Construction and Building Materials 183 (2018) 325–345

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier.com/locate/conbuildmat

State-of-the-art review and future research directions for

FRP-to-masonry bond research: Test methods and techniques
for extraction of bond-slip behaviour
J. Vaculik a, P. Visintin a,⇑, N.G. Burton a, M.C. Griffith a, R. Seracino b
a
School of Civil, Environmental and Mining Engineering, The University of Adelaide, South Australia 5005, Australia
b
Department of Civil, Construction, and Environmental Engineering, North Carolina State University, Raleigh, NC 27695-7908, USA

h i g h l i g h t s

Collates the results of 1583 masonry pull-test results obtained from 56 studies.

Inconsistency in test arrangements, instrumentation methods, and data processing identified.

Lack of reported properties hinders the development of a bond model.

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

Article history: The effectiveness of FRP retrofits is heavily reliant on the shear bond that can be developed between the
Received 8 April 2018 FRP and masonry substrate, which has been the focus of experimental research for almost two decades.
Received in revised form 3 June 2018 This paper collates and critically reviews previous experimental work on the shear bond between FRP
Accepted 12 June 2018
composites and masonry substrates, identifying 1583 individual pull-tests across 56 published studies.
Whilst the pool of existing data is significant in terms of number of tests, it encompasses a rather narrow
range of substrate material, FRP material and retrofit configuration. Most notably, the majority of tests
Keywords:
have been undertaken on clay brick substrates, carbon FRPs and externally-bonded retrofits. By contrast,
Unreinforced masonry (URM)
Clay brick
testing of natural stone substrates and near-surface-mounted retrofits has been limited. Significantly, the
Natural stone review identifies considerable inconsistency in the test arrangements, instrumentation methods, and
Retrofit data processing techniques for extracting local bond-slip properties, which has undoubtedly hindered
Fibre-reinforced polymer (FRP) the development of a unified bond model and codifiable design rules. Methods of extracting bond-slip
Bond-slip behaviour from test data are critically reviewed, and importantly it is shown through numerical examples
Load-slip that without adequate instrumentation it is not possible to reliably extract this behaviour from standard
Bond-strength pull-tests. Finally, suggestions for adequate instrumentation and a framework for undertaking bond-slip
behaviour extraction through inverse analysis are presented. Significantly, the experimental database
compiled as part of this work-thought to be the largest of its kind to date—is made openly available as
an accompanying Data in Brief article with the intent that it will facilitate development of
bond-strength models for FRP bonded to masonry.
Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction
Unreinforced masonry construction is particularly susceptible
to seismic loading due to its low tensile strength and heavy mass,
motivating the need for retrofit. Fibre-reinforced polymer (FRP)
composites (hereafter referred to as ‘plates’) have emerged as a
popular means of strengthening masonry structures due to their
⇑ Corresponding author.
E-mail address: phillip.visintin@adelaide.edu.au (P. Visintin).
high strength and stiffness, low weight, and durability [1]. The role
of FRP plating is to act as tensile reinforcement which can be used
to enhance a wall’s capacity against in-plane shear and/or out-of-
plane flexure [2–6].
The effectiveness of FRP retrofits is largely controlled by the
ability to develop shear load transfer across the FRP-to-masonry
bond. The most common experimental technique for studying
bond behaviour is via the shear pull-test (hereafter referred to sim-
ply as ‘‘pull-test”), shown diagrammatically in Fig. 1. This test
involves adhesively bonding a FRP composite to the masonry

https://doi.org/10.1016/j.conbuildmat.2018.06.103
0950-0618/Ó 2018 Elsevier Ltd. All rights reserved.
326 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345

FRP plate Brick or masonry

substrate
P P
P 2P
P
R R
(a) Single lap (b) Double lap

2P 2P

(c) Two-block double lap

Fig. 1. Various types of pull-test arrangements including (a) single lap, (b) double lap, and (c) two-block double lap. In (a) and (b), the load arrows indicate the more common
push-pull variant where the substrate is subjected to compression. In (c), the load arrows indicate the pull-pull variant.

substrate and applying an increasing slip until the plate eventually
debonds. To date, pull-tests on masonry substrates have covered a
range of:

Retrofit types, including externally-bonded (EB) and near-
surface-mounted (NSM);

Composite materials, including carbon FRP (CFRP), glass FRP
(GFRP), basalt FRP (BFRP), aramid FRP (AFRP), and steel-
reinforced polymers (SRP); and

Substrate materials, including clay brick, concrete block, and
various types of natural stone.
Pull-tests can provide insight into the local behaviour of the
bond under shear deformation, which is typically characterised
in terms a bond-slip (s-d) model relating shear stress to slip
(Fig. 2). The predictive capability of the bond-slip model covers
various aspects of retrofit behaviour including the debonding load,
required anchorage length, axial strain profile, as well as the global
load-slip (P-D) behaviour of the system. From a design perspective
it is particularly useful to further define the bond-slip behaviour in
terms of the mechanical properties of the substrate including its
compressive and/or tensile strength. However, while substantial
testing has been undertaken to date, there are as-yet no standard-
ised guidelines for performing pull-tests. Consequently, a lack of
consistent testing methodology and data analysis technique has
to some extent hindered the definition of generalised material
models necessary for design guidelines (e.g. [7,8], the discussion
of which is the purpose of this paper.
This paper undertakes a state-of-the-art review by bringing
together the results of 1583 individual pull-tests from 56 studies,
in order to:
1. Identify the scope of previous testing in terms of FRP material
properties, masonry material properties, and retrofit types;
2. Discuss the range of common testing methodologies including
instrumentation requirements; and
3. Compare and critique the alternative methods of data process-
ing for characterising the local bond-slip behaviour.
This paper is structured as follows. Section 2 describes the
underlying mechanics governing FRP retrofits and the general
approaches to calibrating bond models. The experimental database
collected as part of this work is presented in Section 3 and further
discussed in Section 4 in terms of scope of tests and experimental
techniques. Recommendations for viable data processing
approaches for extracting local bond-slip behaviour are provided
in Section 5, and properties extracted from the experimental tests
to date are discussed in Section 6.
2. Mechanics of FRP retrofits
2.1. Governing equations

It is commonly accepted that the mechanism of load transfer

Bilinear Curvilinear between FRP reinforcement and brittle substrates is governed by
(e.g. exponential) uniaxial shear lap theory, originally proposed by Volkersen [9].
Linear-brittle Early works involving application of this theory toward structural
Linear- retrofit were in the field of concrete which studied the use of adhe-
f
descending sively bonded steel and FRP plates [10–12]. The theory has since
gained widespread acceptance and formed the foundation for
FRP retrofit predictive models in both concrete [13–16] and unre-
Residual friction inforced masonry [8,17–26].
(NSM) The basis of shear lap theory is that stress transfer across the
bonded interface is controlled by a fundamental bond-slip law that
relates shear stress (s) to slip (d) between the two adherents. This
Gf
law can be represented by a variety of available forms, including
those shown in Fig. 2. By enforcing the conditions of force equilib-
rium and compatibility between the interface slip and elastic
1 f
deformations of the adherents, the following governing equations
Fig. 2. Commonly used bond-slip models. are obtained (e.g. [27,28]):
 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345
2
d d
2
¼ bs ð1Þ
dx
and


1 1
b ¼ Lp þ ð2Þ
E p Ap E m Am
where slip d and shear stress s vary along longitudinal position x as
shown in Fig. 4, EpAp and EmAm are the respective axial stiffness of
the FRP and substrate, and Lp is the bonded perimeter of the plate.
These geometric properties are illustrated in Fig. 3.
Solving these equations for the pull-test arrangements shown in
Fig. 1 involves imposing boundary conditions where the plate
strain is ep = 0 at the free end of the plate and ep = P/EpAp at the
loaded end. The ease with which Eqs. (1) and (2) can be solved
for the global load-slip (P-D) response depends on the complexity
of the adopted bond-slip law and system being modelled. Closed-
form analytical solutions are available for a variety of relatively
simple bond-slip models for a homogenised treatment of the sub-
strate [27–33]. By contrast, solutions for separate treatment of
brick units and mortar joints can be obtained by numerical
finite-difference schemes [34,35], although analytical solutions
for certain types of bond-slip models have also been proposed
[36,37]. Capturing more refined behaviour such as nonlinear
mechanical properties or interaction between mode I (normal
stress) and mode II (shear stress) fracture along the bond interface
by these core mechanics can also be performed through finite-
element modelling [38–43].
Regardless of the functional form of the bond-slip law, as the
bonded length is increased, the debonding load approaches the
capacity PIC (e.g. [12,27,28]:
qffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffi
PIC ¼ 2Gf Ep Ap Lp ð3Þ
where Gf is the interfacial fracture energy defined as the total area
enclosed under the s-d model being considered (shown in grey for
the case of the bilinear model in Fig. 2. Eq. (3) makes the further
assumption that the axial stiffness of the substrate greatly exceeds
that of the FRP (EmAm  EpAp), which is reasonable in most practical
cases. The full load capacity PIC is reached at the critical bond length,
Lc [refer Eq. (4)] beyond which any further increase to the bond
length has no effect on the load capacity. Short and long bonded
lengths generate contrastingly different global load-slip behaviour
as will now be discussed.
2.2. Load-slip behaviour over long and short bonded lengths
To illustrate the mechanism of debonding in a pull-test, con-
sider a FRP plate bonded to a homogenous substrate prism as
shown in Fig. 4. Let us adopt the bilinear bond-slip model
(Fig. 4a), which is typically considered to provide a reasonable
compromise between simplicity and accuracy [8,10,14,16,23,27–
29,33,40,44,46]. The resulting global load-slip (P-D) behaviour
bp bp
Ep, Ap tp Ep, Ap
dp
Lp Lp
Em, Am Em, Am
bs bs
(a) (b)
Fig. 3. Sectional properties for (a) EB plate, (b) NSM plate.
327
shown in Fig. 4b can be explained by the distinct phases of debond-
ing under increasing levels of slip and the associated shear stress
distributions (Fig. 4e–l).
Consider the separate cases of a short and long bonded length as
shown in Fig. 4c and d. In the discussions throughout this paper, a
long bonded length is defined as having the capability to accommo-
date progressive debonding; that is, it can sustain load even if the
loaded-end (LE) slip D exceeds the local slip capacity df. In turn, a
short bonded length becomes fully debonded once LE slip reaches
df. This leads to the introduction of the critical length which for
the bilinear bond-slip model is [29,33]:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p df  d1
Lc ¼ ð4Þ
2 sf b
Eq. (4) defines the limit between a short length (L < Lc) and a
long length (L > Lc). Note that if the linear-descending rule is
adopted (where d1 = 0), Eq. (4) reduces to the well-known expres-
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sion Lc ¼ p=2 df Ep Ap =sf Lp which is used in most simplified design
procedures (e.g. [8]).
In Fig. 4(e) and (f) we initially apply a LE slip of d1 to both
prisms, resulting in the illustrated shear stress distributions. Up
to this point, the slip distribution covers only the rising branch of
the s-d model, and the associated P-D response remains linear-
elastic. The most notable difference between the short and long
bonded lengths is that over the long length, the free-end (FE) slip
remains negligible compared to the loaded-end slip. Consequently,
the short bonded length always resists less load at an equal D.
By increasing the applied LE slip beyond d1, the loaded end
enters the softening s-d region, as shown in Fig. 4(g) and (h). The
short bonded length reaches its peak load resistance somewhere
in this range (d1  D  df), beyond which the load resistance grad-
ually declines and reaches zero once the LE slip reaches df [point
(i)]. Conversely, the load resistance over a long bonded length con-
tinues to rise, achieving its peak capacity at D = df [point (j)].
Importantly, any further increase to the LE slip over the long
bonded length causes progressive debonding where the shear
stress profile shifts from the loaded end toward the free end as
shown in (k). A point of instability is finally reached at point (l)
beyond which the remaining bonded length cannot sustain any
further load and the plate instantaneously debonds. This markedly
different behaviour over long and short bonded lengths has impor-
tant implications toward the interpretation of pull-test results as
will be discussed in Section 5.
2.3. Fitting bond models to test data
Two distinct approaches exist for characterising bond-slip
behaviour from pull-tests:
1. The first is by extraction of bond-slip behaviour using individual
test data by either strain readings or inverse analysis. These
techniques are further discussed in Sections 5.2 and 5.3.
2. The second is by calibration of bond-strength models, which
involves consideration of a large test database and producing
good fit between the load capacity predicted by a postulated
functional form (that often assumes a particular bond-slip
model) and the experimental values across the entire popula-
tion of tests [8,13–26,45,46]. This calibration process allows
for the bond model defining properties such as Gf, sf or df to
be expressed as a function of the fundamental mechanical prop-
erties including the compressive strength (fuc) and/or tensile
strength (fut) of the masonry unit. Whilst this calibration
approach is vital for design purposes, the bond-slip properties
are inferred indirectly. It is intended that the test database com-
piled (Section 3) can be used for this purpose.
328 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345

