of the different
materials controlling the failure (epoxy for Eq. (1) and concrete
 for Eq. (2))andthus the different interfaces that these
two formulae correspond to; the concrete shear strength is
much smaller than that of the adhesive.

3.4.3. Theoretical models for NSM round deformed bars
The bond behaviour of the NSM round deformed bars
is controlled by splitting tensile stresses in the epoxy cover
and the surrounding concrete. Figs. 4a and b illustrate the
approach adopted in [11,17] to model this bond behaviour
in the plane perpendicular to the bar axis. Two simplifying
assumptions are common to these two studies. First, the
frictional coefficient l
(=1/tan c)relating bond shear stresses
 and internal splitting pressures is constant, although it
is known to change during the loading process. Second,
the distribution of the radial pressure is uniform, although
the pressures on the thicker concrete substrate are higher
than those on the thinner cover.

De Lorenzis [11] took c
to be 45 (i.e. l
=1), whereas
Hassan and Rizkalla measured it according to ASTM
G115-98 [29]. However, it should be notedthat the concept
of frictional coefficient used in this context is very different
from that defined by the ASTM specification. The latter
pertains to pure frictional properties between materials,
depending on material and surface characteristics, whereas
the frictional coefficient in the bond context is also influenced
 by ‘‘structural’’ variables such as the cover depth
and the bar diameter in the case of internal reinforcing bars

(e.g. [30]). In the case of NSM reinforcement, even more
variables are involved. For steel bars in concrete, Tepfers
and Olsson [31] performed ‘‘ring pull-out tests’’ in order
to estimate the angle c
of bond stresses at different stages
of loading. They established curves relating the coefficient
of friction to the slip for different types of bars. Similar
measurements for NSM bars would require accurate monitoring
 of strains in epoxy and concrete transverse to the
bar axis.
For rounddeformedbars, Hassan and Rizkalla [19] proposed
 a model for cover splitting bond failure, based on
elastic finite element analysis. They provided two formulae
to predict the local bond strengths for the bar–epoxy and
the epoxy–concrete interfaces respectively, and the one corresponding
 to the smaller local bond strength would
control.

Depending on how they are computed, these local bond
strengths correspond to either the first cracking of the
epoxy cover or the first cracking of the concrete adjacent
to the groove. The two formulae are:

fatl
smax bar–epoxy .e3T

G2

fctl
smax epoxy–concrete .e4T

G1

where fat is the tensile strength of epoxy, and G1 and G2 are
coefficients which were evaluatedby finite element analysis
and are dependent on the groove depth-to-bar diameter

and the groove width-to-bar diameter ratios. Graphs for
G1 and G2 were developed. The range of the first ratio
examined by them is between 2.00 and 2.50, while that of
the second ratiois between 1.50 and 2.50. G1 and G2 range
between 0.58 and 1.3 and between 0.5 and 0.72, respectively.
 Hence, Eq. (4) controls in all cases with values of
0.77–1.72 times lfct. For a concrete cylinder compressive
strength of 20–40 MPa, the local bond strength varies between
 1.4 and 5.2l, which are very low compared with
the test results in Table 2, already with l
=1 instead of

0.5 as originally proposed in Ref. [19].
It should be noted that the so-called local bond strength
in this model is the bond shear stress corresponding to the
initiation of cracking in the epoxy or in the concrete,
whereas the joint can still sustain significant increments
of the appliedload between first cracking andultimate failure.
 Also, the distinction between cracking of epoxy and
cracking of concrete as two independent modes of bond
failure contradicts the experimental evidence that failure
in the concrete usually follows cracking of the epoxy.

An approximate two-dimensional elastic stress analysis
was conducted by De Lorenzis [25] to determine the bond
shear stress corresponding to the cracking of the epoxy
cover. For computation of an upper and a lower bound
to the ultimate radial pressure of the NSM system (and
hence to the local bond strength if the frictional coefficient
is known), different possible failure modes were analyzed.
Auniform tensile stress distribution along the fracture lines
was assumed and justified on the basis of redistribution of
cohesive stresses between the crack faces. The experimental
values of local bond strength were shown to fall within the
computed range. However, no equation for the local bond
strength was proposed.

3.4.4. Theoretical models for NSM bars in pre-formed
grooves
The local bond strength of the epoxy-to-concrete interface
 for pre-formed grooves was shown to decrease almost
linearly with increases in groove size. An equation is given
in Ref. [25] that may be used to compute the local bond
strength for different groove sizes for this case.

3.4.5. Comparison with internal FRP rebars and externally
bonded FRP laminates
Cosenza et al. [32] presented a summary of bond properties
 of FRP bars used as internal reinforcement, obtaining
 an average local bond strength of 2.74 MPa for
sand-blasted bars (with a COV of 52%) and of

11.61MPa for ‘‘ribbed bars’’ (with a COV of 34%). The
local bond strength value for sand-blasted bars is the average
 of results from Ref. [33], conducted on only one type of
bars different from those used by De Lorenzis and Nanni
[21], hencea direct comparisonis not meaningful. The category
 of ‘‘ribbed bars’’ referred to by Ref. [32] encompasses
a wide variety of material and surface configurations as it
covers what are referred to as ribbed bars and spirally wound bars in the present paper, plus some cross-wound

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
bars. The local bond strength values for spirally wound
bars from Ref. [32] vary between 4.76 and 18.04 MPa, with
an average of 11.9 MPa, which is lower than what has been
obtained for NSM systems(Table2). In bond tests of spirally
 wound bars as internal reinforcement, significant
interlocking was not evident, but when used as NSM reinforcement,
 these bars normally produce cover splitting
bond failure before the loss of bond at the bar–epoxy interface.
 The higher local bond strength for an NSM system
can be attributed to a larger resistance at the bar–epoxy
interface than at the bar–concrete interface.

Nanni et al. [34] conducted bond tests on GFRP ribbed
bars of the same type as used by De Lorenzis et al. [26]
(although bars in this study had a larger diameter, equal
to 12.7 mm), and found a local bond strength of 17 MPa
for them as internal reinforcement, which is larger than
the value obtained for an NSM configuration. The bond
failure of these rebars as internal reinforcement occurred
by the shearing-off of the ribs, a mode which was
approached but never attained in NSM bond tests as cover
splitting bond failure always occurs first.

For FRP laminates externally bonded to concrete, Lu
et al. [35] proposed a simple equation for the local bond
strength. Assuming an FRP-to-concrete width ratio of
0.5, this equation yields a local bond strength of 1.5 times
the tensile strength of concrete, which is significantly lower
than the cover splitting local bond strength of NSM bars.
This results from the fact that, in the cover splitting bond

failure mode of an NSM system, the fracture of concrete
relates to the larger perimeter (see Fig. 3, mechanisms
SP-C1 and SP-C2), while the nominal bond strength is
defined using the smaller bar perimeter. In the debonding
failure of an externally bonded laminate, the fracture plane
has approximately the width of the laminate and the nominal
 bond strength is based on the same width.

3.5. Local bond–slip behaviour
Local bond–slip curves were deduced from test data by
De Lorenzis [25] for different types of NSM round bars,
and by Sena Cruz and Barros [27] for NSM strips. De
Lorenzis [25] reported three different types of local bond–
slip behaviour (types I–III), of which the first two are
shown in Fig. 5. The third type (for sand-blasted round
bars) differs from the secondin that the abrupt decay from
the maximum bond stress to the frictional bond stress level
is replaced by a linearly decreasing branch. This third type
could be seen as a special case of the second type. The
equation proposed by Sena Cruz and Barros [27] has the
same form as that of the type I curve shown in Fig. 5.

