You are on page 1of 21

Engineering Structures 24 (2002) 397–417

www.elsevier.com/locate/engstruct

FRP-strengthened RC beams. II: assessment of debonding strength


models
*
S.T. Smith, J.G. Teng
Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, Hong Kong, PR China

Received 11 April 2001; received in revised form 5 September 2001; accepted 12 September 2001

Abstract

Reinforced concrete (RC) beams strengthened in flexure by the bonding of a fiber reinforced polymer (FRP) plate to the tension
face are susceptible to brittle debonding failures. Such failures commonly initiate at or near one of the plate ends at a load below
that to achieve flexural failure of the plated section. For a successful design of flexural strengthening using FRP composites, it is
important to be able to predict such plate end debonding failures. In the first of these two companion papers, 12 plate end debonding
strength models have been reviewed and summarized (Engng Struct 24(4) (2002) 385–395). The aim of the present paper is to
provide a comprehensive assessment of the strengths and weaknesses of all the 12 models. To this end, a large test database
containing the test results of 59 beams reported to have failed by plate end debonding is first presented. This database was carefully
constructed from an extensive survey of the published literature. Both statistical and graphical comparisons between test results
and the predictions of the debonding strength models are next presented. A new simple debonding strength model which is superior
to existing models is also proposed.  2002 Published by Elsevier Science Ltd.

Keywords: Debonding; FRP; RC beams; Retrofitting; Strengthening; Strength models; Test database

1. Introduction comparison of even two distinct debonding strength


models, although disagreement over assumptions and
Reinforced concrete (RC) beams strengthened in approaches for developing debonding strength models
flexure by the bonding of a fiber reinforced polymer clearly exists among researchers (e.g. [2–4]). Further-
(FRP) plate to the tension face are susceptible to brittle more, many of the existing debonding models were pro-
debonding failures. Such failures commonly initiate at posed on the basis of rather limited test data, often only
or near one of the plate ends (simply referred to as the those from tests by the proposers themselves.
plate end hereafter) at a load below that to achieve flex- This paper is aimed at correcting the two deficiencies
ural failure of the plated section. For a successful design mentioned above, namely the lack of information of the
of flexural strengthening using FRP composites, it is performance of each model as assessed by a large
important to be able to predict such plate end debonding reliable test database and the establishment of the rela-
failures. In the first of these two companion papers, 12 tive strengths and weaknesses of these models also using
plate end debonding strength models have been reviewed such a database. To this end, a large test database con-
and summarized [1]. This large number of models is taining the test results of 59 beams reported to have
clear evidence that a great amount of attention has been failed by plate end debonding is first presented. This dat-
devoted to the prediction of plate end debonding failures abase was carefully constructed from an extensive sur-
in plated beams. Surprisingly, however, no study in the vey of the published literature. Both statistical and
published literature has been found which provides a graphical comparisons between test results and the pre-
dictions of the debonding strength models are next
presented. A new simple debonding strength model
* Corresponding author. Tel.: +852-2334-6389; fax: +852-2766- which is superior to existing models is also proposed.
6012. As was pointed out in the companion paper [1], while
E-mail address: cejgteng@polyu.edu.hk (J.G. Teng). the conclusions reached here are applicable to any beams

0141-0296/02/$ - see front matter  2002 Published by Elsevier Science Ltd.


PII: S 0 1 4 1 - 0 2 9 6 ( 0 1 ) 0 0 1 0 6 - 7
398 S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417

with an FRP plate bonded to the tension face, for simpli- of the RC beam and the FRP plate were investigated to
fication of presentation, only simply supported beams are examine their effect on the interfacial stresses. Inspec-
explicitly considered in this paper. All the test data in tion of the data used in their paper revealed that the data-
the published literature which were sufficiently well base included beams bonded with plates with end
documented for present purposes and thus included in anchorage (e.g. U-jacket) as well as beams cracked by
the present database were obtained for simply sup- preloading. In addition, some of the beams included in
ported beams. the database were reported to have failed by intermediate
crack induced interfacial debonding in the original
source. These shortcomings of the database may have
2. Experimental data affected the accuracy of the conclusions reached in
their study.
2.1. Previous databases
2.2. Selection criteria for constructing a new test
The complexity of behavior of FRP-strengthened RC database
structures means that test databases are still the sole veri-
fication tool for the reliability of a strength model. As a In this paper a carefully prepared experimental datab-
result, a number of previous attempts have been made ase is presented which was assembled with clear selec-
to develop test databases for simply supported RC beams tion criteria. For inclusion in this database, the following
strengthened with a soffit FRP plate. These are reviewed requirements must be met: (a) failure of the beam was
below to provide a proper context for the present data- by plate end debonding; (b) the FRP plate was neither
base. prestressed nor anchored in any form at its ends; (c) the
Bonacci [5] reported a database of 64 test results gath- beam never experienced loading before being loaded to
ered from 10 studies for FRP-strengthened RC beams. debonding failure; and (d) sufficient details for various
These beams failed by various modes such as plate end geometric and material parameters were provided to
debonding, FRP rupture, concrete crushing and beam enable the results to be used with confidence. An exten-
shear failure. Bonacci [5] examined trends of certain sive search of the existing literature with these selection
parameters using these test results, including percentages criteria resulted in a database of 59 results from 14 stud-
of occurrence of various failure modes as a function of ies as given in Appendix A. This database, based on its
geometric and material properties of the RC beam and specific selection criteria, is much larger than that used
the FRP plate. The different types of debonding failures in developing any of the existing debonding strength
as discussed in the companion paper [1] were not ident- models for FRP-strengthened RC beams and any of
ified and were therefore lumped into a single category. those reviewed above, and provides a solid basis for
Raoof and Hassanen [6] compared the performance of assessing the performance of all existing debonding
their two “tooth” models for FRP-plated beams using strength models. Criterion (c) above was used for con-
test results of such beams which supposedly failed by sistency in the database as the vast majority of tests were
cover separation. Their “tooth” models, as reviewed in conducted on beams without any preloading. Limited
the companion paper, are for the prediction of concrete research on beams preloaded to cracking and then com-
cover separation failures. The database, however, pletely unloaded before plate bonding appears to suggest
included FRP-plated beams reported to have failed by that such preloading has little effect on the plate end
concrete crushing and FRP rupture in the original debonding load [9]. This aspect should be carefully
sources. In addition, the database included beams investigated in further research but is beyond the scope
bonded with plates with end anchorage or anchored of the present study. A few tests on beams with FRP
under the supports and beams cracked by preloading. plates bonded under sustained loading have also been
The suitability of this database is questionable for carried out [9], but no conclusion can yet be drawn from
assessing their tooth models which are only applicable these tests.
to concrete cover separation failures.
More recently, Mukhopadhyaya and Swamy [7] 2.3. New test database
assembled a database consisting of steel, GFRP and
CFRP-plated RC beams that were reported to have failed The test data in the new database has been grouped
by plate end debonding. The peak interfacial shear and into three sets, the first set for failure by concrete cover
normal stresses at the plate end under the debonding load separation (Tables 7–9) (44 results), the second set for
were determined for these beams using the analytical failure by plate end interfacial debonding (Tables 10–
solution of Roberts [8]. A new approach for determining 12) (8 results), and the third set for failure by a mixed
the interfacial shear stress was also presented by Mukho- concrete cover separation and interfacial debonding
padhyaya and Swamy [7] and compared with Roberts’ mode (Tables 13–15) (7 results). An example of such
method [8]. Various geometric and material properties mixed mode debonding is shown in Fig. 1 [18]. All col-
S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417 399

Fig. 1. FRP-plated RC beam: debonding in a mixed mode (plate end interfacial debonding in combination with concrete cover separation).

lected test results are for simply supported beams rec- 2.4. Assumptions for geometric and material
tangular in cross-section, conventionally reinforced, properties
loaded in three or four point bending, and strengthened
with a glass or a carbon FRP soffit plate. Aramid FRP Many of the references on FRP-plated RC beams
was not used in any of the tests. reviewed for possible inclusion in the database could not
Tables 7, 10 and 13 give the RC beam details where be used. Often the failure mode was poorly reported
bc, h, ds, dsc, f ⬘c, fcu, Ec, and fct denote the width of RC and/or important geometric and material properties of
beam, overall depth of RC beam, distance from beam the RC beam, the FRP plate, or the adhesive layer omit-
compression face to centroid of steel tension reinforce- ted. Only those tests with the failure mode sufficiently
ment, distance from beam compression face to centroid well described and important material and geometric
of steel compression reinforcement, concrete cylinder properties adequately reported could be included in the
compressive strength, concrete cube compressive database. The only geometric and material properties
strength, modulus of elasticity of concrete, and concrete assumed in the database, if not specified in the original
cylinder splitting tensile strength respectively. Tables 8, source, are depth of concrete cover which is assumed be
11 and 14 define the reinforcement properties where 10% of the overall beam depth which is similar to the
Es, Esc and Eyv denote the moduli of elasticity of the average of the different cover depths used for other
steel tension reinforcement, steel compression reinforce- beams in the database, elastic modulus of adhesive,
ment, and steel shear reinforcement (stirrups only) thickness of adhesive layer, and modulus of elasticity
respectively while fys, fyc and fyv denote the correspond- of steel reinforcement (200 GPa). The ACI code [22]
ing yield strengths and As, Asc and Asv the corresponding expressions for the elastic modulus of concrete Ec (in
cross-sectional areas. Also in Tables 8, 11 and 14, s rep- MPa) and the splitting tensile strength of concrete fct (in
resents the spacing of the steel stirrups. Tables 9, 12 MPa) were adopted which are related to the concrete
and 15 define the FRP geometric and material properties cylinder compressive strength f ⬘c (in MPa) as follows
Ec⫽4730冑f ⬘c
where Ea, ta, Efrp, ffrp, tfrp, bfrp and a denote the modulus
(1)
of elasticity of adhesive, thickness of adhesive layer,
modulus of elasticity of FRP, tensile strength of FRP in fct⫽0.53冑f ⬘c (2)
the main fiber direction, FRP thickness, FRP width, and
the distance from the support to the nearer end of the
soffit plate. Tables 9, 12 and 15 also give the details of For beams with only the concrete cube compressive
the loading configuration and the failure load where B, strength fcu reported, the concrete cylinder strength was
L, and Vexp represent the distance from the support to taken to be 0.8fcu [21].
the nearer applied load, the span of the beam, and the For beams with pultruded plates, the adhesive layer
experimental shear force at the plate end to cause thickness was sometimes not reported. In such cases, the
debonding. thickness of the adhesive layer was assumed to be 2 mm,
400 S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417

