Bond Behavior of FRP in Structures ASE 9-6

ASE 9-6 front page
5/12/06
8:39 am
Page a
Advances in
Structural Engineering
An International Journal
ISSN 1369-4332
Editor-in-Chief
Professor J. G. Teng
The Hong Kong Polytechnic University
Volume 9 Number 6 December 2006

Special Issue on Bond Behaviour of FRP in Structures
Guest Editors: J.F. Chen and J.G. Teng
Construction site of a railway project in Hong Kong where GFRP

bars were used to construct an underground diaphragm wall;
photographed by Dr Hon-Ting Wong
A Multi-Science Publication
ASE 9-6 front page
5/12/06
8:39 am
Page b
ASE 9-6 contents
5/12/06
8:39 am
Page a
Advances in
Structural Engineering
An International Journal
CONTENTS
Page
Guest Editorial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
719
Technical Papers
The Importance of Bonding-a Historic Overview and Future Possibilities
B. Tljsten . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 721
FRP Plates Adhesively Bonded to Reinforced Concrete Beams: Generic
Debonding Mechanisms
D. J. Oehlers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 737
Analysis of the Load-Deformation Behaviour and Debonding for FRPStrengthened Concrete Structures
K. W. Neale, U. A. Ebead, H. M. Abdel Baky, W. E. Elsayed and A. Godat 751
Fracture Mechanics Approach to Geometrically Nonlinear Debonding
Problems in RC Beams Strengthened with Composite Materials
Oded Rabinovitch and Yeoshua Frostig . . . . . . . . . . . . . . . . . . . . . . 765
Deterioration of FRP-to-Concrete Bond under Failure Loading
Kent A. Harries, John Aidoo, Andrew Zorn and
Joshua Quattlebaum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
779
Advances in Adhesive Joining of Carbon Fibre/Polymer Composites to

Steel Members for Repair and Rehabilitation of Bridge Structures
L. C. Hollaway, L. Zhang, N. K. Photiou, J. G. Teng and S. S. Zhang . . 791
Bond Behavior of CFRP Strengthened Steel Structures
D. Schnerch, M. Dawood, S. Rizkalla, E. Sumner and K. Stanford . .
805
Experimental and Numerical Investigation of the Behaviour and Strength

of Composite Steel-CFRP Members
Mohammad Al-Emrani and Robert Kliger . . . . . . . . . . . . . . . . . . . . 819
Announcement
Best Paper Award . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
833
ASE 9-6 contents
5/12/06
8:39 am
Page b
ASE 9-6-00_Editorial
4/12/06
5:12 pm
Page 719
Guest Editorial
This special issue contains extended and revised
versions of 8 papers selected from those presented at the
International Symposium on Bond Behaviour of FRP in
Structures (BBFS 2005), which was held on 79
December 2005 in Hong Kong, China. The Symposium
was organised by the Working Group on Bond between
FRP and Concrete (the Bond WG) of the International
Institute for FRP in Construction (IIFC), in conjunction
with the Department of Civil and Structural Engineering,
The Hong Kong Polytechnic University.
The IIFC Bond WG was established in November
2004 with the objectives of:
promoting research collaboration on FRP-toconcrete bond behaviour
providing a forum for exchange, discussion and

consolidation of ideas and results
establishing design models and
developing a state-of-the-art report on the

behaviour and modelling of bond behaviour
between FRP and concrete.
BBFS 2005 was organised as part of the Bond WGs
activities to provide an international forum and a stateof-the-art survey of existing research in the area of bond
behaviour of FRP in structural applications. The scope of
the Symposium covered all aspects of bond performance
of FRP in structures, including bond behaviour
between (external, near surface mounted and internal)
FRP and other structural materials such as concrete,
masonry, metal and timber, as well as between two
FRP elements.
The planning of the symposium began at the end of
2004. A proposal of the symposium was submitted to
and approved by the IIFC Executive Committee in
March 2005. The symposium received 128 abstract
submissions, most of which were accepted after initial
reviews. All full papers underwent a rigorous review
Advances in Structural Engineering Vol. 9 No. 6 2006
process by two expert reviewers selected from the

International Scientific Committee and the International
Organising Committee of BBFS 2005. A total of 6 keynote
papers and 69 contributed papers were finally selected for
publication in the proceedings and for oral presentation.
The papers cover a wide range of bond behaviour issues
including fundamental bond mechanisms, debonding
failures, dynamic and cyclic loadings, fatigue, durability,
and the effects of elevated temperature and fire. They
report the current state-of-the-art and point to future
directions of research in this exciting area.
As Guest Editors of this special issue, we would like
to acknowledge the contributions of many individuals to
the successful publication of this special issue. In
particular, we would like to thank the authors for their
meticulous preparation of the papers and the reviewers
for their prompt and high quality reviews.
This special issue would not have been developed
without the success of the Symposium. As Co-Chairmen
of the Symposium, we would like to thank the
International Organizing Committee and the International
Scientific Committee of BBFS 2005 who not only
reviewed the papers to ensure a quality symposium but
also mobilized widespread support for the Symposium.
Thanks are also due to the Executive Committee of IIFC
for its support to the organization of BBFS 2005, the
Department of Civil and Structural Engineering, The
Hong Kong Polytechnic University, which provided
support in many ways as a Co-Organizer of the
Symposium, all authors of BBFS 2005 for their careful
preparation of manuscripts and all keynote speakers for
sharing their work at the symposium.
GUEST EDITORS:
Jian-Fei Chen, Edinburgh University and Jin-Guang
Teng, The Hong Kong Polytechnic University
719
ASE 9-6-00_Editorial
4/12/06
5:12 pm
Page 720
ASE 9-6-01_Taljesten
4/12/06
5:11 pm
Page 721
The Importance of Bonding A Historic Overview

and Future Possibilities
B. Tljsten*
Department of Civil Engineering, Technical University of Denmark
Abstract: An adhesive is a substance that has the capability to hold materials together.
The manner by which adhesives are able to serve this function is due to a surface
attachment that is resistant to separation. A bond occurs when the adhesive molecules
adsorb onto a solid surface and chemically react with it. Adhesives have a very large
importance for most industries, both for structural and for non-structural use. In this
paper, an overview of the use of adhesives in different industries is presented with
focus on the building industry and on structural use. Not surprisingly, the use of
structural bonding has increased considerably during the last decade. The main reasons
for this are probably a combination of lower cost, more need of lightweight structures,
development of more durable adhesives and improved structural adhesive perf1ormance.
Additionally, adhesives often provide better load transfer than welded structures.
In this paper, an overview of the use of adhesive in the service of man is presented.
The overview is brief, but nonetheless gives an idea of the possibilities of adhering
materials by bonding. The paper briefly discusses the use of adhesives in the
aerospace, automotive and marine industries, but focuses on the construction industry.
Finally some ideas about sticking things together tomorrow are presented.
Key words: adhesives, structural bonding, joints, construction industry, organic.
1. HISTORICAL OVERVIEW
Natural adhesives have been known since pre-historic
times. Two stones dating back at least 36,000 years
have traces of a sticky black substance that was once
used to attach them to a handle. In a chemical analysis
the glue-like material was identified as bitumen. The
oldest discovery of the use of adhesives by humans to
date was made in the Nahal Hemar Cave to the
northwest of Mount Sedom in Israel. When this cave
was excavated in 1983, many of the artefacts unearthed
were found to carry residues of collagen based material
believed to be derived from animal skins. This adhesive
has been carbon dated to over 8,000 years ago (Walker
1998; Bar-Yosef and Schick 1989). Archaeologists
studying burial sites of prehistoric tribes dating to about
4000 B.C. found foodstuffs buried with the deceased in
broken pottery vessels that had been repaired with sticky
resins from tree sap. Archaeologists have also uncovered

statues from Babylonian temples that have ivory
eyeballs glued into eye sockets. This tar-like glue has
been known for almost 6000 years. Egyptian carvings
dating back 3,300 years depict the gluing of a thin piece
of veneer to what appears to be a plank of sycamore.
Flour paste was used to bond together papyrus fibres
that were then used as fabrics. Beeswax, tree pitches
and bitumen were used as protective coatings and
adhesives. Egg whites were used to bind manuscripts at
one time and wooden objects were bonded with glues
made from fish, horn, and cheese. The period of time
between 15001000 B.C. gave further proof that glue
had become a method of assembly. Paintings and
murals showed details of wood gluing operations. A
casket removed from the tomb of King Tutankhamen
shows the use of glue in its construction. A study of
*Corresponding author. Email address: bt@byg.dtu.dk; Fax: +45 45 88 32 82; Tel: +45 45 24 17 73.
721
4/12/06
5:11 pm
Page 722
The Importance of Bonding A Historic Overview and Future Possibilities
history shows the use of glue fell into disuse until about
15001700 A.D. when adhesives were used in the
building of furniture. Some of the greatest furniture and
cabinet makers of all times used adhesives in their
productsnames still recognizable today like Chippendale
and Duncan Phyfe. Another notable name in history
who may owe his notoriety, at least in part to adhesives
in Genghis Khan. Around the year 1000 A.D., Genghis
Khan overcame all attackers because of the exceptional
power and range of the weaponry his men carried. Bows
were made from laminated lemon wood and bullhorn
bonded with an adhesive whose formulation has been
lost to antiquity.
By the 17th century, scientists were beginning to give
consideration to the nature of adhesion itself. Francis
Bacon in his Novum Organum suggested that there is in
all bodies a tendency to avoid breaking up, (Fay 2005).
A little later (Newton 1717) conjectured that, There are
agents in nature able to make the particles of bodies
stick together by very strong attractions. And it is the
business of experimental philosophy to find them out.
A general renaissance in the use of adhesive bonding
began around this time and is clearly demonstrated by
the changing construction methods used for furniture.
During the early days in the history of adhesives, it is
likely that the materials were produced on very small
scale, possible in the kitchen of the individual users.
However, by around the 1700s, the production of
adhesives started to undergo transformation into a major
industry. During the 18th century, the technology of
animal and fish glues advanced. In the 19th century,
rubber and nitrocellulose based cements were introduced.
In about 1700, the widespread use of glue brought about
some rapid changes in the history of adhesives. The first
commercial glue factory was started in Holland to
manufacture animal glue from hides. About 1750, the
first glue patent was issued in Britain for fish glue.
Patents were then rapidly issued for adhesives using
natural rubber, animal bones, fish, starch and milk
protein (casein). By 1900, the U.S. had a number of
factories producing glue from the aforementioned bases.
Although enormous efforts had been made in the
manufacture of animal glue in the period up to
approximately 1925, much bigger developments were
taking place which were to have the most significant
effect on the history of adhesives in thousands of years
the development of synthetic polymers (Fay 2005).
The Industrial Revolution caused an explosion in
technical breakthroughs which resulted in new materials
becoming available for use in formulating adhesives.
The first plastic polymer to be synthesized was cellulose
nitrate, a thermoplastic material derived from the
cellulose of wood. Its first use was in the manufacture of
722
billiard balls, which had been made of ivory. The era of

plastics began with the introduction of Bakelite
phenolic, a thermoset plastic, in 1910. Within a year,
adhesives using phenolic resin were on the market. The
1920s, 30s and 40s saw many new plastics and
rubbers produced synthetically; many out of urgent
necessity were developed during World War II.
Although adhesives have been known for a very long
time, most of the technology of adhesives has been
developed in the last 100 years. Why then use
adhesives? Well, the reason will vary from case to case
and from industry to industry but some advantages
worth to mention can be:
Increased speed of production, in particular for

the automotive and aerospace industries.
Wider selection of materials. It is possible to mix

and match material combination to suit production
costs. Traditional materials may be combined
with new metal alloys, plastics, composites and
ceramics to give distinct product advantages.
Local stress concentrations can be avoided and

the bond is continuous, which also might lead to
stronger and stiffer structures.
Compared to welded structures bonded structures

often has improved fatigue resistance.
2. CLASSIFICATION OF ADHESIVES
Adhesives may be classified as either organic or
inorganic materials in a number of different ways; for
example by origin, by method of bonding, by end use or
by chemical basis (Mays and Hutchinson 1992). Table 1
gives an overview of a general classification of organic
adhesives. Even though focus will be placed on
structural adhesives, some different adhesives are
briefly discussed in general terms below.
Animal Glues are made from the protein extracted
from the bones, hide, hoofs and horns of animals by
boiling. The extract is cooked to form a gelatine
material. The gelatine can then be reliquified with heat,
which gives it quick setting properties. Its major use has
been in the wood and furniture industry. Animal byproducts from meat processing have been the source of
supply for this type of glue. Fish glue is a similar
protein-based glue made from the skins and bones of
fish. An exceptionally clear adhesive can be made from
fish and was the first adhesive used for photographic
emulsions for photo film and photo resistant coatings for
photoengraving processes. Casein glue is made from a
protein isolated from milk. The extraction process
creates an adhesive that is waterproof. Its first use was
in bonding the seam of cigarette paper.
Glues made from vegetable starch, which is a
carbohydrate extracted from vegetable plants such as
4/12/06
5:11 pm
Page 723
B. Tljsten
Table 1. Organic adhesive classification, after (Mays and Hutchinson 1992)

Group
Animal
Vegetable
Mineral
Elastomeric
Thermoplastic
Thermosetting
Type
Gelatin
Casein
Albumen
Starch
Cellulose
Asphalt/bitumen
Natural rubber
Synthetic rubber
PVA
Polystyrene
Cyanoacrylates
Liquid acrylic
Formaldehyde (urea/phenol)
Unsaturated polyesters
Epoxy resins
Polyurethane
Source
Mammals, fish
Milk
Blood
Corn, potatoes, rice
Cellulose
Earths crust
Tree latex
Synthetic
Synthetic
Synthetic
Synthetic
Synthetic
Synthetic
Synthetic
Synthetic
Synthetic
corn, rice, wheat, and potatoes are probably better

known as paste. Major uses of these materials include
bonding of paper and paper products (such as
bookbinding, corrugated boxes, paper bags, wallpaper
paste) and as a sizing in textiles. The laundry uses starch
on your shirt collars, to stiffen and give shape to your
shirt. Cellulose adhesive is made from a natural polymer
found in trees and woody plants; it is the adhesive used
on the cellophane wrapper on cigarette packs and the
adhesive on decals placed on windows.
Elastomeric adhesives are adhesives made by
combining one or more rubbers or elastomers in a
solvent. These solutions are further modified with
additives to improve the tack or stickiness, the degree
of peel strength, flexibility, the viscosity or body.
Rubber-based adhesives are used in a wide variety of
applications such as contact adhesive for plastic
laminates on counter tops, cabinets, desks and tables.
Self sealing envelopes and shipping containers use
rubber cements. Solvent based rubber adhesives have
been the mainstay of the shoe and leather industry.
Furthermore, silicone adhesive is a rubber-like polymer
called polydimethylsiloxanes. Silicone rubber adhesives
are made from a complicated process that turns
elemental silicon metal made from sand (silica) into a
rubbery polymer. Because of its exceptional properties,
silicone adhesive has been used in some exotic
applications such as the soles of the boots worn by the
first astronauts to walk on the moon.
The family of thermoplastics includes cyanoacrylates
which are extremely rapid curing adhesives known as
instant adhesive or super glue. A cyanoacrylate adhesive
is a very rapid curing adhesive also from the acrylic
family tree, but having a completely different cure
Use
Can labels
Plywood, blockboard
Paper, packaging
Leather, wood, china
Road pavements
Carpet making (rare)
Tyre, fabrics, bookbinding
Wood and general
Model making
Plastics, metals, glass, rubber
Structural vehicle assembly
Chipboard and plywood
Glass fiber, resin mortars
Structurally, metal to metal, concrete
Semi-structural uses with plastics,
metals wood and sandwich panel construction
system. When a drop of cyanoacrylate adhesive is put

on the surface of a part, the acid stabilizer molecules
react with water molecules present on the surface of the
part from the relative humidity in the air. The reaction of
the water and acid causes the acid stabilizer to be
neutralized. The cyanoacrylate molecules then react
with each other and form polymer chains without crosslinking. Cyanoacrylates begin to form polymer chains
immediately on contact with the water vapour on the
surface of the part. If parts are moved during initial
contact, the polymerization process and polymer chains
are stopped. The process must start again at a new
catalyzed site. The versatility of these adhesives makes
them highly useful in all industries. Some of the larger
application areas are in electronics for printed circuit
board wires and components, and in medical technology
for disposable plastic medical devices.
Hot melt adhesives are another thermoplastic
polymer that are tough and solid at room temperature,
but are liquid at elevated temperatures. The origin of
hot melts probably started with the use of sealing wax
used to seal documents and letters with a signature ring
or stamp, but the art of hot melts was not pursued until
the 1960s.
Anaerobic adhesives also belong to the family of
thermoplastics. They are derived from methacrylates,
a monomer related to the acrylic more commonly
known as Plexiglas. The basic ingredient in anaerobic
adhesives is a monomer from the acrylic family, a
special liquid of small molecules which can combine
chemically to form a polymer or group of molecules.
Anaerobic adhesives are very versatile and are used in a
wide variety of applications. The very first application
of an anaerobic adhesive was to lock and seal a screw in
723
4/12/06
5:11 pm
Page 724
The Importance of Bonding- A Historic Overview and Future Possibilities
the carburettor of an automobile engine. Today,

anaerobic adhesives not only lock all types of threaded
fasteners, they are used to seal pipe fittings, retain
bearings, pulleys, and gears to shafts as well as sealing
flanged surfaces and replacing gaskets. The major users
of these adhesives are the automotive, truck, construction
and farming equipment companies. The actual market
for these adhesives is anywhere that fasteners, gaskets,
bearings or any mechanical device needs to be secured
or sealed.
Unlike thermoplastics, thermosetting adhesives do
not melt or flow on heating but become rubbery and lose
strength. The molecular chains present in thermosetting
adhesives undergo irreversible cross-linking during
curing (Mays and Hutchinson, 1992). The most familiar
thermosetting adhesives are the family of epoxies.
Epoxy adhesives have only been available commercially
since the Second World War. Epoxies are adhesive
systems made by a complex chemical reaction. Various
resins are made synthetically by reacting two or more
chemicals. The resultant resin can then be reacted or
cured by the addition of another chemical called a
hardener, or catalyst. The basic epoxy resin systems are
further modified to change their physical properties by
the addition of such things as flexibilizers for impact
resistance and flexibility, diluents or solvents to reduce
the viscosity. Fillers and reinforcements like glass fibre,
alumina, silica sand, clay, metal powders and flakes
may be added to change properties such as heat and
electrical resistance, fire resistance or retardance, and
strength and adhesion to certain substrates or materials.
Epoxy adhesives can bond a wide variety of high
strength substrates, particularly metals. They have been
used to replace some traditional metalworking methods
of joining like nuts and bolts, rivets, welding, crimping,
brazing and soldering. High strength epoxies are used
to construct rotor blades of helicopters and attach
aluminium skins to the struts of aircraft wings and
tail sections. In the construction industry they are
commonly used for injection of concrete cracks and
external strengthening of structures.
3. STRUCTURAL ADHESIVES
Adhesives can be classified by the manner in which they
harden. This can be loss of solvent, loss of water, cooling
or chemical reaction, (Comyn 2005). Structural
adhesives are most often adhesives that involve a
chemical reaction. Structural adhesives are distinguished
from other adhesives by being high strength materials that
are designed to support loads. Such adhesives are often
subjected to cycling high and low temperatures and
aggressive environments, fluids or the weather. They are
generally used for the bonding of rigid structures,
724
although a degree of flexibility is required in the adhesive

to counter the effects of movement, impact or vibration.
Materials most commonly bonded with structural
adhesives are metals, glass, ceramics, concrete, plastics
and composites. Structural adhesives include anaerobics,
epoxies, reactive acrylics, polyurethanes, reactive hot melt
polyurethanes, and special formulations of cyanoacrylates.
As structural adhesives, epoxies are the most widely
accepted and used; therefore primarily structural bonding
with epoxies will be discussed in this paper. Epoxies
typically contain several components, the most important
being the resin. In a broad sense, the term epoxy refers to
a chemical group consisting of an oxygen atom bonded
with two carbon atoms already united in some way.
Epoxy resins are a group of polymers with extremely
different chemical, thermal and mechanical properties.
The resins are obtained through refining of petroleum.
The mixing of epoxy resin with a hardener results in an
epoxy adhesive. The properties of epoxy plastics are
mainly dependent on the hardener used. Epoxy adhesives
were first used in 1930 in Germany, USA and
Switzerland. Among the first industries to use epoxies
was the aerospace industry. During the 1950s epoxies
began to be used commercially in the construction
industry. Tremper (1960) describes the use of epoxies in
repairing concrete highways and Gaul and Apton (1959)
for repair on runways and roads. Wakeman et al. (1962)
describe resin injection of cracked pile caps and beams. In
Table 2 typical material properties for epoxies are given.
A very important material property for epoxies is the
glass transition temperature. Most synthetic adhesives are
based on polymeric materials, and as such, they exhibit
properties that are characteristic for polymers. Polymers
change from relatively hard to elastic glass-like to
relatively rubbery materials at a certain temperature.
This temperature level is defined as the glass transition
temperature, and is different for different polymers.
For epoxies the glass transition temperature normally
varies between 45C up to 200C, where cold cured
epoxies exhibit the lowest and heat cured epoxies the
highest. For high tech industries, such as the aerospace
industry, hot cured adhesives dominate. For the
building industry on site applications with cold-cured
epoxies is most common.
4. CONSIDERATIONS FOR SELECTING
THE RIGHT ADHESIVE
The successful performance of a bonded product is
dependent upon adequate adhesion between two or
more constituents. The strength of bonded joints
depends not only on the cohesive strength of the
adhesive, but also on the degree of adhesion to the
bonding surface. One of the disadvantages of adhesive
4/12/06
5:11 pm
Page 725
B. Tljsten
Table 2. Typical material properties for epoxy adhesives