P
h j k
(b) L > Lc
L = Lc l
(a)
f L < Lc
f
e g

Gf
I II III L << Lc

1 f i
1 f

loaded end free end

(c) Short, L < Lc (d) Long, L > Lc

P P

L L

(e) f (f) f = 1

= 1
I = 1
I

f
(g ) (h)
< II < II I
1 f 1 f

f
(i) (j)
III III II I
f f

(k) (l)
f

f III II I II I

Fig. 4. Mechanism of debonding over short and long bonded lengths, including: (a) bilinear bond-slip model, (b) resulting load-slip response; (c) short bonded length and (d)
long bonded length; and (e)-(l) shear stress distribution under increasing loaded-end slip D.

This paper deals with considerations toward both types of
approach.
From the preceding discussion it follows that for test data to be
usable the following minimum information must be available (see
Fig. 3):

Plate axial stiffness EpAp, (or in EB retrofits effective Eptp where
tp is the plate thickness),

Prism width, bs, and plate width bp to account for width effects,

Plate bonded perimeter Lp, determined directly from plate
dimensions bp and tp, and

Bond length, L,

Failure load Pmax together with a description of the failure
mode, and

Substrate mechanical properties fuc and/or fut.
3. Experimental database
An experimental database of 1583 individual pull-tests on
masonry specimens was compiled from across 56 published stud-
ies [17,18,21–26,32,39–43,45,47–86], Kwiecien et al., 2016. This
database is intended for the development and refinement of
bond-strength models and is made available as an accompanying
Data in Brief article. While similar databases have been previously
reported [18–20,24], this is to the authors’ knowledge the most
comprehensive database collected to date.
The database is intended to cover tests under standard condi-
tions involving plates bonded directly to the masonry substrate
and subjected to quasistatic loading. It excludes the following:

Specimens exposed to environmental effects such as tempera-
ture and moisture [32,48,67,69,74,84,87],

The use of additional anchorage between the FRP and substrate,
for example nails, fans or cogs [17,52,61,65,77,79],

Tests in which the FRP was bonded to plaste [17,60–62,76],

Non-quasistatic loading such as impulse loads [81]; and

Textile-reinforced mortars, fibre-reinforced cementitious mor-
tars or similar.
Aside from the above exclusion criteria, any control specimens
found in these studies are still included. Note that some studies
have intentionally opted not to report the failure load for tests
where the mode of failure was not by interfacial debonding—such
tests are still included as long as the mode of failure was reported.
 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345
Where the same tests were identified to be reported in multiple
sources, they are only included once.
The database also includes tests with the following special
conditions:

Tests with confining precompression (30 tests) [25,82],

Curved specimens (10 tests) [77]; and

Specimens that were repaired after an initial test and re-tested
(98 tests) [42,68,88].
4. Review of experimental work to date
A review of the 1583 experimental tests which are included in
the compiled database will now be undertaken. Fig. 5 illustrates
the origin of tests in terms of (a) published source, (b) institution,
and (c) country. It seen that the large-scale studies performed by
Seim and Pfeiffer [25], Valluzzi et al. [86], Rotunno et al. [83], de
Felice et al. [21], and Kwiecień et al. [88] form roughly half of all
tests. Further, the vast majority of tests originate from Italy owing
to high seismicity and large stock of historically significant struc-
tures. Hence there is expected to be a natural bias in these studies
toward brick and stone materials present throughout this region.
Fig. 6 summarises the scope of tests in terms of characteristics
of the test specimens and test method. Tables 1 and 2 demonstrate
the various combinations of the retrofit scheme, FRP material, and
substrate material for EB and NSM retrofits, respectively. Table 3
provides an accompanying overview of each individual study
including the substrate material, prism type, FRP material, FRP
shape, retrofit configuration, test arrangement, type of loading,
and instrumentation.
4.1. Test specimens
4.1.1. Substrate material
Due to its widespread prevalence, clay brick has been the most
widely tested substrate covering 82% of tests (Fig. 6a). Different
varieties of natural stone, most notably limestone and tuff which
are native to Italy and parts of the Mediterranean, have also been
investigated, accounting for 10% of tests [22,32,45,47–49,59,61,
80,84]. Note that ‘limestone’ refers to regular limestone as well
as other lime-based stone types such as Lecce stone and calcaren-
ite. Whilst the lithology covered across these tests is comparable to
some varieties of stone found elsewhere, building stones native
specifically to other regions have not yet been investigated. Con-
crete blockwork has received somewhat limited attention
[17,50,58,63,72,85], which can be explained by the fact that mod-
ern blockwork tends to often be steel-reinforced making it more
resistant toward earthquake loading. Other substrates include cal-
cium silicate bricks and sandstone, collectively accounting for 4% of
tests [25]. The 56 studies cover a total of 95 unique substrate
batches as summarised in Table 4. The difference between the per-
centages in Table 4 (by unique substrate) and Fig. 6a (by individual
pull-test) is mainly due to the large round robin studies [21,86,88]
which predominantly used a single type of clay brick.
The fundamental mechanical properties of the masonry unit are
its compressive strength, fuc, and tensile strength, fut. Compressive
strength is typically determined via crushing tests on unit cubes or
cylinders [89,90]. The tensile strength however can be obtained via
three alternate test methods: direct (coupons or surface pull-off),
flexural EN [91]; ATSM C67-03a [92]) or splitting (ASTM C1006-
07 [93]). The distribution of these properties counted by individual
pull-test is shown in Fig. 7.
The mechanical properties of the unit are a direct determinant
of the bond-slip properties (Gf, sf, df), which is reflected by existing
masonry bond-strength models despite a lack of consensus on the
precise parametric dependence. For example, masonry bond
329
models can be formulated in terms of only fuc [15,17,19,22], only
fut [24,26], or both fuc and fut [8,18,21,23,25,41,44,45]. The inconsis-
tent parametric basis can be partially explained by the lack of
available data on the respective mechanical properties in some
studies as demonstrated by Fig. 7 (last column of each plot) and
Table 4. That is, of the 95 unique material substrates used in the
56 studies, fuc was reported for 88% of substrates (90% of pull-
tests), fut,dir for 17% (54%), fut,flex for 42% (67%), and fut split for 7%
(39%).
The use of alternate testing approaches for the measurement of
tensile strength poses further difficulty since the approaches lead
to different fut values (direct < splitting < flexural), which is an
issue that must be considered in subsequent the calibration of
bond-strength models. Other issues that hinder the usability of
the data include incomplete of reporting of:

Test specimen dimensions and whether the reported values
account for confinement and aspect ratio effects; and

The number of tested specimens and coefficients of variation,
which make it difficult to assess the reliability of the data.
The absence of a consistent approach with regard to mechanical
strength testing is thought to contribute partially to the scatter in
reported bond-slip properties as discussed in Section 6.
4.1.2. Prism type (unit or masonry)
Specimens can be categorised into either unit prisms comprising
individual bricks or blocks, or masonry prisms in which units are
bonded by mortar joints.
The majority of tests (75%) used unit prisms (Fig. 6b). Such tests
generally provide an upper-bound estimate of the debonding force
and a lower-bound estimate of the critical bond length compared
to masonry built with mortar joints, as the interfacial bond along
the unit tends to be stronger than along the mortar. Exceptions
to this can include certain low-strength unit types (e.g. some forms
of tuff), which can sometimes be weaker than the mortar.
The remaining 25% of tests were performed on masonry prisms.
An implicit feature of such tests is that any extracted bond-slip
behaviour becomes effectively homogenised. In the majority of
these studies, the FRP passed consecutively through units and hor-
izontal bed joints; however, a small subset looked at specimens
where the FRP passed through vertical perpend joints [26,82],
which was shown to reduce the debonding load capacity by
approximately 10%. Tests on isolated mortar specimens have been
limited to the work by Carloni and Subramaniam [40], which
means there is at present only limited capability to separately
model the individual masonry constituents [35–37]. A handful of
studies have also investigated various non-standard prisms includ-
ing curved specimens [77], and plates oriented diagonally [55,76]
and horizontally [25,50] with respect to mortar joints
4.1.3. FRP material, shape and retrofit configuration
In terms of composite material, tests on carbon- and glass-FRP
are most numerous, comprising 50% and 31% of total tests respec-
tively (Fig. 6c). The remaining tests include basalt FRP (9%), steel-
reinforced polymers (10%), aramid FRP (1%), and natural flax FRP
(<1%).
The prevalence of different reinforcement shapes and retrofit
configurations is shown in Fig. 6d. For a breakdown of the tested
combinations of reinforcement material and shape, refer to Table 5.
Externally-bonded retrofits cover 92% of tests, and almost entirely
comprise wet-layup sheets. As seen from Table 1, EB retrofits have
been used across the full diversity of FRP materials and substrate
materials. Near-surface-mounted retrofits make up the remaining
8% of tests, and encompass rectangular strips (all CFRP) and circu-
lar rods (mostly GFRP). NSM retrofits have been limited to CFRP
330 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345

(a)

(b)

(c)

Fig. 5. Disaggregation of the 1583 individual tests according to the (a) author, (b) institution of the lead author, and (c) their country. Note that the round robin studies by
Valluzzi et al. [86], de Felice et al. [21] and Kwiecień et al. [88] are disaggregated according to the individual test laboratories.
 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345 331

Fig. 6. Disaggregation of the 1583 individual tests according to characteristics of the test specimens and test arrangement.

Table 1
Number of tests on EB retrofits for different combinations of substrate material and FRP material. For the various substrate materials, ‘u’ denotes unit prisms, ‘m’ denotes masonry
prisms, and ‘other’ includes calcium silicate brick and sandstone.