The type I equation seems to reproduce rather accurately
 experimental curves showing a gradual decrease of
local bond stress after the peak. Such a gradual decrease
exists when bond failure is at an interface (the epoxy–concrete
 or the bar–epoxy interfaces) or by cover splittinggenerated
 by ribbed bars with low rib protrusions. In both

b
Fig. 5. Typical bond–slip curves of NSM FRP reinforcement.

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
cases, a significant amount of post-peak friction develops,
due to interfacial friction in the first case and to aggregate
interlocking of the cracked concrete in the second case.

Conversely, cover splitting failure generated by ribbed
bars with high rib protrusions and spirally wound bars
has a more brittle nature (for a detailed description see
Ref. [26]) with an abrupt decrease in bond stress upon
the attainment of the peak value. However, even after the
complete loss of the cover, a small amount of residual friction
 remains because half of the perimeter of the bar is still
in contact with the epoxy.

The typeI equation, where the bond stress tends to zero
as the slip approaches infinity, cannot reproduce the residual
 frictional branch. This is only a minor drawback if a0 is
larger than 1, as in such cases the bond stress from the
type I equation reaches the frictional plateau only at very
large values of slip and the area underneath the bond–slip
curve is infinite. However, if a0 is smaller than 1, this area
becomes finite, and the joint may be predicted to be unable
to develop the full tensile capacity of the bar (see next subsection),
 which contradicts the better behaviour of this type
of joint compared with type II joints. More work on local
bond–slip curves of NSM systems, both experimental and
theoretical, is still needed, in particular on the post-peak
behaviour which greatly influences the performance of bars
with long embedment lengths such as those used in
practice.

De Lorenzis et al. [26] commented on the variations of
the secant stiffness andthe shape of the curve with test variables.
 In nearly all cases, the slip at peak stress was found
to be in the range of 0.1–0.3 mm. Sena Cruz and Barros

[27] found a value of 0.25 mm for NSM strips.
Based on strain gage and free-end slip readings, Blaschko
 [16] obtained local bond–slip curves of NSM strips,
very similar in shape to the type I curves discussed above.
However, by plotting the local bond–slip curves at different
measurement locations along the bond length, he noted
that the local bond strength tended to be larger close to
the free end and smaller close to the loaded end. This
was attributed to the influence of the transverse displacements
 of the concrete adjacent to the groove, which were
also measured in the same tests. In his analytical model,
the author adopted a local bond–slip curve consisting of
an ascending branch defined by a second-degree parabola
and a horizontal branch at the local bond strength. This
curve was assumed to represent the pure shear stress–strain
behaviour of the epoxy. The local bond strength was taken
as the shear strength of the epoxy, multiplied by an empirical
 function of the transverse displacement of concrete.
The transverse displacements were predicted with a separate
 elastic model. Mohr’s failure criterion was used to
model the effect of local normal stresses, associated with
transverse concrete displacements, on the local bond
strength. Regression analysis was used to maximize the
agreement between measured and predicted strain distributions
 in the strip along the bond length. In summary, his
local bond–slip model considers the response of the NSM

bar and the concrete member as a system so that the influence
 of edge distance can be effectively reflected. However,
the model is rather complex to apply as it requires an iterative
 procedure. Also, it cannot accurately represent the
presence of a frictional asymptote in the local bond–slip
curves obtained from tests and hence it overestimates the
bond failure loads of specimens with long bond lengths.
For this reason, the author relied on the direct calibration
of test results, rather than the output of the bond–slip
model, to obtain Eq. (1).

3.6. Effective bond length, development length
and anchorage length
The maximum stress that can be resisted by a bonded
joint between an NSM bar and the concrete substrate with
a sufficiently long bond lengthis givenby:

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

R

rmax . 2EGf e5T

A

with

Z 1

Gf . sesTds
e6T

0

where Gf, being the area underneath the bond–slip curve, is
the fracture energy of the bonded joint, R
is the perimeter
over which the bond stress acts, A is the cross-sectional
area over which the tensile stress acts, and E is the elastic
modulus of the material on which the tensile stress is applied
 [23]. For round bars bonded to epoxy (or to concrete
in the case of internal rebars), Eq. (5) reduces to

sffiffiffiffiffiffiffiffiffiffiffiffiffiffi

8Eb

rmax . Gf e7Tdb

where Eb is the elastic modulus of the bar, and for laminates
 bonded to concrete (neglecting the thickness of the
adhesive layer), Eq. (5) reduces to

sffiffiffiffiffiffiffiffiffiffiffiffi

2EfGf

rmax .e8T tf

where Ef and tf are the elastic modulus and thickness
respectively of the laminate. If rmax computed by Eq. (5)
is below the tensile strength of the reinforcement, its full
capacity cannot be developed no matter how longthe bond
length is. In this case, a value of bond length exists (‘‘the
effective bond length’’) at which rmax is developed, and beyond
 which a further increase in bond length does not produce
 any benefit. The concept of effective bond length has
been well established for externally bondedFRP laminates
[36,37].If rmax is larger than the tensile strength of the bar
(and in particular for bond–slip curves with infinite values
of Gf), the full capacity of the reinforcement can be developed,
 and the corresponding value of bond length is usually
 termed ‘‘the development length’’.

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
Fig. 6. Idealised bilinear local bond–slip model of FRP laminates
externally bonded to concrete.

For typeI bond–slip curves, Gf has an infinite or a finite
value when a0 P
1or a0 < 1, respectively. For type II
responses, if the post-peak frictional branch is unlimited,
Gf wouldhavean infinite value andthe joint wouldbeable
to reach the tensile capacity of the reinforcement. In the
case of externally bonded laminates, a simplified local
bond–slip model is shown in Fig. 6. The Gf value for such
systems is finite, and its value is usually insufficient for the
reinforcement to develop its full tensile strength; hence no
development length but an effective bond length can be
computed. In the case of internal steel or FRP reinforcing
bars, Gf is usually infinite, due to an unlimited post-peak
frictional branch, hence a development length always
exists. Available evidence for NSM systems shows that
their behaviour is similar to that of internal rebars rather
than that of external laminates. The calibrated values of
a0 given in Refs. [25,27] are larger than 1 in all but one
case, so the drawback of the type I curve outlined in the
previous sub-section has minor practical relevance. Hence,
a development length generally exists for an NSM system.
However, this statement is based on limited experimental
evidence as available bond test results for NSM systems
are still limited.

The development lengthsofNSM roundbarsin saw-cut
grooves computed by De Lorenzis [25] were often impractically
 long. This was partly due to the need for a better calibration
 of the post-peak local bond–slip response, which
displays a significant scatter and often a quite irregular
behaviour particularly in the case of splitting failure. That
is, De Lorenzis [25] may have underestimated the ductility
of the post-peak response, which led to overestimations of
development lengths. For NSM strips, Sena Cruz and Barros
 [27] predicted a development length of about 90 mm,

i.e. less than 10 times the strip height. The failure load-
bond length curves (one for each ae 0 value) proposed by
Blaschko [13,16] are composed of a parabolic portion followed
 by a straight line whose slope is proportional to
the post-peak frictional stress. Full development of the tensile
 capacity of the strip was achieved in the experiments
with a relatively short bond length, which is equal to
approximately 150 mm (7.5 times the strip height) in
absence of edge effects ea0 e P
150 mmT and increases with
a decreasing ae 0 . The good performance of strips results
from the high local bond strength, from the pseudo-ductile
post-peak local bond–slip response and from the large lateral
 surface to cross-sectional area ratio.