Property (at 20C)
Cold-curing epoxy adhesive
Density, (kg/m3)
Youngs modulus, (GPa)
Shear modulus, (GPa)
Poissons ratio
Tensile strength, (MPa)
Shear strength, (MPa)
Compressive strength, (MPa)
Tensile strain at break, (%)
Approximate fracture energy, (Jm2)
Coefficient of thermal expansion, (106/C)
Water absorption: 7 days 25C, (% w/w)
Glass transition temperature, (C)
11001700
0.520
0.28
0.30.4
930
1030
55110
0.55
2001000
2535
0.13
4580
Hot-curing epoxy adhesive

ca 1500
1.04
0.30.35
4060
2030
60120
0.53
4045
<0.2
>180
(a)
(b)
Figure 1. The importance to wet the surface to be bonded
bonding as a method of fastening is that the surfaces

need to be clean and whatever their chemical nature,
coherent if a satisfactory degree of contact, and
therefore adhesion, is to be obtained. This is explained
schematically in Figure 1. If the surface tension value of
the liquid is greater than the surface-free energy value of
the substrate the liquid molecules stay bound together
(a); conversely, when the surface free energy value of
the substrate is higher than that of the liquid it allows the
liquid to uniformly wet the surface (b). The purpose of
surface preparation is to remove contamination and
weak surface layers, to change the substrate surface
geometry, and/or introduce new chemical groups to
provide, at least in the case of metals, an oxide layer
more receptive to the adhesive. Different materials need
different surface treatment, for example when bonding
to concrete the surface is often sandblasted, properly
cleaned and often a primer for the bond system is used
to enhance the bond.
In addition, the design of the bond joint may dictate
the adhesive to use. Depending on the joint design
varying forces will be applied to the bond, i.e.: peel
Tension
Compression
Shear
Cleavage
Peel
Figure 2. Different forces acting on a bonded joint
versus shear, cleavage and compression, as shown

schematically in Figure 2. Also bond thickness and bond
area will affect the ability to transfer the load. A bonded
joint is preferably loaded in shear. It is also important
when choosing the right adhesive to investigate the
environment where the adhesive is going to be used. The
ability of an adhesive to maintain adhesion while exposed
to harsh environments varies as does their chemical
resistance. Consequently; to obtain a good bond between
the adherents the surface treatment is essential.
725
4/12/06
5:11 pm
Page 726
5. SURFACE TREATMENT
To achieve a satisfactory bond between the adherents,
it is usually necessary to carry out some form of surface
treatment. Many treatments are available, ranging from
a simple solvent wipe to the use of series of complex
chemical processes, (Brewis 2005). The method chosen
depends on the nature of the substrate, the conditions to
which the adhesive joint will be subjected, safety,
environmental factors and cost. Treatments may be
divided into physical and chemical methods. The
former includes solvent degreasing and grit blasting.
Physical treatments may remove cohesively weak
layers from a substrate and they may also modify
topography. Chemical treatments, which include the
flame treatment of plastics and anodising procedures
for metals, by definition cause chemical modification to
the surface involved.
It is relatively easy to achieve high initial joint
strengths. However, a satisfactory performance in
service will require careful selection of a pre-treatment.
Water in particular can cause a serious loss of joint
strength in service and pre-treatments vary markedly in
their ability to provide the necessary durability of the
joints involved, (Brewis 2005).
Primers, usually in the form of thin organic coatings,
are often used as an addition or alternative to pretreatments. Primers can provide several advantages. They
usually have much lower viscosities than adhesives and
can therefore achieve greater contact with the substrate.
They can have greater interaction with the substrate and
adhesive. They can contain corrosion inhibitors; this can
be important for metals. Finally, they can protect a surface
until the bonding process is carried out.
To pre-treat metals physical and chemical primers,
often in combination, are used. Etching of aluminium
with chromic acid CAE has been found to give
enhanced performance compared to physical methods.
(Critchlow and Brewis 1996). Much research has also
been carried out on the pre-treatment of titanium alloys.
Considerable success was achieved with an alkaline
peroxide etch. For steel grit blasting, sometimes in
conjunction with a silane, is often used to enhance the
bond properties
To achieve high joint strengths for inorganic
materials, such as glasses, ceramics and concrete, the
outermost layer often needs to be removed. For glass
this weak layer is represented by organic grease and for
concrete by the cement laitance. For glass a solvent
could be used whereas for concrete the weak surface
layer is removed by grit blasting.
Plastics and elastomers have much lower surface
energies than for example glass or metals. The
interaction between polymers and adhesives is therefore
726
generally much lower than between metals and

adhesives (Brewis 2005). To achieve satisfactory
bonding PTFE, polyethylene, polypropylene and similar
plastics it is usually necessary to chemically modify the
surfaces of these polymers, i.e. introduce chemical
groups that can interact relatively strongly with the
adhesives concerned. Such treatments may include
flame treatment, corona treatment, low-pressure plasma
treatment, just to mention a few.
In the next section in the paper a presentation of
adhesives in the aerospace, automotive and marine
industries is provided.
6. AEROSPACE
The earliest structural adhesive applications were made
during the First World War for bonding the wooden
frames of Mosquito aircraft. The strength was adequate
but, by todays standards, moisture resistance was poor.
The motivation in the aerospace industry to replace
mechanical fasteners with adhesives stems from the
desire to prolong aircraft life and to reduce costly
maintenance. Since aerospace structures need to be
reasonably light the use of adhesives reduces the weight
of the structure at same time improving the stiffness and
strengths of connections. In addition, the use of
adhesives prevents corrosion processes when different
materials have to be combined (galvanic corrosion).
Finally, due to uniform, plane load transfer through the
adhesive, layer notch sensitivity is reduced.
There are two basic classes of adhesive bonding
in aerospace structures. One is structural bonding,
with epoxy, phenolic, or acrylic adhesives, that transfers
loads between members. The other is sealants, to protect
against corrosion at interfaces. The stiffness of these
classes of polymers differs greatly, but the two basic needs
are remarkably similar. The first is that the adhesive or
sealant will stay stuck for the life of the structure, in all
service and storage environments, while the second is that
the adhesive will not fail even when the surrounding
structure has been broken, (Hart-Smith 2005).
The use of bonding also provides high potential for
variation in styling due to the possibility of combining
different materials. In modern military aircrafts, as
much as 50% of the airframe may be carbon fibre
reinforced polymers (CFRP), with adhesives being used
for primary structural bonding. An example, the
Swedish military aircraft JAS Gripen, is shown in
Figure 3. A large part of this military plane is built of
composite materials. To fasten the different composite
parts together adhesive has been used. In this particular
case, epoxy as well as cyan ester has been used. It can
also be seen in Figure 4 that the epoxies used for
aerospace applications are considerably stiff.
4/12/06
5:11 pm
Page 727
B. Tljsten
Carbon-fiber composite (CFRP)

Glass-fiber composite (GFRP)
Aramid-fiber composite (AFRP)
Bonding
Bonding
Bonding
Bonding
Bonding
Figure 3. The Swedish Military Aircraft Jas Gripen, (a) Composite Parts, (b) Bonded Parts (SICOMP Workshop, 2004)
Epoxy
Toughened epoxy
Aerospace
Shear strength (MPa)
40
Formula 1
30
Toughened acrylic
Weld bonding for cars
20
Hang-ons for cars

Stiff polyurethane
10
Flexible polyurethane
Plastics for cars
Windscreen
Large body shells

(buses, driver cabs)
25 50 75 100
200
300
400
500
Strain (%)
Figure 4. Shear versus strain for different bonded parts in the automotive industry
7. AUTOMOTIVE
Adhesives have been employed in the automotive
industry since its beginnings, with the use of natural
resins to bond wood and fabric bodies (Mays and
Hutchinson 1992). Recent developments in synthetic
resin technology have resulted in a very wide range of
adhesive materials being available to the design
engineer. In Figure 4, a shear strength versus strain
diagram for various adhesive types used in the
automotive industry is shown. From this diagram it is
clear that bonding is widely used in the automotive
industry both for structural- and non-structural parts.
Adhesives in an automobile body are used to avoid
corrosion, to achieve higher body stiffness and to
enhance resistance to fatigue and crashes. To avoid
corrosion in flanges and hem flanges adhesives have to
fill the gaps to avoid the penetration of water. If there is
no need to enhance the stiffness of the crash resistance

of the parts, plastisols based on a PVC or an acrylic
chemistry are used. To enhance the stiffness of attached
parts and outer panels and to reduce vibration and noise
of these parts so called anti-flutter adhesives are used.
They have to have good adhesion on oily steel sheets
and be cured in the paint oven. Mainly rubber-based
adhesives that are sometimes modified by epoxies are
used, (Dilger 2005).
The drivers for the automotive industry are
lightweight structures, use of mixed materials, long term
performance, crash performance and also styling and
design. Since the adhesive can improve the stiffness and
strength of a joint the weight can be reduced. Adhesive
bonding can furthermore allow the realisation of
combining different structural materials such as FRP,
metals, glasses and ceramics. In Figure 5 an example of
727
4/12/06
5:11 pm
Page 728
Stahl/steel
Aluminium
Magnesium
Kunststoff/plastics
Figure 5. Different adhesively bonded automotive parts for the Mercedes CL-Coupe
different bonded parts and materials for the Mercedes

CL-coupe is shown.
The structural stiffness of a body can be enhanced by
15 to 30% depending on body constructions. Especially
regarding aluminium, a higher fatigue strength compared
to spot- or MIG-welding, clinching and even laser welding
can be achieved, (Dilger 2005). A new generation of
epoxy-based adhesives with high energy absorption at
high velocities is used in car concepts to improve the
crash behaviour of the body.
It is quite clear that many parts of different materials
have to be brought together through bonding, sometimes
together with rivets. In many situations this is preferred
to welding. Similar to aerospace bonding, galvanic
processes can be delayed or prevented when adhesives
are used, in particular when different materials are
joined together. Improvement of crash performance is
possible by the use of substrates and adhesives with a
high potential of energy absorption. Finally, diversity of
styling and design are possible due to the possibility of
combining different materials and components and
joining them together by bonding. Adhesives for
automotive and industrial bonding are modified
acrylics/methacrylates which provide high strength and
elongation properties and also bond to thermoplastics.
Other adhesives are polyurethanes, which are tough and
728
have a high abrasion resistance and good adhesion at

low temperatures. Silicones are also used for bonding to
glass, plastics or other rubbers.
8. MARINE AND OFF-SHORE
The use of adhesives for marine and off-shore
applications need to fulfil some important requirements;
they must be moisture, heat and UV resistance. Tolerate
variations in joint thickness, and be easy to apply to large
structures. Casein and then formaldehyde resin
compositions have been used as adhesives and gap fillers
in wooden boat construction for many years (Mays and
Hutchinson 1992). Many GFRP-hulled boats, both naval
and civilian, now rely significantly on resins for
laminating, stiffening, the fabrication of sandwich panels,
and for bonding attachments. Traditionally the
shipbuilding industry uses welding as the primary process
for joining the different structural parts in a ship. It is well
known however that this process results in induced
stresses during the fabrication stage which in turn lead to
distortion in the shape of structural components and
indeed, the ship itself. Considerable effort has thus to be
expended to mitigate these weld-induced distortions
leading to increased production costs. One additional
problem, in the case of aluminium, is the significant
reduction in the fatigue load capacity in welded structures
4/12/06
5:11 pm
Page 729
B. Tljsten
(Kecsmar 2003). As a consequence, either the structural

topology has to be designed to cope with increased stress
levels or the scantlings of the structure have to be
enhanced. In either case, there is an increase in the weight
of the structure. Since structural weight needs to be
minimised, especially in high speed, high performance
ships, there is a need to investigate alternative joining
techniques for aluminium structures; adhesive bonding is
one of them (Clark 1996; Cantrill 2004).
Adhesive bonding offers the opportunity to replace
welding of steel structures, to reduce distortion,
effectively eliminate residual stress and to improve
fatigue performance when compared to welded
connections. Avoidance of hot-work leads to safer
construction practices in hazardous environments.
Adhesive bonding of composites provides well-distributed
loading and maximises the utilisation of the adherent
materials. Typical adhesives in the marine industry are
polyesters which are less expensive than epoxy and are
widely used in other industrial applications. Polyester,
however, is chemically weaker than epoxies and
experiences a high degree of shrinkage. Vinyl ester, which
provides higher strength, modulus of elasticity, and
elongation than polyesters is still less expensive than
epoxy and often preferable to polyesters. For bonding to
metallic parts epoxies are typically used.
9. BUILDING AND CIVIL ENGINEERING
9.1. General
A wide range of adhesives have been used for many
decades in the construction industry. Historically, many of
the adhesives used were based on natural products and
were highly effective as adhesives in essentially nonstructural applications (Shaw 1990). It has been shown
that structural adhesive bonding has been employed in the
aerospace and other industries for many years and for
many purposes, whereas its use in civil engineering is
relatively recent. However, the application of adhesive
bonded structures has increased dramatically over the past
few years. After the successes achieved in the aviation,
automotive and transport industries, the advantages and
opportunities of structural adhesive bonding are now
acknowledged in the building and construction industry.
The applications include adhesive bonded reinforcements,
sandwich panels, adhesive bonded anchors, composite
aluminium profiles and glass constructions The use of
adhesives in the construction industry may be divided into
a) repair and strengthening; and b) new-built; both these
areas are discussed in more detail below.
9.2. Repair and Strengthening
Repair and strengthening of existing structures using
bonding techniques is mainly associated with concrete
structures. During the 1950s and 1960s an enormous

amount of new constructions were built and, as these
structures age, many faults have become evident. In
addition, the deterioration of concrete structures due to
many different factors has decreased useable remaining
life. A variety of concrete repair methods have been
developed many of which rely on adhesion between a
repair material and the concrete substrate (Mays and
Hutchinson, 1992). When repairing or strengthening an
existing concrete structure the bond between the old
structure and the new substrate is essential for proper
force transfer; this is often dependent on the substrate.
There are many different methods to improve the bond,
but they all have in common that at least the weak
laitance layer is removed. This is preferably carried out
by water or sandblasting. The repair substrate may be
ordinary concrete or polymer concrete, with or without
reinforcement. The concrete may be put in place by
traditional formworks or by spraying. If increased
structural capacity is needed then external plate
bonding can be an alternative. During the 1970s and
up to the end of the 1990s, steel plate bonding was not
unusual, however in the last decade, the use of
advanced composites for external strengthening has
become quite common (see for example Tljsten 2004;
Teng et al. 2003). Also a great number of models for
debonding have been presented (including Vilnay
1988; Roberts 1989; Tljsten 1994, 1997; Chen and
Teng 2001; Gunes 2005).
In the UK a very large number of cast-iron-framed,
brick jack-arch structures were built as public buildings,
textile mills and warehouses. Many of these still stand,
and careful refurbishment allows these structures to be
adapted for alternative modern uses (Cadei et al. 2004).
In addition to this, in the UK 1015% of all bridges are
believed to be metallic and many of these need to be
strengthened. A similar situation also exists in the US
where 52000 structurally deficient steel bridges exist.
Many of these structures can be strengthened by
externally bonded CFRP systems. However, the demand
for strengthening metallic structures with external plate
bonding is not as great as for concrete structures.
Timber structures sometimes also need to be repaired
or strengthened. Strengthening of timber beams by means
of flexural reinforcement of glass or carbon fibres is a
system that enhances wooden beams in several ways.
First of all, a flexural capacity increase of 4060% may
be achieved (Buell and Saadatmanesh 2005; Svecova and
Eden 2004). Furthermore, a stiffness increase of 525%
may also be achieved (Buell and Saadatmanesh 2005).
The most important performance increase relates to the
change of failure mode from a brittle failure on the
tension side of the beam to a ductile failure in
729
4/12/06
5:11 pm
Page 730
compression inflicting crushing of wood. In terms of

increased safety of the structure, the change in failure
mode is important. For in situ upgrading of bridges, for
example, simple and rational installation of the
reinforcement is needed. In the case of an existing
structure it may not be possible to install reinforcement
bars or sheets along the entire length of a beam; therefore,
the establishment of the anchoring length for the
reinforcement is crucial in order to give recommendations
on installation (Johansson and Stenberg 2004). Shear
reinforcement of timber beams (stirrups) might be
necessary to prevent the occurrence of plug shear and
splitting. Such reinforcement should be placed at
locations of maximum moment, not at locations of
maximum shear force. For timber members with a low
span-to-depth ratio, reinforcement to sustain the shear
force at the support might be needed (Svecova and Eden
2004). The increase in bending stiffness is also tempting
to use for members loaded both in bending and
compression e.g. in large bridges or in rafters. Even
heavily loaded columns may be an application. Work is
currently underway in this area. Apart from the possible
occurrence of brittle failure modes, the most pressing
question for future research is the long term reliability and
strength of a reinforcement system. However small, the
reinforcement will always induce stresses in the
surrounding wood and the impact of these stresses is not
known when it comes to shrinking and swelling of wood,
creep and durability of materials.
Reinforced concrete structures are designed so that
the inevitable cracking of the concrete is restricted so
that no cracks at the surface should exceed 0.1 mm in
width. For many reasons, cracking in excess of
acceptable design limits occurs rather too often either
during construction or during the service life of the
structure. If cracks are not sealed or structurally bonded
deterioration may occur. It is possible to restore the
structure to its original un-cracked strength by injection
with low viscosity epoxy resins specially developed for
repairing cracks, providing the bonding surfaces of the
concrete at the crack interface are clean and sound.
There are of course many other situations related to
repair and strengthening which may require a bonded
joint, for example, different types of concrete overlays,
although the principle remains the same the weakest
part fails. In concrete structures this is normally the old
concrete substrate, in metallic structures the weak link is
normally in the adhesive and for timber structures, in the
case of strengthening a combined failure is likely. The
failure also depends on the type of strengthening and
an over-strengthened structure may lead to a changed
failure mode; i.e. over-strengthening a structure in
flexure, may result in a shear or compressive failure.
730
9.3. New-Built
If we discount glulam beams, structural bonding is not
very common in newly-built structures. Adhesive is often
used for filling voids or gaps and to fasten secondary,
non-load carrying elements. Adhesives are also used to
fasten bridge bearing and expansion joint nosing. On steel
decks, skid resistant surfacing materials in the form of
small gravel-filled epoxy or polyester resins are often
used. There are also examples where epoxy resin systems
have been used to bond precast concrete slab units
directly to a steel girder surface for a steel-concrete
composite bridge, (German Bonded Bridge 1979).
Applications of resins as true adhesives in new
constructions have been relatively scarce although
increasing use is being made of epoxies in the joints
between segmental precast, prestressed bridge construction
where the adhesive is being employed as a stress
distributing bonding/sealing layer between the precast
elements. This concept was pioneered in major structures
in France in the construction of the Choisy-le Roi bridge
over the River Seine near Paris in 1962 (Muller 1964).
This bridge was made up of precast concrete box units 2.5
m long. The units were match cast off-site in
continuous lengths; one unit being cast against the next
with a thin plastic film separator to ensure perfect fit and
that the thickness of the epoxy layer was kept to a
minimum. In comparison to mechanical or welded joints,
however, structural bonded joints are few. This might
change, however, with increased use of newly-built FRP
structures for building and civil purposes. By way of
example, a brief description of a FRP bridge structure
built in the UK 2002 having all structural joints made by
epoxy is presented. The ASSET (Advanced Structural
System for Tomorrows Infrastructure) bridge project,
founded by the European Commission, began in 1998
and was completed in the autumn of 2002. The project
mainly covered design, manufacture and prototype
construction of a fully polymer composite road bridge. A
European consortium designed a deck profile. The deck is
pultruded from a combination of roving, woven fabric
and mat. The deck was oriented transverse to the direction
of the traffic bonded together with epoxy. The deck
subassemblies, consisting of 5 or 6 pultruded profiles
bonded together off-site indoors, are then bonded to the
longitudinal composite girders. The mid-span crosssection of the bridge is shown in Figure 6. Below the deck
is a composite support structure consisting of four 300
mm, 11.0m long pultruded glass/polyester box beams
which are spaced 2 m apart and attached to the concrete
abutments. The box beams incorporated unidirectional
carbon fibre on the top and bottom for additional flexural
stiffness. The 10.0m long and 6.8m wide deck structure is
bonded to the box beams with epoxy. This prototype
4/12/06
5:11 pm
Page 731
B. Tljsten
bridge replaced an existing bridge, the West Mill bridge,

in Oxfordshire. The project is described in greater detail
in (Luke et al. 2002). In Figure 7 the bridge is shown
when lifted in position. The material properties of the
FRP materials and adhesives used in the ASSET project
is given in Table 3. A low viscosity adhesive was used in
combination with vacuum infusion to bond the carbon
fibre fabrics to the rectangular girder profiles and a
tixotropic adhesive to bond the deck profiles together and
the main deck to the girders. The average adhesive
thickness between the ASSET profiles was 2.0 mm and
between the deck and the girders approximately 3.0 mm.
The thickness was controlled by distance holders. At site,
in a site factory, the subassemblies were bonded together
to form the bridge deck. Each sub-deck was at the same
time bonded to the composite girders, a photo of the
bonding procedure is shown in Figure 8.
A significant outcome of the ASSET project has been
the ability to develop a complete bridge system, whereby
the advanced composites deck profiles can be connected
to conventional or other advanced composite materials
forming the main longitudinal beams and other relevant
bridge appurtenances by bonding to form a complete
bridge. From the construction work it was concluded that
the bonding procedure chosen was time consuming due
to large amount of manual handling and lack of
experience bonding large structural parts together, but
also that this could be improved in future applications by
a better choice of equipment for handing the composite
deck during bonding. The final installation of the bridge
was easy to carry out. For future applications, it would be
possible to build a bridge (depending on size) off site and
transport and lift it into position at site.
10. CONSIDERATIONS IN STRUCTURAL
BONDING
To obtain a proper bond that can transfer forces over
time in different environments there are many different
factors to consider, in particular when bonding is used in
the construction industry. In the traditional industrialized
business, where the environment to a large extent can be
controlled during the bonding process with regard to
humidity, temperature, cleanness etc. Examples of such
industries are the automotive and aerospace industries,
and one may here expect an increased used of bonded
joints, both for structural and non structural use.
On the contrary, the majority of the assembly
procedures in the construction industry are carried out
outdoors in a non controlled environment, often in
relation to debris, dust, moist or even wet surfaces, at
non controllable temperatures. Also the long time use, a
structure might be designed for a life longer than 100
years, with heavy dead load. This make bonding more
complicated and increases the challenge to increase the

use of structural adhesives in construction. Nevertheless,
the aforementioned examples how structural adhesives
have been and are used in the construction industry show
that it is not only possible but it also bring in new
opportunities. However, some the limitations using
adhesives in the construction industry are:
Often extensive preparation of the surface to be

bonded, time consuming and costly.
Poor short term handleability, often the open

time is less than 1 hour.
Temperature, moisture and cleanness regulations,

both during bonding and long time use. Make it
for example difficult to use outdoor in the
winter time.
Health and safety regulations during handling.