Substrate Material FRP Material Total

CFRP GFRP BFRP SRP AFRP Flax
Clay (u) 491 [8,17,20,21,36, 231 [2,20,23,30,41,44, 91 [20,58,68,69,94] 110 [20,53,58, – – 923
39,40,51,53,54, 50,51,53,56–58,61,64,68,69,89,91,94] 68,69,94,107]
56–58,60,62,68,69,94,106]
Clay (m) 78 [29,31,40,55,60,66,68,69,90,93] 136 [22,29,34,44,55,56,67–69,91] 21 [68,69] 31 [68,69,107] 6 [38] – 272
Concrete (u) 20 [66,106] 26 [50] – – – – 46
Concrete (m) – – – – 6 [38] – 6
Limestone (u) 61 [10,48,49,57,59,87,92] 10 [57,65,87] 6 [10] – – – 77
Limestone (m) 2 [10] 9 [65] 6 [10] – – – 17
Tuff (u) 22 [10,34,57,63,87,92] 15 [10,34,57,63,87] 8 [10,18,63] 3 [107] – 4 [63] 52
Tuff (m) 2 [10] 2 [10] 2 [10] – – – 6
Mortar 3 [40] – – – – – 3
Other (u) 3 [56] 47 [56] – – – – 50
Other (m) – 6 [56] – – – – 6
Total 682 482 134 144 12 4 1458

Table 2
Number of tests on NSM retrofits for different combinations of substrate material and FRP material. For the various substrate materials, ‘u’ denotes unit prisms and ‘m’ denotes
masonry prisms.

Substrate Material FRP Material Total

CFRP GFRP BFRP SRP AFRP Flax
Clay (u) 15 [96] – – – – – 15
Clay (m) 92 [1,13,19,29,52] – – – – – 92
Concrete (u) 3 [47] 15 [7,47] – – – – 18
Concrete (m) – – – – – – 0
Limestone (u) – – – – – – 0
Limestone (m) – – – – – – 0
Tuff (u) – – – – – – 0
Tuff (m) – – – – – – 0
Mortar – – – – – – 0
Other (u) – – – – – – 0
Other (m) – – – – – – 0
Total 110 15 0 0 0 0 125

and GFRP (Table 2 and have also only been applied to clay and con-
crete substrates. No research has yet considered NSM retrofitting
of natural stone.
Fig. 8 shows the distribution of plate properties in EB retrofits
including the elastic modulus Ep and axial stiffness Eptp. Similarly,
Fig. 9 plots the distribution of Ep and EpAp for NSM retrofits. The
property Eptp typically represents an effective stiffness measured
from tensile tests on either the reinforcement fabric or on the com-
posite FRP plate (fibre sheet + adhesive). Some studies however do
not explicitly state which of these cases applies. Further, the
332 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345

Table 3
Details of each experimental program.

ID Reference No. Substrate Prism Type FRP Material FRP Shape Retrofit Laps* Loading** Deformation
Tests Material Config Measurement***
1 [63] 6 Conc Unit CFRP (3) Rod NSM D M SG, FES
GFRP (3)
2 [58] 12 Clay (6) Masonry AFRP Sheet EB D M SG
Conc (6)
3 [85] 12 Conc Unit GFRP Rod NSM D M SG, FES
4 [47] 25 Lime Unit CFRP Sheet EB D M SG, LES
5 [48] 1 Lime Unit CFRP Sheet EB D M SG, LES
6 [72] 50 Clay (24) Unit GFRP Sheet EB S M –
Conc (26)
7 [49] 28 Lime (18) Unit CFRP Sheet EB S (2) D M SG, LES
Tuff (10) (26)
8 [43] 4 Clay Masonry CFRP Sheet EB S M SG, LES
9 [17] 24 Clay (8) Unit CFRP Sheet EB 2D M LES
Conc (16)
10 [53] 18 Clay Unit CFRP Sheet EB D M SG
11 [78] 10 Clay Unit CFRP (5) Sheet EB D M SG
GFRP (5)
12 [71] 6 Clay Masonry CFRP Strip NSM S C SG
13 [82] 18 Clay Masonry CFRP Strip NSM S M SG
14 [26] 29 Clay Masonry CFRP (19) GFRP (10) Sheet (10) EB (14) S M SG
Strip (19) NSM (15)
15 [51] 64 Clay Unit CFRP Sheet EB D M SG
16 [54] 9 Clay Unit CFRP (3) GFRP Sheet EB D M SG
(3) SRP
(3)
17 [41] 8 Clay Unit CFRP Sheet EB S M SG
18 [77] 30 Clay Masonry CFRP (5) GFRP Sheet EB S M SG, LES, FES
(25)
19 [25] 107 Clay (51) Unit (83) CFRP (6) Sheet EB D M LES
CaSi (50) Masonry (24) GFRP (101)
Sand (6)
20 [40] 12 Clay Unit (4) Mortar (3) CFRP Sheet EB S M LES, DIC
Masonry (5)
21 [59] 4 Tuff Unit BFRP Sheet EB S M SG
22 [22] 29 Clay (11) Unit CFRP (15) GFRP (14) Sheet EB 2D M LES
Tuff
(6) Lime
(12)
23 [24] 14 Clay Masonry CFRP Strip NSM S M (12) C SG, LES
(2)
24 [79] 25 Clay Unit CFRP (5) GFRP (10) Sheet EB S (20) D M (5) C SG, LES, FES
BFRP (5) (5) (20)
SRP (5)
25 [32] 1 Lime Unit CFRP Sheet EB D M SG
26 [86] 274 Clay Unit CFRP (68) GFRP (69) Sheet EB S (136) D M SG, LES, FES
BFRP (70) SRP (138)
(67)
27 [55] 16 Clay Unit (6) Masonry GFRP Sheet EB S M SG
(10)
28 [57] 17 Clay Unit (8) Masonry (9) CFRP Sheet EB S M SG
29 [87] 1 Clay Unit GFRP Sheet EB S M LES, FES
30 [52] 5 Clay Unit CFRP Sheet EB S M LES, DIC
31 [64] 39 Clay Masonry CFRP Strip NSM S M SG
32 [65] 6 Clay Unit CFRP Sheet EB S M LES
33 [67] 2 Clay Unit GFRP Sheet EB S M LES, FES
34 [42] 20 Clay Unit CFRP (5) GFRP Sheet EB S M SG, LES, FES
(5) BFRP
(5) SRP
(5)
35 [73] 15 Clay Unit CFRP Strip NSM S M LES
36 [39] 3 Clay Masonry CFRP Sheet EB S M LES
37 [18] 28 Clay Unit CFRP Sheet EB D M SG, DIC
38 [60] 14 Tuff Unit CFRP (4) GFRP (4) BFRP Sheet EB S M SG
(2) Flax (4)
39 [68] 25 Clay Unit GFRP Sheet EB S M DIC
40 [69] 1 Clay Unit GFRP Sheet EB S M LES, FES
41 [70] 14 Clay Masonry CFRP Sheet EB S M DIC
42 [76] 16 Clay Unit (8) Masonry (8) GFRP Sheet EB S M SG
43 [108] 8 Clay Masonry GFRP Sheet EB S M DIC
44 [80] 11 Lime Unit (2) Masonry GFRP Sheet EB S M LES, FES
(9)
45 [83] 222 Clay Unit CFRP Sheet EB S (54) D M SG, LES
(108) 2D
(60)
 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345 333

Table 3 (continued)

ID Reference No. Substrate Prism Type FRP Material FRP Shape Retrofit Laps* Loading** Deformation
Tests Material Config Measurement***
46 [84] 14 Lime (10) Unit CFRP (10) GFRP (4) Sheet EB D M SG
Tuff (4)
47 [50] 12 Clay (8) Unit (4) Masonry CFRP Sheet EB S M SG, LES
Conc (4) (8)
48 [56] 15 Clay Masonry GFRP Sheet EB S M SG
49 [21] 105 Clay Unit (39) Masonry CFRP (24) GFRP (39) Sheet EB S (81) D M SG, LES
(66) BFRP (21) SRP (24)
(21)
50 [88] 97 Clay Unit (54) Masonry CFRP (27) GFRP (33) Sheet EB S (74) D M SG, LES
(43) BFRP (11) SRP (23)
(26)
51 [74] 2 Clay Unit GFRP Sheet EB S M LES, FES
52 [75] 2 Clay Unit GFRP Sheet EB S M LES, FES
53 [45] 17 Clay (14) Unit (8) SRP Sheet EB S M LES, DIC
Tuff (3) Masonry (9)
54 [81] 1 Clay Unit GFRP Sheet EB S M SG, LES, FES,
55 [61] 7 Clay (2) Unit (5) Masonry CFRP (1) GFRP Sheet EB S M SG
Tuff (5) (2) (6)
56 [62] 28 Lime (16) Unit (14) Masonry CFRP (8) GFRP (4) BFRP Sheet EB S M SG, LES
Tuff (12) (14) (16)
*
S = single lap, D = double lap, 2D = two-block double lap.
**
M = monotonic, C = cyclic.
***
SG = strain gauges, LES = loaded-end slip, FES = free-end slip, DIC = digital image correlation. Note that measurement of displacement at the machine grip position (Dm in
Fig. 10 is not considered equivalent to LES if the unbonded length of FRP is greater than zero (Lu > 0).

Table 4
Summary of the number of unique masonry units tested across the 56 studies and number of instances where mechanical properties fuc, fut,dir, fut,split and fut,flex were reported.

Material Total No. unique units No. unique units where property reported
fuc fut dir fut flex fut split
Clay Brick 61 51 9 26 6
Limestone 9 9 – 8 1
Tuff 10 10 – 6 –
Concrete Block 7 6 – – –
Calcium Silicate Brick 5 5 5 – –
Sandstone 1 1 1 – –
Mortar 2 2 1 – –
All Substrates 95 84 16 40 7

dimensions of the plate were not specified in approximately 5% of
EB tests, meaning that this subset of data cannot be meaningfully
used.
In terms of their bond-slip behaviour, EB retrofits tend to exhi-
bit a defined ultimate slip capacity (df in Fig. 2) with the inability to
provide further shear stress transfer beyond this point. In compar-
ison, NSM retrofits have been shown to exhibit improved bond-slip
characteristics due to lateral confinement of the plate, which allow
for the ability to provide residual friction and thus a continual
increase to the resisted load beyond the initiation of debonding
[15,24,33,82].
4.1.4. Adhesive
Epoxy resin is the most commonly used bonding agent in both
EB sheet and NSM bar and strip retrofits, accounting for 93% of
tests (Fig. 6e). The remaining tests investigated cementitious [85]
and polyurethane adhesives [42,68,88].
A basic requirement of experimental design is for the adhesive
to have sufficient strength to avoid premature failure at the
bonded interface. In addition, the global behaviour of the retrofit
can be considerably influenced by the axial and shear deformabil-
ity of the adhesive [94,95].
In EB sheet retrofits, where the epoxy adhesive acts as the com-
posite matrix, the issue of the axial stiffness contribution of the
adhesive is most commonly addressed by measuring the effective
stiffness Eptp of the FRP composite through tensile tests on the
FRP composite (Fig. 8b). In both EB and NSM applications the thick-
ness and stiffness of the adhesive layer can further contribute to
the shear flexibility at the interface which can translate to the
effective bond-slip behaviour. This effect is partially recognised
by the CNR-DT 200 [8] technique in relation to the initial rising
branch in the bond-slip model; however, the adhesive flexibility
is likely to also influence the fracture energy and ultimate slip
[96], which is an issue that is yet to be properly addressed.
As there has in recent years been an increased experimental
focus on the influence of adhesive stiffness [42,68,88], there is a
growing need to explicitly model the shear deformation of the
adhesive in order to adequately describe test observations.
Although this is yet to be substantially investigated, this could be
done by incorporation of a shear deformable layer representing
the adhesive in Eq. (1). Making allowance for these effects through
analysis requires knowledge of the thickness and shear modulus of
the adhesive; however, only approximately half of the 56 studies
reported the E-modulus of the adhesive or the layer thickness. Sim-
ilarly, the size of the groove in NSM tests, which in turn influences
the thickness of the adhesive, is often not reported, making such
allowances difficult.
4.2. Test arrangements and methods
The three main types of setup used for pull-tests are shown in
Fig. 1, which are categorised by the number of lap joints present:
334 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345

Fig. 7. Substrate unit mechanical properties counted by individual pull-tests: (a) compressive strength, (b) direct tensile strength, (c) flexural tensile strength, and (d)
splitting tensile strength. Disaggregated according to substrate type with only the four most common shown. Proportion of tests plotted on vertical axis.