a value similar to the average of those reported for the data of beams with wet lay-up plates and those with pul-
remaining beams with pultruded plates (an average of truded plates are appropriately differentiated from each
1.88 mm from 32 tests). This is believed to be the most other in comparisons.
rational approach one can take without having to exclude In addition, the elastic modulus of the adhesive was
these test data from the database, and to have minimized not reported in many studies. For these cases, the elastic
errors due to this assumed adhesive layer thickness. This modulus was assumed to be 8500 MPa, a value close to
assumption may be further justified by noting that any the average of the values reported for the other cases of
model which is overly sensitive to adhesive layer thick- the database (an average of 8697 MPa from 37 tests).
ness is unlikely to be suitable as design models as Again, this is a rational approach, and is not expected
adhesive layer thicknesses are difficult to control on site to compromise the usefulness of the database.
and have been found to have a limited effect on test
debonding loads [4,26,27].
For beams strengthened with FRP plates constructed 3. Theoretical ultimate moment capacity
in a wet lay-up process, more uncertainty exists. For
many of the tests reported on such beams, the thickness In assessing the debonding strength models, the theor-
of the FRP plate vs the thickness of the adhesive, was etical ultimate moment capacity of the plated section
generally not available. Although in a small number of without considering the possibility of debonding is used
cases, both the fiber sheet thickness alone tsheet and the as a reference value. In this study, this ultimate moment
gross thickness of the fiber sheet plus the applied capacity was evaluated using the design equations given
adhesive tfrp,g were given, only the sheet thickness in Teng et al. [28]. These equations were developed
tsheet was given in most cases. For the former cases, the within the framework of the Bristish code for concrete
adhesive layer thickness and FRP plate thickness were design [29], assuming that the beam fails by either FRP
worked out using the following formulas for a single rupture or concrete crushing without premature failure
layer of FRP: due to debonding. The common assumption of plane sec-
tfrp,g−tsheet tions remaining plane was made in developing these
ta⫽ (3) equations. A key difference between a normal RC sec-
2
tion and an FRP-plated RC section is that the FRP is a
tfrp⫽ta⫹tsheet (4) linear–elastic–brittle material, so when failure occurs by
based on the assumptions that the sheet sits in the middle FRP rupture, the concrete is likely to have not reached
of the adhesive, and the sheet and the outside adhesive its design ultimate strain which according to BS 8110 is
form the plate (Fig. 2). For the latter cases, ta was equal to 0.0035. As a result, the simplified rectangular
assumed to be 0.42 mm based on measurements of plate stress block of the code [29] for the compression con-
samples formed from carbon fiber sheets by wet lay-up crete is not valid. Instead, the idealized stress–strain
in the authors’ laboratory, and the FRP plate thickness, curve for concrete as recommended by BS 8110 was
tfrp, was worked out from Eq. (4) for a single layer of explicitly included in these design equations. All partial
FRP. For cases where the FRP plate was formed from safety factors were set to unity for the present purpose.
a number of layers, the total plate thickness was found
by multiplying the single layer thickness by the number
of layers by assuming that the adhesive thickness 4. Statistical comparisons of debonding strength
between two sheets is ta. The above rational process is models
believed to have minimized the effect of uncertainty in
the existing test data. Furthermore, any debonding 4.1. General
strength model that is overly sensitive to this uncertainty
is unlikely to be suitable for design use for a wet lay- Tables 1–6 present the statistical characteristics of the
up process as adhesive and plate thicknesses are difficult test-to-predicted debonding strength ratios for all 12
to control on site or to define for calculations. Due to debonding strength models reviewed in the companion
this additional uncertainty with wet lay-up plates, test paper [1]. Tables 1–6 are respectively for concrete cover
separation results alone, and all test results combined
(concrete cover separation, plate end interfacial debond-
ing and mixed mode debonding). Tables 1 and 4 are for
pultruded and wet lay-up plates combined, Tables 2 and
5 for pultruded plates alone, and Tables 3 and 6 for wet
lay-up plates only. A separate statistical analysis was not
undertaken for either the plate end interfacial debonding
Fig. 2. Wet lay-up plate thickness. results (8 results) or the mixed mode debonding results
S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417 401

Table 1
Test-to-predicted debonding strength ratios for failure by concrete cover separation: pultruded and wet lay-up plates

Standard Coefficient of Percentage of No. of


Category Model Average
deviation variation exceedence tests

Shear capacity based models Oehlers [36] 2.35 0.49 21% 0.3% 44
Jansze [30] 1.50 0.45 30% 13% 39
Ahmed and van Gemert [14] 1.15 0.25 21% 27% 39
New model (Vdb,end=1.4Vc) 1.67 0.35 21% 2.8% 44
New model (Vdb,end=1.5Vc) 1.56 0.33 21% 4.4% 44
Concrete tooth models Raoof and Zhang: Lower [37] 1.47 0.52 36% 19% 44
Raoof and Zhang: Upper [37] 1.06 0.31 29% 43% 44
Wang and Ling: Lower [33] 1.20 0.35 29% 28% 44
Wang and Ling: Upper [33] 0.88 0.16 18% 78% 44
Raoof and Hassanen I: Lower [6] 1.61 0.65 41% 18% 44
Raoof and Hassanen I: Upper [6] 1.13 0.43 38% 38% 44
Raoof and Hassanen II: Lower [6] 1.57 0.60 38% 17% 44
Raoof and Hassanen II: Upper [6] 1.09 0.35 32% 40% 44
Interfacial stress based models Ziraba et al. I [34] 4.89 3.64 74% 14% 44
Ziraba et al. II [34] 1.63 0.55 34% 13% 44
Varastehpour and Hamelin [35] 0.59 0.20 34% 98% 44
Saadatmanesh and Malek [31] 3.24 2.83 87% 21% 39
Tumialan et al. [32] 1.83 1.28 70% 26% 39

Table 2
Test-to-predicted debonding strength ratios for failure by concrete cover separation: pultruded plates only

Standard Coefficient of Percentage of No. of


Category Model Average
deviation variation exceedence tests

Shear capacity based models Oehlers [36] 2.09 0.27 13% 0% 25


Jansze [30] 1.21 0.35 29% 27% 23
Ahmed and van Gemert [14] 1.06 0.23 22% 40% 23
New model (Vdb,end=1.4Vc) 1.48 0.18 12% 0.4% 25
New model (Vdb,end=1.5Vc) 1.38 0.17 12% 1.2% 25
Concrete tooth models Raoof and Zhang: Lower [37] 1.32 0.38 28% 19% 25
Raoof and Zhang: Upper [37] 0.95 0.26 27% 58% 25
Wang and Ling: Lower [33] 1.12 0.27 24% 32% 25
Wang and Ling: Upper [33] 0.81 0.13 16% 93% 25
Raoof and Hassanen I: Lower [6] 1.41 0.36 25% 12% 25
Raoof and Hassanen I: Upper [6] 0.96 0.21 22% 57% 25
Raoof and Hassanen II: Lower [6] 1.43 0.42 30% 16% 25
Raoof and Hassanen II: Upper [6] 0.96 0.23 24% 57% 25
Interfacial stress based models Ziraba et al. I [34] 3.54 3.25 92% 22% 25
Ziraba et al. II [34] 1.54 0.45 29% 11% 25
Varastehpour and Hamelin [35] 0.57 0.14 25% 100% 25
Saadatmanesh and Malek [31] 2.99 3.30 110% 27% 23
Tumialan et al. [32] 1.63 1.53 94% 34% 23

(7 results) alone due to the small number of such test exceed the theoretical moment capacities of beams with-
results available. out considering the possibility of debonding. For such
It should be noted that the models of Jansze [30], beams, these theoretical moment capacities instead of the
Ahmed and van Gemert [14], Saadatmanesh and Malek predicted debonding moments were used for compari-
[31] and Tumialan et al. [32] cannot be used for plates sons with test data, as has been previously done for these
terminated at the support. These limitations account for models (e.g. [6]).
the smaller number of test data used to assess these mod- Of the 12 debonding strength models, model I of Zir-
els (Tables 1–6). aba et al. [34] and Varastehpour and Hamelin’s model
It should also be noted that many of the upper bound [35] are for plate end interfacial debonding only. All the
predictions and a few of the lower bound predictions of concrete tooth models are for cover separation only,
debonding moments by the tooth models [6,33,37] while the remainder of the interfacial stress based mod-
402 S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417

Table 3
Test-to-predicted debonding strength ratios for failure by concrete cover separation: wet lay-up plates only