Extra protective clothing is needed.
Uncertainty regarding long term performance

and function in different construction related
environments.
Unfamiliarity of using structural adhesives.

In addition, it might be added that the traditional
conservatism that often is found in this industry. Despite
these hurdles there are still many possibilities for the use of
structural adhesives, especially if it can be shown that it is
time and cost saving, that it solve problems that otherwise
can be difficult to overcome, and if enough good long term
properties can be proven. One such successful example is
plate bonding. There is also a strong desire to industrilise
large part of the building industry that means that the
possibilities to control the environment during bonding will
the same as in the automotive and aerospace industries.
11. FUTURE BONDING POSSIBILITIES
Predicting the future is not an easy task and most of
the predictions will sooner or later be proven wrong.
The author does, in addition to this, not expect any new
revolutionary adhesive systems in the foreseeable
future. Most likely improvements will be made to the
already existing ones. If focus is given the construction
industry an increased use of adhesive is to be expected
both for structural and non structural purposes. The
main reason for this is the ongoing trend towards
increased industrialisation mainly for house building
but also for civil structures such as bridges. This means
that the environment will be controllable and industrial
process for structural bonding will be possible to use.
Consequently the quality of bonded joints will be
improved and a larger acceptance of bonding will be
expected. The drivers for the construction industry are
mainly time and cost. An very important issue will be
the long time durability of the adhesive systems used.
Aerospace and automotive industries will be even more
731
Detail 2
500 long plug in asset

profile at holding down
bolt locations
Steel
L/1.00/V
parapet
(80 km/h)
775
400
upstand
500
verge
T
1:40
6800
1750
4 no. composite
longitudinal box beams
each comprising 20 thk.
CFRP top and bottom
flanges and 4 no. 240 240
GFRP S.H.S.
1:40
Asset profiles
spanning
transversely
30 mm at channel
92.5 mm at C
polymer concrete
2500
1750
Figure 6. Mid-span cross-section of West Mill Bridge (sketch by Mouchel, 2002)
1750
Special HB precast
KERB cut to provide
100 mm upstand and
bonded to deck
2 no. 75 mm DIA steel

ducts
Concrete verge
2500
5000 carriageway
6800
400
upstand
775
6 mm HIM surfacing
500
verge
Proposed 150
gas main
500 long plug in asset

profile at holding down
bold locations
Top_line
Bot_line
Level constant at
any transverse section
through bridge deck
225
5:11 pm
SECT
SEC
732
4/12/06
520
Page 732
4/12/06
5:11 pm
Page 733
B. Tljsten
Table 3. Material properties of FRP and adhesives used for the ASSET project
Product
Supplier
Property
Density, (kg/m3)
Viscosity, (mPa s)
Potlife, (min)
EL-modulus, (GPa)
ET-modulus, (GPa)
Tensile Strength, (MPa)
Compressive strength, (MPa)
Strain at failure, (%)
Profiles
Carbon Fibre
Resin
Adhesive
Fiberline A/S
Saertex
T700
1820
230
4800
2.1
NM AB
NM94/93C
1100
500
56160
3
52
100
2.1
BPE Systems AB
BPE 465/464
1498
Tix
90
7
31
103
27
9
1.2
Figure 7. The ASSET Bridge lifted in position
Figure 8. Bonding procedure Adhesive is applied on a girder
733
4/12/06
5:11 pm
Page 734
(a)
(b)
Figure 9. Images of a flat-tailed house gecko and its seta. Top is a photograph of the gecko on a glass covered mirror used in the study.
At the bottom are scanning electron microscope (SEM) images of gecko setae showing the tree-like structure with
a magnification of (a) 900 times and (b) 8500 times (Sun W. et al., 2005)
focused to bring down the weight of their crafts in the

nearby future the main driver for this is the cost for
fuel. This might increased the use of bonding
especially when different materials are bonded together,
such as metals and composites.
Even though I do not see any really big leaps in
bonding technology in the nearest future, it would be
interesting to consider bonding systems used in the
nature will it for example be possible to transfer some
of these extraordinary adhesive capacities to structural
applications? I will here discuss two different adhesion
forces produced in the nature, the gecko lizard and the
blue mussel (Figure 9).
The extraordinary climbing skills of gecko lizards,
shown in Figure 7, have been under investigation for a
long time. (Sun et al. 2005) has measured the adhesion
force produced by an individual gecko setae. As the
gecko force was influenced by the surface hydrophobicity
as well as the presence of water they concluded that the
dominant component of the adhesion force is the
capillary force. This finding epitomizes one of many
intriguing natural phenomena that can possibly be
adopted to improve the technological know-how of
bond without chemical components. The future will
show if this can be transferred to structural applications.
734
Another very interesting finding regarding bonding

comes also from nature. Researchers at Purdue
University in the USA have discovered that iron in
seawater is the key binding agent in the super-strong
glues of the common blue mussel, Mytilus edulis
(Ninan et al. 2003). The knowledge will be used to
develop safer alternatives for surgical and household
glues, among others. Furthermore the discovery could
lead to the creation of unusual new materials with
designed plasticity, strength and adhesion for structural
uses. Figure 8 shows a common blue mussel hanging
tough after a night adhering to otherwise non-stick
Teflon (Figure 10).
12. SUMMARY AND CONCLUSIONS
In this paper it is shown that structural bonding today is
used in many different industries and appears to be
becoming more and more important in the future. The
aerospace and automotive industries are ahead of other
industries, probably due to the need for weight savings and
the need for joining many different materials together. In
the construction industry mechanical joints are most
common and will probably continue to be so in the future.
However, structural bonding in construction is not
uncommon. It has been frequently used in repair and
4/12/06
5:11 pm
Page 735
B. Tljsten
Figure 10. Blue mussel bonding to Teflon, (Ninan et al., 2003)
strengthening of existing structures by the use of external

plate bonding, sprayed concrete, concrete overlays and so
on. The increase of FRP structures in the building and civil
industry probably will also lead to increased use of
structural bonding. However, if structural bonding is to
become more widely accepted in the construction
industry, it is essential that the long term behaviour of
bonded joints can be guaranteed civil structures are
expected to last 50100 years and other factors make them
very different from products in the automotive and
aerospace industries. Finally, our design culture is based
on codes and standards and until structural bonding is
accepted in our codes the use of structural bonding will
still be limited.
ACKNOWLEDGEMENTS
The research presented in this paper has primarily been
funded by the Development fund of the Swedish
Construction Industry (SBUF). Nils Malmgren AB,
Sweden is also acknowledged.
REFERENCES
Bar-Yosef, O. and Schick, T. (1989). Early neolithic organic remains
from Nahal Hemar Cave, National Geographic Research, Vol. 5,
No. 2, pp. 176190.
Brewis, D. (2005). Surfaces: how to treat, Adhesive Bonding
Science, Technology and Applications, Woodhead Publishing
Limited, Cambridge, England, Adams R. D. Ed., ISBN 1 85573 741
8, pp. 7588.
Cadei, J.M.C., Stratford, T.J., Hollaway, L.C. and Duckett, W.G.
(2004). Strengthening Metallic Structures Using Externally
Bonded Fibre-Reinforced Polymers, CIRIA report, Classic House,

174-180, Old Street, London ECV1 9BP, UK, 234 pp.
Cantrill, J., Kapadia, A. and Pugh, D. (2004). Lessons learnt from
designing and producing adhesively bonded structures in a
shipyard, Proceedings of the Institution of Mechanical Engineers
Part M, Vol. 218, No. 4, pp. 267272.
Chen, J.F. and Teng, J.G. (2001). Anchorage strength models for
FRP and steel plates bonded to concrete, Journal of Structural
Engineering, ASCE, Vol. 127, No. 7, pp. 784791.
Clarke, J.L. (1996). Structural design of polymer composites,
Eurocomp Design Code and handbook, E&FN Spon.
Comyn, J. (2005). What are adhesives and sealants and how do they
work?, Adhesive Bonding Science, Technology and Applications,
Woodhead Publishing Limited, Cambridge, England, Adams R.
D. Ed. , ISBN 1 85573 741 8, pp. 2351.
Critschlow, G.W. and Brewis, D.M. (1996). Review of surface pretreatment for aluminium alloys, International Journal of
Adhesion and Adhesives, No. 16, pp 255275.
Dilger, K. (2005). Automobiles, Adhesive Bonding Science,
Technology and Applications, Woodhead Publishing Limited,
Cambridge, England, Adams R. D. Ed., ISBN 1 85573 741 8,
pp. 357385.
Fay, P. A. (2005). History of adhesive bonding, Adhesive Bonding
Science, Technology and Applications, Woodhead Publishing
Limited, Cambridge, England, Adams R. D. Ed., ISBN 1 85573
741 8, pp. 322,
Gaul, R.W. and Apton, A.J. (1959). Epoxy adhesives in concrete
construction, Civil Engineering (New York), Vol. 29, No. 11,
pp. 5052.
German Bonded Bridge (1979). Construction Industry International,
Vol. 5, No. 4, April 1979, pp. 3137.
Gunes, O. (2005). Debonding problems in beams strengthened using
externally bonded FRP reinforcement, COBRAE Conference
Proceedings, April, part 3.
Hart-Smith, L.J. (2005). Aerospace, Adhesive Bonding Science,
Technology and Applications, Woodhead Publishing Limited,
Cambridge, England, Adams R. D. Ed., ISBN 1 85573 741 8,
pp. 489527.
Johansson, T. and Stenberg, J. (2004). Kolfiberkompositfrstrkta
Limtrbalkar Frankring Och Momentkapacitet, Master Thesis,
Lule University of Technology, Department of Civil and
Environmental Engineering, CIV ISSN:1402-1617, 100 pp. (in
Swedish).
Kecsmar, J. (2003). Structural Design of Aluminium Ship Structures,
MPhil thesis, University of Southampton.
Luke, S., Canning, L., Collins, S., Knudsen, E., Brown, P., Tljsten, B.
and Olofsson, I. (2002). Advanced composite bridge decking
system - Project ASSET, Structural Engineering International,
Vol. 12, No. 2, pp. 7679.
Mays, G.C. and Hutchinson, A.R. (1992). Adhesives in Civil
Engineering, Cambridge University Press, ISBN 0 521 32 677 X
Hardback, 333 pp.
735
4/12/06
5:11 pm
Page 736
The Importance of Bonding A Historic Overview and Future Possibilities
Muller, J.N. (1964). Concrete Bridges Built in Cantilever, Ann Inst

Tech Trav Publis, Supple 204.
Newton, I. (1717). Opticks: or, A Treatise of the Reflections,
Refractions, Inflections and Colours of Light, The second
edition, with additions, printed by W. Bowyer for W. Innys at the
Princes Arms in St. Pauls Churchyard, London, 369 pp.
Ninan, L., Monahan, J., Stroshine, R.L., Wilkner, J.J. and Shi, R.
(2003). Adhesive strength of marine mussel extracts on porcine
skin, Biomaterials, Vol. 24, No. 22, pp. 40914099.
Roberts, T.M. (1989). Approximate analysis of shear and
normal stress concentrations in the adhesive layer of plated
RC-beams, The Structural Engineer, Vol. 67. No. 12, pp.
229233.
Shaw, J.D.N. (1990). Adhesives in the construction industry:
materials and case histories, Construction and Building
Materials, Vol. 4, No. 2, pp. 9297.
Sun, W., Neuzil, P, Kustandi, T.S., Oh, S. and Samper, D. (2005).
The nature of the gecko lizard adhesive force, Biophyical
Journal, Vol. 89, pp. L14L17.
SICOMP Workshop (2004). Dimensioneringsfilosofier fr
kompositmaterial inom olika branscher, Februari, olika
presentationer (in Swedish).
736
Teng J.G., Smith S.T., Yao J. and Chen J.F. (2003). Intermediate
crack-induced debonding in RC beams and slabs, Construction
and Building Materials, Vol. 17, No. 6, pp 447462.
Tljsten, B. (1994). Plate Bonding: Strengthening of Existing Concrete
Structures with Epoxy Bonded Plates of Steel or Fibre Reinforced
Plastics, Doctoral thesis, Lule university of technology, Sweden.
Tljsten, B. (1997). Strengthening of beams by plate bonding,
Journal of Materials in Civil Engineering, ASCE, Vol. 9, No. 4,
pp 206212.
Tljsten, B. (2004). FRP Strengthening of Existing Concrete
Structures Design Guideline, ISBN 91-89580-03-6, Lule
University of Technology, Third ed., 228 pp.
Tremper, B. (1960) Repair of damaged concrete with epoxy resins,
Journal of the ACI, Vol. 57, pp 173182.
Vilnay, O. (1988). The analysis of reinforced concrete beams
strengthened by epoxy bonded steel plates, The International
Journal of Cement Composites and Lightweight Concrete, Vol.
10, No. 2, pp. 7378.
Wakeman, C.M., Stover, H.E. and Blye, E.N. (1962). Glue for
concrete repair, A.S.T.M, Bulletin, Vol. 2, No. 2, pp. 9397.
Walker, A.A. (1998). Oldest glue discovered, Archaeology Online
News, May 21, (http://www.archaeology.org).
ASE 9-6-02_Oehlers
4/12/06
5:10 pm
Page 737
FRP Plates Adhesively Bonded to Reinforced Concrete

Beams: Generic Debonding Mechanisms
D. J. Oehlers*
School of Civil and Environmental Engineering, The University of Adelaide, Adelaide, SA5005, Australia
Abstract: Retrofitting reinforced concrete members by adhesively bonding fibre

reinforced polymer plates to their surfaces is now a well established technique as
there has been substantial and much good research on the numerous individual
plate debonding mechanisms which has allowed safe designs. In this paper, it is
shown how the apparently numerous and individual debonding mechanisms can be
categorised into a few distinct generic debonding mechanisms, which should help to
simplify the problem and consequently, in the long term, allow the development of
generic design rules.
Key words: FRP, RC beams, debonding, plate debonding, debonding mechanisms.
1. INTRODUCTION
The main aim of this paper is to identify the generic
plate debonding mechanisms. Reinforced concrete (RC)
beams and slabs are subject to fundamental flexural and
shear deformations that are peculiar to reinforced
concrete members and which are generally well
understood by RC designers and researchers. It is these
fundamental reinforced concrete deformations that
induce fundamental internal forces which cause FRP
plates attached to the beam to debond. Hence, the first
step in understanding and identifying the generic and
fundamental debonding mechanisms is to understand
the fundamental behaviour of RC beams which is
explained for flexural, shear and axial deformations.
These generic reinforced concrete deformations, that is
flexural, shear and axial, control the plate debonding
mechanism. However, the capacity to resist the plate
debonding mechanism is controlled by the generic
material properties which are covered next. The generic
plate debonding mechanisms are then described in
terms of the reinforced concrete flexural, shear and axial
deformations.
2. GENERIC FLEXURAL
DEFORMATION MECHANISMS
A standard cross-sectional flexural analysis of an RC
beam is shown in Figure 1 (Oehlers and Seracino 2004).
For unplated RC beams with ductile reinforcing bars,
failure is always caused by concrete crushing at a strain
c. In contrast, a plated beam can also fail by plate
debonding at a strain db or plate fracture at a strain fr.
Hence, in plated beams three material strain failure
have to be considered, as shown in Figure 1(b), instead of
the one strain failure of c in unplated reinforced concrete
beams. In the analysis depicted in Figure 1, the strain
profile is then pivoted about the failure strain as shown in
Figure 1(c) where in this example plate debonding is
assumed to cause failure. The curvature of the strain
profile is then varied until from the stress distribution in
Figure 1(d), the force distribution in Figure 1(e) is in
longitudinal equilibrium. It is worth noting, that if plate
debonding strain or plate fracture strain is found to
precede the concrete crushing strain then the pivotal
strain must be changed and allowance must be made for
the concrete to remain at least partly elastic as shown in
*Corresponding author. Email address: doehlers@civeng.adelaide.edu.au.
737
ASE 9-6-02_Oehlers
4/12/06
5:10 pm
Page 738
FRP Plates Adhesively Bonded to Reinforced Concrete Beams: Generic Debonding Mechanisms
(pivotal strains)
(stress)
(strain)
F (force)
F1 (y1)
F2 (y2)
kud
Plate stress
resultants
Tension
face plate
db
(a)
frac
(b)
F3 (y3)
F4 (y4)
db
Mplate = 0
Pplate = F4
Pivotal point
(c)
(d)
(e)
(f)
Figure 1. Generic flexural analysis of tension face plated beam
(stress)
(strain)
F (force)
Plate stress
resultants
F1
F2
Plastic zone
Elastic zone
Mplate
Neutral axis
Plate centroid
bar
F5 F4
Pplate
db
(b)
(a)
F3
Pivotal point
(c)
(d)
(e)
Figure 2. Flexural analysis of side plated beam
Point of
contraflexure
Position of
maximum moment
Vplate
Side plate
Point of
contraflexure
Continuous T-beam
Pplate
Pplate
Figure 3. Generic stress resultants
Figure 1(d). It is also worth noting that the stresses in the

plate induce the fundamental stress resultants in the plate
in Figure 1(f), that is the axial force Pplate and a moment
Mplate. For tension face plates as in Figure 1(a), the plate
moment Mplate may be quite small.
A clearer indication of the development of the plate
moment is shown in Figure 2 for a side plated beam
which also represents a tension face near surface
mounted (NSM) plate as this NSM plate will also be bent
about its major axis. It can now be seen that the nonuniform stress distribution in the plate in Figure 2(c)
induces a moment about the plate centroid as well as the
axial force through the plate centroid as in Figure 2(e).
The plate moment Mplate in Figure 2 is resisted by
transverse forces Vplate shown in Figure 3. Therefore,
the adhesive bond between the plate and the RC beam
must be designed to resist both longitudinal forces Pplate
as well as transverse forces Vplate induced by curvature
as these are generic stress resultants.
738
3. GENERIC SHEAR
DEFORMATION MECHANISMS
The behaviour of RC beams under shear is much more
complex than that under flexure (Oehlers and Seracino
2004). What is known is that the forces in the shear
reinforcement, that is stirrups, are minimal and can be
ignored until a critical diagonal crack (CDC) forms. An
example of a CDC in a beam without stirrups is shown
in Figure 4, and it is this failure of a beam without
stirrups which is directly associated with the concrete
component of the shear capacity of the beam Vc.
The presence of stirrups tends to hide the importance
of the initial CDC as shown in Figure 5 where a series
of diagonal cracks can be seen. However, it should be
remembered that the formation of the first or initial
CDC in a beam with stirrups occurs at the same shear
load Vc as that of a beam without stirrups. Furthermore,
the formation of the first CDC often causes debonding
of longitudinal plates in beams with or without stirrups.
ASE 9-6-02_Oehlers
4/12/06
5:10 pm
Page 739
D. J. Oehlers
Applied load
Edge of
applied load
Critical diagonal crack

Ripping along bars
Flexural-shear cracks
Support
Figure 4. CDC in a beam without stirrups

Concentrated load
Edge of applied load

O
Diagonal
crack focal
point
A
B
Critical diagonal cracks
Support
Figure 5. CDC in a beam with stirrups
The rigid body shear deformation across the CDC then

induces axial forces in the reinforcement spanning the
CDC, this reinforcement could be in the form of
longitudinal reinforcing bars, longitudinal plates,
transverse stirrups or transverse plates. Critical diagonal
cracks should not be confused with the flexural/shear
cracks, shown in Figure 4, where rigid body displacement
does not occur. Critical diagonal cracks are associated
with rigid body displacements across the crack, are
directly associated with the concrete component of the
shear capacity of the beam Vc, and are required to stretch
the internal stirrups to induce their contribution Vs.
Existing procedures for quantifying the shear
capacity of RC beams simplify this extremely complex
problem of shear by assuming that the longitudinal
reinforcing bars contribute indirectly to the concrete
component of the shear capacity Vc and that the
transverse bars or stirrups contribute directly to the
shear capacity Vs. Following this simplistic but
convenient approach, longitudinal FRP plates can be
assumed to indirectly increase Vc, as shown in Figure 6,
by holding the two faces of the CDC together and,
thereby, through the axial forces in the longitudinal
plates, increase the shear resisted by Vc as in Eqn 1.
Furthermore, transverse plates can be considered to act
in the same way as the internal steel stirrups (Vs), that is
these external stirrups directly resist the vertical shear
through their axial forces Vtp = Pplate as shown in
Figure 7 where Pplate is the theoretical maximum axial
force in a transverse plate that crosses a critical diagonal
crack and which could be the debonding resistance or
fracture strength.
Reinforcing
bar area s
Prestressing
tendon
Vc + Vc
(Pplate)tfp
Longitudial tension face

Critical diagonal
in hogging
(Pplate)sp
V
RC beam
Longitudinal side
Figure 6. Increase in Vc due to longitudinal plates
Vcap = Vc + Vc + ksVs + ktpVtp