Table 5
Number of tests performed on combinations of composite material versus shape and configuration.

CFRP GFRP BFRP SRP AFRP Flax

Sheet (EB) 678 482 134 144 12 4
Strip (EB) 4 – – – – –
Strip (NSM) 107 – – – – –
Rod (NSM) 3 15 – – – –

Fig. 8. EB plate properties counted by individual pull-tests disaggregated according to composite material (C = CFRP, G = GFRP, B = BFRP, S = SRP, A = AFRP, F = Flax).
Proportion of tests plotted on vertical axis.
 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345
Fig. 9. NSM plate properties counted by individual pull-tests disaggregated according to composite material (C = CFRP, G = GFRP). Proportion of tests plotted on vertical axis.
single-lap, double-lap, or two-block double-lap. Each of these can
be performed in either a push-pull or pull-pull arrangement. Note
that these arrangements are also known by alternate terminology-
for instance, single- and double-lap tests are sometimes referred to
as single-shear and double-shear tests, respectively; and the push-
pull and pull-pull arrangements are sometimes respectively
referred to as near-end-supported and far-end-supported [97].
4.2.1. Single versus double lap
The single-lap test shown in Fig. 1a has been the slightly more
commonly used setup (53%, Fig. 6f). This test is generally per-
formed in the push-pull variant, where the plate is pulled while
the prism is subjected to compression by an abutting reaction
block. The appeal of the test lies in its simplicity and that it con-
sumes only a single FRP plate. The main drawback of the test is that
the loading eccentricity produces internal shear in the prism with
the potential for premature failure, and also generates a flexural
strain profile through the thickness. These conditions are however
also representative of walls under flexure, which are a common
target of FRP retrofits. A practical consideration of this setup is
ensuring that the loading mechanism is aligned with the axis of
the FRP to limit normal stresses at the interface. The position of
the restraint block is also likely to have an influence on overall
response [97]; however, there is presently no standardised proto-
col for this aspect of the arrangement.
The double-lap test accounts for 47% of tests and can be per-
formed using either a single- or double-block setup (Fig. 1b and
c), with the former being more common. In the double-lap test,
FRP plates are bonded on opposite sides of the prism and pulled
simultaneously. The main benefit of this technique is that it
removes loading eccentricity and thus minimises flexural strain
through the section. The test however can cause other unfavour-
able effects. For example, the single-block arrangement commonly
uses a single continuous fibre sheet looped over a pulley, which
creates equal load in each lap. However, stochastic variability of
mechanical properties means that one side fails before the other,
and thus the specimen needs to be rotationally restrained to avoid
peeling [98] on the non-failed side. Furthermore, it is standard to
quantify the failure load as half of the total applied load. While this
is technically correct, the fact that the test only ever measures the
strength of the weaker side introduces conservative bias into the
results. Eliminating this bias would require application of order
statistics in the processing of the results, but to the authors’ knowl-
edge this is not generally done. In the pulley arrangement, the fact
that the stronger side begins to unload once the weaker side
reaches its peak load also leads to the inability to accurately mea-
sure the loaded-end slip in the post-peak range.
335
A small subset of double-lap tests used a two-block arrange-
ment, in which two double-lap prisms were either pushed or
pulled apart (Fig. 1c). The benefit of this technique is that abutting
the two blocks minimises the unbonded length of plate (Lu in
Fig. 9) which creates arguably the most realistic representation of
a flexural crack and ensures alignment of the plate with the prism
surface. It also acts to mitigate the effect of any snap-through due
to the build-up of strain energy in the unbonded portion of the
plate (as discussed later). However, since the arrangement mea-
sures the weakest of the four individual laps, it introduces an even
greater statistical bias into the measured load capacity.
4.2.2. Push-pull versus pull-pull
Fig. 1 depicts the single-block, single-lap and double-lap tests (a
and b) in the more commonly used push-pull variant where the
substrate prism is subjected to compression using a reaction block.
In the two-block test (Fig. 1c) a compressive load becomes more
difficult to apply, but can be achieved either by using full-length
anchors through the prism or by introducing a gap between the
blocks and using an internal jack to push them apart.
An alternate option for each of these three arrangements is the
pull-pull variant which subjects the prism to tension; however this
is less common. The pull-pull condition allows for the possibility of
global cracking in the substrate and hence arguably a more realistic
representation of the state of stress in the debonding zone of a
flexural member. In terms of the governing mechanics of lap joints
[Eqs. (1) and (2)], the push-pull and pull-pull arrangements
become equivalent as long as flexibility of the prism remains
negligible relative to the plate and the tensile strength of the sub-
strate is sufficient to avoid cracking.
4.2.3. Loading
Monotonic loading was used in 98% tests (Fig. 6g). Only three
studies used cyclic loading histories and in each one the loading
phases consisted of loading, unloading and further reloading in
the same slip direction [24,71,79]. Whilst monotonic loading tests
are adequate for the investigation of static strength, the fact that
FRP retrofits are most commonly used to improve seismic perfor-
mance highlights a potential gap for future research, particularly
if ductility under cyclic loading is to be allowed for in the design.
It is worth noting that some progress on analytical simulation of
cyclic load-slip behaviour from cyclic bond-slip models has been
made in previous work by [99,100], which demonstrated cyclic
loading to influence the deformation capacity and ductility of the
retrofit. Results of the cyclic tests identified as part of this review
have further indicated that some strength degradation does occur
336 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345
upon reversal of slip [71]. However, additional study is required to
quantify the extent of cyclic degradation in bond-slip behaviour.
The potential for the FRP reinforcement to buckle in compres-
sion also warrants consideration; however, investigation of this
phenomenon may be better achieved through tests of full-height
walls rather than pull-tests (e.g. [2]).
4.2.4. Instrumentation
The various forms of uniaxial deformation measurement
include strain in the FRP, ep, loaded-end (LE) slip, D, and free-end
(FE) slip, Df, as shown in Fig. 10. Across the 56 studies there has
not been a consistent approach to the type of instrumentation used
as indicated in the last column of Table 3.
The most commonly used method of measuring longitudinal
strain in the FRP is using conventional foil strain gauges attached
to the plate. The main role of strain readings is that they allow the
shear stress and slip profiles to be obtained by differentiation and
integration, respectively, and therefore provide a direct means to
determine the local bond-slip behaviour. However as will be shown
in Section 5.2, the accuracy of processing strain gauge readings is
dependent on the measurement of a reference slip along the speci-
men or by taking alternate measures to ensure that the FE slip
remains zero.
Slip is defined as the relative displacement between the plate
and the prism. This can be measured in two-ways:
1. Either directly by attaching a displacement transducer to the
prism at the same longitudinal positon the slip being measured,
or
2. By measuring the displacement of the plate with respect to a
stationary reference point (e.g. test frame), but also measuring
the displacement of the prism relative to the same point to
allow for its deformation.
Measurement of loaded-end slip is necessary to generate P-D
curves required for extraction of bond-slip parameters using
inverse analysis, as described in Section 5.3. LE slip is defined as
the slip at the location where the bond initiates as indicated by
D in Fig. 10.
It is preferable to plot curves using LE slip in order to ensure a
standardised convention for reporting as well as consistency with
available analytical solutions for load-slip response (e.g., [27–
31,33,101]. However, in the works reviewed, some inconsistency
was found in terms of the location of slip measurement. While
most works used the LE slip (D), some others used the displace-
(a) Machine grip Loaded end Free end
Lu = Lu1 + Lu2 Lu2 Lu1 Lb
(b)
m f
P
Fig. 10. Measurement of longitudinal deformation; (a) definitions, and (b) direct
displacement measurement.
ment at the machine grip (Dm). A transformation between Dm
and D can be made by allowing for the elastic elongation over
the total unbonded length between the two locations (Lu in
Fig. 10 at any imposed load. Furthermore, as it is common to pro-
vide a short unbonded length at the loaded end of the prism (Lu1),
the elastic elongation over this length can sometimes be significant
and should also be allowed for.
In recent years, more sophisticated techniques such as digital
image correlation (DIC) or particle image velocimetry have also
emerged [18,40,52,68,70,76,81]. These methods alleviate the
issues raised above as they are capable of continuous deformation
measurements.
4.2.5. Snap-through effect
To allow for the observation of the post-peak behaviour, care
should be taken to minimise the overall flexibility of the test setup.
The main sources of flexibility can include the test rig and also the
plate over its total unbonded length, Lu, as shown in Fig. 10. There-
fore, it is preferable to minimise any unbonded length Lu2 as far as
possible. Excessive setup flexibility can give rise to a snap-through
instability caused by the accumulation and sudden release of the
elastic strain energy at the point of debonding. This generates a
sudden jump forward in the LE slip which make it difficult to reli-
ably estimate the ultimate slip capacity in the bond-slip relation-
ship (df in Fig. 2).
4.3. Failure modes
For a test to be usable toward the extraction of bond properties,
failure must occur by interfacial debonding, which involves the
detachment of the plate along with a layer of the substrate. Alter-
native failure modes include tensile rupture of the FRP, failure
along the adhesive, or failure of the prism by either shear, tension
or crushing. The prevalence of different failure modes observed in
individual tests is shown in Fig. 11, with debonding occurring in
86% of tests and debonding in addition to another failure in a fur-
ther 2%. The remaining 12% of tests exhibited FRP rupture, adhesive
failure, prism failure, or a combination of these.
It should be noted that whilst the practice of discarding tests
where failure is by a mode other than debonding is commonplace,
this potentially introduces conservative bias into the data pool
Strain
gauges
Fig. 11. Failure modes reported across all individual tests. ‘Other’ refers to non-
debonding failure modes, that is, either adhesive failure, FRP rupture, or prism
failure.
 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345
since tests with a high debonding capacity can become disre-
garded. Prevention of failure modes other than debonding can, to
some degree, be managed through the experimental design. For
example, due to scaling effects, a thin EB sheet is more prone to
tensile rupture as opposed to debonding than a thick sheet
[through Eq. (3)]. Similarly, a small diameter NSM bar is more
prone to rupture compared to a larger diameter.
5. Approaches for analysis of experimental data
Approximately one third of the studies reviewed undertook
some form of bond property extraction. These included two tiers
of data analysis. The simpler of these involves estimation of the
fracture energy (Gf), whose predictive ability in turn becomes
restricted to the debonding load capacity PIC without a direct
means to quantify the required bond length (Lc). The higher tier
involves estimation of the entire set of properties defining the
bond-slip model which is considerably more rigorous but allows
for prediction of PIC, Lc, load-slip response, and the plate’s stress
and strain under a particular load. The two main approaches to
extracting bond-slip behaviour are through strain measurements
or by inverse analysis. Of these, extraction via strain measurements
was found to be by far the more common approach in the studies
reviewed; however as both methods are viable they will both be
discussed.
5.1. Extraction of fracture energy
The simplest form of data analysis involves quantifying the
bond fracture energy from the measured debonding force (Pmax)
using the following rearrangement of Eq. (3):
Gf ¼ P2max =ð2Ep Ap Lp Þ
This can be done without the necessity to establish the precise
shape of the bond-slip model (see Fig. 2) and by using only the
measurement of the peak load. This approach however makes is
the a priori assumption that the tested bond length (L) exceeds
Lc. In cases where L < Lc the quantified Gf underestimates the actual
value. This assumption can be verified using several different
ways:

By performing tests over increasing bonded lengths until the
debonding force no longer increases;

By ensuring that measured FE slip (Fig. 10) remains negligible at
the instance that the maximum load is reached; or

By establishing the shear stress distribution from strain gauge
readings (refer to Section 5.2) and ensuring that Pmax was
achieved prior to shear stress reaching the free end.
However, the review undertaken has identified a small number
of works providing insufficient detail to ascertain whether any of
these verification approaches were undertaken.
5.2. Bond-slip extraction using strain gauge readings
The main role of strain readings is that they allow the shear
stress and slip profiles to be obtained by differentiation and inte-
gration, respectively, and therefore provide a direct means to
determine the local bond-slip behaviour. The standard way to
implement this procedure is as follows.
The slip profile d(x) is defined as the integral of plate strain such
that
Z x Z x
dðxÞ ¼ dref þ ðep  em Þdx  dref þ ep dx
xref xref
337
where dref is a known reference slip at location x = xref. For discrete
strain gauge readings at nodes i and i + 1, strain is assumed to vary
linearly within the interval. From this, slip along the interval xi  x
 xi+1 is formulated as


1 ep;iþ1  ep;i
dðxÞ ¼ di  ep;i ðx  xi Þ  ðx  xi Þ2 ð7Þ
2 xiþ1  xi
where di is the known slip at node i; and ep,i and ep,i+1 are the mea-
sured strains. Using Eq. (7) the slip at each gauge location is com-
puted by starting at a reference location where the slip is known
by direct measurement (usually at the loaded or free end) and step-
ping along the prism. The slip at the midpoint between each pair of
gauges, di+1/2, is obtained by substituting x = (xi + xi+1)/2 into Eq. (7).
The shear stress profile is in turn defined as the derivative the
plate strain such that
dep Ep Ap
sðxÞ ¼  ð8Þ
dx Lp
and thus, the average shear stress along each gauge interval is


ep;iþ1  ep;iþ1 Ep Ap
siþ1=2 ¼  ð9Þ
xiþ1  xiþ1 Lp
Using this technique, the experimental bond-slip curve is gen-
erated by plotting si+1/2 versus di+1/2 at each load level of applied
load.
The appeal of this approach lies in the fact that it does not
require any prior knowledge of the shape of the s-d relationship,
and thus it is well suited to investigating new retrofitting schemes
or technologies. Despite its conceptual simplicity, however, this
process typically produces bond-slip curves that are highly irregu-
lar in shape, and often the same test can produce completely differ-
ð5Þ ent curves at different levels of applied load. Consequently,
substantial user interpretation can still be required to select a par-
ticular bond-slip law to fit to the data (e.g. blinear), and also to
select a suitable metric for quantifying goodness-of-fit in model
parameter identification. To enable equitable comparison between
properties extracted from different studies, a consistent protocol is
required by the research community with regard to these issues
(e.g. [102]).
The potential sources of error are:
1. Distorted strain readings due to local substrate heterogeneity
and presence of micro-cracks,
2. Excessive strain gauge spacing,
3. Cumulative error in slip introduced by strain integration (due to
1 and 2),
4. Scatter in shear stress introduced by strain differentiation (due
to 1 and 2), and
5. The often-made and potentially incorrect assumption of zero
free-end slip in calculating the d profile.
Of the aforementioned error sources, items 2 and 5 are control-
lable by the user and therefore, let us consider their effect on the
observable bond-slip through a synthetic example.
Consider an example in which we use basic uniaxial shear lap
theory [Eqs. (1) and (2)] to synthesise idealised data representing
a pull-test. To generate the data, assume that the underlying
bond-slip law is defined by the bilinear model (Fig. 4a) with sf =
6 MPa, df = 0.4 mm, d1 = 0.1 mm and the plate is an externally-
bonded 50  2 mm FRP strip with Ap = 100 mm2, Lp = 50 mm, and
Ep = 150,000 MPa. For this system, Eqs. (3) and (4) give PIC = 42.4
kN and Lc = 192 mm. For a range of bonded lengths, including
400, 240, 160, and 80 mm, a numerical analysis was used solve
ð6Þ Eqs. (1) and (2) and generate the continuous distribution of strain
ep along the prism at monotonically increasing LE slip. The associ-
338 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345
ated load-slip curves at each bonded length are shown in Fig. 12a.
The analysis was performed using the numerical technique
described by Haskett et al. [34], but other equivalent analytical
approaches such as those by Caggiano et al. [29] and Vaculik
et al. [33] could also be used.
The continuous distribution of plate strain was sampled at 80,
40 and 10 mm intervals to represent gauge readings. From these,
bond-slip curves were generated using the conventional approach
[by Eqs. (7) and (9)] and are plotted in Fig. 13. Each graph further
compares observable behaviour by taking the reference slip [dref in
Eq (6)] as the actual ‘measured’ loaded-end slip (‘o’ markers) to the
assumption that free-end slip always remains zero (‘x’ markers) as
is often done [96].
The following observations can be made:

The assumption of zero FE slip can lead to significant underes-
timation of the true slip, irrespective of gauge spacing. Accept-
able agreement with actual behaviour occurs only over long
bonded lengths (here 400 mm) and only before debonding
propagates toward the free end. However the validity of these
conditions cannot be known a priori, thus highlighting the need
to either measure a reference slip or to manually examine the
shear stress profile to ensure that shear stress (and therefore
slip) remains zero at the free end. Notably, a number of the
works reviewed performed bond-slip extraction without mea-
suring a reference slip, and thus it is not clear how these consid-
erations were dealt with.

Strain gauge spacing affects predominantly the ability to cap-
ture the sharpness of the true bond-slip relationship. In this sce-
nario, 10 and 40 mm spacing captures the true behaviour
comparatively well, whereas the 80 mm spacing leads to a loss
of resolution near d1 and df. The adequacy of gauge spacing is
expected to depend on the ratio of gauge spacing to Lc, which
in this example is 0.05, 0.21 and 0.42 (for 10, 40 and 80 mm).
It should be noted that this idealised scenario does not consider
distortion of strain readings due to heterogeneity of the mate-
rial, which would introduce additional error that could be mit-
igated by closer gauge spacing.
Some more recent studies have addressed the discontinuity of
strain measurements by using techniques such as DIC which are
capable of continuous readings [18,40,52,68,70,76,81]. The tech-
Fig. 12. Theoretical load-slip behaviour obtained by uniaxial shear lap theory, including: (a) load-slip curves for different bonded lengths, (b) relationship between peak load
and bonded length, and (c) influence of bonded length on yield slip and slip at debonding. The examples shown assume constant bond-slip properties defined using the
bilinear model with sf = 6 MPa, df = 0.4 mm, d1 = 0.1 mm, and an externally-bonded 50  2 mm FRP strip with Ep = 150,000 MPa.
niques for processing such data are dealt with in other works
(e.g., [66,103]).
It should be noted that direct measurement of a reference slip is
not the only viable way of guaranteeing the method’s accuracy. An
alternate strategy is to examine the free-end shear stress through
differentiated strain readings and ensuring that FE stress is suffi-
ciently small to validate the assumption of zero FE slip. However,
this approach requires the user to assume a reasonable acceptance
tolerance, which is in turn dependent on the bonded length and the
bond-slip characteristics, with the latter being unknown a priori.
Thus the direct measurement of a reference slip is the more reli-
able approach.
5.3. Bond-slip extraction using load-slip inverse analysis
As an alternative to using strain readings, bond-slip parameter
identification can be performed by the method of inverse analysis
[33,34,96,104]. This method involves postulating a specific type of
bond-slip law (see Fig. 2) and varying values of the defining inputs
(e.g. sf, df and d1 in the bilinear model) until good agreement is
obtained between the measured and predicted P-D curves. This
process relies only on the measurement of the LE slip and thus sig-
nificantly reduces the amount of instrumentation required. How-
ever, it also requires more intense post-processing and
interpretation of results. A proposed framework for performing
inverse analysis as described in Vaculik et al. [33] will now be
demonstrated through a synthetic example. A novel aspect of this
framework is that it does not require a priori knowledge of whether
the test was performed over a long or short bonded length and can
thus be applied generically.
5.4. Example using synthetic experimental data
As before (Section 5.2), let us generate synthetic experimental
data using the bilinear bond-slip law at the parameter values
sf = 6 MPa, df = 0.4 mm, d1 = 0.1 mm for constant values of Ep, Ap,
and Lp. Consider bonded lengths of 400, 240, 160, and 80 mm,
noting that the critical length for this set of parameter inputs is
Lc = 192 mm, which defines the shortest length to get progressive
debonding (that is, Ddb > df; with Ddb defined in Fig. 12a inset).
The resulting load-slip curves, shown in Fig. 12a, represent exper-
imental data in which we wish to identify the underlying bond-slip
 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345 339

Fig. 13. Bond-slip curves obtained by applying the strain gauge extraction approach to synthesised data. The cases considered use the same bilinear bond-slip properties as
the reference example in Fig. 12. The approach is applied to different bonded lengths of 400, 240, 160, and 80 mm (left to right), and strain gauge spacing of 10, 40, and 80 mm
(top to bottom). Each graph further compares observable behaviour obtained using an uncorrected slip profile based on the assumption zero FE slip (‘x’ markers) to a
corrected slip profile that sets the LE slip to its actual value (‘o’ markers).

parameters (sf, df and d1). For convenience and computational effi- of the influence of these parameters on the shape of the load-slip
ciency, load-slip curves were generated using the analytical formu- curves are available in other works [33,105].
lation described in Vaculik et al. [33]; however, other equivalent
techniques could also be used (e.g. [29,34]). 5.6. Effect of varying sf and d1 at constant df

First, consider the effect of varying sf and d1 at constant df = 0.4

5.5. Goodness-of-fit metrics
mm. Fig. 14 shows the results of this study in terms of error maps
for err(rms), err(Pmax) and err(Ko) as well as contour plots of Pmax
The goodness-of-fit between a hypothesised analytical P-D
and Ko. Regions where good fit is achieved for the different proper-
curve and the experimental one can be quantified using alternate
ties are indicated by dark zones in the associated error maps. This
metrics.
means that unique estimates of both independent parameters with
A common approach for quantifying the overall similarity
respect to a particular error metric can only be obtained where the
between two curves is with the root-mean-squared (RMS)
error becomes minimised over a small localised patch rather than
error:
an elongated streak.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
R1 The following observations can be made with regard to using
0
ðPref  Phyp Þ2 dD the various error measures to identify parameters sf and d1 over
errðrmsÞ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
R1 2 ð10Þ
P ref dD different bonded lengths (L):
0

RMS error is capable of identifying sf over any L, but can identify

where Pref and Phyp are the load ordinates of the reference and
hypothesised curves.
In addition, consider various key points along the P-D curve
including peak strength, Pmax, initial stiffness, Ko, debonding slip,
Ddb, and ‘yield’ slip Dy (taken here as the slip at which 90% of Pmax
is achieved). These properties are shown inset within Fig. 12a.
Define the error in any of these properties as
errðXÞ ¼ jX ref  X hyp j=X ref
where Xref and Xhyp are the respective property values on the refer-
ence curve and hypothesised curve.
The inverse analysis technique relies on minimising the error
between the hypothesised (theoretical) curve and reference
(experimental) curve. Thus, let us consider the parametric influ-
ence of sf, d1 and df on the various error metrics. Visual examples
d1 only over short L (Fig. 14a)

Peak strength Pmax can be used to identify sf within all ranges of
L, but it cannot be used to identify d1 (Fig. 14b and c).