Standard Coefficient of Percentage of No. of


Category Model Average
deviation variation exceedence tests

Shear capacity based models Oehlers [36] 2.69 0.51 19% 0% 19


Jansze [30] 1.91 0.18 9.7% 0% 16
Ahmed and van Gemert [14] 1.28 0.21 16% 9.3% 16
New model (Vdb,end=1.4Vc) 1.92 0.36 19% 0.6% 19
New model (Vdb,end=1.5Vc) 1.79 0.34 19% 1.0% 19
Concrete tooth models Raoof and Zhang: Lower [37]] 1.66 0.63 38% 15% 19
Raoof and Zhang: Upper [37] 1.20 0.33 27% 28% 19
Wang and Ling: Lower [33] 1.30 0.41 32% 24% 19
Wang and Ling: Upper [33] 0.97 0.15 15% 57% 19
Raoof and Hassanen I: Lower [6] 1.87 0.85 46% 15% 19
Raoof and Hassanen I: Upper [6] 1.35 0.54 40% 26% 19
Raoof and Hassanen II: Lower [6] 1.76 0.74 42% 15% 19
Raoof and Hassanen II: Upper [6] 1.25 0.41 33% 27% 19
Interfacial stress based models Ziraba et al. I [34] 6.67 3.43 51% 4.9% 19
Ziraba et al. II [34] 1.74 0.66 38% 13% 19
Varastehpour and Hamelin [35] 0.60 0.26 43% 93% 19
Saadatmanesh and Malek [31] 3.59 2.02 56% 9.9% 16
Tumialan et al. [32] 2.10 0.76 36% 7.4% 16

Table 4
Test-to-predicted debonding strength ratios for all plate end failures: pultruded and wet lay-up plates

Standard Coefficient of Percentage of No. of


Category Model Average
deviation variation exceedence tests

Shear capacity based models Oehlers [36] 2.25 0.53 24% 1.0% 59
Jansze [30] 1.41 0.46 33% 19% 54
Ahmed and van Gemert [14] 1.08 0.29 27% 40% 53
New model (Vdb,end=1.4Vc) 1.60 0.38 24% 5.8% 59
New model (Vdb,end=1.5Vc) 1.49 0.36 24% 8.3% 59
Concrete tooth models Raoof and Zhang: Lower [37] 1.44 0.48 33% 18% 59
Raoof and Zhang: Upper [37] 1.04 0.28 27% 44% 59
Wang and Ling: Lower [33] 1.14 0.34 30% 34% 59
Wang and Ling: Upper [33] 0.87 0.16 19% 79% 59
Raoof and Hassanen I: Lower [6] 1.55 0.59 38% 18% 59
Raoof and Hassanen I: Upper [6] 1.09 0.38 35% 41% 59
Raoof and Hassanen II: Lower [6] 1.49 0.55 37% 18% 59
Raoof and Hassanen II: Upper [6] 1.05 0.32 31% 44% 59
Interfacial stress based models Ziraba et al. I [34] 4.36 3.43 79% 16% 59
Ziraba et al. II [34] 1.57 0.52 33% 14% 59
Varastehpour and Hamelin [35] 0.57 0.19 33% 99% 59
Saadatmanesh and Malek [31] 2.75 2.59 94% 25% 54
Tumialan et al. [32] 1.58 1.20 76% 32% 54

els and all the shear capacity based models appear to be dictions, are given by Oehlers [36], Ziraba et al. (model
suitable for both cover separation and plate end interfa- II) [34], Jansze [30], Raoof and Zhang (lower bound
cial debonding, although their applicability to plate end predictions) [37] and Raoof and Hassanen (lower bound
interfacial debonding has not been found to be clearly predictions of both of their models) [6].
stated in the descriptions of these models. For the combined data of pultruded plates and wet
lay-up plates (Table 1), the most un-conservative of the
4.2. Concrete cover separation: better models better performing models is Raoof and Zhang’s model
[37] which provides an 81 percentile lower bound (i.e.
The statistically better performing models for concrete 19% of exceedence), while the most conservative is that
cover separation failures, based on the results given in given by Oehlers [36] which has a percentage of exceed-
Tables 1–3 and in terms of providing safe and close pre- ence being 0.3%. Oehlers’ model [36] also shows the
S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417 403

Table 5
Test-to-predicted debonding strength ratios for all plate end failures: pultruded plates only

Standard Coefficient of Percentage of No. of


Category Model Average
deviation variation exceedence tests

Shear capacity based models Oehlers [36] 2.00 0.32 16% 0.1% 37
Jansze [30] 1.15 0.31 27% 31% 35
Ahmed and van Gemert [14] 0.99 0.28 28% 52% 34
New model (Vdb,end=1.4Vc) 1.41 0.22 16% 3.0% 37
New model (Vdb,end=1.5Vc) 1.32 0.21 16% 6.0% 37
Concrete tooth models Raoof and Zhang: Lower [37] 1.31 0.35 27% 19% 37
Raoof and Zhang: Upper [37] 0.95 0.23 24% 59% 37
Wang and Ling: Lower [33] 1.06 0.27 26% 42% 37
Wang and Ling: Upper [33] 0.81 0.13 16% 92% 37
Raoof and Hassanen I: Lower [6] 1.39 0.34 25% 13% 37
Raoof and Hassanen I: Upper [6] 0.96 0.20 21% 58% 37
Raoof and Hassanen II: Lower [6] 1.38 0.38 28% 16% 37
Raoof and Hassanen II: Upper [6] 0.94 0.21 22% 61% 37
Interfacial stress based models Ziraba et al. I [34] 3.06 2.82 92% 23% 37
Ziraba et al. II [34] 1.46 0.41 28% 13% 37
Varastehpour and Hamelin [35] 0.57 0.14 25% 100% 37
Saadatmanesh and Malek [31] 2.37 2.83 119% 31% 35
Tumialan et al. [32] 1.34 1.32 99% 40% 35

Table 6
Test-to-predicted debonding strength ratios for all plate end failures: wet lay-up plates only

Standard Coefficient of Percentage of No. of


Category Model Average
deviation variation exceedence tests

Shear capacity based models Oehlers [36] 2.68 0.55 21% 0.1% 22
Jansze [30] 1.89 0.26 14% 0% 19
Ahmed and van Gemert [14] 1.24 0.24 19% 16% 19
New model (Vdb,end=1.4Vc) 1.91 0.39 21% 1.0% 22
New model (Vdb,end=1.5Vc) 1.78 0.37 21% 1.6% 22
Concrete tooth models Raoof and Zhang: Lower [37] 1.65 0.59 36% 13% 22
Raoof and Zhang: Upper [37] 1.20 0.30 25% 26% 22
Wang and Ling: Lower [33] 1.28 0.40 31% 25% 22
Wang and Ling: Upper [33] 0.97 0.16 17% 58% 22
Raoof and Hassanen I: Lower [6] 1.80 0.81 45% 16% 22
Raoof and Hassanen I: Upper [6] 1.31 0.51 39% 27% 22
Raoof and Hassanen II: Lower [6] 1.69 0.71 42% 17% 22
Raoof and Hassanen II: Upper [6] 1.22 0.40 32% 29% 22
Interfacial stress based models Ziraba et al. I [34] 6.55 3.28 50% 4.5% 22
Ziraba et al. II [34] 1.75 0.63 36% 12% 22
Varastehpour and Hamelin [35] 0.58 0.25 43% 95% 22
Saadatmanesh and Malek [31] 3.45 1.95 57% 10% 19
Tumialan et al. [32] 2.01 0.77 38% 9.4% 19

least scatter based on the coefficient of variation, fol- among the other better performing models, with coef-
lowed by Jansze’s model [30]. The predictions of both ficients of variation in the range of 22%–30%. Jansze’s
models of Raoof and Hassanen (lower bound model [30] is the most un-conservative for these beams,
predictions) [6] are seen to be the most scattered, with with a percentage of exceedence being 27%.
coefficients of variation being about 40%. If wet lay-up plates are examined alone (Table 3),
If pultruded plates are examined alone (Table 2), the Jansze’s model [30] becomes the best performing model,
same 6 models already identified above remain the better with the lowest coefficient of variation (9.7%) and con-
ones. Predictions by Oehlers’ model [36] remain the servative predictions for all cases. Oehlers’ model [36] is
most conservative and least scattered. For these beams, also conservative for all cases and has the second lowest
the predictions of this model are conservative in all coefficient of variation. The rest of the better performing
cases. The degree of scatter now does not vary too much models all show a high degree of scatter with coefficients
404 S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417

of variation being in the range of 38%–46%, with the lin’s model [35] performs worse for wet lay-up plates
percentages of exceedence being either 13% or 15%. than for pultruded plates as the percentages of exceed-
Based on the statistical characteristics given in Tables ence are similar for the two types of plates, but the
1–3, it can be concluded that Oehlers’ model [36] is the degree of scatter is greater for the former. The other
best performing model in terms of providing safe and interfacial stress based models show much better per-
reliable predictions, although it may be considered formance for wet lay-up plates, with both the coefficient
overly conservative as indicated by its relatively high of variation and the percentage of exceedence substan-
average ratios between the test and predicted debonding tially reduced from those for pultruded plates. In fact for
loads (Tables 1–3). wet lay-up plates, Tumialan et al.’s model [32] and Zir-
aba et al.’s model I [34] are comparable to some of the
4.3. Concrete cover separation: poorer models better performing models.