This already complex problem of shear is further
complicated by the fact that the external FRP plates are
made of a brittle material and, furthermore, the adhesive
bond between the plate and the concrete is also known
to behave in a brittle fashion. Hence the transverse FRP
plates in Figure 7 may debond before the maximum
axial force in each transverse plate Pplate is achieved
(Teng et al. 2002); in which case the interaction factor
ktp in Eqn 1 would be less than unity. Furthermore in
contrast to the FRP plates which have a brittle
behaviour, the internal steel stirrups have a ductile
behaviour as they can yield and a ductile bond because
they are fully anchored at their ends. Hence the
transverse plates may debond before all the steel stirrups
crossing a critical diagonal crack have yielded such that
ks in Eqn 1 is less than unity.
An example of the variation of the interaction factors
k in Eqn 1 is illustrated in Figure 8 for a beam with
transverse near surface mounted (NSM) plates
(Mohammed Ali et al. 2006). The CDC width w in
Figure 7 is gradually widened and this width is given in
739
ASE 9-6-02_Oehlers
4/12/06
5:10 pm
Page 740
Side plate
Critical diagonal shear

w
Steel
stirrup
U-jacket
(PAIC)1
(PAIC)3
Vs-2
Transverse
plate
Vs-1(PAIC)2
Figure 7. Shear resisted by transverse plates
Interaction factor (k)
1.1
1
0.9
Stirrups
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Direction of IC interface
crack propagation
Intermediate crack
63%
55%
Total
ki
ke
kt
IC interface crack
NSM plates
PIC (IC)
0
0.5
1
1.5
2
2.5
3
Uniform crack width (mm)
3.5
Figure 9. Intermediate crack debonding mechanism
Figure 8. Interaction factors for wet lay-up U-jackets
the abscissa in Figure 8. As the width of the CDC is

increased, the transverse force or shear force is given
in the ordinate in Figure 8 where: ke is the force in the
transverse plates crossing the CDC as a proportion of
the sum of their individual peak strengths Pplate that is
Vtp in Eqn 1; ki is the force in the transverse steel
stirrups crossing the CDC as a proportion of the sum of
their individual yield strengths Vs; and kt is the total
shear force as a proportion of Vtp + Vs. As the CDC
width increases in Figure 8, the peak force in the FRP
plates reaches 55% of its maximum value and then
decreases gradually until the stirrups yield when the
shear force is 63% of its theoretical maximum value.
It can be seen that the peak plate force does not coincide
with the peak stirrup force.
4. GENERIC PLATE INTERFACE
DEBONDING MATERIAL FAILURE
Fundamental to the whole concept of plate debonding
are the bond characteristics at the interface between the
plate and the concrete (Oehlers et al. 2006). The
interface is subjected to both interface shear stresses and
stresses normal to the interface due to the fundamental
deformations of RC beams that has been described
previously. It will be assumed in this section that the
adhesive is strong enough to ensure failure within the
concrete. This is usually the case as the adhesive and its
740
B
P
Slip
Figure 10. Pull-test (show slip distribution)
bond are normally several times stronger than the tensile

strength of the concrete.
4.1. Aggregate Interlock Resistance
When a crack intercepts a plate as in Figure 9, infinite
strains need to be induced in the plate to accommodate
the crack which, of course, cannot occur. Hence the
stress concentrations induced by the crack intercepting
the plate cause the IC interface cracking in the concrete
adjacent to the interface. Shear can still be transferred
across this crack through aggregate interlock which
requires localised forces or stresses normal to the
interface. This aggregate interlock debonding material
behaviour is often determined directly from pull-tests as
shown in Figure 10 where the front of the block is
equivalent to a crack face and where the material shear
() slip () characteristics are either determined directly
or indirectly. The interface IC debonding / material
characteristics are of profound importance to the
ASE 9-6-02_Oehlers
4/12/06
5:10 pm
Page 741
D. J. Oehlers
Elastic
Micro-cracking
required for the aggregate interlock resistance cannot be

generated. Hence the / interface characteristics in
Figure 11 are not applicable to the interface shear
stresses in the vicinity of the plate end so that the
interface strength may be directly related to the tensile
strength of the concrete.
Away from stress concentrations at the intercept of a
crack or at the plate-end due to the discontinuity of the
plate, the interface shear stress is given by the well
known elementary structural mechanics equations
of VAy/Ib or VQ/Ib. This form of debonding rarely
controls design so that it is suggested these shear
stresses can also be directly related to the tensile strength
of the concrete as is often assumed in national standards.
Debonding
Figure 11. Idealised / material characteristics
behaviour of plated structures and there has been

recently much very good research in quantifying this
material characteristic which is often idealised as bi-linear
as shown in Figure 11.
5. DEBONDING MECHANISMS
IN EXISTING GUIDELINES
Most national guidelines, such as those listed in the
table in Figure 13 (Oehlers and Seracino 2004), deal
with tension face plated beams and it is felt that within
these guidelines there is already a general consensus as
to the major debonding mechanisms, even though each
guideline may use its own terminology as listed in the
table in Figure 13. All the guidelines acknowledge IC
(intermediate crack) debonding due to flexural cracks in
the beam and do distinguish between the behaviours of
IC debonding at the anchorage zone shown in column 1
and IC debonding due to a succession of flexural and
flexural/shear cracks as in columns 2 and 3; this form of
debonding is controlled in the guidelines through
limitations to the plate strains.
All the guidelines acknowledge CDC (critical
diagonal crack) debonding due to rigid body shear
deformations as in column 4 in Figure 13 and often
quantify this form of debonding by adjusting the
concrete component of the shear capacity Vc. Most
guidelines acknowledge PE (plate end) debonding due
to curvature, as in column 5, but may not give design
rules. For the European approach this is understandable,
4.2. Concrete Interface Resistance

The plate/concrete interface can also fail in the absence
of the aggregate interlock that is required when
debonding propagates from the stress concentrations
induced by cracks intercepting plates as described in the
previous section. Take for example the stresses that are
induced at the plate-end in Figure 12, due to curvature
in the beam, and where, because the plate is simply
trying to stay straight, debonding propagates from the
plate-end inwards. In this case, there are large stresses
normal to the plate/concrete interface in the vicinity of
the plate-end which can be assumed to be directly
resisted by the direct tensile strength of the concrete as
opposed to the aggregate-interlock shear resistance
depicted in Figure 11. There are also interface shear
stresses which are at a maximum near the plate end.
However it is felt that the mechanism that occurs in
developing the aggregate interlock material resistance
in Figure 11 does not occur when debonding initiates
from the plate end, even when the interface shear stresses
are dominant, as the normal force to the interface
Direction of debonding
crack propagation
Interface shear
PE debonding crack
Nplate
Pplate
Nplate
Plate strain = 0
= 0 at plate end
Mplate
Figure 12. Plate end debonding mechanism
741
ASE 9-6-02_Oehlers
4/12/06
5:10 pm
Page 742
Debonding mechanisms
p.o.c.
Uncracked section
p.o.c.
VC
Cracked section
Guidelines
[1]
AUST: IC
debonding
EUR mode 1:
peeling off in
uncracked
anchorage
BRIT:
debonding
failure
HK:
IC induced
debonding
USA: referenced.
[2]
AUST: IC
debonding
EUR mode 2:
peeling off at
flexural crack
BRIT:
debonding
failure
HK:
IC induced
debonding
USA: refer.
IC
[3]
VAy
[6]
AUST: IC
debonding
EUR mode 2:
peeling off at
flexural crack
BRIT:
debonding failure,
peeling failure
HK:
IC induced
debonding
USA: refer.
CDC
PE
[4]
[5]
AUST: CDC
debonding
EUR mode 3:
peeling off
caused by
shear cracks
EUR (mode 5):
plate-end
shear failure
BRIT: refer.
HK: plate end
interfacial
USA: refer.
AUST: PE
debonding
EUR (mode 6):
concrete rip-off
BRIT:
peeling failure
HK:
concrete cover
separation
BRIT: refer.
USA: not
recognised
Figure 13. Tension face plate guidelines
as the European approach requires the plate to be

terminated in an uncracked region where the curvature
is low and, furthermore, only deals with thin externally
bonded FRP plates in the tension region which are being
bent about their minor axis; both requirements reduce
the probability of PE debonding. Few guidelines
directly acknowledge VAy/Ib debonding between stress
concentrations shown in the block marked 6.
6. GENERIC DEBONDING MECHANISMS
AXIAL DEFORMATIONS
Research has shown (Liu et al. 2004) that the axial
intermediate crack (AIC) debonding resistance in
pull-tests as in Figure 10 (where the plate is only
subjected to axial deformations, subjected to the
equivalence of one crack, and debonding is
characterised by the / material properties in Figure
11) is a lower bound or equal to the flexural IC
intermediate crack (FIC) debonding resistance in
beams. Hence the AIC debonding resistance in pulltests is a fundamental property.
The governing mathematical equations based on the
/ characteristics in Figure 11 have been developed at
Hong Kong Polytechnic University for predicting the
partial-interaction AIC debonding resistance in pulltests. Partial-interaction theory based on the rising
branch in Figure 11 has been well established for over
fifty years in the analysis of composite steel and
concrete structures (Oehlers and Bradford 1995).
However, the research at Polytechnic University
developed partial-interaction solutions to the falling
branch for the first time and is, hence, outstanding
ground breaking research. This research was then used
742
at Adelaide University in Australia to predict the AIC

debonding resistance in fully anchored plates based on
the falling branch only, as this allowed a closed form
solution (Seracino et al. 2006; Oehlers et al. 2006). In
this research, the failure plane was also separated from
the geometry of the plate to give the following
fundamental AIC debonding resistance which can be
applied to both externally bonded (EB) plates and near
surface mounted (NSM) plates as in Figure 14.
PAIC = f f
L per ( EA) p
(1)
where is twice the area under the idealised /

characteristics in Figure 11, Lper is the length of the failure
plane in Figure 14 and (EA)p is the axial rigidity of the
plate which can also include the axial stiffness of the
adhesive bonding the plate. This is a convenient
mathematical form as it no longer depends directly on the
plate thicknesses tp, plate width bp nor glue thicknesses
which may be difficult to measure particularly for the wet
lay-up procedure and, furthermore, the same equation
applies to both EB and NSM plates.
FLEXURAL DEFORMATIONS
7.1. FIC Debonding
Flexural intermediate crack debonding is illustrated in
Figure 9 for the case of a single dominant flexural crack;
this is also equivalent to AIC debonding as described
previously (Liu et al. 2006). An example of flexural
intermediate crack (FIC) debonding of an externally
bonded (EB) plate due to flexural deformations in a
ASE 9-6-02_Oehlers
4/12/06
5:10 pm
Page 743
D. J. Oehlers
bp tp
2 mm
Concrete surface
Ap of EB plate
df
2 mm
bf
Failure plane EB plate
Ap of NSM plate
Lper
Concrete
element
Lper
bp
df
tp
2 mm
Failure plane NSM plate
2 mm
bf
Figure 14. EB and NSM plates

Full-interaction
region
Partial-interaction
region
Full-interaction
region
Flexural
crack
(a) Plated
beam
D
(b) Slip
FA-B
(c) Resultant
bond force
FB-C
Figure 15. FIC debonding mechanism
beam is shown in Figure 13 adjacent to the region

marked IC for the case where there are several flexural
cracks intercepting the plate. The distribution of slip
across the plate/concrete interface is shown in Figure 15
for the case of a beam with three cracks. Typically,
between the three cracks in region B-B, the slip reverses
in direction as shown and beyond the outermost flexural
cracks, regions B-C, the slip is in one direction only.
Regions B-C represent the pull test in Figure10 as the
distributions of slip are identical. The FIC debonding
resistance depends on the sum of the resulting bond
shear forces FA-B and FB-C in regions A-B and B-C.
The problem is complex as it is a brittle problem such
that the resulting bond shear forces do not necessarily

reach their peak values simultaneously, which is
analogous to the problem illustrated in Figure 8 for
transverse plates. However, what is known is that when
there is only one crack the FIC debonding resistance is
equal to the AIC debonding resistance and when there
are several cracks the FIC debonding resistance is
greater than or equal to the AIC debonding resistance.
Figure 16 shows FIC debonding of a externally
bonded plate on the tension face of the beam and
Figure 17 shows FIC debonding of a near surface
mounted plate also on the tension face of a beam. Figure 17
shows the classical herringbone formation of cracks
associated with shear failure of concrete. The flexural
cracks (vertical in the figure) first form and then between
these cracks the plate gradually debonds as can be seen
by the herringbone cracks (inclined cracks) on either side
of the NSM plates (horizontal). Importantly, debonding
(herringbone cracks) starts from the flexural cracks and
from where the strains are at their maximum and spread
outwards. Towards the centre of the beam and between
flexural cracks, these herringbone cracks are inclined
in opposite directions clearly showing the opposing
directions of the interface shear forces also indicated by
the direction of the slip in Figure 15(b). In some regions
they are in both directions (crossing each other) showing
that the shear forces start in opposite directions and then
change to the same direction due to slip between the
concrete teeth between flexural cracks. Towards the ends
of the herringbone crack formation, the herringbone
cracks are in one direction as in a pull test, that is as in
AIC debonding. Figure 18 shows FIC debonding of a
NSM plate on a side face where the classical formation
of herringbone cracks can also be seen.
Each concrete tooth (block of concrete between
flexural cracks) in Figure 17 can be considered to act as
a shear connector (as in composite steel and concrete
beams) imposing some force on the plate. The ends of
the herringbone formation are anchorages equivalent to
pull tests. This mechanism, of concrete-teeth and end
anchorages is a further explanation of why the FIC
Figure 16. FIC debonding of EB plate on tension face
743
ASE 9-6-02_Oehlers
4/12/06
5:11 pm
Page 744
Near surface mounted plate
Herringbone IC debonding
Figure 17. FIC debonding of NSM plate on tension face
Figure 18. FIC debonding of NSM plate on side face
Figure 19. PE debonding of EB tension face plate
debonding resistance in beams can be greater than the

AIC debonding resistance in pull-tests. However, the
problem is complex. If the beam were under high
vertical shear forces, then only one flexural crack would
occur, hence, the FIC resistance would tend to the AIC
resistance. If the beam were subjected to a constant
moment region in which the vertical shear is zero, then
within the constant moment the concrete teeth would
have a net zero shear force (as the force in the
reinforcing bars on either side of the concrete tooth are
the same) so the strength may also tend to the AIC
debonding resistance depending on the conditions
beyond the constant moment region. In between these
two extremities, the FIC resistance may be greater
than the AIC resistance. There is much research on
744
quantifying this increase but it is important to realise

that the AIC resistance is a lower bound or equal to the
FIC debonding resistance in a beam.
7.2. PE Debonding
The plate end debonding mechanism is illustrated in
Figure 12. The plate simply tries to stay straight
inducing stress concentrations at the plate-end which
cause debonding from the plate-end inwards (Oehlers
et al. 2000). An example of PE debonding of an EB
plate is shown in Figure 19. In this example, the plated
section is in a constant moment region so that it is
only the curvature that is inducing debonding as the
plate tries to stay straight inducing large interface
normal tensile stresses at its ends. Importantly in this
ASE 9-6-02_Oehlers
4/12/06
5:11 pm
Page 745
D. J. Oehlers
PE debonding
Figure 20. PE debonding of EB side face plate
Figure 21. PE debonding of an angle section
Plate end
Plate end
Compression
face plate
PE debonding
Figure 22. PE debonding EB compression face plate
form of debonding, the cracks start at the plate-end

and work inwards and for tension face plated beams
they propagate along the level of the tension
reinforcement as this is a zone of weakness for the
forces in Figure 12 that are normal to the plate interface
Nplate. Exactly the same mechanism as in Figure 19
occurs in side plates as in Figure 20 where debonding
starts at the plate-ends and works inwards, but in this
case the interface stress that causes debonding is
shear. A further example of PE debonding is shown in
Figure 21 for an angle section bonded to both the side
and tension faces.
PE debonding also occurs in compression face plates as
in Figure 22, but in this case the signs of the interface
normal stresses or forces Nplate in Figure 12 reverse in
direction making the resistance larger. Plate end
debonding also occurs in NSM plates on the tension face
as in Figure 23. In the three cases shown in Figures 2022,
PE debonding has not caused concrete cover separation as
in the tension face plate in Figure 19.

7.3. VAy/Ib Debonding
Tests have shown that debonding due to the elementary
interface shear between stress concentrations VAy/Ib
seldom controls debonding but it is felt that it should
be included in design as it may control debonding for
prestressed beams where flexural cracks are suppressed
and for serviceability where thick plates may be used to
control deflections when there are few flexural cracks.
TRANSVERSE SHEAR DEFORMATIONS
8.1. CDC Debonding
The critical diagonal crack debonding mechanism is
illustrated in Figure 24 (Oehlers et al. 2004; Mohamed
Ali et al. 2005). In this case the rigid body shear
deformation across the CDC widens the CDC which
induces axial plate forces Pplate that are limited by the
745
ASE 9-6-02_Oehlers
4/12/06
5:11 pm
Page 746
PE debonding
IC debonding
Figure 23. PE debonding of NSM plate

IC interface crack
Pplate
Critical diagonal crack
Pplate
Rigid body displacement
Figure 24. Critical diagonal crack debonding mechanism
AIC debonding resistance of the plates.

An example of critical diagonal crack (CDC)
debonding due to the shear deformations across a CDC
in a tension face plated beam with stirrups is shown in
Figure 25. Debonding is induced by the CDCs starting
near the support and working inwards. It can be seen
Debonding crack from right to

left but spreading to the right
A
B
2
that the debonding cracks propagate outwards, as shown

by the arrows, but sometimes progress inwards, as
shown by the sequence numbers, and for tension face
plated beams the debonding crack follows the weak
zone along the level of the tension reinforcement.
Figure 26 shows the sudden formation of a CDC at
188 kN that has caused the tension face plate to debond.
Figure 27 shows the gradual propagation of IC interface
cracks and then the sudden formation of a CDC that has
caused debonding. Figures 26 and 27 help illustrate the
rapid nature of CDC debonding which is often
associated with transverse shear failure.
CDC debonding can also be seen in the side plated
beam without stirrups in Figure 28 where because of the
absence of stirrups only one CDC forms. For EB plates,
CDC debonding occurs before the stirrups can be
stretched. CDC debonding can also occur in NSM tension
face plated beams as in Figure 29 and in EB
Sequence of crack
propagation
C
3
E
4
CDC debonding crack

Plate end
Tension face plate
Figure 25. CDC debonding EB tension face plate
Diagonal
crack (100 kN)
Critical diagonal
crack (188 kN)
CDC debonding crack
Tension face plate
CDC propagation
Figure 26. Sudden formation of CDC
746
ASE 9-6-02_Oehlers
4/12/06
5:11 pm
Page 747
D. J. Oehlers
IC interface
cracking
Intermediate
crack
CDC debonding crack

Critical
diagonal crack
Figure 27. CDC debonding of tension face plate in hogging region
Figure 28. CDC debonding EB side face plate
CDC debonding
Critical
diagonal
crack
Figure 29. CDC deb. NSM tension face plate
compression face plated beams in Figure 30. In the latter

three cases, Figures 2830, there is no concrete cover
separation as in the tension face plate in Figure 25.
A further example of a CDC can be seen in the push
test Figure 31 where there are three horizontal NSM
plates that cross a vertical critical diagonal crack induced
in the push test. The rigid body shear deformation across
the CDC causes the crack faces to separate through
aggregate interlock inducing axial forces in the plates
that induce the herringbone formation of cracks. As we
are now dealing with a single crack intercepting each
plate, the axial resistances of the NSM plates is their AIC
debonding resistance as can be determined from pulltests. This photograph clearly shows the dependence of
the CDC resistance on the AIC debonding resistance.
The difference between FIC (flexural intermediate
crack) and CDC (critical diagonal crack) debonding
is that in the former the crack is widened by flexure
and in the latter by aggregate interlock through

shear deformations. The AIC force across the crack
increases the concrete component of the shear
capacity Vc.
Unlike EB plates where debonding always occurs on
the formation of the CDC crack and before the stirrups
are stretched, it is worth noting in Figure 31 that the
shear deformation has not actually caused the NSM
plate to debond even though they may increase the shear
capacity Vc. This is in contrast to the NSM plates on
the tension face in Figure 29 where debonding has
occurred because it is being deformed about the plates
major axis compared with the deformations in Figure 31
that occur about the plates minor axis. It would appear
that that placing NSM plates on the sides of beams
subjected to large shear forces as in Figure 32 would be
an appropriate design solution.
747
ASE 9-6-02_Oehlers
4/12/06
5:11 pm
Page 748
Compression face plate
Applied
load
CDC debonding
Support
Figure 30. CDC deb. EB compression face plate
IC debonding
Critical
diagonal crack
Figure 31. CDC deb. NSM side plated push-test
Figure 33. SIC debonding of EB transverse plated beam
(PAIC)2
Inclined
plate
(PAIC)1
Internal
steel stirrup
CDC
Figure 32. CDC deb. NSM side face plate
Figure 34. Shear increase due to inclined plates
8.2. SIC Debonding

Transverse plates act as external stirrups to increase the
vertical shear capacity directly as in Figure 33 (Teng et al.
2002). This is an example of a plate crossing a single crack
so that the debonding resistance of an individual plate is its
AIC debonding resistance as shown in Figure 10.
The debonding resistance of each plate in Figure 33
depends on its minimum anchorage lengths on either
side of the CDC which can be seen to vary, and also on
the / characteristics in Figure 11 (Mohamed Ali et al.
2006). Hence as the CDC widens each plate achieves its
maximum capacity at differing crack spacings so that the
plates with short anchorage lengths will achieve their
maximum capacities and debond before plates with

larger anchorages achieve theirs as described previously.
The increase in the shear capacity due to inclined
plates as in Figure 34 can now be accommodated
by allowing the horizontal component of the AIC
debonding resistance to increase the concrete component
Vc in Eqn 1 and the vertical component to act as
stirrups contributing ktpVtp, although allowance must be
made for any reduction in the stirrup contribution ksVs.
748
9. CONCRETE COVER SEPARATION

Concrete cover separation is a frequently used term to
describe debonding failure. It is felt that even though this
ASE 9-6-02_Oehlers
4/12/06
5:11 pm
Page 749
D. J. Oehlers
ve or hogging region
+ve or sagging region
ve or hogging region
(a) Tension face plates

PE
EB
FIC
CDC
CDC
PE
FIC
VAy
NSM
Compression face
(b) Compression face plates
PE
CDC
SIC
VAy
PE
FIC
(c) Side plates