Initial stiffness Ko cannot be used to identify unique values of sf
nor d1, but it can be used to identify their ratio, sf/d1 (Fig. 14d
and e). Notably, Ko is independent of df and therefore sf/d1 can
be identified from the outset of the analysis.
ð11Þ
5.7. Effect of varying sf and df at constant d1
Having established the ratio sf/d1, now consider the effect of fix-
ing its value (60 MPa/mm in this example) and varying the remain-
ing parameters sf and df. Fig. 15 shows the corresponding results
including error maps for err(rms), err(Pmax), err(Ddb), and err(Dy)
as well as contour plots of Pmax, Ddb, and Dy.
340 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345

Fig. 14. Effect of varying sf and d1 under constant df = 0.4 mm shown in terms of contour plots for Pmax and Ko, and error maps for err(P-D rms), err(Pmax) and err(Ko). Error
maps are calculated with respect to the reference example in Fig. 12, based on sf = 6 MPa, df = 0.4 mm, d1 = 0.1 mm. Regions of good fit are indicated by dark zones on the
respective error maps. Dashed line indicates the critical bonded length (Lc) which forms the boundary between long length above and short length below.

The following observations can be made with regard to identi-
fying sf and df:

As seen from Fig. 15a, RMS error is capable of identifying both
parameters sf and df but only over short L. Over long L, the error
is minimised over a zone where sf / 1/df, which means that
the error is sensitive to the fracture energy Gf (since Gf = ½ sf
df), hence being able to identify Gf but not unique values of sf
and df.

Similarly, peak strength can be used to identify Gf over long L
only, and sf over short L only (Fig. 15b and c).

Debonding slip Ddb can be used to identify Gf over long L only,
and df over short L only.

Yield slip Dy can be used to identify df over long L only, and sf
over short L only.
The combined error is plotted in the bottom row of Fig. 15,
where it is seen that minimisation of the error only occurs in close
proximity to the reference parameter values.
In order to demonstrate the main trends, the preceding example
considered an idealised scenario where experimental data was
synthesised using the same type of bond-slip law as that used in
the inverse analysis process. Applying this technique to real exper-
imental data however requires consideration of various practical
limitations, including:

smoothness of the data,

the fact it will generally not be possible to minimise all error
metrics simultaneously, and

system flexibility which can cause a jump in measured slip at
the point of debonding (snap-through, refer to Section 4.2).

The above observations demonstrate that no single property
(Pmax, Ddb or Dy) is sufficient to be able to establish either sf or df,
due to the behavioural difference over short and long bonded
lengths. Importantly however, these properties are shown to be
complimentary in that sf and df can be simultaneously identified
by finding zones of overlap where errors err(Pmax), err(Ddb) and
err(Dy) each become small. This is demonstrated by devising a
combined error metric that considers each of these error simulta-
neously, such that
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
errðPmax ; Ddb ; Dy Þ ¼ ðerrðPmax Þ2 þ errðDdb Þ2 þ errðDy Þ2 Þ
3 for approximately 10% of the total number of tests. The majority of
Application of this method to published experimental data is
also somewhat limited due to incomplete reporting of P-D curves.
A further discussion these difficulties is provided in Vaculik et al.
[33].
6. Reported bond-slip properties
Sixteen of the 56 studies reviewed undertook some form of
bond-slip property extraction from individual test data. From
ð12Þ these, values of extracted properties (any of Gf, sf, df) were reported
these studies used the strain gauge technique [24,26,40,53–56,58,
 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345 341

Fig. 15. Effect of varying sf and df under constant sf/d1 = 60 MPa/mm shown in terms of contour plots for Pmax, Ddb and Dy, and error maps for err(P-D rms), err(Pmax), err(Ddb),
err(Dy), and err(Pmax, Ddb, Dy). Errors maps are calculated with respect to the reference example in Fig. 12, based on sf = 6 MPa, df = 0.4 mm, d1 = 0.1 mm. Dashed line indicates
the critical bonded length (Lc) which forms the boundary between long length above and short length below.

61,71,77,78,82,84] whilst extraction from DIC deformation measure-
ments was also undertaken in one instance [68]. Extraction by load-
slip inverse analysis was limited only to the work by Barbieri et al.
[39] and Ceroni et al., 2014b [44]. It is not entirely clear why this
data processing technique has been somewhat underutilised in
masonry bond research, since it is a viable alternative to strain read-
ings that has gained widespread acceptance in concrete bond
research [34,96,104,106,107].
It is commonplace for FRP bond models to express properties sf,
df and Gf as a function of the masonry unit’s compressive strength
fuc and/or flexural tensile strength fut. To visualise this dependency
across the aggregated dataset, Fig. 16 plots each bond property
versus each fundamental mechanical property based on the
reported values. The bottom part of the figure also plots Gf divided
by the width factor kb, as per the CNR-DT 200 [8] formulation,
p
where kb = ((3-bp/bs)/(1 + bp/bs)). The figure demonstrates consid-
erable scatter to the extent that no clear trends can be readily iden-
tified between any of the properties.
It is important to note that data in Fig. 16 have been plotted
without consideration of the rigour used to apply the method in
the individual studies (e.g. integration of strain gauge readings
with/without measurement of a reference slip). This approach
has been taken to demonstrate that the scatter of results in the
existing work may result not only due to variability in mechanical
properties (due to natural variability and inconsistency in testing
methods) but also due to the possible error introduced in the
extraction process (refer to Section 5). Together, these issues high-
light the need for careful consideration of experimental design in
342 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345
Fig. 16. Extracted bond-slip properties sf, df, Gf, and Gf /kb as reported throughout individual studies, plotted against: (a) substrate unit compressive strength, fuc, and (b)
substrate unit tensile strength, fut, determined by flexure.
conjunction with the method of data analysis—and perhaps, the
need to develop a standardised methodology-in order to maximise
the reliability of data toward the development of material models.
7. Summary and recommendations
The design of FRP retrofits in masonry structures requires the
development of material models for the local bond behaviour
between the FRP and the substrate, so that they can be incorpo-
rated into design guides (e.g. [7,8]). To this end, this paper has pre-
sented a state-of-the-art review of past experimental
investigations involving pull-tests in order to identify the scope
of testing and critically review the experimental arrangements
and data extraction techniques. The review has collated the results
of 56 studies comprising a total of 1583 individual tests which
have been incorporated into a publically available database (see
accompanying Data in Brief article).
Of the total number of individual tests performed:

The predominant substrates include clay brick (82% of tests)
and natural stone (10%);

Specimens comprise both unit (75%) and masonry (25%) prisms;

Retrofit configurations include externally-bonded (92%) and
NSM (8%); and

Composite material comprise mainly CFRP (50%), GFRP (31%),
SRP (9%), and BFRP (9%).
From this review it is evident that the primary gaps in lie in fur-
ther investigation of the following:

The retrofit of different varieties of stone masonry, particularly
as such structures are among the most earthquake-prone;

The impact of mortar joints and their effect on required anchor-
age lengths; and
 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345

The influence of cyclic loading, especially considering that a key
driver for retrofitting masonry structures is to improve seismic
performance.
A major finding of this review is the need to standardise test
and instrumentation arrangements, reporting guidelines, and
bond-slip parameter extraction techniques. The following are the
main recommendations:

In the authors’ opinion, as long as proper alignment between
the plate and prism can be provided, the single-lap test is pre-
ferred to the double-lap test, as the latter introduces statistical
bias into the observed load capacity.

To enable for extraction of bond-slip properties using inverse
analysis, it is required that load-slip curves are reported up to
the point of full debonding with slip measured precisely at
the loaded end.

Any unbonded length of plate between the prism and machine
grip should be minimised to limit the influence of snap-through
and provide the potential for observation of the post-peak
response.

Certain bond-slip property extraction strategies, including
quantification of fracture energy by Eq. (5), rely on the assump-
tion of having a long bonded length (exceeding critical length),
and this assumption needs to be verified to ensure the reliabil-
ity of the quantified values.

The use of strain gauges for bond-slip extraction should be com-
plemented by the measurement of a reference slip.

Robustness of bond-slip parameter identification by inverse
analysis can be improved by considering multiple key points
along the load-slip curve.
Further limiting the ability to use existing test data for develop-
ment of bond-strength models are inconsistencies relating to
accompanying mechanical property tests on the substrate unit.
To this end, the following is recommended:

Both the compressive and tensile strength should be quantified
wherever possible.

There is a strong need to standardise the type of mechanical
property tests performed with pull-tests. The most common
techniques in studies to date have been cube tests to obtain com-
pressive strength and flexural tests to obtain tensile strength.