The models of Raoof and Zhang (upper bound 4.4. All test results
predictions) [37], Varastehpour and Hamelin [35], Saa-
daatmanesh and Malek [31], Wang and Ling (both lower The statistical performance of all 12 models as
and upper bound predictions) [33], Ahmed and van assessed using all the test results in the database which
Gemert [14], Tumialan et al. [32], Raoof and Hassanen’s includes 8 results for plate end interfacial debonding and
(upper bound predictions) [6], and model I of Ziraba et another 7 results for mixed mode debonding is shown
al. [34] are identified to be the worse performing models in Tables 4–6. Twelve of these fifteen results are for
in terms of giving close and safe predictions (Tables pultruded plates with only three for wet lay-up plates.
1–3). The inclusion of these additional results has no or very
When the combined data for the two types of plates limited effect on the statistical characteristics of the
are examined (Table 1), Varastehpour and Hamelin’s models, with some showing a small deterioration in per-
model [35] suffers from great un-conservativeness with formance (e.g. Oehlers’ model). The better and worse
the percentage of exceedence being 98% which means performing models for all plate end debonding failures
that its predictions are un-conservative in almost all combined, based on the statistics in Tables 4–6, are the
cases. On the other hand, Ziraba et al.’s model I [34] same as those for concrete cover separation (Tables 1–3).
suffers from a great scatter, with its coefficient of vari- It should be noted that one of the predictions by
ation being 74%. These two models developed for plate Ahmed and van Gemert’s model [14] is in fact negative
end interfacial debonding are thus not suitable for cover ([25], Beam B7, Tables 13–15, mixed mode failure) and
separation failures as can be expected. The upper bound was therefore neglected from both the statistical analyses
predictions of all tooth models are unsafe as they are (Tables 4 and 5) presented here and the graphical com-
expected to be. The interfacial stress based cover separ- parison presented in the next section. Geometrical and/or
ation models of Saadaatmanesh and Malek [31] and material limits appear to exist for Ahmed and van
Tumialan et al. [32] suffer from great scatters, with Saa- Gemert’s model [14], which were not stated in the orig-
daatmanesh and Malek’s model [31] having the highest inal source.
coefficient of variation of all models (87%). The model
of Ahmed and van Gemert [14] and the lower bound
predictions of Wang and Ling’s model [33] show more 5. Graphical comparisons of debonding strength
reasonable performance. models
When the worse performing models are examined
using the data for pultruded plates (Table 2) and those 5.1. General
for wet lay-up plates (Table 3) separately, the models
are found to perform differently for the two types of More illustrative comparisons of the test data with all
plates. Ahmed and van Gemert’s model [14] based on of the debonding strength models are given in Figs. 3–
the shear capacity approach, which was modified from 14, where the test data are differentiated based on both
Jansze’s model [30], is rather un-conservative for pul- plate types and failure modes. In these figures, Mu is the
truded plates (Table 2) as evidenced by the high percent- theoretical ultimate moment capacity of the plated sec-
ages of exceedence, but is satisfactory for wet lay-up tion without the influence of debonding, Mexp is the
plates alone (Table 3). The worse performing tooth mod- maximum experimental moment in the beam at debond-
els generally show greater scatters (except the upper ing, and Mdb is the corresponding debonding moment
bound predictions of Wang and Ling’s model) but predicted by the debonding strength models. The aver-
reduced percentages of exceedence for wet lay-up plates ages of the test-to-predicted debonding strength ratios
than for pultruded plates but they are still too un-con- are shown for all failure modes lumped together
servative for design use for wet lay-up plates. Of the (concrete cover separation, plate end interfacial debond-
interfacial stress based models, Varastehpour and Hame- ing and mixed mode debonding), and concrete cover sep-
S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417 405

Fig. 3. Test results vs predictions of Oehlers’ model [36]. Fig. 5. Test results vs predictions of Ahmed and van Gemert’s
model. [14].

aration alone, without distinction between pultruded and Ziraba et al. (model I) [34] (Fig. 10), Ziraba et al. (model
wet lay-up plates. II) [34] (Fig. 11), Varastehpour and Hamelin [35] (Fig.
The illustrative comparison of results is presented in 12), Saadatmanesh and Malek [31] (Fig. 13), and Tumia-
order of the classification of the model as given in the lan et al. [32] (Fig. 14).
companion paper [1]. Such order is given by (a) shear
capacity based models, namely those of Oehlers [36] 5.2. Shear capacity based models
(Fig. 3), Jansze [30] (Fig. 4), and Ahmed and van
Gemert [14] (Fig. 5), (b) concrete tooth models, namely Oehlers’ model [36] (Fig. 3) is more conservative for
those of Raoof and Zhang (both lower and upper bound wet lay-up plates than for pultruded plates. Concrete
predictions) [37] (Fig. 6), Wang and Ling (both lower cover failures are also generally more conservatively
and upper bound predictions) [33] (Fig. 7), Raoof and predicted than plate end interfacial debonding failures.
Hassanen (model I, both lower and upper bound Oehlers’ model [36] is overly conservative, as even the
predictions) [6] (Fig. 8), and Raoof and Hassanen (model lowest test result is about 40% higher than the prediction
II, both lower and upper bound predictions) [6] (Fig. 9), from this model.
and (c) interfacial stress based models, namely those of Jansze’s model [30] (Fig. 4) is also more conservative
for wet lay-up plates than for pultruded plates. Its predic-
tions are also more conservative for cover separation
alone than for all plate end failures combined. Many test
results are lower than predictions from this model, with
the lowest test result being about 65% of the prediction.
However, the predictions are safe for wet lay-up plates
alone.
Predictions from Ahmed and van Gemert’s model [14]
are unsafe for all failure modes and both plate types
(Fig. 5).

5.3. Concrete tooth models

Fig. 6 shows that for both Raoof and Zhang’s lower


[37] (Fig. 6a) and upper (Fig. 6b) bound predictions, no
clear patterns appear to exist for different plate types
and failure modes. Furthermore, the averages of test-to-
predicted debonding strength ratios for concrete cover
separation and all plate end failures combined are simi-
Fig. 4. Test results vs predictions of Jansze’s model [30]. lar. Therefore, the fact that the model was developed for
406 S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417

Fig. 7. (a) Test results vs predictions of Wang and Ling’s model:


lower bound [33]. (b) Test results vs predictions of Wang and Ling’s
Fig. 6. (a) Test results vs predictions of Raoof and Zhang’s model: model: upper bound [33].
lower bound [37]. (b) Test results vs predictions of Raoof and Zhang’s
model: upper bound [37].
lead to an apparent trend which however should not be
mistaken as any measure of the reliability of the model.
cover separation only is not consistent with its perform- The predictions for all failure modes and both plate
ance shown in Fig. 6. The model suffers from over pre- types are scattered for both models of Raoof and Has-
dictions which exist regardless of failure modes or plate sanen (both lower and upper bound predictions) [6]
types, although the lower bound predictions are un-con- (Figs. 8 and 9). For a few beams with wet lay-up plates,
servative only for a limited number of cases. both the lower and upper bound predictions from these
Wang and Ling’s lower bound predictions [33] (Fig. two models are very conservative. The upper bound pre-
7a) are generally more conservative for concrete cover dictions of both models are lower than the test results
separation and mixed mode debonding than for plate end for many cases. The lower bound predictions give sig-
interfacial debonding. Most of the upper bound predic- nificantly unsafe predictions only for two test results. It
tions of Wang and Ling’s model [33] (Fig. 7b) are higher should be noted that some of the upper bound predictions
than the theoretical ultimate moment capacity of the sec- of both models of Raoof and Hassanen (Figs. 8b and 9b)
tion without debonding. As a result, the ultimate moment [6] exceed the theoretical ultimate moment capacity of
capacities are shown in Fig. 7b for such beams, and these the plated section without debonding. For these beams,
S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417 407

Fig. 9. (a) Test results vs predictions of Raoof and Hassanen’s model


II: lower bound [6]. (b) Test results vs predictions of Raoof and Has-
Fig. 8. (a) Test results vs predictions of Raoof and Hassanen’s model sanen’s model II: upper bound [6].
I: lower bound [6]. (b) Test results versus predictions of Raoof and
Hassanen’s model I: upper bound [6].
plates, as for both plate end interfacial and mixed mode
debonding, the two very conservative predictions are for
the theoretical ultimate moment capacities are again wet lay-up plates. This may be attributed to the fact that
shown in Figs. 8b and 9b. the model was based on test results of beams bonded
with steel plates, which resemble more beams bonded
5.4. Interfacial stress based models with FRP pultruded plates than those with wet lay-up
plates.
The predictions of Ziraba et al.’s model I [34] are Predictions of Ziraba et al.’s model II [34] (Fig. 11)
clearly very widely scattered (Fig. 10) for cover separ- are less scattered than those of Ziraba et al.’s model I
ation failures. Its performance for its intended failure [34] but are very unsafe for 5 cases, raising doubt about
mode of plate end interfacial debonding is substantially the reliability of the model. For all other cases, test
better and generally conservative, with the scatter being results are conservatively predicted. For concrete cover
much smaller. It also shows reasonable performance separation failures, Ziraba et al.’s model II [34] provides
when compared with test data of mixed mode debond- more conservative predictions for beams with wet lay-
ing. The model appears to be more suitable for pultruded up plates than for those with pultruded plates. All plate
408 S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417

Fig. 10. Test results vs predictions of Ziraba et al.’s model I [34]. Fig. 12. Test results vs predictions of Varastehpour and Hamelin’s
model [35].

Fig. 11. Test results vs predictions of Ziraba et al.’s model II [34].

end interfacial debonding failures are closely predicted. Fig. 13. Test results vs predictions of Saadatmanesh and Malek’s
It is interesting to note that for plate end interfacial model [31].
debonding, their model II provides better predictions
than their model I developed specifically for plate end
interfacial debonding (Fig. 10).
All of Varastehpour and Hamelin’s [35] predictions particular, for beams with the plate terminated at the sup-
(Fig. 12) have a ratio of Mexp to Mdb below 1.0 and are port, Saadatmanesh and Malek [31] and Tumialan et al.’s
therefore un-conservative. The predictions are also scat- models [32] predict debonding loads which are unrealist-
tered for all failure modes and both plate types. ically high. This is due to the fact that even though the
Both Saadatmanesh and Malek [31] (Fig. 13) and interfacial shear stress t and normal sy stress at the plate
Tumialan et al.’s [32] (Fig. 14) results are widely scat- end are non-zero, the longitudinal stress at the base of
tered. They both provide conservative predictions when the beam sx is zero. Without the influence of sx in
only beams with wet lay-up plates are considered. For determining the maximum principal stress, the load
beams with pultruded plates, about half of the predic- required to achieve a principal stress equal to the failure
tions from these two models are un-conservative. In criterion of concrete becomes unrealistically high.
S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417 409

Fig. 14. Test results vs predictions of Tumialan et al.’s model [32]. Fig. 15. Comparison of a new debonding strength model with test
data.