From their tests (see the section on flexural strengthening),
 Hassan and Rizkalla [19] found an effective bond
length to exist for NSM round ribbed bars. They computed
this effective bond length by assuming a uniform distribution
 of bond stress equal to the lower of the two local bond
strengths given by Eqs. (3) and (4). However, in deriving
these equations, bond failure was equated to the imminent
cracking of epoxy and concrete, when no significant bond
stress redistribution is expected to have occurred and hence
the assumption of a uniform bond distribution is unjustified.
 Indeed, the assumption of a uniform bond stress distribution
 contradicts the existence of an effective bond
length as the latter implies a limited possibility of bond
stress redistribution. These authors did not provide an
equation for the maximum tensile stress that can be sustained
 by the bar, but suggested to obtain it from testing
or finite element modelling.

The local bond–slip curves of NSM FRP bars have been
used to determine their bond failure load as a function of
the bond length by solving the governing differential equation
 [27,21,25]. It is important to note that, while the local
bond strength smax dictates the ultimate load of a short
NSM FRP-to-concrete joint, the ultimate load of a long
joint depends more on the shape of the local bond–slip
relationship, particularly its post-peak descending branch
which controls the ability of the joint to redistribute stresses
 along its bond length. The shortest bond length
required for an FRP bar to resist a given load is herein
termed the anchorage length for that load. Note that the
anchorage lengthdepends on the load level, and obviously,
the anchorage length corresponding to the FRP bar rupture
 failure load is equal to the development length. In
design, it is also important to ensure that the anchorage
length provided satisfies serviceability requirements: the
state of the bond stress along the bar should stay within
the ascending portion of the bond–slip curve and the free
end of the bar should not show slip under service loads.
Both De Lorenzis [25] and Sena Cruz and Barros [27]
provided curves of load at onset of free-end slip versus
bond length and outlined a design procedure complying
with both ultimate and serviceability limit state
requirements.

3.7. Bond test methods
The most common types of bond tests used for NSM
reinforcement are the direct pull-out test and the beam
pull-out test.While detailed descriptions of the various test
arrangements can be retrieved from the original papers
(see Table 1), some of the issues of concern are discussed
below.

A number of practical disadvantages exist with beam
pull-out tests [14,21]. For example, the specimen size is

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
large, especially if long bond lengths are tested; it is difficult
 to conduct the test in slip-control mode; and it is
difficultto visually inspect the behaviour of the joint during
loading, especially the initiation and propagation of
cracks.

Direct pull-out tests overcome the drawbacks of beam
pull-out tests mentioned above. The simplest direct pullout
 test specimen may be composed of a square/rectangular
 concrete block embedded with an NSM bar on one of
the sides, but in this set-up, the NSM bar leads to eccentric
loading of the concrete block. The use of two bars on two
opposite sides [38] or even four bars on all four sides [3] has
been attempted to overcome this problem. The multiple-
bar specimen has its own problem: any small deviations
of the groove/bar positions can induce flexural effects,
significantly altering test results.

De Lorenzis et al. [23] introduced a C-shaped
block where a single NSM bar was placed at the centre
of gravity of the block. The set-up performed well, but
the specimen dimensions had to be specifically designed
for each groove depth. This set-up is also not suitable for
studying edge effects due to the presence of two thick
flanges.

Based on the above discussions, a direct shear test on a
single NSM FRP bar embedded in a concrete block, where
the tensile force applied on the bar is reacted by compressive
 stresses on the concrete block at the loaded end, is
probably a good choice that combines simplicity with reliability.
A similar test set-up has been popular in studies on
externally bonded laminates (e.g. [39]). Blaschko [16] used
such a set-up, in which a steel plate was used to provide
the reaction to the concrete block. The steel plate had a
central hole of 80-mm diameter to avoid reactive stresses
on the immediate vicinity of the groove.

To minimize the transverse friction generated by the
bearing pressure, which could delay the initiation of splitting
 cracks as generally observed in pull-out tests of steel
rebars in concrete, layers of PTFE or similar materials
can be placed between the bearing plate and the concrete
block.

Three main methods are available for obtaining the
local bond–slip curve of a bonded joint:

 Approximate it with the curve relating the average bond
stress to the loaded-end slip (or the free-end slip, or the
average of the two) from specimens with a short bond
length (SBL).
 Obtain bond stresses and slips from free-end slip and
strain measurements at discrete points along the bond
length. This method is usually adopted when long bond
length (LBL) specimens are used, as strain gages on a
SBL specimen are likely to significantly affect the bond
performance.
 Calibrate the unknown parameters in the local bond–
slip equation, whose form needs to be known or
assumed in advance, from loaded-end slip and free-end
slip measurements.
Each of the above methods has its advantages and disadvantages.
 The first method does not require the use of
strain gages, which simplifies specimen preparation and
does not alter the bond properties between the bar and
the epoxy. However, the bond length still needs to be long
enough for the specimen to behave as a representative sample
 (e.g. for ribbed or spirally wound bars, a minimum
number of deformations should be included)and to reduce
the influence of end effects. The bond–slip curve obtained
represents the average performance of the chosen bond
length of the specimen.

The second method, more onerous and altering to
some extent the bar–epoxy interface, has the advantage
that the bond performance over a longer and hence more
representative portion of the reinforcement can be
studied. A local bond–slip curve can be retrieved at each
load level, or alternatively at each measurement point, so
that more data are obtained from such a test than from
one on a SBL specimen. It is worth noting that even
for tests conducted with load control on LBL
specimens, strain measurements allow the descending
branch of the local bond–slip relationship to be obtained
as a result of stress redistribution over the bond
length, although these measurements tend to show a notable
 scatter, due to damage to the strain gages resulting
from slips of the reinforcement. Teng et al. [20] presented
tests where strain gauges were sandwiched between two
CFRP strips so that the strain gauges were well protected
and did not affect the interfacial properties. Wang et al.

[40] recently explored the use of fiber optic sensors
embedded inside FRP bars for the measurement of
strains so that the use of strain gauges on the bar surface
 can be avoided. These and similar techniques
for strain measurement are highly desirable for bond
tests on LBL specimens to minimize damage to the
strain sensors and interference with the interfacial
behaviour.
The third method has the advantage that it allows the
local bond–slip curve to be obtained from LBL specimens
without the need for strain measurements. The disadvantages
 are that the form of the bond–slip equation must
be known a priori, and the accuracy of the deduced equation
 may be compromised if the assumed form is inappropriate
 in some way. If many different forms of equations
are tested to find the most suitable form, this approach
involves a much more onerous process than the other
two approaches.

In all cases, the specimen must be carefully designed
to ensure that the failure mode and load are not significantly
 influenced by the specimen dimensions. An example
of inadequate specimen size can be found in Ref. [3],
which reports shear fracture failure of the concrete involving
 the entire specimen cross-section due to its limited
section size. When the NSM bar is close to the edge of
the concrete block, test results have indicated a strong
influence of the transverse stiffness of the concrete on the
measured bond–slip curve [13], which then also varies

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
along the bond length as a result of the variation of the
bond stresses. This effect should either be a parameter to
be considered or be eliminated by the use of a sufficiently
wide specimen.

In addition to section size, the bondedportion of the bar
should start at a suitable distance from the loaded end of
the concrete block.Ifthis distance is insufficient, the behaviour
 of a specimen with a short bond length may be similar
to that of a fastener (see e.g. [38]), or the failure mode of
the specimen is unduly influenced by the cracking of concrete
 at the loaded end [16].

4.
Flexural
strengthening
4.1. Summary of existing work
Asummary of the existing experimentaldatais reported
in Tables 3 and 4.