Reporting should include the test methods, specimen dimen-
sions, number of specimens, result variability (COV), and clarity
regarding allowance for size and aspect ratio effects in the
reported values.
Conflict of interest
None.
Acknowledgements
The authors gratefully acknowledge the financial support from
the ARC through its Discovery Grant Program (DP140102695) and
the Bushfire and Natural Hazards CRC. The findings and views
expressed are those of the authors and not necessarily those of
the sponsors.
References
[1] Burgoyne, C. (2009). ‘‘Fibre reinforced polymers–strengths, weaknesses,
opportunities and threats.” Proc., 9th International Symposium on Fiber
Reinforced Polymer Reinforcement for Concrete Structures (FRPRCS-9).
343
[2] M.C. Griffith, J. Kashyap, M.S. Mohamed Ali, Flexural displacement response of
NSM FRP retrofitted masonry walls, Constr. Build. Mater. 49 (2013) 1032–
1040.
[3] K.M.C. Konthesingha, M.J. Masia, R.B. Petersen, A.W. Page, Experimental
evaluation of static cyclic in-plane shear behavior of unreinforced masonry
walls strengthened with NSM FRP strips, J. Compos. Constr. 19 (3) (2015).
[4] M.D. Kuzik, A.E. Elwi, J.J.R. Cheng, Cyclic flexure tests of masonry walls
reinforced with glass fiber reinforced polymer sheets, J. Compos. Constr. 7 (1)
(2003) 20–30.
[5] V. Turco, S. Secondin, A. Morbin, M.R. Valluzzi, C. Modena, Flexural and shear
strengthening of un-reinforced masonry with FRP bars, Compos. Sci. Technol.
66 (2) (2006) 289–296.
[6] C.R. Willis, R. Seracino, M.C. Griffith, Out-of-plane strength of brick masonry
retrofitted with horizontal NSM CFRP strips, Eng. Struct. 32 (2) (2010) 547–
555.
[7] ACI 440.7R-10, Guide for the Design and Construction of Externally Bonded
fibre-Reinforced Polymer Systems for Strengthening Unreinforced Masonry
Structures, American Concrete Institute, Farmington Hills, MI, US, 2010.
[8] CNR-DT 200 R1-2013, Guide for the Design and Construction of Externally
Bonded FRP Systems for Strengthening Existing Structures, National Research
Council, Italy, 2013.
[9] O. Volkersen, Die Nietkrafverteilung in zugbeanspruchten Nietverbindungen
mit konstanten Laschenquerschnitten, Luftfahrtvorschung 15 (1938) 41–47.
[10] P. Holzenkämpfer, Ingenieurmodelle des Verbunds geklebter Bewehrung für
Betonbauteile PhD Thesis, TU Braunschweig, Germany, (in German), 1994.
[11] U. Neubauer, F.S. Rostasy, Design Aspects of Concrete Structures
Strengthened with Externally Bonded CFRP-Plates, Proc., 7th International
Conference on Structural Faults and Repair, 1997, pp. 109–118.
[12] B. Täljsten, Strengthening of concrete prisms using the plate-bonding
technique, Int. J. Fract. 82 (3) (1996) 253–266.
[13] J.F. Chen, J.G. Teng, Anchorage strength models for FRP and steel plates
bonded to concrete, J. Struct. Eng. 127 (7) (2001) 784–791.
[14] X.Z. Lu, J.G. Teng, L.P. Ye, J.J. Jiang, Bond-slip models for FRP sheets/plates
bonded to concrete, Eng. Struct. 27 (6) (2005) 920–937.
[15] D.J. Oehlers, R. Rashid, R. Seracino, IC debonding resistance of groups of FRP
NSM strips in reinforced concrete beams, Constr. Build. Mater. 22 (7) (2008)
1574–1582.
[16] R. Seracino, M.R. Raizal Saifulnaz, D.J. Oehlers, Generic debonding resistance of
EB and NSM plate-to-concrete joints, J. Compos. Constr. 11 (1) (2007) 62–70.
[17] U.S. Camli, B. Binici, Strength of carbon fiber reinforced polymers bonded to
concrete and masonry, Constr. Build. Mater. 21 (7) (2007) 1431–1446.
[18] F.G. Carozzi, P. Colombi, C. Poggi, Calibration of end-debonding strength
model for FRP-reinforced masonry, Compos. Struct. 120 (2015) 366–377.
[19] P. Carrara, F. Freddi, Statistical assessment of a design formula for the
debonding resistance of FRP reinforcements externally glued on masonry
units, Compos. B Eng. 66 (2014) 65–82.
[20] F. Ceroni, B. Ferracuti, M. Pecce, M. Savoia, Assessment of a bond strength
model for FRP reinforcement externally bonded over masonry blocks,
Compos. B Eng. 61 (2014) 147–161.
[21] G. de Felice, M.A. Aiello, A. Bellini, F. Ceroni, S. De Santis, E. Garbin, M. Leone,
G.P. Lignola, M. Malena, C. Mazzotti, M. Panizza, M.R. Valluzzi, Experimental
characterization of composite-to-brick masonry shear bond, Mater. Struct. 49
(7) (2016) 2581–2596.
[22] C. Faella, G. Camorani, E. Martinelli, S.O. Paciello, F. Perri, Bond behaviour of
FRP strips glued on masonry: Experimental investigation and empirical
formulation, Constr. Build. Mater. 31 (2012) 353–363.
[23] E. Grande, M. Imbimbo, E. Sacco, Simple model for bond behavior of masonry
elements strengthened with FRP, J. Compos. Constr. 15 (3) (2011) 354–363.
[24] J. Kashyap, C.R. Willis, M.C. Griffith, J.M. Ingham, M.J. Masia, Debonding
resistance of FRP-to-clay brick masonry joints, Eng. Struct. 41 (2012) 186–
198.
[25] W. Seim, U. Pfeiffer, Local post-strengthening of masonry structures with
fiber-reinforced polymers (FRPs), Constr. Build. Mater. 25 (8) (2011) 3393–
3403.
[26] C.R. Willis, Q. Yang, R. Seracino, M.C. Griffith, Bond behaviour of FRP-to-clay
brick masonry joints, Eng. Struct. 31 (11) (2009) 2580–2587.
[27] M.S. Mohamed Ali, D.J. Oehlers, M.C. Griffith, R. Seracino, Interfacial stress
transfer of near surface-mounted FRP-to-concrete joints, Eng. Struct. 30 (7)
(2008) 1861–1868.
[28] H. Yuan, J.G. Teng, R. Seracino, Z.S. Wu, J. Yao, Full-range behavior of FRP-to-
concrete bonded joints, Eng. Struct. 26 (5) (2004) 553–565.
[29] A. Caggiano, E. Martinelli, C. Faella, A fully-analytical approach for modelling
the response of FRP plates bonded to a brittle substrate, Int. J. Solids Struct. 49
(17) (2012) 2291–2300.
[30] P. Cornetti, A. Carpinteri, Modelling the FRP-concrete delamination by means
of an exponential softening law, Eng. Struct. 33 (6) (2011) 1988–2001.
[31] J. Dai, T. Ueda, Y. Sato, Unified analytical approaches for determining shear
bond characteristics of FRP-concrete interfaces through pullout tests, J. Adv.
Concr. Technol. 4 (1) (2006) 133–145.
[32] M.S. Sciolti, M.A. Aiello, M. Frigione, Influence of water on bond behavior
between CFRP sheet and natural calcareous stones, Compos. B Eng. 43 (8)
(2012) 3239–3250.
[33] J. Vaculik, A.B. Sturm, P. Visintin, M.C. Griffith, Modelling FRP-to-substrate
joints using the bilinear bond-slip rule with allowance for friction – Full-
range analytical solutions for long and short bonded lengths, Int. J. Solids
Struct. 135 (2018) 245–260.
344 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345
[34] M. Haskett, D.J. Oehlers, M.S. Mohamed Ali, Local and global bond
characteristics of steel reinforcing bars, Eng. Struct. 30 (2) (2008) 376–383.
[35] J. Kashyap, M.C. Griffith, M.S. Mohamed Ali, D.J. Oehlers, Prediction of load-slip
behavior of FRP retrofitted masonry, J. Compos. Constr. 15 (6) (2011)
943–951.
[36] F. Focacci, C. Carloni, Periodic variation of the transferable load at the FRP-
masonry interface, Compos. Struct. 129 (2015) 90–100.
[37] M. Malena, F. Focacci, C. Carloni, G. de Felice, The effect of the shape of the
cohesive material law on the stress transfer at the FRP-masonry interface,
Compos. B Eng. 110 (2017) 368–380.
[38] M. Accardi, C. Cucchiara, L. La Mendola, Bond Behavior Between CFRP Strips
and Calcarenite Stone, Proc., 6th International Conference on Fracture
Mechanics of Concrete and Concrete Structures, 2007, pp. 17–22.
[39] G. Barbieri, L. Biolzi, M. Bocciarelli, S. Cattaneo, Pull out of FRP reinforcement
from masonry pillars: experimental and numerical results, Compos. B Eng. 69
(2015) 516–525.
[40] C. Carloni, K.V. Subramaniam, FRP-masonry debonding: numerical and
experimental study of the role of mortar joints, J. Compos. Constr. 16 (5)
(2012) 581–589.
[41] E. Grande, M. Imbimbo, E. Sacco, Bond behaviour of CFRP laminates glued on
clay bricks: experimental and numerical study, Compos. B Eng. 42 (2) (2011)
330–340.
[42] A. Kwiecień, Shear bond of composites-to-brick applied with highly
deformable, in relation to resin epoxy, interface materials, Mater. Struct./
Materiaux et Constr. 47 (12) (2014) 2005–2020.
[43] S. Xia, D. Oehlers, Debonding mechanisms in FRP plated unreinforced
masonry under out-of-plane loading, Adv. Struct. Eng. 9 (5) (2006)
619–637.
[44] F. Ceroni, G. de Felice, E. Grande, M. Malena, C. Mazzotti, F. Murgo, E. Sacco, M.
R. Valluzzi, Analytical and numerical modeling of composite-to-brick bond,
Mater. Struct. 47 (12) (2014) 1987–2003.
[45] A. Napoli, G. de Felice, S. De Santis, R. Realfonzo, Bond behaviour of Steel
Reinforced Polymer strengthening systems, Compos. Struct. 152 (2016) 499–
515.
[46] R. Seracino, N.M. Jones, M. Mohamed Ali, M.W. Page, D.J. Oehlers, Bond
strength of near-surface mounted FRP strip-to-concrete joints, J. Compos.
Constr. 11 (4) (2007) 401–409.
[47] M. Accardi, L. La Mendola, Stress Transfer at the Interface of Bonded Joints
Between FRP and Calcarenite Natural Stone, Proc., 4th International Seminar
of Structural Analysis of Historical Constructions, 2004, pp. 867–874.
[48] M.A. Aiello, M.S. Sciolti, Influence of Environmental Agents on Bond Between
FRP Reinforcement and Calcarenite Ashlars, Proc., 4th International
conference of Structural Analysis of Historical Constructions, 2004, pp.
875–881.
[49] M.A. Aiello, M.S. Sciolti, Bond analysis of masonry structures strengthened
with CFRP sheets, Constr. Build. Mater. 20 (1–2) (2006) 90–100.
[50] O. Anil, C. Durucan, S.W. Din, Experimental study on the stress distribution at
the interface between CFRP and three different types of masonry units,
Compos. B Eng. 92 (2016) 63–73.
[51] S. Briccoli Bati, M. Fagone, An Analysis of CFRP-Brick Bonded Joints, Proc.,
XVIII Conference of the Italian Group of Computational Mechanics (GIMC),
2010, pp. 22–24.
[52] C. Caggegi, V. Pensee, M. Fagone, M. Cuomo, L. Chevalier, Experimental global
analysis of the efficiency of carbon fiber anchors applied over CFRP
strengthened bricks, Constr. Build. Mater. 53 (2014) 203–212.
[53] R. Capozucca, Behaviour of CFRP Sheets Bonded to Historical Masonry, Proc.,
Asia-Pacific Conference on FRP in Structures, 2007, pp. 723–728.
[54] R. Capozucca, Experimental FRP/SRP-historic masonry delamination, Compos.