6. A simple new model This plot shows that no interaction appears to exist
between the moment and the shear force, and the shear
6.1. Oehlers’ model force at failure is always greater than Vdb,s(Vc) as pre-
dicted by Eq. (2) of the companion paper (Fig. 15). As
Of the six better performing models, Oehlers’ model a result, the following simple debonding strength model
[36] is the only one which gives safe but overly con- is proposed
servative predictions for design use when either cover
Vdb,end⫽hVc (5)
separation data are examined alone or when all test data
are considered together (Tables 1–6), despite the fact
that it was developed for steel plated RC beams. It also This equation with h=1.5 and with Vc given by Eq.
generally has the smallest scatter so a new model is (2) of the companion paper provides a 95.6 percentile
therefore proposed here by modifying Oehlers’ model lower bound prediction to the cover separation data
[36]. It is also worth noting that of the six better per- (Table 1) and is shown as the dashed line in Fig. 15.
forming models, three of them [30,34,36] relate the When all plate end failures are considered but with the
debonding strength to the shear strength of concrete in test data of Ross et al. [38] excluded, h=1.4 provides a
an RC beam, while the other three are concrete tooth 94.0 percentile lower bound prediction (Table 4) and is
models from the same group with only small differences shown as the solid line in Fig. 15. In practical design,
from each other. The shear capacity based approach since it is difficult to ascertain whether cover separation
independently proposed by three different groups will be the failure mode, Eq. (5) with h=1.4 is rec-
appears to be a promising approach, which is another ommended for use to design against all forms of plate
reason for building a new model on the basis of Oehlers’ end debonding failures. It is reassuring to note that the
model [36]. Oehlers’ model [36] is given by Eqs. (1)– test results of Ross et al. [38], for beams which failed
(3) in the companion paper. by plate end interfacial debonding in the adhesive, lie
near or below the lower bound design recommendation.
6.2. A simple new model In practical application, it is important to ensure that
debonding failure occurs in the concrete by proper sur-
In Fig. 15, Eq. (3) of the companion paper is plotted face preparation and the use of strong adhesives, other-
against the test data of FRP-plated beams where the test wise the present design recommendation can lead to
data are differentiated based on failure by either concrete unsafe predictions.
cover separation, plate end interfacial debonding, and At this stage, it is necessary to limit the range of appli-
mixed mode debonding. In addition, eight test results cability of the model to that of the collected data in terms
from Ross et al. [38] (Tables 16–18) are shown on this of the magnitude of the moment at the plate end. It is
plot for beams which were reported to have failed by suggested that the above proposed model can only be
plate end interfacial debonding in the adhesive. Further used when Mdb,end/Muⱕ0.67 as this approximates the
comments on these results are made later in this section. upper limit of the test data. The choice of this ratio as
410 S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417

Table 7
Database: concrete cover separationa

Reference Beam designation Concrete beam

bc h ds dsc fc⬘ fcu Ec fct


(mm) (mm) (mm) (mm) (MPa) (MPa) (MPa) (MPa)

[10] C 152 305 251 0 39.8 49.8c 22754 3.3e


D 152 305 251 0 39.8 49.8c 22754 3.3e
G 152 305 251 0 43.0 53.8c 25512 3.5e
I 152 305 251 0 39.8 49.8c 22754 3.3e
M 152 305 251 0 43.0 53.8c 25512 3.5e
[11] B2 100 100 84 16 42.4b 53.0 34000 3.5e
B4 100 100 84 16 42.4b 53.0 34000 3.5e
B6 100 100 84 16 42.4b 53.0 34000 3.5e
[12] 1Au 100 100 84 16 47.3b 59.1 39900 4.2
1Bu 100 100 84 16 47.3b 59.1 39900 4.2
1B2u 100 100 84 16 47.3b 59.1 39900 4.2
1Cu 100 100 84 16 47.3b 59.1 39900 4.2
2Bu 100 100 84 16 47.3b 59.1 39900 4.2
2Cu 100 100 84 16 47.3b 59.1 39900 4.2
[13] B1u,1.0 100 100 84 16 43.2b 54.0 31089d 3.5e
B2u,1.0 100 100 84 16 43.2b 54.0 31089d 3.5e
B1u,2.3 130 230 206 25 37.6b 47.0 29004d 3.2e
[14] DF.2 125 225 193a 32a 46.0 57.5c 30000 3.6e
DF.3 125 225 193a 32a 46.0 57.5c 30000 3.6e
DF.4 125 225 193a 32a 46.0 57.5c 30000 3.6e
[15] AF3 125 225 193a 32a 46.0 57.5c 30000 3.6e
CF2-1 125 225 193a 32a 46.0 57.5c 30000 3.6e
CF3-1 125 225 193a 32a 46.0 57.5c 30000 3.6e
CF4-1 125 225 193a 32a 46.0 57.5c 30000 3.6e
[16] VR5 120 250 214a 34a 33.6 42.0c 27418d 3.1e
VR6 120 250 214a 34a 33.6 42.0c 27418d 3.1e
VR7 120 250 214a 34a 33.6 42.0c 27418d 3.1e
VR8 120 250 214a 34a 33.6 42.0c 27418d 3.1e
VR9 120 250 214a 34a 33.6 42.0c 27418d 3.1e
VR10 120 250 214a 34a 33.6 42.0c 27418d 3.1e
[17] P2 150 300 257a 0 40.0 50.0c 29915d 3.4e
P3 150 300 257a 0 40.0 50.0c 29915d 3.4e
P4 150 300 257a 0 40.0 50.0c 29915d 3.4e
P5 150 300 257a 0 40.0 50.0c 29915d 3.4e
[18] 2 150 250 205 45 35.4b 44.3 28158d 3.2e
6 150 250 205 45 39.9b 49.9 29885d 3.3e
7 150 250 205 45 37.6b 47.0 29004d 3.2e
[19] A3 150 300 250 0 51.7 64.6c 34010d 3.8e
A8 150 300 250 0 51.7 64.6c 34010d 3.8e
C2 150 300 250 0 51.7 64.6c 34010d 3.8e
[20] A950 120 150 120 34 25.7b 32.1 23970d 2.7e
A1100 120 150 120 34 25.7b 32.1 23970d 2.7e
A1150 120 150 120 34 25.7b 32.1 23970d 2.7e
B2 120 150 120 34 35.7b 44.6 28254d 3.2e

a
Notes: Values followed by lowercase letters have the following meaning. a — cover assumed to be 10% of h; b — converted from cube
strength using f ⬘c=0.8fcu [21]; c — converted from cylinder strength using fcu=f ⬘c/0.8 [21]; d — modulus of elasticity of concrete calculated from
Ec=4730√f ⬘c [22]; e — splitting tensile strength of concrete calculated from fct=0.53√f ⬘c [22].

a parameter means that little extra work is required as 6.3. Advantages of the new model
Mu is expected to be calculated in any flexural strength-
ening design. The use of the alternative ratio of This new model has a number of advantages over all
Mdb,end/Mdb,f may be considered with its value limited to the models assessed in the previous section and is rec-
0.78 (Fig. 15), but this would require the extra work of ommended for use in design. These include: (a) the new
evaluating Mdb,f. The use of Mdb,end/Mu thus appears to model was calibrated with a large and reliable test datab-
be preferable unless evidence to the contrary emerges. ase of FRP-strengthened beams and provides a suitable
S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417 411

Table 8
Database: concrete cover separationa

Reference Beam designation Reinforcement

Es fys As Reo* Esc fyc Asc Eyv fyv Asv s


(GPa) (MPa) (mm2) (tension) (GPa) (MPa) (mm2) (GPa) (MPa) (mm2) (mm)

[10] C 200f 414 253 2–12.7 0 0 0 200f 414 99 102


D 200f 414 253 2–12.7 0 0 0 200f 414 99 102
G 200f 414 253 2–12.7 0 0 0 200f 414 99 102
I 200f 414 253 2–12.7 0 0 0 200f 414 99 102
M 200f 414 253 2–12.7 0 0 0 200f 414 99 102
[11] B2 215 350 85 3–6 215 350 57 215 350 14 50
B4 215 350 85 3–6 215 350 57 215 350 14 50
B6 215 350 85 3–6 215 350 57 215 350 14 50
[12] 1Au 215 350 85 3–6 215 350 57 215 350 14 51
1Bu 215 350 85 3–6 215 350 57 215 350 14 51
1B2u 215 350 85 3–6 215 350 57 215 350 14 51
1Cu 215 350 85 3–6 215 350 57 215 350 14 51
2Bu 215 350 85 3–6 215 350 57 215 350 14 51
2Cu 215 350 85 3–6 215 350 57 215 350 14 51
[13] B1u,1.0 215 350 85 3–6 215 350 57 215 350 14 51
B2u,1.0 215 350 85 3–6 215 350 57 215 350 14 51
B1u,2.3 220 556 236 3–10 220 556 101 215 350 57 150
[14] DF.2 185 568 151 3–8 195 553 57 195 553 57 100
DF.3 185 568 151 3–8 195 553 57 195 553 57 100
DF.4 185 568 151 3–8 195 553 57 195 553 57 100
[15] AF3 185 568 101 2–8 195 553 57 195 553 57 71
CF2-1 185 568 129 2–8,1–6 195 553 57 195 553 57 71
CF3-1 185 568 151 3-8 195 553 57 195 553 57 71
CF4-1 183 586 207 2–10,1–8 195 553 57 195 553 57 71
[16] VR5 200f 565 157 2–10 200f 738 57 200f 738 57 110
VR6 200f 565 157 2–10 200f 738 57 200f 738 57 110
VR7 200f 565 157 2–10 200f 738 57 200f 738 57 110
VR8 200f 565 157 2–10 200f 738 57 200f 738 57 110
VR9 200f 565 157 2–10 200f 738 57 200f 738 57 110
VR10 200f 565 157 2–10 200f 738 57 200f 738 57 110
[17] P2 200f 500 308 2–14 0 0 0 200f 500 57 140
P3 200f 500 308 2–14 0 0 0 200f 500 57 140
P4 200f 500 308 2–14 0 0 0 200f 500 57 140
P5 200f 500 308 2–14 0 0 0 200f 500 57 140
[18] 2 231 537 157 2–10 231 537 157 231 537 157 100
6 231 537 157 2–10 231 537 157 231 537 157 100
7 231 537 157 2–10 231 537 157 231 537 157 100
[19] A3 207 427 792 4–15.9 0 0 0 207 427 143 125
A8 207 427 792 4–15.9 0 0 0 207 427 143 125
C2 207 427 792 4–15.9 0 0 0 207 427 143 250
[20] A950 200f 384 236 3–10 200f 400 57 200f 400 57 50
A1100 200f 384 236 3–10 200f 400 57 200f 400 57 50
A1150 200f 384 236 3–10 200f 400 57 200f 400 57 50
B2 200f 466 628 2–20 200f 400 57 200f 400 57 50