Some existing studies were conducted on beams
strengthened with NSM bars of limited embedment lengths
(Table3). Although such tests were intended to study bond
failure mechanisms, they are not ‘‘pure’’ bond tests as the
bond performance is affected by flexural cracking. Moreover,
 the NSM FRP bars in such tests generally extend into
the shear spans, where part of the interfacial shear stress is
directly dependent on the transverse shear force in the
beam.

Hassan and Rizkalla [17,19] conducted flexural tests on
RC beams with NSM CFRP round ribbed bars and strips
of varying embedment length. Failure of beams with
NSM round ribbed bars occurred by splitting of the concrete
 cover followedbythe complete debonding of the bars
in all cases. These authors concluded that the tensile rupture
of this type of bars is unlikely to occur, regardless of the
embedment length, that the maximum usable strain of these
bars should be limited to 0.7–0.8%, and that the anchorage
length should not be shorter than 800 mm. In the case of
beams with NSM strips, rupture of the strips occurred when
the embedment length was larger than 850 mm.

Teng et al. [20] conducted flexural tests on RC beams
with NSM strips of varying embedment length. As the
embedment length increased, the failure mode changed
from concrete cover separation starting from the cut-off
section, to concrete crushing followed by secondary cover
separation close to the maximum moment region. In the
beams with the two longest embedment lengths, secondary
debonding mechanisms were also observed.

All existing test results of strengthened beams, slabs, and
columns(Tables3 and4)indicate that the NSM reinforcement
 improved the ultimate load and the load at the yielding
 of steel reinforcement, as well as the post-cracking
stiffness. Some test programs included identical beams
strengthened with equivalent amounts of FRP provided
as either externally bonded or NSM reinforcement. In all
cases, the NSM reinforcement performed more efficiently,
as debonding of the NSM reinforcement occurred at a
higher strain or did not occur [17,41–43].

One study [41] has compared equivalent amounts of
NSM reinforcement provided as round bars or strips. As
expected, strips performed better and failed by tensile rupture
 as opposed to debonding of the round bars, as a result
of the higher local bond strength and larger lateral surface
to cross-sectional area ratio of NSM strips.

4.2. Failure modes of flexurally-strengthened beams
The possible failure modes of beams flexurally-strengthened
 with NSM FRP reinforcement are of two types: those
of conventional RC beams, including concrete crushing or
FRP rupture generally after the yielding of internal steel
bars, for which the composite action between the original
beam and the NSM FRP is practically maintained up to
failure, and ‘‘premature’’ debonding failure modes which
involve the loss of this composite action. Although debonding
 failures are less likely a problem with NSM FRP
compared with externally bonded FRP, they may still significantly
 limit the efficiency of this technology.

The likeliness of a debonding failure depends on several
parameters, among which the internal steel reinforcement
ratio, the FRP reinforcement ratio, the cross-sectional
shape and the surface configuration of the NSM reinforcement,
 and the tensile strengths of both the epoxy and the
concrete. Some researchers [11,10] extended the NSM
FRP reinforcement over the beam supports to simulate
anchorage in adjacent members. Despite this anchorage,
debonding failures can still occur [11]. The beam reported
in Ref. [10] failed by FRP rupture, as opposed to debonding
 observed in an identical beam with the NSM reinforcement
 terminated away from the supports. Blaschko [16]
reported the results of two beam tests: the first one failed
by concrete cover separation starting from the cut-off section
 but the second beam, which was provided with a steel
U-jacket bondedto the cut-off section, failedby the rupture
of the FRP strips. In the same study it was observed that
fatigue loading to two million load cycles did not affect
the residual beam capacity.

There is still limited understanding of the mechanics of
debonding in beams strengthened with NSM systems.
Descriptions of failure modes in the existing literature are
often not sufficiently detailed to understand the progression
 of the failure process. Based on the available experimental
 evidence, the possible failure modes of beams
flexurally-strengthened with NSM FRP reinforcement are
classified in Fig. 7 and described below. The interactions
between the main failure modes described below and the
‘‘secondary’’ failure modes are still unclear and deserve further
 investigations.

4.2.1. Bar–epoxy interfacial debonding
This mode involves interfacial debonding between a bar
and the epoxy and hasbeen observed for sand-blasted round
bars [44]. This mode correlates well with the failure mode
observed in bond tests on the same type of bars (see the
previous section). However, unlike in a bond specimen, the

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
Fig. 7. Debonding failure modes of NSM bars and strips observed in tests on flexurally-strengthened beams: (a) debonding at the bar–epoxy interface;(b)
separation of concrete cover between two cracks in the maximum moment region; (c) separation of concrete cover over a large length of the beam; (d)
separation of concrete cover starting from a cutoff section; (e) separation of concrete cover along the edge; (f) secondary loss of bond between epoxy and
concrete; (g) secondary splitting of the epoxy cover.

epoxy cover in the beam was intersected by flexural ting cracks and hence accelerated interfacial debonding
cracks which facilitated the initiation of longitudinal split-(Fig.7a).

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
4.2.2. Concrete cover separation
The formation of bond cracks on the soffit of the
beam has been observed in tests [11,16,20], and these bond
cracks are inclined at approximately 45 [20] to the beam
axis. Upon reaching the edges of the beam soffit, these
cracks may propagate upwards on the beam sides maintaining
 a 45 inclination within the cover thickness, and
then propagate horizontally at the level of the steel tension
bars. Debonding may next occur in different forms,
depending on the subsequent evolution of the crack
pattern:

(a) Bar end cover separation. If the NSM FRP reinforcement
 is terminated at a significant distance from the
supports, separation of concrete cover typically starts
from the cut-off section and propagates inwards
[16,20] (Fig. 7d). This mode is similar to the cover
separation failure mode observed in RC beams with
an externally bonded FRP laminate [37,45,46].
(b) Localized cover separation. Bond cracks within or
close to the maximum moment region, together with
the pre-existing flexural and flexural-shear cracks,
may isolate triangular or trapezoidal concrete
wedges, of which one or more are eventually split
off(Fig.7b). This mode can be identified from photos
of failed beams given in Refs. [11,15,20].
(c) Flexural crack-induced cover separation. Separation
of the concrete cover occurs almost simultaneously
over a long portion of the NSM reinforcement, often
involving one of the shear spans and the maximum
moment region (Fig. 7c) [15,44]. This mode was
observed by De Lorenzis et al. [44] to start from the
maximum moment region, whereas the location of
initiation was not made clear in Ref. [15]. This mode
is similar to the intermediate crack-induced debonding
 failure mode observed in RC beams with an externally
 bonded FRP laminate [37,47,48].
(d) Beam edge cover separation. NSM bars located near
the edges may generate detachment of the concrete
coveralong the edges(Fig.7e).
4.2.3. Epoxy–concrete interfacial debonding
For beams with NSM strips of a limited embedment
length, Hassan and Rizkalla [17] reported cohesive shear
failure in the concrete at the epoxy–concrete interface starting
 from the cut-off section. Unfortunately no picture of
the failed specimens was provided. This mode is believed
to be similar to the plate end interfacial debonding failure
of RC beams with externally bonded FRP laminate
described in Refs. [37,45,46].

4.2.4. Secondary debonding failure mechanisms
Other debonding mechanisms have also been observed.
They are herein classified as ‘‘secondary’’ failure modes
and the role they play in the context of debonding failures
is still unclear. It has been observed [20] that upon the formation
 ofthe bond cracks, the opening-up of these inclined

bond cracks was restrained by the dowel action of NSM
reinforcement which in turn tended to cause the detachment
 of the NSM FRP reinforcement from the soffit of
the beam. After failure, the prism formed by the CFRP
strip and surrounding epoxy was found to retain a thin
concrete layer of variable thickness on the sides(Fig.7f),
indicating that a strong epoxy–concrete bond existed.
Moreover, localized splitting occurred in the epoxy cover,
exposing the internal CFRP strip(Fig.7g). Similar observations
 had been reported previously [11].