CDC
CDC
FIC
FIC
PE
VAy
PE
NSM
(d) Combinations
Angle section
CDC
Wrapped plate
PE
Figure 35. Generic debonding of a range of plates
term is correct in describing a visually seen form of

debonding, it can lead to confusion. This is because
concrete cover separation occurs in PE debonding of
tension face plates as can be seen in Figure 19 which was
tested in a region of zero shear. However, concrete cover
separation also occurs in the CDC debonding of tension
face plates as in Figure 25 which occurs in regions of high
shear. Hence, concrete cover separation occurs in two
distinct forms of debonding of tension face plated beams
that are induced by two distinct deformations that of either
curvature or that of shear. Furthermore, concrete cover
separation only applies to tension face plates and does not
describe the same form of debonding in side or
compression face plates. For example, the mechanism that
causes PE debonding in Figure 19 also occurs in side plates
in Figure 20, compression face plates in Figure 22 and
NSM plates in Figure 23 none of which show concrete
cover separation. The same can be said for CDC debonding
in Figure 25 for tension face plates, which also applies to
side plates in Figure 28 and to compression face plates in
Figure 30 in which there is no concrete cover separation.
10. SUMMARY GENERIC DEBONDING
MECHANISMS
Generic debonding mechanisms have been identified for
all forms of plating as illustrated in Figure 35. The
eventual quantification of these generic debonding

mechanisms will allow engineers to use all forms of
plating and hence give them the freedom to develop
their own retrofitting techniques.
11. CONCLUSIONS
It has been shown that plated RC beams and slabs are
subjected to the same generic flexural and shear
deformations as in unplated RC beams and that these
generic deformations induce stress resultants that can
lead to generic debonding mechanisms that, because of
their generic nature, encompass all forms of plating. It is
suggested that it is this understanding of the generic
response of plated RC beams which will allow generic
national standards to be developed in line with those for
unplated RC beams.
REFERENCES
Liu, I., Oehlers, D.J. and Seracino, R. (2004). Parametric study of
intermediate crack (IC) debonding on adhesively plated beams,
Proceedings, The Second International Conference on FRP
Composites in Civil Engineering, CICE 2004, December.
Liu, I.S.T., Oehlers, D.J. and Seracino, R. (2006). Study of
intermediate crack debonding on FRP plated beams. Journal of
Composites for Construction, ASCE, in press.
Mohamed Ali, M.S., Oehlers, D.J. and Seracino, R. (2006). Vertical
749
ASE 9-6-02_Oehlers
4/12/06
5:11 pm
Page 750
shear interaction model between external FRP transverse plates

and internal steel stirrups, Engineering Structures, Vol. 28,
No. 3, pp. 381389.
Mohamed Ali, M.S., Oehlers D.J. and Bradford, M.A. (2005).
Debonding of steel plates bonded to the compression faces of
RC beams, Construction and Building Materials, Vol. 19, No. 6,
pp. 413422.
Oehlers, D.J., Nguyen, N.T. and Bradford, M.A. (2000).
Retrofitting by adhesive bonding steel plates to the sides of R.C.
beams. Part 1: debonding of plates due to flexure, Structural
Engineering and Mechanics, Vol. 9, No. 5, pp. 491504.
Oehlers, D.J. and Bradford, M.A. (1995). Composite Steel and
Concrete Structural Members: Fundamental Behaviour,
Pergamon Press, Oxford.
Oehlers, D.J. and Seracino, R. (2004). Design of FRP and Steel
Plated RC Structures: Retrofitting Beams and Slabs for Strength,
Stiffness and Ductility, Elsevier.
Oehlers, D.J., Seracino, R. and Smith, S. (2006). Design Guideline
for RC Structures Retrofitted with FRP and Metal Plates:
Beams and Slabs, SAI Global Limited Standards Australia, 108
pages, in press.
Seracino, R., Raizal Saifulnaz, M.R. and Oehlers, D.J. (2006).
Generic intermediate crack debonding resistance of EB and
NSM plate-to-concrete joints, Journal of Composites for
Construction, ASCE, in press.
Teng, J.G., Chen, J.F., Smith, S.T. and Lam, L. (2002). FRP
Strengthened RC Structures, John Wiley and Sons Ltd.
Chichester, England.
APPENDIX: NOTATION
Ap
cross-sectional area of plate
AIC
axial intermediate crack
AUST Australia
bf
width of failure plane of perimeter length Lper
bp
width of plate
BRIT British
CDC critical diagonal crack
d
effective depth of beam
df
depth of failure plane of perimeter length Lper
(EA)p axial rigidity of plate
EB
externally bonded
EUR
Europe
F
component of longitudinal or axial force;
force profile
FIC
flexural intermediate crack
FRP
fibre reinforced polymer
HK
Hong Kong
IC
intermediate crack
k
interaction factor
ke
total axial force in transverse plates crossing
CDC as a proportion of Vtp
ki
total axial force in stirrups crossing CDC as a
proportion of Vs
750
ks
kt
proportion of Vs at transverse plate debonding

total force in stirrups and transverse plates
crossing CDC as proportion of Vs+Vtp
ktp
proportion of Vtp at transverse plate debonding
ku
neutral axis depth factor; depth of neutral
axis as a proportion of d
Lper
length of perimeter of failure plane
Mplate moment in plate
Nplate force normal to plate interface
NSM near surface mounted
P
axial force
PAIC
AIC debonding capacity of fully anchored plate;
maximum AIC debonding capacity
PIC
axial force in plate at intermediate crack
Pplate
axial force in plate
PE
plate end
p.o.c
point of contraflexure
RC
reinforced concrete
t
plate thickness
V
transverse shear
Vc
concrete component of shear capacity of beam
Vcap
transverse shear capacity
Vplate transverse force in plate
Vs
stirrup component of shear capacity of beam;
yield strength of stirrups crossing CDC
Vtp
maximum axial force in transverse plate
VAy/Ib interface shear stress in region beyond stress
concentrations
VQ/Ib interface shear stress in region beyond stress
concentrations
y
distance of F from some datum point
Vc
increase in Vc due to longitudinal plate
shear slip
f
shear slip capacity
strain; strain profile

bar
strain in reinforcing bar
c
concrete crushing strain
db
plate debonding strain
fr
plate fracture strain
IC
strain in plate at intermediate crack
stress; stress profile
shear stress
f
maximum shear strength
twice fracture energy

Subscripts:
AIC
axial intermediate crack
NSM near surface mounted plate
p
Plate
sp
side plate
tfp
tension face plate
ASE 9-6-03_Neale
4/12/06
5:10 pm
Page 751
Analysis of the Load-Deformation Behaviour

and Debonding for FRP-Strengthened
Concrete Structures
K. W. Neale*, U. A. Ebead, H. M. Abdel Baky, W. E. Elsayed and A. Godat
Department of Civil Engineering, University of Sherbrooke,
Sherbrooke, Quebec, Canada J1K 2R1
Abstract: Results from nonlinear finite element analyses of fibre reinforced polymer
(FRP)-strengthened concrete beams and slabs are presented. The direct shear test, a
basic application that provides insight into FRP-concrete interfacial behaviour, is also
considered. The motivation for this work is the fact that, although there is a large
amount of experimental data available on the FRP strengthening of concrete structures,
a full understanding of the various loaddeformation behaviours and debonding
phenomenon is still lacking. The numerical models presented in this paper adopt a
displacementcontrolled solution and are capable of simulating FRP-strengthened
beams either in shear or in flexure, as well as slabs strengthened using either passive or
prestressed FRP laminates. Results of the different applications are presented and
compared with published test data, and a very good agreement in terms of the ultimate
load carrying capacities, loaddeflection behaviour and modes of failure, is obtained.
Key words: FRP, strengthening, numerical modelling, reinforced concrete, slabs, beams, flexure, shear, direct shear.
1. INTRODUCTION
The strengthening of reinforced concrete structures by
means of externally bonded fibre reinforced polymers
(FRPs) is now routinely considered as an effective method
for load capacity enhancement. However, with this
technique there is often a concern that possible failures
may occur due to debonding of the FRP from the concrete.
Laboratory investigations on FRP-strengthened concrete
beams in flexure and shear, for example, have shown that
despite the capability of achieving considerable increases
in strength capacities, premature failures by debonding
often limit the effectiveness of the strengthening schemes.
At present, our basic understanding of the mechanics of
the bond and failure between the FRP and the concrete for
such applications is somewhat limited.
As far as numerical studies on FRP-strengthened
beams are concerned, researchers have recently
attempted to simulate the behaviour of the beams using

finite element techniques. The complex behaviour of
the strengthened structures first led researchers to use a
linear elastic analysis to address the interfacial
behaviour before cracking (Ehsani and Saadatmanesh
1990). A more advanced finite element model, the
layer-by-layer numerical technique, was introduced to
take into consideration the material nonlinearities of the
concrete before and after cracking, and also to include
the effect of tension stiffening (Takahashi et al. 1997;
Nitereka and Neale 1999; Ebead and Marzouk 2005).
These analytical models were proposed to predict the
loaddeflection behaviour and the ultimate load
carrying capacities of FRP-strengthened concrete
beams; however, they did not simulate the debonding
failure modes or specifically address the interfacial
behaviour. It has been demonstrated that debonding
*Corresponding author. Email address: Kenneth.Neale@USherbrooke.ca; Fax: +1-819-821-7974; Tel: +1-819-821-7752.
751
ASE 9-6-03_Neale
4/12/06
5:10 pm
Page 752
Analysis of the Load-Deformation Behaviour and Debonding for FRP-Strengthened Concrete Structures
often prevents FRP-strengthened beams from attaining

their ultimate flexural capacity (Brena et al. 2003;
Thomsen et al. 2004). Hence, to accurately predict the
ultimate load carrying capacities and capture the
observed debonding failures, it is necessary to properly
model the bond behaviour at the FRP/concrete interface.
For this purpose, interface elements that are able to
properly represent the local shearslip characteristics
and failure are required.
Shear strengthening of reinforced concrete beams is
required when the beam is deficient in shear, or when its
shear capacity falls below its flexural capacity after
flexural strengthening. An accepted technique for the
shear strengthening of reinforced concrete beams is to
provide an additional FRP web reinforcement in the
form of externally bonded FRP laminates. Quite a few
numerical studies concerning the shear strengthening of
reinforced concrete members have been published
(Malek and Saadatmanesh 1998a, 1998b; Kachlakev and
Miller 2001; Al-Mahaidi et al. 2001; Wong and Vecchio
2003). However, the available numerical algorithms in
the literature do not explicitly simulate the interfacial
behaviour of the strengthened beams. To our knowledge,
the analysis presented here is the first that directly models
the bondslip behaviour for these applications.
With regard to reinforced concrete slabs, FRPs have
been used for the flexural strengthening of one-way
slabs (Erki and Heffernan 1995; Seim et al. 2001). In
addition, FRPs have been used for the flexural
strengthening of two-way slabs using FRP laminates
bonded along the middle of the slab (Ebead and
Marzouk 2004; Harajli and Soudki 2003), or distributed
along the slab width (Mosallam and Mosalam 2003;
Limam et al. 2003). In another application, prestressed
FRP laminates have been used for the strengthening of
two-way slabs (Longworth et al. 2004). In general, there
have been limited analytical studies on FRP-strengthened
slabs compared to those on beams.
In this paper, the finite element formulation for the
basic application of the FRP/concrete direct shear test is
first presented. In addition, the results of the nonlinear
finite element analyses for reinforced concrete beams
strengthened in flexure and in shear are presented. For
these applications, the finite element models address the
FRP/concrete interfacial responses and are capable of
simulating the various failure modes, including FRP
debonding at either the plate end or at intermediate
cracks. Results of the numerical modelling of FRPstrengthened two-way slabs are also presented. Various
FRP configurations, with both passive as well as
prestressed FRP strengthening, are considered. The
results of these applications are presented in terms of
loaddeflection relationships, ultimate load capacities,
752
failure modes, and interfacial slip and stress

distributions. The numerical predictions are validated
against experimental test data, and very good agreements
are obtained.
2. NONLINEAR FINITE ELEMENT MODELS
The numerical analyses are carried out using the finite
element software package ADINA (2004a). The entire
nonlinear loaddeformation behaviour of the structure is
simulated under displacement-controlled loading
conditions. In the analyses, the software formulations
for the concrete, steel and FRP are employed. These are
described in detail in the ADINA theory and modelling
guide (ADINA 2004b), and are summarized below. To
model the FRP/concrete interface, appropriate bond
stressslip relations via truss elements for the direct
shear and flexurally-strengthened beam applications,
and spring elements for the slabs and shear-strengthened
beam applications are introduced. These truss and
spring elements link the FRP laminates to the concrete.
2.1. Material Models for Concrete, Steel and FRP
The constitutive law used to model the concrete has the
following features (ADINA 2004b): (i) a nonlinear
stressstrain relation to allow for the weakening of the
material under increasing compressive stresses, (ii)
failure envelopes that define both failure in tension and
crushing in compression, and (iii) a strategy to model the
post-cracking and post-crushing behaviour of the
material. The general multiaxial stressstrain relations are
derived from a nonlinear uniaxial stressstrain relation.
The cracked concrete is assumed to be orthotropic, with
the directions of orthotropy being defined by the principal
stress directions. Failure envelopes are used to establish
the uniaxial stressstrain law accounting for multiaxial
stress conditions, and to identify whether tensile or crushing
failure of the concrete has occurred. The post failure
material behaviours account for post-tensile cracking,
post-compression crushing, and strain-softening.
For the steel reinforcement, a uniaxial elastic-plastic
stressstrain law is employed. A linear elastic orthotropic
constitutive relation is adopted for the FRP composites.
2.2. FRP/Concrete Interface Constitutive Model
One of the most accurate bond stressslip models that
can be incorporated into a finite element analysis is that
proposed by Lu et al. (2005a). With this, the mechanical
behaviour of the FRP/concrete interface is represented
by a relationship between the local shear stress, , and
the relative displacement, s, between the FRP laminate
and the concrete. The area under the s curve
represents the interfacial fracture energy, Gf, which
corresponds to the energy per unit bond area required
ASE 9-6-03_Neale
4/12/06
5:10 pm
Page 753
K. W. Neale, U. A. Ebead, H. M. Abdel Baky, W. E. Elsayed and A. Godat
In Eqn 3
Bond
stress
se = max K 0
max
(5)
The value of K0 is given by

K0
K0 = Ka Kc
(Ka + Kc ) ,
(6)
where
Slip
S0
K a = Ga ta
(7)
K c = Gc tc ,
(8)
Figure 1. Bondslip model
for complete debonding. Lu et al. (2005a) have

proposed three different models according to their level
of sophistication; namely, the precise, the simplified,
and the bilinear models. In the current study, a slight
modification to the precise model will be used, as it
gives the flexibility of altering the stiffness of the
adhesive, which is believed to be an important factor
to consider.
The modification of the original model employed
here consists of imposing a linear relationship between
the shear stress and the slip in the pre-peak zone, as
given in the bilinear model of Lu et al. (2005a). The
descending part of the relationship is the same as that in
their precise model. The model is shown in Figure 1 and
formulated as follows:
Considering max to be the maximum bond stress and
s0 the corresponding slip, then for the ascending part;
i.e., s s0
= max
(1)
max = 1.5w ft
(2)
while the value of s0 is related to ft as follows:

s0 = 0.0195 w ft + se
(3)
The factor w refers to the FRP width factor; it is related

to the ratio between the width of the FRP, bf, and that of
the concrete structural element, bc, through the following
equation:
(2.25 b
bc
) (1.25 + b
= max exp ( s / s0 1 )
(9)
where
=
1
Gf
max s0 3
(10)
The interfacial energy Gf is obtained from:
s
,
s0
The value of max is related to the tensile strength of the

concrete, ft, according to the following equation:
w =
Here Ga and ta, and Gc and tc are the shear modulus and
the thickness of the adhesive and concrete, respectively.
The value of tc is taken as 5 mm; this corresponds to the
effective thickness whose deformation forms part of the
interfacial slip (Lu et al. 2005b).
For the descending part, i.e., s > s0
bc
),
(4)
G f = 0.308 w2 ft f ( K a )
(11)
The function f(Ka) proposed in the original model by Lu

et al. (2005a) is equal to 1 for normal adhesives, for
which Ka 2.5 GPa/mm. In our analyses of the beam
and slab applications, the value of f(Ka) was taken as 1. For
the direct shear applications, the value of this parameter
(ranging from 1 to 2) was determined on the basis of the
stiffness of the adhesive, in the manner suggested by Lu
et al. (2005a).
3. GEOMETRICAL MODELLING
In order to characterize the interfacial behaviour, the
FRP nodes are connected to the concrete nodes by using
nonlinear interface elements, as shown in Figure 2. It is
necessary to emphasize that these elements represent the
entire interfacial behaviour between the concrete and the
FRP, and not the adhesive as such. The aforementioned
constitutive model (the bondslip model) thus represents
753
ASE 9-6-03_Neale
4/12/06
5:10 pm
Page 754
Point 1 (master)- concrete node

2-D node interface element
Point 1 (slave)
Point 2 (master)
Point 2 (slave)- FRP node
Figure 2. Interface element
the contribution of the FRP composites, adhesive and

concrete. The relative displacement between Point 1 and
Point 2 of the 2-node interface element (Figure 2)
represents the interfacial slip, and the stress in the element
represents the interfacial shear stress. Full strain
compatibility is assumed between the FRP nodes and the
Truss element
(attached FRP)
Interface
element
concrete nodes in the normal direction by enforcing

suitable constraint equations in that direction between the
FRP laminate and the concrete.
For the direct shear and the flexurally-strengthened
beam applications, the out-of-plane stresses are assumed
negligible compared to the in-plane stresses, hence plane
stress elements are used to simulate the concrete. Typical
geometrical representations and the type of elements used
for the direct shear and the beams are shown in Figures
3(a) and 3(b), respectively. Quadrilateral 9-node plane
stress elements are used to represent the plain concrete.
The FRP strengthening laminates are simulated as 3-node
truss elements. To precisely investigate the interfacial
behaviour along the FRP/concrete interface and consider
the influence of crack propagation, a finer mesh is used
close to the FRP/concrete interface. For the beams, the
steel reinforcement bars are represented using 3-node truss
C.L.
Applied load
Applied load
9-node
plane stress
(concrete)
9-node
plane stress
(concrete)
Truss element
(steel)
Support edges
Interface
element
Truss element
(attached FRP)
(a) Direct shear test
(b) Beams in flexure
C.L.
Applied load
C.L.
Applied load
C.L.
Attached FRP
Support edges
A s'
Support
Attached
U-shape FRP
Concrete
node
FRP
node
As
Beam cross section
Detail A
(c) Beams in shear
Truss element (steel)

3-D solid element
(concrete)
concrete node
Interface element
FRP node
Shell element (FRP)

Detail B
Detail C
(d) Slabs in flexure
Figure 3. Finite element models
754
ASE 9-6-03_Neale
4/12/06
5:10 pm
Page 755
elements due to their small flexural inertia compared to

that of the beam section. The element size for the beam
application at the interface was 12.5 12.5 mm.
For the shear-strengthened beams and the strengthened
slabs, 3-D analyses are employed as shown in Figures
3(c) and 3(d) where typical geometrical representations
and the types of elements used are depicted. Due to the
geometrical and the loading symmetry, only one half
of the beam or one quarter of the slab is modelled.
27-node solid elements are used to represent the plain
concrete. Using such elements satisfies shear and
bending deformations due to their quadratic interpolation
functions. Nine-node thin shell elements are used to
represent the FRP composites. The orthotropy of each
FRP strip is accounted for in the constitutive material
properties of the corresponding elements depending on
the fibre orientation. The steel reinforcement bars
embedded in the concrete are modelled using 3-node
truss elements. Symmetrical boundary conditions are
placed along the planes of symmetry. For one of the
slabs tested by Longworth et al. (2004), prestressing is
applied to the FRP laminates. These prestressing forces
are simulated numerically by applying appropriate
initial strains to the FRPs.
4. NUMERICAL RESULTS
AND DISCUSSION
The results presented in the subsequent sections are in
terms of the ultimate load carrying capacities,
loaddeflection relationships and failure modes for the
different applications simulated in this study. Special
emphasis is placed on the results of the interfacial
behaviour between the FRP laminates and the concrete
in terms of the interfacial stress distributions and slip
profiles. The specimen notations here correspond to
those employed in the original references.
4.1. Ultimate Load Carrying Capacities
For the direct shear tests, the finite element analysis
results are compared to 92 specimens to validate the use
of the aforementioned interface element and the
associated bondslip model. The database used in this
validation is versatile in terms of the FRP bonded length
and width, and the properties of the adhesive, the FRP,
and the concrete. The finite element model and the
bondslip model are very general in a way that
allows accommodating all of the variations in the test
characteristics and parameters. Comparisons of the finite
element predictions with the corresponding experimental
data show a good agreement, as can be seen from the
results depicted in Figure 4(a). For the series of
specimens for the direct shear case tested by Tan (2002),
Zhao et al. (2000), Takeo et al. (1997), Ren (2003),
Bizindavyi and Neale (1999) and Chajes et al. (1996),

the average numerical-to-experimental load capacity
ratio is 98.9% with a standard deviation of 14.5%.
The numerical-to-experimental ratios of the load
capacities for the beams strengthened in flexure and in
shear are depicted in Figures 4(b) and 4(c), respectively.
The experimental results of 25 flexurally-strengthened
beams are used to assess the validity of the finite
element model (Brena et al. 2003; Ross et al. 1999;
Kaminska and Kotynia 2000; Ritchie et al. 1991;
MBazaa 1995; Chicoine 1997). In addition, 15 FRP
shear-strengthened beams are considered (Adhikary
and Mutsuyoshi 2004; Pellegrino and Modena 2002;
Khalifa and Nanni 2000; Chaallal et al. 1998). As can
be seen in these figures, there is a fairly good agreement
between the numerical predictions of the ultimate load
capacities and the experimental data. The average
numerical-to-experimental load capacity ratio for the
flexurally-strengthened beam applications is 100.3% with
a standard deviation of 6.5%. The corresponding values
of the average and the standard deviation in the case of
the FRP shear-strengthened beams are 102% and 2.31%,
respectively, thus indicating an excellent agreement.
The numerical analysis can be applied to arbitrary
configurations for FRP-strengthened slabs, and can also
accommodate both passive as well as prestressed FRP
strengthening schemes. In Figure 4(d) the comparison is
made for the slab applications, showing a very good
agreement with an average numerical-to-experimental
load capacity ratio of 97% with a standard deviation of
7.3% for the 12 specimens tested by Mosallam and
Mosalam (2003), Harajli and Soudki (2003), Ebead and
Marzouk (2004) and Longworth et al. (2004).
4.2. LoadDeflection Relationships
and Failure Modes
The proposed models are able to simulate the entire
loaddeflection relationships, including the descending
and post failure profiles in view of the displacementcontrolled solution adopted in these analyses. The
numerical results shown in Figures 5(a) to 5(c) are for
the numerical versus experimental comparisons in terms
of the loaddeflection relationships for selected beam
specimens (MBazaa 1995; Brena et al. 2003 and
Chicoine 1997). Debonding of the FRP laminates off the
concrete surface caused most of the failures that were
observed experimentally. As can be seen from Figure 5(a) for
Specimen P0 tested by MBazaa (1995), we were able not
only to capture the debonding load but also the complete
post-debonding plateau until complete failure. Generally
speaking, experimental post-failure measurements of the
loads and the associated deflections are not very
accurate. This explains the discrepancy in the
755
ASE 9-6-03_Neale
4/12/06
5:10 pm
Page 756
1.6
1.4
1.4
1.2
1
1
0.8
Tan (2002)
Zhao et al. (2000)
Takeo et al. (1997)
Ren (2003)
Bizindavyi and Neale (1999)
Chajes et al. (1996)
0.6
0.4
0.2
0
0
20
40
60
Specimen reference
80
Pnum /Pexp
Pnum /Pexp
1.2
0.8
Brena (2003)
Ross (1999)
Kaminska and Kotynia (2000)
Ritchie (1991)
M'Bazaa (1995)
Chicoine (1997)
0.6
0.4
0.2
0
0
100
10
15
20
25
30
Specimen reference
(a) Direct shear
(b) Beams in flexure