a
Notes: *Number of tensile steel reinforcing bars and bar diameter (all bars are positioned at the same effective depth ds). Values followed by
lowercase letters have the following meaning. f — modulus of elasticity of steel reinforcement assumed to be 200 GPa.

lower bound for design use (Tables 7–15); (b) the new intuitive feel of the possibility of plate end debonding
model is simpler to apply as the shear capacity of the as it is directly related to the shear capacity of the con-
concrete in the RC beam is either known or has to be crete in the beam, particularly when steel shear
evaluated in flexural strengthening design so little extra reinforcement is not present as often is the case for slabs.
work is required for checking plate end debonding; (c) Although this new model is the best proposal available
in the new model, Vc may be evaluated according to any for design use, the predicted debonding strength, being
national or international design code, and this facilitates only 40% higher than the shear capacity of the concrete,
the inclusion of the model in any national code on FRP can become a limiting factor to FRP flexural strengthen-
strengthening; (d) the new model gives designers a better ing of beams. Specialist advice shall be sought and/or
412 S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417

Table 9
Database: concrete cover separationa

Beam
Reference Adhesive Soffit plate Loading Vexp (kN)
designation

Ea ta Type Efrp ffrp tfrp bfrp a B L


(MPa) (mm) (k) (GPa) (MPa) (mm) (mm) (mm) (mm) (mm)

[10] C 8500g 2.0j G–P 11.722 161 4.76 152 203 914 2438 55.4
D 8500g 2.0j G–P 11.722 161 4.76 151 203 914 2438 59.6
G 8500g 2.0j G–P 10.343 184 4.19 152 0 914 2438 62.9
I 8500g 2.0j C/G–P 27.58 319 4.06 150 203 914 2438 50.6
M 8500g 2.0j C–P 117.905 1489 1.27 152 0 914 2438 72.1
[11] B2 11560h 2.0 G–P 49 1078 1.2 80 20 300 900 17.0
B4 11560h 2.0 G–P 49 1078 1.6 60 20 300 900 17.5
B6 11560h 2.0 C–P 118.5 987 1.2 80 20 300 900 20.4
[12] 1Au 11560 2.0 C–P 111 1273 0.5 90 20n 300 900 19.8
1Bu 11560 2.0 C–P 111 1273 0.7 65 20n 300 900 18.3
1B2u 11560 2.0 C–P 111 1273 0.7 65 20n 300 900 18.2
1Cu 11560 2.0 C–P 111 1273 1.0 45 20n 300 900 16.0
2Bu 11560 2.0 C–P 111 1273 0.7 65 20n 340 900 17.0
2Cu 11560 2.0 C–P 111 1273 1.0 45 20n 340 900 17.8
[13] B1u,1.0 8600 2.0 C–P 111 1414 0.82 67 20 300 900 18.3
B2u.1.0 8600 2.0 C–P 111 1414 0.82 67 20 300 900 16.0
B1u,2.3 8600 2.0 C–P 115 1284 1.28 90 40 844 2200 50.2
[14] DF.2 7200 — C–W 240 3500 0.167 (2)m 75 50 500 1500 60.3
DF.3 7200 — C–W 240 3500 0.167 (3)m 75 50 500 1500 60.0
DF.4 7200 — C–W 240 3500 0.167 (4)m 75 50 500 1500 62.8
[15] AF3 7200i — C–W 240 3500 0.167 (2)m 75 100 500 1500 48.3
CF2-1 7200i — C–W 240 3500 0.167 (2)m 75 100 500 1500 52.4
CF3-1 7200i — C–W 240 3500 0.167 (2)m 75 100 500 1500 59.1
CF4-1 7200i — C–W 240 3500 0.167 (2)m 75 100 500 1500 70.1
[16] VR5 8500g — C–W 230 3400 0.11 (4)m 120 75 783 2350 51.1
VR6 8500g — C–W 230 3400 0.11 (4)m 120 75 783 2350 50.3
VR7 8500g — C–W 230 3400 0.11 (7)m 120 75 783 2350 62.1
VR8 8500g — C–W 230 3400 0.11 (7)m 120 75 783 2350 62.0
VR9 8500g — C–W 230 3400 0.11 (10)m 120 75 783 2350 64.8
VR10 8500g — C–W 230 3400 0.11 (10)m 120 75 783 2350 68.5
[17] P2 8500g 1.0 C–P 150 2400 1.2 100 200 933 2800 68.0
P3 8500g 1.0 C–P 150 2400 1.2 100 200 933 2800 71.1
P4 8500g 1.0 C–P 150 2400 2.4 100 200 933 2800 78.0
P5 8500g 1.0 C–P 150 2400 2.4 100 200 933 2800 79.5
[18] 2 3260 0.37 G–W 19.723 259 1.32 150 350 500 1500 53.0
6 3260 0.37 G–W 19.723 259 1.32 150 200 500 1500 63.1
7 3260 0.37 G–W 19.723 259 1.32 150 350 500 1500 53.9
[19] A3 2000 — C–W 230 3400 0.165 (3)m 150 0 1065p 2130 86.1
A8 2000 — C–W 230 3400 0.165 (6)m 75 0 1065p 2130 98.2
C2 2000 — C–W 230 3400 0.165 (3)m 150 0 1065p 2130 79.3
[20] A950 12800 1.5 C–P 181 3140 1.2 80 190 440 1330 28.1
A1100 12800 1.5 C–P 181 3140 1.2 80 115 440 1330 28.7
A1150 12800 1.5 C–P 181 3140 1.2 80 90 440 1330 29.5
B2 12800 1.5 C–P 181 3140 1.2 80 115 440 1330 65.1

a
Notes: Values followed by lowercase letters have the following meaning. g — modulus of elasticity of adhesive assumed to be 8500 MPa;
h — assumed from [12] of the same research group; i — assumed from [15] of the same research group; j — thickness of adhesive layer for
pultruded plates assumed to be 2.0 mm; k — G=GFRP, C=CFRP, P=pultruded plate, W=wet lay-up plate; m — thickness of a single fiber sheet
and number of layers of FRP; n — assumed from [11] and [13] of the same research group; p — three point bending (remaining beams four
point bending).

tests shall be conducted if a more liberal design is 7. Conclusions


desired. Further research is urgently needed to develop
a more accurate debonding strength model to reduce the This paper has presented a large reliable test database
conservativeness of this new model. consisting of 59 test results for plate debonding failures
S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417 413

Table 10
Database: plate end interfacial debondinga

Reference Beam designation Concrete beam

bc h ds dsc f ⬘c fcu Ec fct


(mm) (mm) (mm) (mm) (MPa) (MPa) (MPa) (MPa)

[23] B 205 455 400 55 35.0 43.8c 27983d 3.1e


[11] B3 100 100 84 16 42.4b 53.0 34000 3.5e
[24] SB1 200 300 252a 48a 51.2b 64.0 33845d 3.8e
SB2 200 300 252a 48a 52.0b 65.0 34109d 3.8e
SB3 200 300 252a 48a 52.0b 65.0 34109d 3.8e
MB1 200 300 252a 48a 56.0b 70.0 35396d 4.0e
HB1 200 300 252a 48a 56.0b 70.0 35396d 4.0e
FB1 200 300 252a 48a 51.2b 64.0 33845d 3.8e

a
Notes: Values followed by lowercase letters have the following meaning. a — cover assumed to be 10% of h; b — converted from cube
strength using f ⬘c=0.8fcu [21]; c — converted from cylinder strength using fcu=f ⬘c/0.8 [21]; d — modulus of elasticity of concrete calculated from
Ec=4730√f ⬘c [22]; e — splitting tensile strength of concrete calculated from fct=0.53√f ⬘c [22].

Table 11
Database: plate end interfacial debondinga

Reference Beam designation Reinforcement

Es fys As Reo* Esc fsc Asc Eyv fyv Asv s


(GPa) (MPa) (mm2) (tension) (GPa) (MPa) (mm2) (GPa) (MPa) (mm2) (mm)

[23] B 200f 456 1013 2–25.4 200f 456 253 200f 456 253 150
[11] B3 215 350 85 3–6 215 350 57 215 350 14 50
[24] SB1 200f 527 402 2–16 200f 527 402 200f 527 157 75
SB2 200f 527 402 2–16 200f 527 402 200f 527 157 75
SB3 200f 527 402 2–16 200f 527 402 200f 527 157 75
MB1 200f 527 402 2–16 200f 527 402 200f 527 157 75
HB1 200f 527 402 2–16 200f 527 402 200f 527 157 75
FB1 200f 527 402 2–16 200f 527 402 200f 527 157 75

a
Notes: *Number of tensile steel reinforcing bars and bar diameter (all bars are positioned at the same effective depth ds). Values followed by
lowercase letters have the following meaning. f — modulus of elasticity of steel reinforcement assumed to be 200 GPa.