4.3. Prediction of ultimate loads and load–deflection
behaviour
For the safe design of an NSM FRP system for the flexural
 strengthening of an RC beam, the foremost issue is the
prediction of the ultimate load. If the failure of a strengthened
 beam does not involve debonding, then the failure
load can be easily predicted using equations developed
for externally bonded FRP based on the plane section
assumption (e.g. [37]) and with the difference in position
between the two types of reinforcement duly taken into
account. Accurate predictions of debonding failure loads
are much more challenging.

In the past few years, numerous studies have examined
debonding failures of beams with externally bonded steel
plates and FRP laminates [37,45–49]. Research on RC
beams flexurally-strengthened with NSM FRP has been
much more limited. The few theoretical models developed
so far are extensions of approaches developed for externally
 bonded FRP laminates. For instance, Hassan and
Rizkalla [17] proposed a theoretical model for NSM strips
(valid only for concrete shear failure at the epoxy–concrete
interface at the cut-off section) which is an extension of the
interfacial stress-based approach proposed by Malek et al.
[45]. The model has been compared with a very limited
database. Moreover, the failure mode assumed in the
model of Hassan and Rizkalla [17] was only observed in
the tests by these authors but has not been observed in
other tests.

Of the reported failure modes, the most critical appears
to be the mode of concrete cover separation starting from
the maximum moment region. Provided that bars with a
reasonable degree of surface deformation are used, failure
at the bar–epoxy interface is unlikely. Moreover, in cases
where the NSM reinforcement is needed over the entire
length of a member, NSM bars can be easily anchored
to adjacent members so that debonding failure at a cutoff
 section can be prevented. For similar reasons, the
intermediate crack-induced debonding failure mode has
been recognised as the most important mode for RC
beams strengthened with externally bonded FRP laminates
 [47].

For design purposes, the simplest approach to the prediction
 of debonding is to establish a bond reduction factor
for the ultimate tensile strain of the reinforcement. Such an
approach is currently adopted by ACI-440 [6] for debond

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
ing of externally bonded laminates. The predictive models
in Refs. [47,48] also follow this approach. Alternative
approaches [5,50] for this type of debonding failure of
externally bonded FRP laminates consider stress gradients
in the FRP between two adjacent cracks instead of a simple
bond reduction factor [51].

Development of reliable predictive models for debonding
 failures requires a thorough understanding of the
mechanics of debonding failures, and of the qualitative
and quantitative roles of relevant variables. The most
challenging aspect in tackling this problem appears to
be the lack of a direct correlation between the bond failure
 modes in bond specimens and the debonding failure
modes in flexurally-strengthened beams. The possible
reasons are the presence of flexural and flexural-shear
cracks which alter the bond stress distribution, the curvature
 of the beam, the dowel action of the FRP bars
restraining the opening-up of inclined bond cracks, phenomena
 which are all absent in a bond specimen. The
same problem has been encountered in predicting intermediate
 crack debonding failures in RC beams with
externally bonded FRP laminates, although to a lesser
extent [48].

The load–deflection behaviour of beams strengthened
with NSM reinforcement can be predicted with reasonable
 accuracy by the conventional sectional approach
neglecting tension stiffening and assuming a perfect bond
for both steel and NSM reinforcement [11,16]. More
refined approaches where tension stiffening is taken into
account (with different laws for un-strengthened and
strengthened beams) [15] and where slips of steel and
FRP reinforcement are modelled using experimentally
determined bond–slip equations [52] have delivered more
accurate predictions of experimental load–deflection
curves.

4.4. Strengthening with pre-stressed NSM FRP
NSM FRP bars or strips can be more easilypre-stressed
and anchored than externally bondedlaminates, so flexural
strengthening with pre-stressed NSM FRP is a promising
technique. Nordin and Taljsten [53] have explored the use
of this technique by tensioning NSM bars to 20% of their
tensile strength, filling the grooves with epoxy, and releasing
 the pre-stressing force upon hardening of the epoxy.
The expected gains in the cracking load and the stiffness
of the beam as a result of pre-stressing were achieved,
and the failure mode was the tensile rupture of the FRP
bars in all cases. The method used by these authors cannot
yet be implemented in a real strengthening project as their
procedure of tensioning and anchoring the bars requires
access to the ends of the beam, which is generally not possible
 in reality. For this reason, a tensioning–anchoring
device for NSM bars was proposed by De Lorenzis et al.
[54].

5.
Shear
strengthening
5.1. Summary of existing work
The use of NSM FRP reinforcement is also effective in
enhancing the shear capacity of RC beams. For this purpose,
 the bars are embedded in grooves cut on the sides
of the member at a desired angle to the beam axis.

Only three studies appear to have been published on the
use of NSM FRP bars for shear strengthening of RC
beams. De Lorenzis and Nanni [55] carried out eight tests
on large size T-beams, of which six had no internal stirrups.
CFRP ribbed round bars in epoxy-filled grooves were used
as NSM shear reinforcement. The test variables included
bar spacing, inclination angle and anchorage of the bars
in the flange. The NSM reinforcement produced a shear
strength increase which is as high as 106% in the absence
of steel stirrups, and still significantin presenceofa limited
amount of internal shear reinforcement.

Barros and Dias [56] tested beams of different sizes with
no internal stirrups. Some of these beams were strengthened
 with NSM CFRP strips ofdifferent inclinations, while
the rest were strengthened with equivalent amounts of
externally bonded FRP shear reinforcement. The reported
strength increases ranged from 22% to 77%, and werein all
cases larger than those obtained with externally bonded
FRP. Although failure modes were not described, based
on the reported load–deflection curves, at least some of
the beams are believed to have failed in bending.

Nanni et al. [57] reported the test results of a single full-
scale PC girder taken from a bridge and shear-strengthened
with NSM CFRP strips. The beam failed in flexure at a
shear force close to the shear resistance predicted by the
model given in Ref. [55].

5.2. Failure modes
Two different failure modes were identified by De
Lorenzis and Nanni [55]. The first was debonding of the
FRP bars by splitting of the epoxy cover and cracking of
the surrounding concrete, associated with the diagonal tension
 failure of concrete(Fig.8a). This failure mode may be
prevented by providing better anchorage of the NSM bars
crossing the critical shear crack, by either anchoring the
bars in the beam flange or the use of inclined (e.g. 45)bars
atasufficiently close spacing to achievea longer total bond
length. Once this mechanism was prevented, separation of
the concrete cover of the steel longitudinal reinforcement
became the controlling failure mode in the tests presented
in Ref. [55](Fig.8b). Unlike internal steel stirrups, NSM
shear reinforcement does not exert a restraining action on
the longitudinal reinforcement subjected to dowel forces.
These forces, in conjunction with the normal pressures generated
 by the bond action of the steel longitudinal reinforcement,
 create considerable tensile stresses in the cover

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
Fig. 8. Debonding failure modes of NSM bars observed in tests on shear-strengthened beams: (a) debonding of NSM bars by splitting of epoxy cover; (b)
local separation of concrete cover.

which may eventually lead to cover separation failure. This
second mode, however, may be attributed to the fact that
no or very limited steel stirrups were present in these
beams, and is unlikely in beams with a significant amount
of steel stirrups. The most important failure mode is thus
debonding of the FRP bars. Although it has not been
observed so far, tensile rupture of the NSM reinforcement
is another possible failure mode.