Struct. 92 (4) (2010) 891–903.
[55] R. Capozucca, Effects of mortar layers in the delamination of GFRP bonded to
historic masonry, Compos. B Eng. 44 (1) (2013) 639–649.
[56] R. Capozucca, V. Ricci, Bond of GFRP strips on modern and historic brickwork
masonry, Compos. Struct. 140 (2016) 540–555.
[57] P. Carrara, D. Ferretti, F. Freddi, Debonding behavior of ancient masonry
elements strengthened with CFRP sheets, Compos. B Eng. 45 (1) (2013)
800–810.
[58] M. Casareto, A. Oliveri, A. Romelli, S. Lagomarsino, Bond Behavior of FRP
Laminates Adhered to Masonry, Proc., International Conference Advancing
with Composites, 2003.
[59] F. Ceroni, M. Pecce, A. Garofano, Bond Tests on Masonry Elements Externally
Strengthened with FRP Materials, Proc., 6th International Conference on FRP
Composites in Civil Engineering (CICE), 2012, pp. 13–15.
[60] F. Ceroni, A. Garofano, M. Pecce, Bond tests on tuff elements externally
bonded with FRP materials, Mater. Struct./Materiaux et Construct. 48 (7)
(2015) 2093–2110.
[61] F. Ceroni, Bond tests to evaluate the effectiveness of anchoring devices for
CFRP sheets epoxy bonded over masonry elements, Compos. B Eng. 113
(2017) 317–330.
[62] F. Ceroni, M. Leone, V. Rizzo, A. Bellini, C. Mazzotti, Influence of mortar joints
on the behaviour of FRP materials bonded to different masonry substrates,
Eng. Struct. 153 (2017) 550–568.
[63] L. De Lorenzis, D. Tinazzi, A. Nanni, Near Surface Mounted FRP rods for
Masonry Strengthening: Bond and Flexural Testing, Proc., International
Conference on Composite Engineering, 2000.
[64] D. Dizhur, M.C. Griffith, J.M. Ingham, Pullout strength of NSM CFRP strips
bonded to vintage clay brick masonry, Eng. Struct. 69 (2014) 25–36.
[65] M. Fagone, G. Ranocchiai, C. Caggegi, S. Briccoli Bati, M. Cuomo, The efficiency
of mechanical anchors in CFRP strengthening of masonry: an experimental
analysis, Compos. B Eng. 64 (2014) 1–15.
[66] B. Ghiassi, J. Xavier, D.V. Oliveira, P.B. Lourenço, Application of digital image
correlation in investigating the bond between FRP and masonry, Compos.
Struct. 106 (2013) 340–349.
[67] B. Ghiassi, D.V. Oliveira, P.B. Lourenço, Hygrothermal durability of bond in
FRP-strengthened masonry, Mater. Struct./Materiaux et Construct. 47 (12)
(2014) 2039–2050.
[68] B. Ghiassi, J. Xavier, D.V. Oliveira, A. Kwiecien, P.B. Lourenço, B. Zaja˛c,
Evaluation of the bond performance in FRP-brick components re-bonded
after initial delamination, Compos. Struct. 123 (2015) 271–281.
[69] B. Ghiassi, P.B. Lourenço, D.V. Oliveira, Accelerated hygrothermal aging of
bond in FRP-masonry systems, J. Compos. Constr. 19 (3) (2015).
[70] A. Hosseini, D. Mostofinejad, M. Emami, Influence of bonding technique on
bond behavior of CFRP-to-clay brick masonry joints: experimental study
using particle image velocimetry (PIV), Int. J. Adhes. Adhes. 59 (2015)
27–39.
[71] K.M.C. Konthesingha, M.J. Masia, R.B. Petersen, A.W. Page, Bond Behaviour of
NSM FRP Strips to Modern clay Brick Masonry Prisms Under Cyclic Loading,
Proc., 11th Canadian Masonry Symposium, 2009.
[72] Liu, Y., Dawe, J., and McInerney, J. (2005). ‘‘Behaviour of GFRP sheets
bonded to masonry walls.” Proc., International Symposium on Bond
Behaviour of FRP in Structures, International Institute for FRP in
Construction (IIFC), 473-480.
[73] H. Maljaee, B. Ghiassi, P.B. Lourenço, D.V. Oliveira, Bond Performance in NSM-
Strengthened Masonry Brick, Proc., 9th National Conference of Experimental
Mechanics, 2014.
[74] H. Maljaee, B. Ghiassi, P.B. Lourenço, D.V. Oliveira, FRP-brick masonry bond
degradation under hygrothermal conditions, Compos. Struct. 147 (2016)
143–154.
[75] H. Maljaee, B. Ghiassi, P.B. Lourenço, D.V. Oliveira, Moisture-induced
degradation of interfacial bond in FRP-strengthened masonry, Compos. B
Eng. 87 (2016) 47–58.
[76] C. Mazzotti, B. Ferracuti, A. Bellini, Experimental bond tests on masonry
panels strengthened by FRP, Compos. B Eng. 80 (2015) 223–237.
[77] D.V. Oliveira, I. Basilio, P.B. Lourenço, Experimental bond behavior of FRP
sheets glued on brick masonry, J. Compos. Constr. 15 (1) (2011) 32–41.
[78] M. Panizza, E. Garbin, M.R. Valluzzi, C. Modena, Bond behaviour of CFRP and
GFRP laminates on brick masonry, in: Proc., 6th International Conference on
Structural Analysis of Historic Construction (SAHC08), CRC Press, 2008, pp.
763–770.
[79] M. Panizza, E. Garbin, M.R. Valluzzi, C. Modena, Experimental Comparison of
Various types of Specimens Subjected to SL and DL Shear Bond Tests on EB
Composites Applied to Bricks, Proc., 6th Int. Conf. on FRP Composites in Civil
Engineering–CICE, 2012, pp. 13–15.
[80] M. Panizza, E. Garbin, M.R. Valluzzi, C. Modena, Experimental study of the
bond of FRP applied to natural stones and masonry prisms, Key Eng. Mater.
624 (2015) 453–460.
[81] J.M. Pereira, P.B. Lourenço, Experimental bond behaviour of GFRP and
masonry bricks under impulsive loading, Mater. Struct./Materiaux et
Construct. 49 (11) (2016) 4799–4811.
[82] R.B. Petersen, M.J. Masia, R. Seracino, Bond behavior of near-surface mounted
FRP strips bonded to modern clay brick masonry prisms: influence of strip
orientation and compression perpendicular to the strip, J. Compos. Constr. 13
(3) (2009) 169–178.
[83] T. Rotunno, L. Rovero, U. Tonietti, S. Briccoli Bati, Experimental study of bond
behavior of CFRP-to-brick joints, J. Compos. Constr. 19 (3) (2015).
[84] M.S. Sciolti, M.A. Aiello, M. Frigione, Effect of thermo-hygrometric exposure
on FRP, natural stone and their adhesive interface, Compos. B Eng. 80 (2015)
162–176.
[85] V. Turco, N. Galati, L. De Lorenzis, C. Modena, A. Nanni, Bond Between Near
Surface Mounted FRP rods and Masonry in Structural Strengthening, Proc.,
Advancing with composites, Plast, 2003, pp. 7–9.
[86] M.R. Valluzzi, D.V. Oliveira, A. Caratelli, G. Castori, M. Corradi, G. De Felice, E.
Garbin, D. Garcia, L. Garmendia, E. Grande, U. Ianniruberto, A. Kwiecien, M.
Leone, G.P. Lignola, P.B. Lourenço, M. Malena, F. Micelli, M. Panizza, C.G.
Papanicolaou, A. Prota, E. Sacco, T.C. Triantafillou, A. Viskovic, B. Zaja˛c, G.
Zuccarino, Round Robin Test for composite-to-brick shear bond
characterization, Mater. Struct./Materiaux et Construct. 45 (12) (2012)
1761–1791.
[87] B. Ghiassi, G. Marcari, D.V. Oliveira, P.B. Lourenço, Water degrading effects on
the bond behavior in FRP-strengthened masonry, Compos. B Eng. 54 (1)
(2013) 11–19.
[88] A. Kwiecień, G. de Felice, D.V. Oliveira, B. Zaja˛c, A. Bellini, S. De Santis, B.
Ghiassi, G.P. Lignola, P.B. Lourenço, C. Mazzotti, A. Prota, Repair of composite-
to-masonry bond using flexible matrix, Mater. Struct./Materiaux et Construct.
49 (7) (2016) 2563–2580.
[89] EN 772-1:2011.‘‘ Methods of test for masonry units - Part 1: Determination of
compressive strength.
[90] EN 1926:2006, Natural stone test methods - Determination of uniaxial
compressive strength.
[91] EN 12372:2006, Natural stone test methods - Determination of flexural
strength under concentrated load.
[92] ASTM C67-03a, Standard Test Methods for Sampling and Testing Brick and
Structural Clay Tile, ASTM International, West Conshohocken, PA, 2003.
 J. Vaculik et al. / Construction and Building Materials 183 (2018) 325–345
[93] ASTM C1006-07(2013), Test method for splitting tensile strength of masonry
units, ASTM International, West Conshohocken, PA, 2013.
[94] L.F.M. da Silva, T.N.S.S. Rodrigues, M.A.V. Figueiredo, M.F.S.F. de Moura, J.A.G.
Chousal, Effect of adhesive type and thickness on the lap shear strength, J.
Adhes. 82 (11) (2006) 1091–1115.
[95] A. Li, T. Assih, Y. Delmas, Influence of the adhesive thickness and steel plate
thickness on the behaviour of strengthened concrete beams, J. Adhes. Sci.
Technol. 14 (13) (2000) 1639–1656.
[96] J. Dai, T. Ueda, Y. Sato, Development of the nonlinear bond stress-slip model
of fiber reinforced plastics sheet-concrete interfaces with a simple method, J.
Compos. Constr. 9 (1) (2005) 52–62.
[97] J. Yao, J.G. Teng, J.F. Chen, Experimental study on FRP-to-concrete bonded
joints, Compos. B Eng. 36 (2) (2005) 99–113.
[98] J. Pan, C.K.Y. Leung, Debonding along the FRP-concrete interface under
combined pulling/peeling effects, Eng. Fract. Mech. 74 (1–2) (2007) 132–150.
[99] P. Visintin, D.J. Oehlers, C. Wu, M.C. Griffith, The reinforcement contribution
to the cyclic behaviour of reinforced concrete beam hinges, Earthquake Eng.
Struct. Dyn. 41 (12) (2012) 1591–1608.
[100] P. Visintin, M.C. Griffith, M.S. Mohamed Ali, Bond of FRP Near Surface
Mounted to Clay Brick Masonry Under Monotonic and Cyclic Loads, Proc.,
7th International Conference on FRP Composites in Civil Engineering
(CICE), 2014.
345
[101] F.F. Ren, Z.J. Yang, J.F. Chen, W.W. Chen, An analytical analysis of the full-
range behaviour of grouted rockbolts based on a tri-linear bond-slip model,
Constr. Build. Mater. 24 (3) (2010) 361–370.
[102] M. Savoia, B. Ferracuti, L. Vincenzi, Inverse analysis for the calibration of FRP
– concrete interface law, Adv. Struct. Eng. 12 (5) (2009) 613–625.
[103] M. Tekieli, S. De Santis, G. de Felice, A. Kwiecien, F. Roscini, Application of
Digital Image Correlation to composite reinforcements testing, Compos.
Struct. 160 (2017) 670–688.
[104] J.M. Sena Cruz, J.A.O. Barros, Bond between near-surface mounted carbon-
fiber-reinforced polymer laminate strips and concrete, J. Compos. Constr. 8
(6) (2004) 519–527.
[105] H. Zhang, S.T. Smith, R.J. Gravina, Analysis of FRP-to-concrete interfaces using
a displacement driven partial interaction model, Int. J. Mech. Sci. 117 (2016)
210–217.
[106] H. Aydin, R.J. Gravina, P. Visintin, A partial-interaction approach for
extracting FRP-to-concrete bond characteristics from environmentally
loaded flexural tests, Compos. B Eng. 132 (2018) 214–228.
[107] R.J. Gravina, H. Aydin, P. Visintin, Extraction and analysis of bond-slip
characteristics in deteriorated FRP-to-concrete joints using a mechanics-
based approach, J. Mater. Civ. Eng. 29 (6) (2017).
[108] C. Mazzotti, B. Ferracuti, A. Bellini, Experimental study on masonry panels
strengthened by GFRP: the role of inclination between mortar joints and
GFRP sheets, Key Eng. Mater. 624 (2015) 559–566.