Table 12
Database: plate end interfacial debondinga

Beam
Reference Adhesive Soffit plate Loading Vexp (kN)
designation

Ea ta Type Efrp ffrp tfrp bfrp a B L


(MPa) (mm) (i) (GPa) (MPa) (mm) (mm) (mm) (mm) (mm)

[23] B 8500g 1.5 G–P 37.23 400 6.0 152 155 1983 4575 125.0
[11] B3 11560h 2.0 G–P 49 1078 1.2 30 20 300 900 12.3
[24] SB1 8500g 2.1 C–P 155 2400 1.4 120 150 1300 3600 71.4
SB2 8500g 2.4 C–P 155 2400 1.4 120 200 1300 3600 75.5
SB3 8500g 3.0 C–P 155 2400 1.4 120 300 1300 3600 73.9
MB1 8500g 2.4 C–P 210 2000 1.4 120 150 1300 3600 79.6
HB1 8500g 2.1 C–P 300 1400 1.4 100 150 1300 3600 80.1
FB1 8500g 0.4 C–W 95 1800 2.4 150 150 1300 3600 74.4

a
Notes: Values followed by lowercase letters have the following meaning. g — modulus of elasticity of adhesive assumed to be 8500 MPa;
h — assumed from [12] of the same research group; i — G=GFRP, C=CFRP, P=pultruded plate, W=wet lay-up plate.
414 S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417

Table 13
Database: mixed mode debondinga

Reference Beam designation Concrete beam

bc h ds dsc f ⬘c fcu Ec fct


(mm) (mm) (mm) (mm) (MPa) (MPa) (MPa) (MPa)

[12] 2Au 100 100 84 16 47.3b 59.1 39900 4.2


3Au 100 100 84 16 47.3b 59.1 39900 4.2
3Bu 100 100 84 16 47.3b 59.1 39900 4.2
3Cu 100 100 84 16 47.3b 59.1 39900 4.2
[25] B7 75 150 131a 22a 37.0 46.3c 28772d 3.2e
[18] 4 150 250 205 45 36.2b 45.3 28474d 3.45
5 150 250 205 45 40.6b 50.8 30154d 3.4e

a
Notes: Values followed by lowercase letters have the following meaning. a — cover assumed to be 10% of h; b — converted from cube
strength using f ⬘c=0.8fcu [21]; c — converted from cylinder strength using fcu=f ⬘c/0.8 [21]; d — modulus of elasticity of concrete calculated from
Ec=4730√f ⬘c [22]; e — splitting tensile strength of concrete calculated from fct=0.53√f ⬘c [22].

Table 14
Database: mixed mode debondinga

Beam Concrete beam


Reference
designation

Es fys As Reo* Esc fyc Asc Eyv fyv Asv s


(GPa) (MPa) (mm2) (tension) (GPa) (MPa) (mm2) (GPa) (MPa) (mm2) (mm)

[12] 2Au 215 350 85 3–6 215 350 57 215 350 14 51


3Au 215 350 85 3–6 215 350 57 215 350 14 51
3Bu 215 350 85 3–6 215 350 57 215 350 14 51
3Cu 215 350 85 3–6 215 350 57 215 350 14 51
[25] B7 200f 190 14 2–3 200f 470 151 200f 190 14 60
[18] 4 231 537 157 2–10 231 537 157 231 537 157 100
5 231 537 157 2–10 231 537 157 231 537 157 100

a
Notes: *Number of tensile steel reinforcing bars and bar diameter (all bars are positioned at the same effective depth ds). Values followed by
lowercase letters have the following meaning. f — modulus of elasticity of steel reinforcement assumed to be 200 GPa.

Table 15
Database: mixed mode debondinga

Beam Vexp
Reference Adhesive Soffit plate Loading
designation (kN)

Ea ta Type Efrp ffrp tfrp bfrp a B L


(MPa) (mm) (g) (GPa) (MPa) (mm) (mm) (mm) (mm) (mm)

[12] 2Au 11560 2.0 C–P 111 1273 0.5 90 20h 340 900 19.3
3Au 11560 2.0 C–P 111 1273 0.5 90 20h 400 900 19.5
3Bu 11560 2.0 C–P 111 1273 0.7 65 20h 400 900 17.3
3Cu 11560 2.0 C–P 111 1273 1.0 45 20h 400 900 15.4
[25] B7 10250 2.5 C–P 150 2400 1.2 50 10 650 1500 12.5
[18] 4 3260 0.37 G–W 19.723 259 1.32 150 200 500 1500 65.4
5 3260 0.37 G–W 19.723 259 2.64 150 50 500 1500 79.4

a
Notes: Values followed by lowercase letters have the following meaning. g — G=GFRP, C=CFRP, P=pultruded plate, W=wet lay-up plate;
h — assumed from [11] and [13] of the same research group.

in FRP-strengthened simply-supported RC beams and 1. Of the two distinct modes of plate end debonding fail-
assessed the performance of all 12 debonding strength ures, concrete cover separation is far more common
models reviewed in the companion paper [1] using this than plate end interfacial debonding. Further studies
database. Based on this study, the following conclusions should be carried out to determine the circumstances
and observations can be made: under which plate end interfacial debonding occurs.
S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417 415

Table 16
Database: plate end interfacial debonding (failure in adhesive)a

Reference Beam designation Concrete beam

bc h ds dsc f ⬘c fcu Ec fct


(mm) (mm) (mm) (mm) (MPa) (MPa) (MPa) (MPa)

[38] 1B 200 200 152 48 54.8 68.5a 34500 3.9b


1C 200 200 152 48 54.8 68.5a 34500 3.9b
2B 200 200 152 48 54.8 68.5a 34500 3.9b
2C 200 200 152 48 54.8 68.5a 34500 3.9b
2D 200 200 152 48 54.8 68.5a 34500 3.9b
3B 200 200 152 48 54.8 68.5a 34500 3.9b
3C 200 200 152 48 54.8 68.5a 34500 3.9b
3D 200 200 152 48 54.8 68.5a 34500 3.9b

a
Notes: Values followed by lowercase letters have the following meaning. a — converted from cylinder strength using fcu=f ⬘c/0.8 [21]; b —
splitting tensile strength of concrete calculated from fct=0.53√f ⬘c [22].

Table 17
Database: plate end interfacial debonding (failure in adhesive)a

Beam Reinforcement
Reference
designation

Es fys As Reo* Esc fsc Asc Eyv fyv Asv s


(GPa) (MPa) (mm2) (tension) (GPa) (MPa) (mm2) (GPa) (MPa) (mm2) (mm)

[38] 1B 200 410 143 2–9.5 200 410 143 200 410 143 102
1C 200 410 143 2–9.5 200 410 143 200 410 143 102
2B 200 410 253 2–12.7 200 410 143 200 410 143 102
2C 200 410 253 2–12.7 200 410 143 200 410 143 102
2D 200 410 253 2–12.7 200 410 143 200 410 143 102
3B 200 410 396 2–15.9 200 410 143 200 410 143 102
3C 200 410 396 2–15.9 200 410 143 200 410 143 102
3D 200 410 396 2–15.9 200 410 143 200 410 143 102

a
Notes: *Number of tensile steel reinforcing bars and bar diameter (all bars are positioned at the same effective depth ds).

Table 18
Database: plate end interfacial debonding (failure in adhesive)a

Beam
Reference Adhesive Soffit Plate Loading Vexp (kN)
designation

Ea ta Type Efrp ffrp tfrp bfrp a B L


(MPa) (mm) (e) (GPa) (MPa) (mm) (mm) (mm) (mm) (mm)

[38] 1B 8500c 2.0d C–P 138 2206 0.45 200 1 914 2742 40.1
1C 8500c 2.0d C–P 138 2206 0.45 200 1 914 2742 35.6
2B 8500c 2.0d C–P 138 2206 0.45 200 1 914 2742 49.0
2C 8500c 2.0d C–P 138 2206 0.45 200 1 914 2742 35.6
2D 8500c 2.0d C–P 138 2206 0.45 200 1 914 2742 40.1
3B 8500c 2.0d C–P 138 2206 0.45 200 1 914 2742 54.5
3C 8500c 2.0d C–P 138 2206 0.45 200 1 914 2742 54.1
3D 8500c 2.0d C–P 138 2206 0.45 200 1 914 2742 54.3

a
Notes: Values followed by lowercase letters have the following meaning. c — modulus of elasticity of adhesive assumed to be 8500 MPa;
d — thickness of adhesive layer for pultruded plates assumed to be 2.0 mm; e — C=CFRP, P=pultruded plate.
416 S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417