5.3. Prediction of ultimate loads
The truss analogy was used by De Lorenzis and Nanni

[55] to compute the shear capacity of a member strengthened
 in shear with NSM FRP reinforcement, and in particular
 the load at diagonal tension failure of concrete
involving debonding of the NSM bars. The basic assumption
 of their approach is that, at the instant of failure,
the bond stresses are evenly distributed along the bars
crossed by the critical shear crack, and are equal to the
local bond strength. This assumption is acceptable if the
bond–slip behaviour is ductile enough to allow a substantial
 redistribution of bond stresses along the bars crossed
by the shear crack. This basic assumption yields easily
the tensile stresses in the bars crossed by the shear crack
and hence the corresponding shear force. This approach
was shown to compare favourably with test results
[55,57]. Further research is obviously needed for the assessment
 and improvement of the model.
Unlike the case of flexurally-strengthened beams, existing
 evidence indicates that the debonding failure mode of
NSM bars in shear-strengthened beams is similar to the
bond failure mode of the same bars in bond specimens.
Further work is needed to confirm this observation as it
is important for the modelling of debonding failures of
shear-strengthened RC beams. If confirmed, this observation
 means that local bond–slip relationships developed
from bond tests can be directly used in predicting debond

ing failures of RC beams shear-strengthened with NSM
FRP bars. With such an approach, the truss model [55]
can be easily generalized to the case of debonding failure
by incorporating an appropriate bond–slip curve instead
of the simple ideally plastic curve assumed in the original
model. Similarly, local bond–slip relationships obtained
from bond tests can also be directly used in the numerical
modelling of debonding failures.

6.
Strengthening
of
beam–column
joints
Recently, Prota et al. [58] proposed the combined use of
FRP laminates and NSM bars for upgrading RC beam–column
 connections. NSM bars were installed on the column
prior to wrapping and anchored through the beam (i.e.
the bars passed through the beam). This provided additional
 reinforcement which is fully anchored and effective
in the maximum moment region of the column. The presence
 of FRP wraps prevented the NSM reinforcement from
becoming ineffective as a result ofload reversals. The installation
 of NSM bars enabled the transition of the failure
mode from the column to the shear failure of the joint. In
further specimens, additional strengthening was provided
to the joint to suppress joint shear failure. For this purpose,
the FRP reinforcement was placed in two directions: (a)
along the beam axis, NSM FRP bars were provided and
transverselyconfined by single-plyCFRPU-jackets bonded
to the beams; (b) along the column axis, a one-ply FRP laminate
 was provided, and this laminate was terminated below
the base of the upper column to simulate the field condition
of the presence of a slab. With this strengthening scheme,
failure shifted to the column–joint interface at the termination
 of the FRP laminate. The upgrading of the joint zone
increased its deformability andhence provided a significant
contribution to the ductility of the system.

This study [58] showed that the combination of
NSM bars with externally bonded laminates enabled the

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
advantages of both techniques to be exploited in a complementary
 manner. This topic deserves further investigation
as similar advantages may be realised by suitable combinations
 of the two techniques in solving other strengthening
problems.

7.
Research
needs
This paper has provided a comprehensive and critical
assessment of existing research on the structural behaviour
of RC structures strengthened with NSM FRP reinforcement.
 On the basis of this review, the main research needs
in this area for the immediate future are outlined as
follows.

7.1. Bond behaviour
There is a lack of experimentalevidence on the effects of
the many variables that are likely to have a significant effect
on the bond behaviour of NSM FRP bars. The number of
variables associated with NSM FRP systems is much larger
than that for externally bonded FRP systems, but the
amount of work available on the former is much more limited
 than that on the latter. Therefore, extensive further
testing is obviously required.

Based on existing knowledge and experience, the preferred
 bond test set-up is believed to be a simple pull-out
test where an NSM FRP bar subjected to tension is bonded
to a concrete block which is supported at the loaded end.
Using such a set-up, effects of factors such as the distance
between adjacent grooves (i.e. groove spacing), and the distance
 between the member edge andthe nearest groove (i.e.
edge distance) should be examined in the near future. In
such bond tests, specimens with either long bond lengths
or short bond lengths may be used, and the local bond–slip
curves obtained from them need to be compared to understand
 the advantages and disadvantages of using short and
long bond lengths. To obtain reliable strain measurements
without interference with the interfacial behaviour, the
development and application of innovative strain sensors
such as fiber optic sensors embeddedin the FRP bar should
also be given due attention.

The most important outcome of the bond tests should
be local bond–slip curves. Further tests are required to
assess the existing local bond–slip equations, and more
importantly to explore the possibility of developing general
local bond–slip models with parameters expressed as functions
 of geometry-and material-related properties. To
achieve this, the tests should preferably be assisted by
numerical modelling of the bond behaviour. For externally
bonded FRP laminates, an accurate meso-scale finite element
 method has recently been developed by Lu et al.

[59] which produced numerical results for use with test
results in the development of a set of accurate bond–slip
models [35]. A similar approach should be explored for
NSM FRP reinforcement. The model by Lundgren [60]
which successfully predicted the bond behaviour of steel
bars in concrete, coupled with appropriate modelling of
the epoxy–concrete interface and constitutive modelling
of the materials [26], also seems to be a promising
approach.

Analytical modelling also has a significant role to play in
the modellingofthe bond behaviourofNSMFRPbars.In
existing bond tests, splitting of the epoxy cover of NSM
FRP bars has been identified as an important failure mode.
Here, the frictional coefficients of different deformed bars
with different groove , shapes and dimensions should be
measured as a function of the slip in order to further clarify
the splitting bond failure mechanism and develop a splitting
 bond strength model to be used in splitting-critical
cases. The approach in Ref. [25] can be considered a first
step towards the definition of a simple splitting bond
strength formula accounting for geometric parameters
and material properties. For this purpose, the simplifying
assumptionsofa friction coefficientequalto1andofa uniform
 distribution of radial pressures need to be removed,
on the basis of experimental measurements of the friction
coefficient.

7.2. Flexural strengthening
Obviously, given the larger number of parameters that
can affect the flexural behaviour of RC beams with NSM
FRP reinforcement, a great deal of further experimental
and theoretical work is required.In particular, the debonding
 failure mechanisms in beams strengthened with NSM
reinforcement need to be clarified through further testing.
The relationship between concrete cover separation and
other modes of debonding ‘‘local’’ to the NSM FRP–concrete
 joint such as fracture at the epoxy–concrete interface
and splitting of the epoxy cover needs further research.
Furthermore, the behaviour of pre-damaged beams
strengthened with NSM FRP is of significant practical
interest, as cracking and damage to the cover of the steel
reinforcement may have a significant effect on the debonding
 failure process.

The relationship between bond failure mechanisms in
bond test specimens and debonding failure mechanisms
in flexurally-strengthened beams needs to be clarified by
detailed experimentalstudies as well as rigorous theoretical
modelling. Here, the study of the interaction between flex-
ural/flexural-shear cracking and bond stresses is of crucial
importance. Once this relationship is clarified, it will then
be possible to develop numerical and analytical models
for predicting debonding failures.

7.3. Shear strengthening
More tests need to be conducted to further clarify the
failure modes of strengthened beams and to evaluate the
effects of various significant factors. More tests are also
needed to confirm the applicability of local bond–slip models
 from bond tests in predicting debonding failures of

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
shear-strengthened beams. This confirmation will facilitate
the development of accurate numerical and analytical models
 for RC beams shear-strengthened with NSM FRP.

7.4. Other issues
As mentioned earlier in the paper, the combined use of
NSM FRP reinforcement in conjunction with externally
bonded FRP reinforcement has been found to be effective
in strengthening beam–column joints [58]. As externally
bondedFRP reinforcement alone has met with only limited
success in strengthening beam–column joints, this combined
 approach definitely deserves further work. This combined
 use to take advantages of both techniques should
also be explored in solving other strengthening problems.