1.2
1.6
1
1.4
0.8
1
0.8
Adhikary and Mutsuyoshi (2004)
Pelligrino and Modena (2002)
Khalifa and Nanni (2000)
Chaallal et al. (1998)
0.6
0.4
0.2
Pnum /Pexp
Pnum /Pexp
1.2
0.6
Mosallam & Mosalam (2003)
Harajli & Soudki (2003)
Ebead & Marzouk (2004)
Longworth et al. (2004)
0.4
0.2
0
0
0
5
10
Specimen reference
15
8
10
6
Specimen reference
12
14
(d) Slabs
(c) Beams in shear
Figure 4. Ratios of Pnum/Pexp for the various applications
post-failure region for Specimen P3 in the work of

Chicione (1997) (Figure 5(c)), for example, in which the
failure occurred due to the rupture of the FRP (the failure
mode predicted in our analysis).
Figures 6(a) and 6(b) present the numerical versus
experimental comparisons of the loaddeflection
relationships for the shear-strengthened specimens
tested by Pellegrino and Modena (2002) and Adhikary
and Mutsuyoshi (2004), respectively. The specimens
tested by Pellegrino and Modena used side-bonded FRP
laminates,
whereas
the
specimens
tested
by Adhikary and Mutsuyoshi were strengthened with
U-shaped FRP laminates. An excellent agreement is
observed between our predictions and the experimental
results. The predicted failure mode in these analyses is
the delamination of the FRPs as the bond interface failed
in shear; this is what was observed experimentally.
With regard to the slabs, the comparisons are depicted
in Figure 7(a) for Specimens SA1F5 and SA1F10 tested
by Harajli and Soudki (2003), and in Figure 7(b) for
Specimens B2-SL1 (reference specimen) and B2-SL4
(using prestressed FRP laminates) tested by Longworth et
al. (2004). Specimen SA1F10 had a larger laminate width
(100 mm) compared to that of the Specimen SA1F5
756
(50mm). When increasing the FRP width, the ductility of

the strengthened slab is reduced. As seen in Figure 7(b), an
excellent agreement is obtained between our predictions
and the experimental data. We successfully simulated the
case of prestressed FRP laminates for Specimen B2-SL4,
and were able to predict the post-peak behaviour for all
the specimens. The failure modes predicted numerically
for all the slabs were in agreement with what was
reported in experiments. A more detailed discussion on
the modes of failure, as determined from the slip profiles
at the FRP/concrete interfaces, are presented below.
4.3. Interfacial Shear Stresses and Slip Profiles
A distinct advantage of having reliable numerical tools
is that they can provide valuable insight into phenomena
that are very difficult to assess experimentally. For
example, knowing the values of the interfacial stresses and
slips between the bonded FRPs and concrete can be very
helpful for a better understanding of the FRP/concrete
interfacial behaviour and bond performance. In the
following discussions, the capabilities of our models to
capture FRP/concrete interfacial stresses and/or slips
(and thus debonding behaviour) for the various
applications are illustrated.
ASE 9-6-03_Neale
4/12/06
5:10 pm
Page 757
120
160
100
140
Debonding
Load (kN)
Load (kN)
120
80
60
40
Control-exp
Control-num
Po-exp
Po-num
20
100
Control-exp
Control-num
A1-exp
A1-num
A2-exp
A2-num
A3-exp
A3-num
80
60
40
20
0
0
20
40
60
80
100
10
20
30
40
Deflection (mm)
Deflection (mm)
(a) M'Bazaa (1995)
(b) Brena et al. (2003)
50
60
180
Rupture
160
Debonding
Load (kN)
140
120
100
80
Po-exp
Po-num
P1-exp
P1-num
P3-exp
P3-num
60
40
20
0
0
20
40
60
80
Deflection (mm)
(c) Chicoine (1997)
Figure 5. Loaddeflection relationships for beam specimens (flexural strengthening)

200
400
Applied load (kN)
Applied load (kN)
500
300
TR30D3-exp
TR30D3-num
TR30D4-exp
TR30D4-num
TR30D2-exp
TR30D2-num
200
100
150
100
B-8-exp
50
B-8-num
0
0
12
16
12
Deflection (mm)
Deflection (mm)
(a) Pellegrino and Modena (2002)
(b) Adhikary and Mutsuyoshi (2004)
16
Figure 6. Loaddeflection relationships for beam specimens (shear strengthening)
4.4. Interfacial Stress Distributions

The interfacial shear stresses play an important role in
the debonding of the FRP laminates. The numerical
results presented here are for the direct shear application
for the specimens tested by Chajes et al. (1996) and the
FRP-strengthened beams; they illustrate the interfacial
behaviour before and after cracking.
4.4.1. Direct shear application

As shown in Figures 8(a)8(d), there is a certain trend in
the shear stress distributions along a bonded joint. At a
certain load level, the shear stress reaches a peak value
near the loaded end, after which the stress values descend
progressively. As the load is further increased, the
location of the peak values on the shear stress curves
757
ASE 9-6-03_Neale
4/12/06
5:10 pm
Page 758
70
250
60
200
Load (kN)
Load (kN)
50
40
SA1F10-exp
30
SA1F10-num
20
150
100
SA1F5-exp
10
B2-SL1-exp
B2-SL1-num
B2-SL4-exp
B2-SL4-num
50
SA1F5-num
0
0
10
15
20
20
40
60
80
100
Deflection (mm)
Deflection (mm)
(b) Longworth et al. (2003)
(a) Harajli and Soudki (2003)
1 kN
2.28 kN
3.56 kN
4.84 kN
6.12 kN
7.4 kN
2
1
0
Interfacial shear stress (MPa)
10
20
30
40
50
1 kN
3.2 kN
5.4 kN
7.6 kN
9.8 kN
12 kN
2
1
0
0
60
20
40
60
80
100
Distance along the interface (mm)
(a) Bonded length = 50 mm (S1)
(b) Bonded length = 100 mm (S2)
1 kN
3.2 kN
5.4 kN
5
7.6 kN
9.8 kN
12 kN
2
1
0
0
50
100
150
200
(c) Bonded length = 150 mm (S3)
Figure 7. Loaddeflection relationships for slab specimens
120
1kN
3.24 kN
5.48 kN
7.72 kN
9.96 kN
12.2 kN
2
1
0
0
50
100
150
200
250
(d) Bonded length = 200 mm (S4)
Figure 8. Interfacial shear stress profiles
moves towards the unloaded end of the laminate. The

decreased values of the shear stress, whether post-peak at
a certain load level or pre-peak for the subsequent load
levels, indicate a weakening due to the cracking in these
zones. On the other hand, the attainment of a peak in itself
indicates that the load is being transferred starting at that
point forward at that particular load value. The stress
profile along the interface is such that the transfer length
758
is reached after which no further load is transferred to the

concrete, thus indicating the onset of failure.
From Figure 8(a), it can be concluded that the bond
length for Specimen S1 (50mm) is not sufficient to
transfer the load from the FRP laminate to the concrete.
This is obvious since the values of the shear stress at the
unloaded end are still high and since the slope of the
descending part of the curve is very mild. The slope of
ASE 9-6-03_Neale
4/12/06
5:10 pm
Page 759
1.2
maximum value occurring at the plate end. The interfacial

shear stress distribution before cracking is similar to that
observed for the direct shear tests, where the maximum
shear stress occurs at the plate cut-off point.
C. L
0.8
b
a
0.6
0.4
0.2
0
0.2 0
200
400
600
c
800
1000
b
1200
1400
P = 30% Pcr
P = 60% Pcr
P = Pcr
Distance along the interface (mm), l
0.4
0.6
0.8
Figure 9. Interfacial shear stress distribution at cracking load for

Specimen A3 (Brena et al. 2003)
the descending part of Specimen S2 (Figure 8(b)) is

steeper than that of S1, resulting in low stress values at
the fixed end. Relatively low shear stress values are
observed for Specimens S3 and S4 (Figures 8(c), 8(d)),
where the bonded lengths are obviously sufficient to
transfer the stresses. The shear stress profiles can be
used to determine the transfer lengths of the different
specimens (Chajes et al. 1996).
4.4.2. Flexurally-strengthened beams
before cracking
Figure 9 depicts the shear stress distribution along the
interface for an FRP-strengthened beam tested by Brena
et al. (2003). At stress levels that are less than the
cracking stress, the interfacial shear stress concentration
occurs at the plate end and descends to a constant value
within a short length ranging from 1.5 to 2.5% of the
laminate length measured from the beam centre-line.
With an increase of the applied stress up to the cracking
stress, the interfacial shear stress linearly increases with a
4.4.3. Flexurally-strengthened beams

after cracking
The crack initiation and propagation mainly controls the
interfacial behaviour, while the crack width causes an
abrupt slip at the interface. Hence, the slip value is
controlled by several factors; namely, the crack width,
spacing, and number of cracks. The interfacial shear
stress fluctuates from negative to positive values, as can
be seen in Figures 10(a) and 10(b). With an increase of
the applied load up to the yield load, the flexural cracks
tend to open; therefore, the maximum bond stress shifts
from the plate-end to the mid-span where the flexural
cracks have propagated (Figure 10(a)). At the failure
load and close to it, the flexural cracks become wider and
the shear cracks occur near the plate end. As a result, the
shear stress dramatically increases in the regions of the
flexural cracks near the concentrated load and at the
vicinity of the shear cracks close to the plate end, thus
causing an end-plate debonding (Figure 10(b)).
4.5. Slip Profiles Along the FRP/Concrete
Interface
4.5.1. Direct shear application
Figures 11(a)11(d) show the slip profiles along the
FRP/concrete interfaces; these are important quantities
that are difficult to measure directly in the laboratory.
The slip values are consistent with the overall
behaviour described earlier in terms of the shear stress
transfer from the FRP laminate to the concrete. The slip
distributions show that, at failure, the slip rapidly
propagates causing debonding between the FRP and
concrete. It is noticed that, when the maximum shear
8
C. L
3
c
b
a
2
1
0
0
200
400
600
c
800
a
1000
b
1200
1
2
3
1400
6
c
b
a
4
2
0
0
200
400
600
c
800
1000
b
1200
1400
2
4

4
C. L

6
(a) At yield load
(b) At failure load
Figure 10. Interfacial shear stress distributions for Specimen A3 (Brena et al. 2003)
759
ASE 9-6-03_Neale
4/12/06
5:10 pm
Page 760
1 kN
2.28 kN
3.56 kN
4.84 kN
6.12 kN
7.4 kN
Interfacial slip (mm)
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
0
10
20
30
40
50
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
1 kN
3.2 kN
5.4 kN
7.6 kN
9.8 kN
12 kN
60
120
(b) Bonded length = 100 mm (S2)
1 kN
3.2 kN
5.4 kN
7.6 kN
9.8 kN
12 kN
0.2
0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
1 kN
3.24 kN
5.48 kN
7.72 kN
9.96 kN
12.2 kN
0.25
(a) Bonded length = 50 mm (S1)
20
40
60
100
80
0.2
0.15
0.1
0.05
0
50
100
150
200
(c) Bonded length = 150 mm (S3)
50
100
150
200
250
(d) Bonded length = 200 mm (S4)
Figure 11. Interfacial slip profiles for the direct shear application
200
188 kN
291 kN
332 kN
410 kN
200
100
Beam depth (mm)
Beam depth (mm)
300
0
0.0
0.4
0.8
Slip (mm)
1.2
1.6
(a) Pellegrino and Modena (2002)

side-bonded
150
100
59 kN
109 kN
150 kN
50
166 kN
0
0.4
0.0
Slip (mm)
0.4
(b) Adhikary and Mutsuyoshi (2004)

U-shaped
Figure 12. Interfacial slip profiles along the FRP laminate depth for shear-strengthened beams
stress is reached, an abrupt increase in the slip values

occurs causing the debonding of the FRP laminates.
4.5.2. Shear-strengthened beams
The numerical results shown in Figures 12(a) and 12(b)
are for the slip profiles along the beam depth for a sidebonded shear-strengthened beam tested by Pellegrino
and Modena (2002) and a U-shaped FRP configuration
tested by Adhikary and Mutsuyoshi (2004). The slip
760
profiles depicted in these figures correspond to load

increments up to failure for a section midway in the
shear span. As seen in Figure 12(a) for the side bonded
beam, the predicted slip values are significantly higher
at the bottom of the beam than those at the top. This
suggests that this specimen will experience debonding
of the FRP laminate at the bottom edge of the beam,
which is in accordance with the experimental results.
For the specimen strengthened using U-shaped FRP
ASE 9-6-03_Neale
4/12/06
5:10 pm
Page 761
0.7
0.40
0.25
0.5
slip (mm)
Slip (mm)
0.30
P = 206.5 kN
P = 275.4 kN
P = 303.0 kN
P = 316.7 kN
P = 330.5 kN
0.6
P = 26.4 kN
P = 39.6 kN
P = 52.8 kN
P = 58.0 kN
P = 66.0 kN
0.35
0.20
0.15
0.4
0.3
0.2
0.10
0.1
0.05
0.00
0a
100
200
300 b
400
Distance from slab end towards the centre (mm)
0
0
b
200
400
600
800
1000
(a) Specimen SA1F15
(b) Specimen GFRP-F-035%
0.45
0.40
Zones of intersecting FRP
0.35
P = 11.1 kPa
P = 27.0 kPa
P = 45.8 kPa
P = 58.0 kPa
Slip (mm)
0.30
0.25
0.20
0.15
0.10
0.05
0.00
0 a
200
400
600
800
1000
1200 b1400
(c) Specimen C-REP-R1
Figure 13. Slip profiles for Specimens SA1F15, GFRP-F-035% and C-REP-R1
wraps, the interfacial slips are higher at the top edge and
shift to negative values around the mid-depth. With an
increase of the applied shear force up to failure, the
interfacial positive and negative values of slip are
increased. It is thus obvious that using U-shaped FRP
wraps is more efficient in mitigating debonding, as can
be seen from comparing these slip values with those for
the specimen with side-bonded laminates.
4.5.3. Two-way slabs
The predicted slip profiles along the FRP/concrete
interfaces, at different load levels, are shown in Figure 13
for the Specimen SA1F15 tested by Harajli and Soudki
(2003), Specimen GFRP-F-0.35% tested by Ebead and
Marzouk (2004), and Specimen C-REP-R1 tested by
Mosallam and Mosalam (2003).
In Figure 13(a) it can be observed that the slip values
at the centre of the slab are significantly higher than
those at the end of the laminate. This suggests a punching
shear failure associated with debonding of the FRP
material in the central region of the slab. By contrast, the
slip profiles for Specimen GFRP-F-0.35% indicate that
the main failure mode is debonding of the FRP plate. The

profiles are such that the maximum slips occur near the
centres of the slabs, and decrease monotonically towards
the ends. This implies that failure should correspond to
debonding initiating in the central zone, and propagating
to the ends of the laminate. This is indeed what was
reported in the experiments.
A comparison of the results shown in Figure 13(a)
(Harajli and Soudki 2003) against those found in Figure
13(b) (Ebead and Marzouk 2004) shows that the predicted
slip values for the slabs tested by Harajli and Soudki are
less than those for the specimens of Ebead and
Marzouk. This is simply an indication that the former
slabs failed in punching shear while the latter were
characterized by a more typical flexural mode of
failure.
In the predicted slip profiles of Specimen C-REF-R1
tested by Mosallam and Mosalam (2003) (Figure
13(c)), there exist certain zones where the slips
decrease relative to those in the adjacent regions.
These, in fact, correspond to the areas where the FRP
laminates overlap, and where one would indeed expect
761
ASE 9-6-03_Neale
4/12/06
5:10 pm
Page 762
reductions in the interfacial shear stresses and slips.

This result suggests that, in general, the addition of
transverse anchorage strips at the ends of the FRP
laminates should be quite effective to mitigate
debonding failures in these regions.
5. CONCLUSION
Finite element analyses have been carried out to address
the interfacial behaviour of FRP strengthened reinforced
concrete structures. A nonlinear constitutive model was
incorporated to represent the interfacial behaviour
between the FRP laminates and concrete. In order to
investigate the validity of the numerical model, the
finite element results were evaluated against available
experimental data. The comparisons between the
numerical predictions and the available test results
showed very good correlations in terms of the ultimate
carrying capacities, loaddeflection relationships, and
failure modes. Our studies have shown that accounting
for bondslip characteristics at the FRP/concrete
interface is viable, and that the model proposed by
Lu et al. (2005a) leads to excellent results.
ACKNOWLEDGEMENTS
This research was funded by the Natural Sciences and
Engineering Research Council of Canada (NSERC)
and the Canadian Network of Centres of Excellence on
Intelligent Sensing for Innovative Structures (ISIS
Canada). KWN is Canada Research Chair in Advanced
Engineered Material Systems, and the support of this
program is gratefully acknowledged.
REFERENCES
Adhikary, B. and Mutsuyoshi, H. (2004). Behaviour of concrete
beams strengthened in shear with carbon-fibre sheets,
Journal of Composites for Construction, ASCE, Vol. 8, No. 3,
pp. 258264.
ADINA (2004a). Automatic Dynamic Incremental Nonlinear
Analysis, Finite Element Software, Version 8.2. ADINA R&D
Inc., Watertown, MA, USA.
ADINA (2004b). Theory and Modeling Guide, Vol. I, Chapter 3,
Version 8.2. ADINA R&D Inc., Watertown, MA, USA.
Al-Mahaidi, R., Lee, K. and Taplin, G. (2001). Behaviour and
Analysis of RC T-Beams Partially Damaged in Shear and Repaired
with CFRP Laminates, Internal Report, Structural Engineering,
Monash University, Australia, 8p.
Bizindavyi, L. and Neale, K.W. (1999). Transfer lengths and bond
strengths for composites bonded to concrete, Journal of
Composites for Construction, ASCE, Vol. 3, No. 4, pp. 153159.
Brena, S.F., Bramblett, R.M., Wood, S.L. and Kreger, M.E. (2003).
Increasing flexural capacity of reinforced concrete beams using
carbon fibre-reinforced polymer composites, ACI Structural
Journal, Vol. 100, No. 1, pp. 3646.
762
Chaallal, O., Nollet, M. and Perraton, D. (1998). Shear