2. Of the 12 debonding strength models, those by 12) (8 results), and the third set for failure by a mixed
Oehlers [36], Ziraba et al. [34], Jansze [30], Raoof concrete cover separation and interfacial debonding
and Zhang [37], and Raoof and Hassanen [6] are the mode (Tables 13–15) (7 results). In addition test results
more accurate ones. Surprisingly, all but those of from Ross et al. [38] are shown for beams which were
Raoof and Hassanen [6] were developed for steel reported to have failed by plate end interfacial debonding
plated beams, and even Raoof and Hassanen’s models in the adhesive (Tables 16–18).
[6] were modified from that of Raoof and Zhang [37]
for steel plated beams. All models specifically
References
developed for FRP-plated beams give poorer predic-
tions. These conclusions remain basically unchanged [1] Smith ST, Teng JG. FRP-strengthened RC beams. I: review of
whether cover separation test data only or all plate debonding strength models. Engng Struct 2001;24(4):385–395.
end debonding test data are considered. [2] Oehlers DJ. Discussion on “Theoretical study of the behaviour
3. Of the three approaches employed for developing of reinforced concrete beams strengthened by externally bonded
steel plates” by Roberts TM and Haji-Kazemi H. Proc Inst Civil
debonding strength models, the shear capacity based Engnrs Part 2 1989;87:651–63.
models appears to be the most robust, followed by the [3] Oehlers DJ. Discussion on “Prediction of peeling failure of
concrete tooth models, with interfacial stress based reinforced concrete beams with externally bonded plates” by
models being the least reliable. The interfacial stress Zhang S, Raoof M, Wood LA. Proc Inst Civil Engnrs Struct Build
based models by Saadatmanesh and Malek [31] and 1995;122:493–6.
[4] Oehlers DJ, Ali M. Discussion on “Predictions of the maximum
Tumialan et al. [32] however provide safe predictions plate end stresses of FRP strengthened beams: Part II” by Quan-
for beams with wet lay-up plates. trill RJ, Hollaway LC, Thorne AM. Mag Concrete Res
4. Only Oehlers’ model [36] can be used directly in 1998;50(1):91–2.
design as it provides safe predictions although its pre- [5] Bonacci JF. Strength, failure mode and deformability of concrete
dictions are overly conservative. All other models do beams strengthened externally with advanced composites. In: El-
Badry MM, editor. Advanced composite materials in bridges and
not provide predictions of test debonding strengths structures, Proceedings of the Second International Conference,
with a sufficient degree of conservativeness for direct Quebec, Canada, 1996:419–26.
design use. [6] Raoof M, Hassanen MAH. Peeling failure of reinforced concrete
5. The interaction between the shear force and the bend- beams with fibre-reinforced plastic or steel plates glued to their
ing moment at the plate end at debonding included in soffits. Proc Inst Civil Engnrs Struct Build
2000;140(August):291–305.
Oehlers’ model [36] does not appear to exist for FRP- [7] Mukhopahdyaya P, Swamy RN. Interface shear stress: a new
plated beams, at least according to the test data avail- design criterion for plate debonding. J Composites Construct,
able so far. Based on this observation, a simple new ASCE 2001;5(1):35–43.
model which is superior to all existing models in pro- [8] Roberts TM. Approximate analysis of shear and normal stress
viding safe and accurate predictions for design use concentrations in the adhesive layer of plated RC beams. Struct
Engnr 1989;67(12):229–33.
has been proposed by modifying Oehlers’ model [36]. [9] Arduini M, Nanni A. Behavior of precracked RC beams strength-
This new model is suitable for direct application in ened with carbon FRP sheets. J Composites Construct, ASCE
design, and can be easily included in any national 1997;1(2):63–70.
code. [10] Ritchie PA, Thomas DA, Lu LW, Connely GM. External
reinforcement of concrete beams using fiber reinforced plastics.
ACI Struct J 1991;88(4):490–500.
[11] Quantrill RJ, Hollaway LC, Thorne AM. Experimental and ana-
lytical investigation of FRP strengthened beam response: Part I.
Acknowledgements Mag Concrete Res 1996;48(177):331–42.
[12] Garden HN, Hollaway LC, Thorne AM. A preliminary evaluation
Both authors wish to thank The Hong Kong Polytech- of carbon fibre reinforced polymer plates for strengthening
reinforced concrete members. Proc Inst Civil Engnrs Struct Build
nic University for the provision of a postdoctoral fellow- 1997;123(May):127–42.
ship to the first author. Dr J.F. Bonacci of the Depart- [13] Garden HN, Quantrill RJ, Hollaway LC, Thorne AM, Parke
ment of Civil Engineering at the University of Toronto, GAR. An experimental study of the anchorage length of carbon
Canada, is gratefully acknowledged for providing the fibre composite plates used to strengthen reinforced concrete
authors with a copy of his test database. beams. Construct Build Mater 1998;12:203–19.
[14] Ahmed O, van Gemert D. Effect of longitudinal carbon fiber
reinforced plastic laminates on shear capacity of reinforced con-
crete beams. In: Dolan CW, Rizkalla SH, Nanni A, editors. Pro-
Appendix A. Experimental database ceedings of the Fourth International Symposium on Fiber
Reinforced Polymer Reinforcement for Reinforced Concrete
The test data in the new database have been grouped Structures, Maryland, USA, 1999:933–43.
[15] Ahmed O, Van Gemert D. Behaviour of RC beams strengthened
into three sets, the first set for failure by concrete cover in bending by CFRP laminates. In: Forde MC, editor. Proceedings
separation (Tables 7–9) (44 results), the second set for of the Eighth International Conference on Advanced Composites
failure by plate end interfacial debonding (Tables 10– for Concrete Repair, London, UK, 1999.
S.T. Smith, J.G. Teng / Engineering Structures 24 (2002) 397–417 417

[16] Beber AJ, Filho AC, Campagnolo JL. Flexural strengthening of [27] Quantrill RJ, Hollaway LC, Thorne AM. Predictions of the
R/C beams with CFRP sheets. In: Forde MC, editor. Proceedings maximum plate end stresses of FRP strengthened beams: Part II.
of the Eighth International Conference on Advanced Composites Mag Concrete Res 1996;48(177):343–51.
for Concrete Repair, London, UK, 1999. [28] Teng JG, Chen JF, Smith ST, Lam L. RC structures strengthened
[17] David E, Djelal C, Ragneau E, Bodin FB. Use of FRP to with FRP composites. Hong Kong, China: Research Centre for
strengthen and repair RC beams: experimental study and numeri- Advanced Technology in Structural Engineering, Department of
cal simulations. In: Forde MC, editor. Proceedings of the Eighth Civil and Structural Engineering, The Hong Kong Polytechnic
International Conference on Advanced Composites for Concrete University, 2000.
Repair, London, UK, 1999. [29] BS 8110. Structural use of concrete, Part 1. Code of practice for
[18] Hau KM. Experiments on concrete beams strengthened by bond- design and construction. London, UK: British Standards Insti-
ing fibre reinforced plastic sheets. MSc in Civil Engineering The- tution; 1997.
sis. Hong Kong, China: The Hong Kong Polytechnic Univer- [30] Jansze W. Strengthening of RC members in bending by externally
sity; 1999. bonded steel plates. PhD Thesis. Delft: Delft University of Tech-
[19] Tumialan G, Serra P, Nanni A, Belarbi A. Concrete cover delami- nology; 1997.
nation in reinforced concrete beams strengthened with carbon [31] Saadatmanesh H, Malek AM. Design guidelines for flexural
fiber reinforced polymer sheets. In: Dolan CW, Rizkalla SH, strengthening of RC beams with FRP plates. J Composites Con-
Nanni A, editors. Proceedings of the Fourth International Sym- struct, ASCE 1998;2(4):158–64.
posium on Fiber Reinforced Polymer Reinforcement for [32] Tumialan G, Belarbi A, Nanni A. Reinforced concrete beams
Reinforced Concrete Structures, Maryland, USA, 1999:725–35. strengthened with CFRP composites: Failure due to concrete
[20] Nguyen DM, Chan TK, Cheong HK. Brittle failure and bond cover delamination. Report No. CIES-99/01 USA: Department of
development length of CFRP–concrete beams. J Composites Con-
Civil Engineering, Center for Infrastructure Engineering Studies,
struct, ASCE 2001;5(1):12–7.
University of Missouri-Rolla, 1999.
[21] Kong FK, Evans RH. Reinforced and prestressed concrete. 3rd
[33] Wang CY, Ling FS. Prediction model for the debonding failure
ed London: Chapman & Hall, 1987.
of cracked RC beams with externally bonded FRP sheets. In:
[22] ACI 318-95. Building code requirements for structural concrete
Proceedings of the Second International Conference of Com-
(318-95) and commentary (318R-95), 5th printing. Michigan,
USA: American Concrete Institute (ACI); 1999. posites in Infrastructure (ICCI 98), Arizona, USA, 1998:548–62.
[23] Saadatmanesh H, Ehsani MR. RC Beams strengthened with [34] Ziraba YN, Baluch MH, Basunbul IA, Sharif AM, Azad AK, Al-
GFRP plates. I: Experimental study. J Struct Engng Sulaimani GJ. Guidelines towards the design of reinforced con-
1991;117(11):3417–33. crete beams with external plates. ACI Struct J 1994;91(6):639–
[24] Täljsten B. Concrete beams strengthened for bending using 46.
CFRP-sheets. In: Forde MC, editor. Proceedings of the Eighth [35] Varastehpour H, Hamelin P. Strengthening of concrete beams
International Conference on Advanced Composites for Concrete using fiber-reinforced plastics. Mater Struct 1997;30:160–6.
Repair, London, UK, 1999. [36] Oehlers DJ. Reinforced concrete beams with plates glued to their
[25] Juvandes L, Figueiras JA, Marques AT. Performance of concrete soffits. J Struct Engng, ASCE 1992;118(8):2023–38.
beams strengthened with CFRP laminates. Proceedings of the [37] Raoof M, Zhang S. An insight into the structural behaviour of
Second International Conference on Composites in Infrastructure reinforced concrete beams with externally bonded plates. Proc
(ICCI), Arizona, USA. 1998. Inst Civil Engnrs: Struct Build 1997;122(November):477–92.
[26] Swamy RN, Jones R, Bloxham JW. Structural behaviour of [38] Ross CA, Jerome DM, Tedesco JW, Hughes ML. Strengthening
reinforced concrete beams strengthened by epoxy-bonded steel of reinforced concrete beams with externally bonded composite
plates. Struct Engnr 1987;65A(2):59–68. laminates. ACI Struct J 1999;96(2):212–20.

You might also like