It is widely accepted that pre-stressing the FRP refinement
 before bonding it to concrete structures for strengthening
 purposes is often desirable, both to improve the
serviceability of the structure and to make more efficient
use of the FRP material. Pre-stressing externally bonded
FRP reinforcement has had little success in practice so
far because it is difficultto tension and anchor FRP laminates
 on site, particularly when they are formed by the
wet lay-up process. NSM FRP reinforcement has a much
better chance to succeed: NSM FRP bars can be tensioned
much more easily, particularly when compared with dry
fiber sheets, and are much better anchored than externally
bonded laminates.

The use of cement grout to replace epoxy as the groove
filler has been explored by a limited amount of work
[9,18,23,26]. There are benefits with the use of a cementitious
 groove filler as discussed earlier in the paper.
Research is needed to provide a better understanding of
the performance of cement grout as a groove filler and to
formulate stronger cementitious groove fillers.

8.
Concluding
remarks
Strengthening of structures with NSM FRP reinforcement
 is a technique that has attracted a considerable attention
 as a feasible and economic alternative to the technique
of strengthening structures with externally bonded FRP
reinforcement. The former technique offers some significantadvantages
 over the latter, including the more efficient
use of the FRP material due to a reduced risk of debonding
failure and the better protection of the FRP material from
external sources of damage.

Research on the strengthening of structures using NSM
FRP reinforcement startedonly afew yearsagobuthasby
now attracted worldwide attention. A significant amount
of research has been conducted on this emerging technique,
particularly on the application of this technique in the
strengthening of concrete structures. This paper has provided
 a detailed and critical review of existing research on
the structural behaviour of concrete structures strengthened
with NSM FRP reinforcement. This review has shown that

the existing work is still limited in both scope and depth,
and many questions remain to be answered before the technique
 can be widely acceptedbypracticing engineers. Based
on this review, the more urgentresearch needs have been outlined
 for NSM FRP strengthening of concrete structures.

This paper has been limited to the short-term structural
behaviour of concrete structures strengthened with NSM
FRP reinforcement, so the use of NSM FRP reinforcement
for the strengthening of masonryand timber structures and
long-term aspects have not been addressed. Obviously,
these additional aspects are equally important, andexisting
work on them is even more limited.Itis thus clear that this
emerging strengtheningtechnique poses manychallenges as
well as opportunities for the international research
community.

Acknowledgements

The work presented in this paper has received financial
support from the Research Grants Council of the Hong
Kong SAR (PolyU 5173/04E) and The Hong Kong Polytechnic
 University (G31 YD 61). The authors are grateful
to both organizations for their financial support.

References

[1] Asplund SO. Strengthening bridge slabs with grouted reinforcement.
J Am Concr Inst 1949;20(6):397–406.
[2] Warren
GE. Waterfront repair and upgrade, advanced technology
demonstration site No. 2: Pier 12, NAVSTA San Diego. Site Specific
Report SSR-2419-SHR, Naval Facilities Engineering Service Center,
Port Hueneme (CA), 1998.
[3] Warren
GE. Waterfront repair and upgrade, advanced technology
demonstration site No. 3: NAVSTA Bravo 25, Pearl Harbour. Site
Specific Report SSR-2567-SHR, Naval Facilities Engineering Service
Center, Port Hueneme (CA), 2000.
[4] Garrity
SW. Near-surface reinforcement of masonry arch highway
bridges. In: Proceedings of the 9th Canadian masonry symposium,
Fredericton (Canada), CD-ROM, 2001.
[5] Fib TG9.3. Externally bonded FRP reinforcement for RC structures.
Technical report on the design and use of externally bonded fibre
reinforcedpolymer reinforcement (FRP EBR) for reinforced concrete
structures. International Federation for Structural Concrete, Lausanne,
 2001.
[6] American Concrete Institute Technical Committee 440. Guide for the
design and construction of externally bonded FRP systems for
strengthening concrete structures. ACI 440.2R-02, 2002.
[7] De Lorenzis L, Teng JG. Near-surface mounted FRP reinforcement:
an emerging technique for structural strengthening. Research Report
of Hong Kong Polytechnic University, in preparation.
[8] American Concrete Institute Technical Committee 440. Guide for the
design and construction of concrete reinforced with FRP bars. ACI
440.1R-03, 2003.
[9] Nordin H, Taljsten B. Concrete beams strengthened with CFRP. A
study of anchor lengths. In: Proceedings 10th conference on
structural faults and repair, London (UK), 2003.
[10] Taljsten B, Carolin A, Nordin H. Concrete
structures strengthened
with near surface mounted reinforcement of CFRP. Adv Struct Eng
2003;6(3):201–13.
[11] De Lorenzis L. Strengthening of RC structures
with near surface
mounted FRP rods. PhD Thesis, Department of Innovation Engineering,
 University of Lecce, Italy, 2002.