strengthening of reinforced concrete beams by externally bonded
side CFRP strips, Journal of Composites for Construction,
ASCE, Vol. 2, No. 2, pp. 111113.
Chajes, M.J., Finch, W.W., Jaunszka, T.F. and Thomson, A. (1996).
Bond and force transfer for composite material plates bonded to
concrete, ACI Structural Journal, Vol. 93, No. 2, pp. 208217.
Chicoine, T. (1997). Conception et analyse dancrages sur les poutres
renforces en flexion laide de matriaux composites, M.A.Sc.
Thesis, University of Sherbrooke, Sherbrooke, Quebec, Canada.
Ebead, U.A. and Marzouk, H. (2004). Fiber-reinforced polymer
strengthening of two-way slabs, ACI Structural Journal, Vol.
101, No. 5, pp. 650659.
Ebead, U.A. and Marzouk, H. (2005). Tension-stiffening model for
FRP-strengthened concrete two-way slabs, Journal of Material
and Structures/Materiaux et Constructions, RILEM, Paris, Vol.
38, pp. 193200.
Ehsani, M.R. and Saadatmanesh, H. (1990). Fibre composite plates
for strengthening bridge beams, Composite Structures, Vol. 15,
No. 4, pp. 343355.
Erki, M.A. and Heffernan, P.J. (1995). Reinforced concrete slabs
externally strengthened with fibre-reinforced plastic materials,
Proceedings of the Second International RILEM Symposium,
FRPRCS-2, L. Taerwe, Ed., University of Ghent, Ghent,
Belgium, pp. 509516.
Harajli, M.H. and Soudki, K.A. (2003). Shear strengthening of
interior slab-column connections using carbon fibre-reinforced
polymer sheets, Journal of Composites for Construction, ASCE,
Vol. 7, No. 2, pp. 145153.
Kachlakev, D. and Miller, T. (2001). Finite Element Modeling of
Reinforced Concrete Structures Strengthened with FRP
Laminate, Final Report SPR 316, Federal Highway
Administration, Washington DC, USA, 99 pp.
Kaminska, M.E. and Kotynia, R. (2000). Experimental Research on
RC Beams Strengthened with CFRP Strips, Technical Report,
ISSN. 12306010, Department of Concrete Structures, University
of Lodz, Poland.
Khalifa, A. and Nanni, A. (2000). Improving shear capacity of
existing RC T-section beams using CFRP composites, Cement
and Concrete Composites, Vol. 22, No. 3, pp. 165174.
Limam, O., Foret, G. and Ehrlacher, A. (2003). RC two-way slabs
strengthened with CFRP strips: experimental study and a limit
analysis approach, Composite Structures, Vol. 60, No. 4, pp.
467471.
Longworth, J., Bizindavyi, L., Wight, R.G. and Erki, A. (2004).
Prestressed CFRP sheets for strengthening two-way slabs in
flexure, Proceeding of the Forth International Conference on
Advanced Composite Materials in Bridges and Structures, M. ElBadry, and L. Dunaszegi, eds., Canadian Society for Civil
Engineering, Calgary, July, pp. 146.
Lu, X.Z., Teng, J.G., Ye, L.P. and Jiang, J.J. (2005a). Bondslip
models for sheets/plates bonded to concrete, Engineering
Structures, Vol. 27, No. 6, pp. 938950.
ASE 9-6-03_Neale
4/12/06
5:10 pm
Page 763
Lu, X.Z., Jiang J.J., Teng, J.G. and Ye, L.P. (2006). Finite element
simulation of debonding in FRP-to-concrete bonded joints,
Construction and Building Materials, Vol. 20, No. 6, pp. 412424.
Malek, A. and Saadatmanesh, H. (1998a). Ultimate shear capacity
of reinforced concrete beams strengthened with web-bonded fibre
reinforced plastic plates, ACI Structural Journal, Vol. 95, No. 4,
pp. 391399.
Malek, A. and Saadatmanesh, H. (1998b). Analytical study of
reinforced concrete beams strengthened with web-bonded
fibre reinforced plastic plates or fabrics, ACI Structural Journal,
Vol. 95, No. 3, pp. 343352.
MBazaa, I. (1995). Renforcement en flexion de poutres en bton
arm a laide de lamelles en matriaux composites: Optimisation
de la longueur des lamelles, M.A.Sc. Thesis, University of
Sherbrooke, Sherbrooke, Quebec, Canada.
Mosallam, A.S. and Mosalam, K.M. (2003). Strengthening of twoway concrete slabs with FRP composite laminates, Construction
and Building Materials, Vol. 17, No. 1, pp. 4354.
Nitereka, C. and Neale, K.W. (1999). Analysis of reinforced concrete
beams strengthened in flexure with composite laminates,
Canadian Journal of Civil Engineering, Vol. 26, pp. 646654.
Pellegrino, C. and Modena, C. (2002). Fibre reinforced polymer
shear strengthening of reinforced concrete beams with transverse
steel reinforcement, Journal of Composites for Construction,
ASCE, Vol. 6, No. 2, pp. 104111.
Ren, H.T. (2003). Study on Basic Theories and Long Time Behavior
of Concrete Structures Strengthened by Fiber Reinforced
Polymers, Ph.D. Thesis, Dalian University of Technology, China.
Ritchie, P.A., Thomas, D., Lu, L-W. and Connelly, G. (1991).
External reinforcement of concrete beams using fibre
reinforced plastics, ACI Structural Journal, Vol. 88, No. 4, pp.

490500.
Ross, C.A, Jerome, M.D., Tedesco, J.W. and Hughes, M.L. (1999).
Strengthening of reinforced concrete beams with externally
bonded composite laminates, ACI Structural Journal, Vol. 96,
No. 2, pp. 212219.
Seim, W., Horman, M., Karbhari, V. and Seible, F. (2001). External
FRP post-strengthening of scaled concrete slabs, Journal of
Takahashi, Y., Sato, Y., Ueda, T., Maeda, T. and Kobayashi, A.
(1997). Flexural behaviour of RC beams with externally bonded
carbon fibre sheet, Non-Metallic (FRP) Reinforcement for
Concrete Structures, Japan Concrete Institute, Vol. 1, pp. 327334.
Takeo, K., Matsushita, K., Makizumi T. and Nagashima, G. (1997).
Bond characteristics of CFRP sheets in the CFRP bonding
technique, Proceedings of Japan Concrete Institute, Vol. 19,
No. 2, pp. 15991604.
Tan, Z. (2002). Experimental Research for RC Beam Strengthened
with GFRP. M.A.Sc Thesis, Tsinghua University, Beijing, China.
Thomsen, H., Spacone, E., Limkatanyu, S. and Camata, G. (2004).
Failure mode analyses of reinforced concrete beams strengthened
in flexure with externally bonded fibre-reinforced polymers,
Journal of Composites for Construction, ASCE, Vol. 8, No. 2, pp.
123131.
Wong, R. and Vecchio, F. (2003). Towards modeling of reinforced
concrete members with externally bonded fibre-reinforced polymer
composites, ACI Structural Journal, Vol. 100, No. 1, pp. 4755.
Zhao, H.D., Zhang, Y. and Zhao, M. (2000). Research on the bond
performance between CFRP plate and concrete, Proceedings of the
First Conference on FRP-Concrete Structures of China, pp. 247253.
763
ASE 9-6-03_Neale
4/12/06
5:10 pm
Page 764
ASE 9-6-04_Frostig
4/12/06
5:10 pm
Page 765
Fracture Mechanics Approach to Geometrically

Nonlinear Debonding Problems in RC Beams
Strengthened with Composite Materials
Oded Rabinovitch* and Yeoshua Frostig
Faculty of Civil and Environmental Engineering, Technion Israel Institute of Technology, Haifa, 32000, Israel
Abstract: The geometrically nonlinear debonding mechanism in reinforced concrete

beams strengthened with externally bonded FRP strips is analytically investigated. The
paper focuses on cases in which various loading scenarios yield compressive stresses
in the bonded strip. Under these conditions, the FRP strip tends to buckle, and, in turn,
to trigger and accelerate the unstable growth of the debonded region. A high order
geometrically nonlinear model of the strengthened beam is developed and combined
with the fracture mechanics concept of the energy release rate to provide a quantitative
criterion for the initiation, propagation, and stability or instability of the debonding
growth process. Numerical results that demonstrate the capabilities of the proposed
modeling approach and throw light on some of the physical phenomena associated
with the structural response of the strengthened beam are presented. The analysis
reveals the response through the loading process and through the debonding growth
process, highlights the localized effects near the debonded region, quantifies its
tendency to grow, and describes the stability characteristics of the debonding mode of
failure. A summary and conclusions are included in the sequel.
Key words: Analytical modelling, debonding, energy release rate, fracture mechanics, FRP, geometrically nonlinear
analysis, RC beams, strengthening.
1. INTRODUCTION
Strengthening, upgrading, and retrofitting of existing
structures with externally bonded fiber reinforced polymer
(FRP) composite materials have gained a worldwide
acceptance in the past two decades (Neale and
Laboissiere 1997; Bakis et al. 2002). The versatility of
the strengthening method, its applicability to a wide
range of structural problems, the ease of installation
of the composite systems on site, and the improved
mechanical properties of the composite materials are
among the main advantages of the method. On the other
hand, it is also characterized by the formation of brittle
modes of failure, and, especially, by the development
of unstable and sudden debonding failure modes
(Buyukozturk and Hearing 1998). The two most dominant
debonding modes of failure are illustrated in Figure 1.

The first mode includes debonding that initiates at the
edge of the composite material strip and rapidly
propagates towards mid-span, see Figure 1(a). The
second characteristic debonding mode of failure is
governed by the formation and the unstable growth of a
debonded region along the strengthened beam. Such
regions may result from flexural and flexure-shear
cracking of the concrete beam (Figure 1(b)), from
localized damage to the bonding layer, or from
improper placement and fastening of the FRP strip
(Figure 1(c)). In all cases, the debonding failure is
triggered and governed by the stress concentrations
near the cut-off section (Rabinovitch and Frostig 2000;
Lau et al. 2001; Smith and Teng 2001, 2002a, 2002b;
*Corresponding author. Email address: cvoded@tx.technion.ac.il; Fax: +972-4-8295697; Tel: +972-4-8293047.
765
ASE 9-6-04_Frostig
4/12/06
5:10 pm
Page 766
Fracture Mechanics Approach to Geometrically Nonlinear Debonding Problems in RC Beams Strengthened with Composite Materials
Debonding growth
Unbonded regions
Debonding Shear flexure cracks
(a)
(b)
(c)
Figure 1. Debonding failure in RC beams strengthened with composite materials: (a) Edge debonding; (b) Debonding near shear-flexure
cracks; (c) Voids and debonded regions in a peeled-off FRP strip
Teng et al. 2002; Buyukozturk et al. 2004) or near the

debonded regions throughout the beam (Rabinovitch
and Frostig 2001a). Furthermore, the development of
the debonded regions, and, especially, their rapid
growth process, are considerably accelerated by the
formation of compressive stresses in the FRP strip.
Externally bonded FRP systems are generally
designed to provide the existing structure with
supplementary tensile reinforcement. As such, they are
designed to carry tensile stresses while their use under
compressive stresses is not recommended (ACI 440.2R02 2002), CEB-FIP (2001)). However, in some cases,
the exposure of the FRP laminate to compressive
stresses is inevitable. Among these cases, moment and
curvature reversal under different loading cases; cyclic,
sign-reversing, and dynamic loads; combined axial and
flexural loading; unloading of members strengthened
while carrying loads; and strengthening on two faces of
the element are the most common ones (see Rabinovitch
2004a). In these cases, the development of compressive
stresses in the FRP layer in a debonded region may
trigger a localized buckling or wrinkling of the thin FRP
strip near the debonded region (ACI 440.2R-02 2002).
Although the contribution of the compressed FRP strip
can be neglected for design purposes, the formation of
the localized buckling phenomena and the unstable
growth of the debonded region may lead to separation of
the bonded FRP strip, to failure of the strengthening
system, or even to failure of the entire element. Hence,
the local buckling of the FRP strip and the tendency of
the debonded region to grow critically affect the
functionality of the strengthening system and the safety
of the strengthened element.
The strengthening and upgrading of existing
structures with externally bonded FRP systems have
attracted the attention and research efforts in the past
years. However, the number of research works that
focus on the performance of the FRP system under the
inevitable development of compressive stresses is rather
766
limited. The behavior of RC beams and columns

strengthened with FRP sheets bonded to the side faces
of the element and subjected to combined bending and
compression were studied by Uy et al. (2002). The
experimental observations revealed that in spite of
microbuckling, air voids in the resin materix, and
misalignment of fibers, the bonded composite system
may develop compressive stresses. Yet, these localized
phenomena affect the compressive strength and the
equivalent elastic modulus of the FRP system, which are
reduced by about 25%-35% with respect to the
unidirectional tensile properties. The numerical finite
strip analyses conducted by Uy et al. (2002) addressed
the possibility of local buckling of the bonded FRP
system. However, due to the use of presumed stress
patterns, the unclear stress transfer mechanism between
the bonded FRP plate and the concrete element, and the
use of anisotropy calibration factors, the completeness
and the generality of this model are limited.
A linear and a geometrically nonlinear high order
models for the stress analysis of the strengthened beam
were presented by Rabinovitch and Frostig (2000) and
Rabinovitch (2004a). However, these models did not
consider the initiation and growth of the debonded region.
A combined high-order and fracture approach was used
by Rabinovitch and Frostig (2001a) and Rabinovitch
(2004b) for the case of edge debonding modes and for the
analysis of debonding modes driven by cracking of the
RC beam. The buckling-driven debonding mode, which
requires a comprehensive geometrically nonlinear and
fracture analyses, was not studied.
Deuring (1993) tested a reinforced concrete (RC)
beam with FRP laminates bonded in the compression
zone. In this test, the beam was designed to fail due to
yielding of the tension steel, thus buckling of the
laminate was triggered by failure of the concrete in
compression. (also see CEB-FIP 2001). However, the
post failure mechanism revealed the propagation of a
debonding mechanism governed by the post buckling
ASE 9-6-04_Frostig
4/12/06
5:10 pm
Page 767
Oded Rabinovitch and Yeoshua Frostig
behavior of the laminate. Another experimental study on

RC beams strengthened with steel plates bonded to the
compression face of bending elements is presented by
Mohamed Ali et al. (2002). In this study, the externally
bonded tensile reinforcement is extended into the
compression region or initially placed on the upper
compressed face of the beams. The experiments revealed
the cracking and debonding characteristics of the
strengthened beam whereas buckling of the external
reinforcement was prevented by using relatively thick and
stiff steel plates and stiff adhesives. Grace et al. (1999)
experimentally studied continuous beams strengthened in
the positive and negative moment regions. While in this
case the strengthening system has not been extended into
the compression zone, the effects associated with
development of compression forces in the bonded strip
due to variable load cases have not been addressed.
In terms of codes, standards, and design guides, ACI
440.2R-02 (2002) strongly discourages the use of
externally bonded FRP systems to resist compressive
forces. However, it does not address the cases where the
development of compressive stresses in the FRP system
is inevitable. Notice that the draft of the design
guidelines for RC structures retrofitted with FRP and
metal plates proposed by Oehlers (2005) addresses the
problem of buckling of the compressed external
reinforcement. In the cases of adhesively bonded
reinforcement, it assumes that the adhesive layer and the
stiff concrete substrate are sufficient to support the FRP
layer and to avoid buckling. However, the problem of
buckling in the vicinity of the debonded region, and the
debonding growth process triggered by the localized
buckling are not addressed. CEB-FIP (2001) recognizes
the possibility of the formation of compressive stresses
in FRP system as well as the possibility of local buckling
of the thin compressed laminate. However, it does not
provide guidelines for the quantitative assessment of this
effect. Furthermore, it highlights the lack of an adequate
analytical approach that addresses this problem and the
lack of reliable design parameters for its prevention.
Here, a general analytical approach for quantifying
and characterizing the nonlinear behavior, buckling, and
buckling-driven growth of the debonded region is
developed. For that purpose, a modeling approach that
combines two fundamental analytical methodologies is
derived. The first methodology uses the high order
analytical model of the strengthened RC beam
(Rabinovitch and Frostig 2000), which is extended to
account for the geometrically nonlinear aspects of the
structural response (Rabinovitch 2004a) and for
the formation of debonded regions. The second
methodology that is employed uses the fracture
mechanics concept of the energy release rate as the
criterion for the development and stable or unstable

growth of the debonded region. The fracture mechanics
criterion is preferred over the standard stress criteria
(Varastehpour and Hamelin 1997; Malek et al. 1998;
Mukhopadhyaya and Swamy 2001) due to the brittle
character of the debonding crack (also see Rabinovitch
and Frostig 2001a).
The mathematical modeling approach and the
energy-release-rate (ERR) criteria for the growth and
stability of the debonded region are presented in the
mathematical formulation section. Emphasis is placed
on the existence of fully bonded and the debonded
regions along the beam and on the unique stress and
displacement fields in each typical region. The analytical
model uses the large displacements, moderate rotations,
and small strains kinematic relations for the mathematical
representation of the RC beam and the FRP strip.
Thus, it allows the consideration of the geometrically
nonlinear effects. The adhesive layer is modeled as a 2D
elastic medium that resists shear and vertical normal
stresses, while its inplane rigidity is neglected (see
Rabinovitch and Frostig 2000). The closed form stress
and displacement fields of the adhesive are based on
2D linear elasticity and they take into account the
compatibility or debonding conditions at the interfaces
of the adhesive layer. The beam and lamination theories
are employed for the RC beam and the FRP strip,
respectively. Finally, it is assumed that the external
loads are exerted at the RC beam only and that the stress
and deformation fields are uniform through the width of
each component.
Following the mathematical formulation, a numerical
study that highlights the geometrically nonlinear
behavior of a strengthened beam with an inner
debonding and quantitatively examines the tendency of
the debonded region to grow is presented. In the sequel,
a summary and conclusions are presented.
2. MATHEMATICAL MODEL
The mathematical formulation focuses on the
geometrically nonlinear analytical model for the
strengthened beam as well as on the formulation and
evaluation of the fracture mechanics criterion for the
growth of the debonded region. The general layout of the
strengthened beam, along with the notation for
displacements, stresses, stress resultants, and coordinate
systems, appears in Figure 2 and it reveals that the
strengthened beam is comprised of three characteristic
regions. The first one is an un-strengthened region near
the supports; the second one is a fully-bonded region,
where the adhesive-FRP and adhesive-concrete interfaces
maintain compatibility of deformations and transfer shear
and vertical normal stresses; and the third one is a
767
ASE 9-6-04_Frostig
4/12/06
5:10 pm
Page 768
x,uorc
zrc,wrc
c.g
ua
za,wa
rc
Yrc d
uofrp
Mfrp
xx
RC Beam
frp
Nxx
ca
zfrp,wfrp
Qfrp
xx
dfrp
azz(x,0)
(a)
_
Prc
j
nrc
qrc
_
Mrc
j
azz(x,ca)
axz(x,ca)
FRP Strip
x = xj
Debonded
region
axz(x,0)
Adhesive
Layer
_ rc
Nj
Mfrp
xx
Nfrp
xx
frp
Qxx
(b)
(c)
Figure 2. Notation and sign conventions: (a) Deflections and coordinates; (b) Loads; (c) Stresses and stress resultants
debonded region, where the debonded interfaces are

free of shear stresses and can slip with respect to each
other. A further distinction is made between debonded
sub-regions in which the debonded interfaces are in
contact and debonded sub-regions in which they are not in
contact. In the first case (debonding with contact)
the interface maintains compatibility of the vertical
deformations and thus accumulates vertical normal
compressive stresses while in the second case (debonding
without contact) the debonded interface is free of
stresses. In order to account for this unique topology of the
beam, the mathematical formulation makes a distinction
between the various characteristic regions and subregions. The formulation presented here focuses on the
fully bonded and debonded strengthened regions. The
model for the un-strengthened region is achieved through
degeneration of the ones developed here.
2.1. Variational Principle and Kinematic
Relations
The equilibrium equations and the boundary and
continuity conditions are derived using the variational
principle of virtual work, which equals:
(U + V) = 0
rc
xx
xxrc dvrc
v rc
frp
frp
frp
a
a
a
a
a
xx xx dv xz xx + zz zz dv
v frp
(2)
va
where the superscripts rc, frp, and a refer to the RC

beam, the FRP strip, and the adhesive layer,
respectively; ixx and ixx are the longitudinal normal
stresses and strains in the RC beam (i = rc) and the
768
u i ( x, zi ) = uoi z i w,ix ;
w i ( x, z i ) = w i ;
=u
i
xx
i
o, x
( )
1
+ w,ix
2
(3a-c)
2
z w
i
i
, xx
where wi and uio are the vertical and longitudinal

displacements at the reference line of the RC beam (i=rc)
or the FRP strip (i=frp), zi is measured from the reference
line of each component downwards (Figure 2(a)), and
( ),x denotes a partial derivative with respect to x.
The kinematic relations for the adhesive are based on
2D linear elasticity as follows:
(1)
where U is the strain energy, V is the potential of the

external loads, and is the variational operator. The first
variation of the strain energy is:
U =
FRP strip (i = frp); axz and azz are the shear and vertical
normal stress fields in the adhesive layer, respectively;
axz and azz are the shear angle and the vertical normal
strain fields in the adhesive layer, respectively; and vi
(i = rc,a,frp) is the volume of each component.
The kinematic relations for the RC beam and the
FRP strip independently follow the Bernoulli-Euler
assumption and the theory of large displacements,
moderate rotations, and small strains as follows:
zz = w,z;
a
xz = u,az + w,ax
a
(4a,b)
where wa and ua are the vertical and longitudinal

displacements of the adhesive layer and ( ),z denotes a
partial derivative with respect to z.
The first variation of the potential of the external
loads equals:
x = L
rc
V= ( nrc uorc + qrc wrc + m rc w , x )
x = 0
NC rc x = L
dx +
j =1 x = 0
rc
( N j uorc
rc
P j wrc
rc
Mj
rc
, x ) D ( x
x j )dx
(5)
ASE 9-6-04_Frostig
4/12/06
5:10 pm
Page 769
where qrc, nrc, and mrc are distributed loads and bending
rc
rc
rc
moments, respectively, N j , P j , and M j are
concentrated loads and bending moments at x=xj, D is
the Dirac function, and NC is the number of the
concentrated loads.
2.2. Interfacial Compatibility and Debonding
Conditions
The compatibility conditions at the fully bonded interfaces
of the adhesive layer are:
a
a
rc
rc
rc
rc rc
u ( x, z = 0 ) = u ( x, Y ) = u o Y w , x
(6)
a
rc
rc
a
w ( x, z = 0 ) = w ( x ) = w
(7)
a
a
frp
u ( x, z = c ) = u ( x,
a
d frp
d frp frp
) = u ofrp +
u
2
2 ,x
w ( x, z = c ) = w ( x ) = w
a
frp
frp
(8)
(10)
If the debonded interfaces are in contact, the conditions

of compatibility of the vertical displacements, Eqn 7 and
Eqn 9 are imposed. If vertical contact does not exist, the
debonded interface is free of shear and vertical normal
stresses. Thus, in case the adhesive-concrete interface is
debonded and contact does not exist, Eqn 6 is replaced
with Eqn 10 and Eqn 7 is replaced with the following
condition:
(11)
2.3. Equilibrium Equations