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
[12] Parretti
R, Nanni A. Strengthening of RC members using near-
surface mounted FRP composites: design overview. Adv Struct Eng
2004;7(6):469–83.
[13] Blaschko
M. Bond behaviour of CFRP strips glued into slits. In:
Proceedings FRPRCS-6. Singapore: World Scientific; 2003. p.
205–14.
[14] Sena Cruz JM, Barros JAO. Bond behavior of carbon laminate strips
into concrete by pull-out bending tests. In: Proceedings of the
international symposium ‘‘Bond in concrete – from research to
standards’’, Budapest (Hungary), 2002. p. 614–21.
[15] Barros
JAO, Fortes AS. Flexural strengthening of concrete beams
with CFRP laminates bonded into slits. Cement Concr Compos
2005;27(4):471–80.
[16] Blaschko
M. Zum tragverhalten von betonbauteilen mit in schlitze
eingeklebten CFK-lamellen. Bericht 8/2001 aus dem Konstruktiven
Ingenieurbau, TU Mu¨ nchen, 2001. 147 pp [in German].
[17] Hassan T, Rizkalla S.
Investigation of bond in concrete structures
strengthened with near surface mounted carbon fiber reinforced
polymer strips. ASCE J Compos Constr 2003;7(3):248–57.
[18] Carolin
A, Taljsten B. Behaviour of concrete beams strengthened
with near-surface mounted reinforcement, NSMR. In: Proceedings
ACIC2002, Southampton (UK), 2002. p. 177–84.
[19] Hassan
T, Rizkalla S. Bond mechanism of near-surface-mounted
fiber-reinforced polymer bars for flexural strengthening of concrete
structures. ACI Struct J 2004;101(6):830–9.
[20] Teng JG, De Lorenzis L, Wang B, Rong L, Wong TN, Lam L.
Debonding failures of RC beams strengthened with near-surface
mounted CFRP strips. J Compos Constr, ASCE 2006;10(2):92–105.
[21] De
Lorenzis L, Nanni A. Bond between NSM fiber-reinforced
polymer rods and concrete in structural strengthening. ACI Struct J
2002;99(2):123–32.
[22] Al-Zahrani MM. Bond
behavior of fiber reinforced plastic (FRP)
reinforcements with concrete. PhD Thesis, Department of Civil and
Environmental Engineering, The Pennsylvania State University,
University Park (PE), 1995. p. 263.
[23] De Lorenzis L, Rizzo A, La Tegola A. A modified pull-out test for
bond of near-surface mounted FRP rods in concrete. Compos—Part
B: Eng 2002;33(8):589–603.
[24] Tepfers R. A theory of bond applied to overlapped tensile reinforcement
 splices for deformed bars. Publication 73:2, Division of
Concrete Structures, Chalmers University of Technology, Gothenburg
 (Sweden), 1973. 328 pp.
[25] De Lorenzis L. Anchorage length of near-surface mounted FRP bars
for concrete strengthening – analytical modeling. ACI Struct J
2004;101(3):375–86.
[26] De Lorenzis
L, Lundgren K, Rizzo A. Anchorage length of near-
surface mounted FRP bars for concrete strengthening – experimental
investigation and numerical modeling. ACI Struct J
2004;101(2):269–78.
[27] Sena Cruz JM, Barros JAO. Modeling of bond between near-surface
mounted CFRP laminate strips and concrete. Comput Struct
2004;82(17–19):1513–21.
[28] MacGregor JC. Reinforced
concrete – mechanics and design. 3rd
ed. Upper Saddle River (NJ): Prentice-Hall, Inc.; 1997, 939 pp.
[29] ASTM G115-98. Standard guide for measuringand reporting friction
coefficients. ASTM International, West Conshohocken (PA), 2001.
11 pp.
[30]
van der Veen C. Cryogenic bond stress–slip relationship. Delft
University of Technology, Doctoral Thesis, Delft (The Netherlands),
1990. 111 pp.
[31] Tepfers R, Olsson PA. Ring test for evaluation of bond properties of
reinforcing bars. In: Proceedings of the international conference on
bond in concrete – from research to practice, Riga, Latvia, vol. 1,
1992. p. 89–99.
[32] Cosenza
E, Manfredi G, Realfonzo R. Behavior and modeling of
bond of FRP rebars to concrete. ASCE J Compos Constr
1997;1(2):40–51.
[33] Alunno Rossetti V, Galeota D, Giammatteo MM. Local bond stress–
slip relationships of glass fibre reinforced plastic bars embedded in
concrete. Mater Struct 1995;28(180).
[34] Nanni A, Conrad JO, Bakis CE, Uchno J. Material properties of C-
Bar reinforcing bar. Civil Engineering Department, Pennsylvania
State University, PA, 1997.
[35] Lu XZ, Teng JG, Ye LP, Jiang JJ. Bond–slip models for FRP sheets/
plates bonded to concrete. Eng Struct 2005;27(6):920–37.
[36] Chen JF, Teng JG. Anchorage strength models for FRP and steel
plates bonded to concrete. J Struct Eng, ASCE 2001;127(7):784–91.
[37] Teng
JG, Chen JF, Smith ST, Lam L. FRP strengthened RC
structures. England: John Wiley & Sons, Ltd; 2002.
[38] Yan X, Miller B, Nanni A, Bakis CE. Characterization of CFRP bars
used as near-surface mounted reinforcement. In: Proceedings 8th
international structural faults and repair conference, Edinburgh
(Scotland), CD-ROM version, 1999.
[39] Yao J, Teng JG, Chen JF. Experimental study
on FRP-to-concrete
bonded joints. Compos—Part B: Eng 2005;36(2):99–113.
[40] Wang B, Teng JG, De Lorenzis L, Zhou LM, Ou JP, Jin W, et al.
Strain Monitoring of RC Beams Strengthened with Smart NSM FRP
Bars, in preparation.
[41] El-Hacha
R, Rizkalla SH. Near-surface-mounted fiber-reinforced
polymer reinforcements for flexural strengthening of concrete structures.
 ACI Struct J 2004;101(5):717–26.
[42] Alkhrdaji T, Nanni A, Chen G, Barker M. Upgrading the transportation
 infrastructure: solid RC decks strengthened with FRP. Concr
Int 1999;21(10):37–41.
[43] Hassan
T, Rizkalla S. Flexural strengthening of prestressed bridge
slabs with FRP systems. PCI J 2002;47:76–93.
[44] De
Lorenzis L, Nanni A, La Tegola A. Flexural and shear
strengthening of reinforced concrete structures with near surface
mounted FRP rods. In: Proceedings ACMBS III, Ottawa (Canada),
2000. p. 521–8.
[45] Smith
ST, Teng JG. FRP-strengthened RC beams. I: review of
debonding strength models. Eng Struct 2002;24(4):385–95.
[46] Smith ST, Teng JG. FRP-strengthened RC beams. II: assessment of
debonding strength models. Eng Struct 2002;24(4):397–417.
[47] Teng JG, Smith
ST, Yao J, Chen JF. Intermediate crack-induced
debonding in beams and slabs. Constr Build Mater 2003;17(6–7):
447–62.
[48] Lu XZ, Teng JG, Ye LP, Jiang JJ. Intermediate crack debonding in
FRP strengthened RC beams: FE analysis and strength model,
submitted for publication.
[49] Oehlers
DJ, Seracino R. Design of FRP and steel plated RC
structures: retrofitting beams and slabs for strength, stiffness and
ductility. Elsevier; 2004.
[50] JSCE. Recommendations for upgrading
of concrete structures with
use of continuous fiber sheets. Japan Society of Civil Engineers, 2001.
[51] Yao J, Teng JG, Lam L. Experimental study on intermediate crack
debonding in FRP-strengthened RC flexural members. Adv Struct
Eng 2005;8(4):365–96.
[52] Aiello MA, De Lorenzis
L, Micelli F. Load–deflection behavior of
RC beams strengthened with near-surface mounted FRP bars. In:
Proceedings CCC2003, Cosenza (Italy), September 2003.
[53] Nordin H, Taljsten B. Concrete beams strengthened with prestressed
near surface mounted reinforcement. In: Proceedings FRPRCS-6,
Singapore, 2003. p. 1077–86.
[54] De Lorenzis
L, Micelli F, La Tegola A. Passive and active near-
surface mounted FRP rods for flexural strengthening of RC beams.
In: Proceedings ICCI-02, San Francisco (CA), CD-ROM version,
2002.
[55] De Lorenzis L, Nanni A. Shear strengthening of reinforced concrete
beams with NSM fiber-reinforced polymer rods. ACI Struct J
2001;98(1):60–8.
[56] Barros
JAO, Dias S. Shear strengthening of reinforced concrete
beams with laminate strips of CFRP. In: Proceedings CCC2003,
Cosenza (Italy), 2003. p. 289–94.

L. De Lorenzis, J.G. Teng/Composites: PartB38 (2007) 119–143
[57] Nanni A, Di Ludovico M, Parretti R. Shear strengthening of a PC
bridge with NSM CFRP rectangular bars. Adv Struct Eng
2004;7(4):97–109.
[58] Prota
A, Nanni A, Manfredi G, Cosenza E. Selective upgrade of
underdesigned RC beam–column joints using CFRP. ACI Struct J
2004;101(5):699–707.
[59] Lu XZ, Ye LP, Teng JG, Jiang JJ. Meso-scale finite element model
for FRP sheets/plates bonded to concrete. Eng Struct
2005;27(4):564–75.
[60] Lundgren
K. Bond between ribbed bars and concrete: Part 1.
Modified model. Mag Concr Res, in press.
[61] Crasto A, Kim R, Ragland W. Private communication, University of
Dayton Research Institute, 1999.
[62] Yost
JR, Gross SP, Dinehart DW, Mildenberg J. Near surface
mounted CFRP reinforcement for the structural retrofit of concrete
flexural members. In: Proceedings ACMBS-IV, Calgary (Canada),
July 2004, CD-ROM.
[63] Arduini
M, Romagnolo M, Camomilla G, Nanni A. Influence of
concrete tensile softening on the performance of FRP strengthened
RC beams: experiments. In: Proceedings ACMBS-IV, Calgary
(Canada), July 2004, CD-ROM.