The nonlinear equilibrium equations for the strengthened
(bonded or debonded) regions are formulated using the
variational principle (Eqns 12 and 5), the kinematic
(12)
N xxfrp,x ca b xza ( x, z a = c a ) = 0
(13)
rc rc
rc
rc a
a
M xx
)
, xx + ( N xx w , x ), x + 0 bY xz , x ( x, z = 0)
+ 0 b zza ( x, z a = 0 ) = q rc + m,rcx
(14)
d frp a
( x, z a = c a )
2 xz , x
(15)
a
a
a
ca b zz ( x, z = c ) = 0
xza ,x + zza ,z = 0
(16)
xza ,z = 0
(17)
(9)
where
is the height of the center of area axis
(reference axis) in the RC beam, dfrp is the thickness of
the FRP strip, ca is the thickness of the adhesive layer,
and za is the vertical coordinate of the adhesive
layer, measured from the adhesive-concrete interface
downwards, see Figure 2(b).
In the debonded regions, the damaged interface
cannot transfer shear or vertical tensile stresses. Hence,
the requirements of compatible longitudinal deformations
are replaced with the conditions of stress free surfaces.
For example, if the adhesive-concrete interface is
debonded, Eqn 6 is replaced with:
zza ( x, z a = 0 ) = 0
N xxrc ,x + 0 b xza ( x, z a = 0 ) = n rc
M xxfrp, xx + ( N xxfrp w ,frp

x ), x + ca b
Yrc
xza ( x, z a = 0 ) = 0
relations (Eqns 34), and the interfacial compatibility/

debonding conditions (Eqns 611) and read:
where Nixx and Mixx are the in-plane stress resultant and
the bending moment in the RC beam (i = rc) and the
FRP strip (i = frp), b is the width of the adhesive layer,
i is a flag that defines the bonding condition at the
adhesive concrete interface (j = 0) or the adhesive-FRP
interface (j = ca), thus j = 1 for a fully bonded interface
and j=0 for a delaminated interface; and i is a flag that
defines the contact conditions at the adhesive concrete
interface (j = 0) or the adhesive-FRP interface (j = ca)
and equals 1 for debonding with contact and 0 for
debonding without contact. In the fully bonded region,
0 = ac = 0 = ac = 1. Thus, by using the proper
combination of flags, Eqns 1217 are valid for all
characteristic regions and sub-region.
2.4. Constitutive Relations
The constitutive relations for the RC beam are:
1
N xxrc = EArc uorc,x + ( w,rcx )2 ,
M xxrc = EI rc w,rcxx (18)
where EArc and EIrc are the longitudinal and flexural

rigidities of the beam. These rigidities are determined
bases on the properties of the un-cracked section (EAci,
EIci), the properties of the cracked section (EAr, EIr), or
an interpolation formula. A more inclusive model, in
which the rigidities are iteratively determined
considering the exact nonlinear constitutive laws for
concrete and the internal steel reinforcement, appears in
Rabinovitch and Frostig (2001b). For simplicity and
769
ASE 9-6-04_Frostig
4/12/06
5:10 pm
Page 770
brevity, these effects and their influence on the

geometrically nonlinear and crack growth behavior are
not included here.
The constitutive relations for the FRP strip use the
classical lamination theory and read:
= B (u
frp
11
frp
o,x
(19)
1
)2 ) D11i w,fxxrp
+ ( w,frp
2 x
frp
frp
frp
where A11 , B11 , D11 are the extensional, coupling, and
flexural rigidities of the FRP strip (Vinson and
Sierakowski 1986) multiplied by its width, b. Note that in
most practical cases, a symmetric (and balanced) layup,
in which the coupling rigidity vanish, is used. Hence, in
frp
the formulation presented next B11 = 0 is assumed.
The constitutive relations of the adhesive layer are:
=E ;
a
zz
a
zz
=G
a
xz
a
xz
2.5. Adhesive Layer Stress

and Displacement Fields
In the fully bonded region, the 2D elasticity
equilibrium equations of the adhesive, Eqns 1617,
along with the compatibility requirements (Eqns 67
and 9) and the kinematic and constitutive relations
(Eqns 4 and 20) yield the following stress and
displacement fields of the adhesive layer (Rabinovitch
and Frostig 2000):
( x, z ) = ( x ) =
a
a
xz
(21)
w frp w rc a
a c
( x, z ) =
E
+
za
,
x
a
c
2
a
zz
w a ( x, z a ) =
770
a 2
(22)
a a ( ) ( a)
z
6
Ga
E 4
a
, xx
a
(z )
w,frp
w,rcx w,rcx z a + uorc Y rc w,rcx
x
a
2c
0 ca E a ( w frp w rc )
( x, z ) =
ca
(26)
2.6. Governing Equations, Continuity

Conditions and Solutions
The governing equations for the strengthened (fully
bonded or debonded) regions are derived using the field
equations Eqns 1316, the constitutive relations (Eqns
1819), the compatibility requirements (Eqn 8), and the
stress and deformation fields of the adhesive layers (Eqns
2126). The governing equations are stated in terms of the
vertical deflections (wrc and wfrp), the longitudinal
deflections (uorc and uofrp) and the shear stress a as follows:
rc rc
rc rc rc
a
rc
EA u o, xx + EA w, x w, xx + 0 ca b = n
(27)
frp frp
frp frp frp
frp
a
A11 u o, xx + A11 w, x w, xx 0 ca b = n
(28)
EI rc w ,rcxxxx + EArcu rco, x w ,rcxx

+
3
rc 2 rc
rc
EA ( w, x ) w, xx
2
+ EArcu rco, x w ,rcx + 0 ca b a + Y rc ,ax
2
+ 0 ca
(24)
(29)
bEa
( w frp w rc ) = q rc
ca
3 frp frp 2 frp

A11 ( w, x ) w, xx
2
ca + d frp a
+ A11frpu ofrp, xx w ,frp
+
x
0 ca
2 ,x
frp
frp frp
D11frp w ,frp
xxxx + A11 u o, x w , xx +
0 ca
,xa z a ( z a c a ) ( w frp w rc ) z a
+
+ w rc (23)
2E a
ca
u a ( x, z a ) =
(25)
(20)
where Ga and Ea are the shear and the elastic moduli of

the adhesive in the vertical direction, respectively.
a
xz ,z
xza ( x, z a ) = 0 ;
a
zz
1
)2 ) B11i w,fxxrp
N xxfrp = A11frp (uofrp, x + ( w,frp
x
2
frp
xx
The stress fields for the adhesive layer in the debonded

regions reduce to:
(30)
bEa
( w frp w rc ) = q frp
ca
c a a (c a) 3 a
0 ca u rco u ofrp +

12 Ea , xx
Ga
c a + d frp frp
ca
+ Y rc w ,rcx
w =0
2 ,x
(31)
Note that Eqn 31 results from the compatibility

conditions of the longitudinal deformations at the
ASE 9-6-04_Frostig
4/12/06
5:10 pm
Page 771
adhesive-FRP interface (Eqn 8). This equation is valid

in the fully bonded regions only and it vanishes in the
debonded regions. Correspondingly, the shear stress
also vanishes in the debonded regions (see Eqn 25), and
the order of the governing equations reduces from 14 in
the fully bonded region to 12 in the debonded ones. This
requires a unique set of continuity conditions at the
mutual joints between the two characteristic regions. In
the RC beam (i=rc) and the FRP strip (i=frp), these
conditions require continuity of the deflections and
equilibrium of the internal resultants and the external
loads as follows:
i( )
o
=u
i( + )
o ;
i( )
=w
i( + )
i
; w
i( )
,x
=w
i( + )
,x
(32a-c)
i
i( )
i( + )
N xx N xx = N j , M ixx( ) + M ixx( + ) = M ij
(M
(M
i( )
xx , x
+ ( N ixx( )w ,ix( ) ) + b a ( )Y i
i( + )
xx , x
+ (N
i( + )
xx
i( + )
,x ) +
a( + )
Y = Pj
i
(33a,b)
G = J = Wdy T ds
(35)
where is the path surrounding the crack tip, T is the

traction vector along the path , u is the displacement
vector, ds is a differential length element along and W
is the strain energy density function. Alternatively, the
ERR can be directly evaluated through the analytical or
the numerical derivative of the total potential energy:
(33c)
where () and (+) refer to quantities left and right of the

connection point at x = xj and the external concentrated
loads (if present) are included in the conditions imposed
on the RC beam only. In the adhesive layer, the
connectivity conditions require that the shear stress in
the fully bonded region at x = xj equals to zero. Thus, for
example, at the left edge of the delaminated region
shown in Figure 2(b), the condition imposed on the
shear stresses in the adhesive layer reduces to:
(34)
a( ) = 0
The continuity conditions within the debonded region
itself include Eqns 3233 without the terms that
include the vanishing shear stresses a. The boundary
conditions at the edges of the beam, the unique
boundary/continuity conditions at the cut-off section
(the connection between the un-strengthened and the
strengthened region), and the continuity conditions
within the fully bonded region itself appear in
Rabinovitch (2004a). For brevity, these conditions are
not presented here.
Eqns 2731 form a set of nonlinear ordinary
differential equations with constant coefficients. Due to
the nonlinearity of the model, the governing equations
do not have a general closed form analytical solution.
Alternatively, they are solved numerically using a
multiple shooting algorithm (Stoer and Bulirsch 1993).
2.7. Debonding Growth Fracture Criterion
The criterion for the propagation of the interfacial crack
and the growth of the debonded region adopts the
fracture mechanics Griffith concept of the energy

release rate (also see Rabinovitch and Frostig 2001a).
Following this concept, the propagation of the
debonding crack occurs if G, the amount of the energy
released by the structure due to an infinitesimal crack
growth of area dA (i.e., the energy release rate (ERR)) is
larger than the critical fracture energy Gc, see Broek
(1974). G is evaluated using the J integral (Rice 1968):
G=
1
1 (U + V )
=
=
A
b a
b
a
(36)
where = U+V is the total potential energy and a is the

length of the interfacial crack and the debonded region.
A comparison between the J-integral formulation, Eqn
35, and the energy derivative, Eqn 36, and a study of
their applicability to the evaluation of ERR criterion for
the edge debonding using a geometrically linear theory
appear in Rabinovitch (2004b).
The stability of the crack growth process is governed
by the derivatives of the ERR with respect to the crack
length (also see Kardomateas and Pelegri 1994). Thus,
the crack growth process is (locally) stable if:
G
<0
(37)
a
whereas the debonding process is (locally) unstable if:
G
>0
(38)
a
In the numerical study presented next, the ERR
associated with the growth of the inner debonded region
under geometrically nonlinear conditions and the
stability characteristics of the debonding growth process
are investigated.
3. NUMERICAL STUDY
The geometry, mechanical properties, and loading
scheme of the examined beam appear in Figure 3. The
response of the strengthened beam with a 50 mm long
debonded region at midspan through the loading
process is examined first. (In this case, it is assumed
771
ASE 9-6-04_Frostig
4/12/06
5:10 pm
Page 772
qrc = q
E = 30 GPa
250
Ea = 0.7 GPa
Ga = 0.28 GPa
Debonded
region
161
a
CFRP Laminate
E11 = 120 GPa
161
L = 2400 mm
5.0
0.8
100
(a)
(b)
Figure 3. Geometry, loading, and mechanical properties: (a) Geometrical layout; (b) Cross section.
100
q (kN/m)
90
max(|wrc|)
80
70
max(|wfrp|)
60
50
40
30
20
10
w (mm)
0
2
4
6
8
10
12
100
q (kN/m)
90
80
70
60
50
40
30
20
10
0
0.05
w
Dw
w, Dw ()
0.1
(a)
0.15
(b)
0.2
0.25
100
90
80
70
60
50
40
30
20
10
0
25
q (kN/m)
Nfrp
xx (x = L/2) (kN)
20
15
10
(c)
Figure 4. External load versus structural response, a=50 mm: (a) Peak vertical deflections; (b) Relative deflection and waviness parameters;
(c) Inplane force in the FRP strip at midspan.
that the un-bonded region exists prior to loading). Next,

the response through the debonding growth process
under a constant load level of qrc=-q=-35kN/m is
studied. In all cases, it is assumed that the debonding
results from improper fastening of the FRP strip and the
formation of large voids in the adhesive layer (see
Figure 1(c)). Thus, the debonding without contact
conditions are adopted.
The nonlinear behavior of the strengthened beam
under the uplifting load is studied in Figure 4. The
development of the peak deflections in the RC beam and
the FRP strip through the loading process appear in
Figure 4(a). Also, in order to investigate the local
buckling effects, the results are examined in terms of a
relative deflection parameter, w, and a waviness
parameter, Dw, that are defined by:
L
w = w frp w rc dx
0
Dw = w ,fxrp w ,rcx dx
0
772
frp
rc
w + w dx ;
0
frp
rc
w, x + w, x dx
0
(39)
The development of these normalized parameters

through the loading process appears in Figure 4(b). The
load Vs. deflection curve and the normalized relative
deflection and waviness parameters reveal that buckling
of the debonded FRP strip occurs at a load level of about
62 kN/m. This phenomenon is characterized by the
formation of increased relative displacements between
the FRP strip and the RC surface and increased
waviness of the buckled FRP strip. The development
of the inplane stress resultant in the FRP strip at
midspan with the increase of the load appears in Figure
4(c). This curve shows that the magnitude of the axial
compressive force in the strip at midspan beyond the
buckling load, i.e. in the post-buckling stage, remains
constant. As a result, the bending moments due to loads
beyond the buckling load level are solely carried by the
RC section without further increase of the contribution
of the strengthening system.
The distributions of the vertical deflections along the
RC beam and the FRP strip and the distribution of the
longitudinal normal stresses in the inner and outer faces
of the FRP strip under a load level of 70 kN/m appear in
ASE 9-6-04_Frostig
4/12/06
5:10 pm
Page 773
10
8.5
w (mm)
wfrp
wrc
wfrp
wfrp
7.5
wrc wfrp
2
0
w (mm)
x (mm)
0
500
1000
1500
2000
2500
6.5
900
x (mm)
1000
1100
(a)
50
50
0
50
50
100
100
150
150
200
200
250
250
300
300
350
350
400
450
Upper face
Lower face
500
1000
1300
1400
1500
(b)
xx (MPa)
1200
xx (MPa)
400
x (mm)
1500
2000
2500
(c)
450
1150
Upper face
Lower face
1170
x (mm)
1210
1190
1230
1250
(d)
Figure 5. Vertical deflections and longitudinal stresses in the FRP strip with q=70 kN/m and a=50 mm: Vertical deflections; (a) Along the
beam; (b) zoom on the debonded region; Longitudinal normal stresses in the FRP strip: (c) Along the beam; (d) zoom on the debonded region.
Figure 5. Figures 5(a) and 5(b) clearly show the localized

wrinkling of the compressed FRP strip within the
debonded region. This effect also yields a significant
increase in the localized bending moments that develop in
the FRP strip and thus an increase in the longitudinal
normal stresses in its inner and outer faces, see Figures
5(c) and 5(d). In quantitative terms, the magnitudes of the
normal stresses associated with the local buckling are of
the same order of magnitude or even larger than the ones
associated with the global action of the strip, see Figure
5(c). Yet, the magnitudes of the relative displacement are
relatively small and do not exceed 12 mm.
The distributions of the shear stresses in the adhesive
layer and the vertical normal stresses in the adhesiveconcrete and adhesive-FRP interfaces under a load level
of 70 kN/m appear in Figure 6. These results reveal that
the buckling of the debonded strip is associated with
severe stress concentrations near the edges of the
debonded region. The magnitudes of the shear stresses
that develop near the tips of the debonded region are
only slightly smaller than the ones observed near the
edges of the bonded strip, see Figure 6(a). However, the
magnitudes of the vertical normal stresses in this point
exceed the ones observed near the cut-off section, see
Figure 6(c). Hence, it is anticipated that under these
conditions, a debonding failure of the beam would
initiate at the edges of the debonded region at mid-span.

The results also clarify that the shear stress distribution
strictly satisfies the null shear conditions at the edges of
the debonded region and at the edges of the FRP strip
and that the both the vertical normal and the shear
stresses vanish within the debonded region.
The development of the peak shear and vertical
normal stresses near the interfacial crack tip with the
increase of the external load is studied in Figures 7(a)
and 7(b), respectively. These curves reveal that the
development of the stress concentration near the edge of
the debonded region is significantly accelerated by the
localized buckling of the FRP strip and by the nonlinear
response of the structure. The rapid increase in the
magnitudes of the shear and, especially, the vertical
normal stresses in this region imply that the debonded
region may tend to grow. This effect is examined using
the fracture mechanics criterion in Figure 7(c).
The response of the debonded strengthened beam in
terms of the energy release rate (ERR) due to growth of
the 50 mm long debonded region appears in Figure
7(c). This curve reveals that the buckling of the FRP
strip significantly amplifies the tendency of the
debonded region to grow. In quantitative terms, the
energy release rate is increased by up to 3 orders of
magnitudes in the post-buckling stage. Adopting a
773
ASE 9-6-04_Frostig
4/12/06
5:10 pm
Page 774
a (MPa)
1.5
1.5
0.4
0.8
a (MPa)
1.2
0.5
0.5
1.6
100
200
300
0.5
0.5
1.5
1.5
x (mm)
2
0
500
1000
1500
2000
2500
2
1150
x (mm)
1170
1190
(a)
4
1210
1230
1250
(b)
2
azz (MPa)
azz (MPa)
AdhesiveConcrete
Interface
4
6
6
8
AdhesiveFRP
Interface
10
10
12
12
x (mm)
14
0
500
1000
1500
2000
2500
(c)
14
1150
x (mm)
1170
1190
1210
1230
1250
(d)
Figure 6. Shear and vertical normal stresses in the adhesive layer under q=70 kN/m and a=50 mm: Shear stresses: (a) Along the beam;
(b) Zoom on the debonded region; Vertical normal stresses: (c) Along the beam; (d) Zoom on the debonded region.
characteristic value of Gc = 65 J/m2 (Rabinovitch and

Frostig 2003) this observation means that once the
buckling load is reached, a further increase in the load
would immediately lead to a spontaneous growth of the
interfacial debonding crack. As a result, the localized
buckling of the FRP strip near the debonded region may
directly lead to a progressive or total debonding failure
of the strengthening system. This observation
highlights the critical role that the localized effects play
in the overall response of the strengthened beam.
Hence, it emphasizes the necessity of an analytical
approach that is capable of describing and quantifying
these effects.
Another aspect of the response of the debonded
structure is the stability of the crack growth and the
possibility of crack arrest. This aspect is studied through
the response of the structure during the debonding crack
growth under a constant load level. The curves showing
the development of the peak vertical deflections, the
774
waviness parameters, the axial force in the FRP strip at

midspan, the peak stresses in the adhesive layer, and the
ERR with the growth of the debonded region under
qrc = 35 kN/m appear in Figure 8. This figure reveals
that debonded regions shorter than about 65 mm do not
tend to buckle. Correspondingly, the magnitudes of the
localized deflections, the stress near the debonded
region, and the ERR associated with debonded regions
shorter than 65 mm are not significant. Beyond this
critical crack length, buckling of the FRP strip, relative
vertical deflection between the FRP strip and the RC
beam, and a reduction in the compressive force carried
by the FRP strip are observed (see Figures 8(a)-(c)). The
localized buckling of the FRP strip at a critical
debonding length of 65 mm also yields a considerable
increase in the vertical normal and shear stresses in the
adhesive layer near the debonded region (Figures 8(d)
and 8(e)). Also here, these results quantitatively clarifies
that the buckling of the debonded FRP strip under the
ASE 9-6-04_Frostig
4/12/06
5:10 pm
Page 775
100
q (kN/m)
90
80
70
60
50
min (zz [za = 0])
40
max (zz [za = 0])
30
min (zz [za = Ca])
20
max (zz [za = Ca])
10
zz (MPa)
0
50 40 30 20 10
0
10
100
90 q (kN/m)
80
70
60
50
40
30
20
10
0
0.5 1
a (MPa)
1.5 2
(a)
2.5 3
3.5 4
100
q (kN/m)
90
80
70
60
50
40
30
20
10
G (J/m2)
0
100 200 300 400 500 600 700 800 900
(b)
(c)
Figure 7. External load versus structural response, a=50 mm: (a) Peak vertical normal stresses (za=0 adhesive-concrete interface;
za=ca adhesive-FRP interface); (b) Peak shear stresses; (c) Energy release rate.
300
300
a (mm)
300
a (mm)
a (mm)
max(|wrc|)
200
200
200
max(|wfrp|)
Dw
100
w, Dw ()
w (mm)
0
10
0.05 0.1
0.15
(a)
300
0.2 0.25
0.3 0.35
0
12
Nfrp
xx (x = L/2) (kN)
10
(b)
300
a (mm)
300
a (mm)
a (mm)
200
200
200
a3cr =185 mm
100
100
100
a2cr = 105 mm
a1cr = 65 mm
0
14 12 10 8
zz (MPa)
6
a (MPa)
0
0.5
(d)
1.5
(e)
(c)
2.5
Gc165 J/m2
100
100
G (J/m2)
10 20 30 40 50 60 70 80 90
(f)
Figure 8. Debonded length versus structural response at q=35 kN/m: (a) Peak vertical deflections; (b) Relative deflection and waviness
parameters; (c) Inplane force in the FRP strip at midspan; (d) Peak vertical normal stresses at adhesive layer interfaces (za=0 adhesiveconcrete; za=ca adhesive-FRP); (e) Peak shear stresses; (f) Energy release rate.
critical combination of load and debonded length

critically affects the response of the structure.
The response in terms of the development of the ERR
with the growth of the interfacial debonding crack
appears in Figure 8(f) and reveals a special and unique
phenomena. This curve shows that the buckling point
under a critical length of 65 mm is associated with a
rapid and prominent increase of the ERR. By adopting
the value Gc=65 J/m2, Figure 8(f) shows that the ERR
exceeds the critical fracture energy very fast and very

close to the buckling point. At the initial stages of the
interfacial crack growth process, the ERR is further
increased with the growth of the debonded region. As a
result, in this range, the debonding propagation process
is spontaneous and unstable. However, at a crack length
of about 105 mm, the ERR attains its peak value.
Beyond this point, it decays and at a crack length of
about 185 mm it drops below the critical fracture
775
ASE 9-6-04_Frostig
4/12/06
5:10 pm
Page 776
energy. In physical terms, this observation means that

beyond the peak at 105 mm, the crack growth process
becomes stable. Beyond a=185 mm, the debonding
crack may be arrested, and, if the dynamic effects are
minor, the debonding process may stop. In other words,
the ERR curve quantitatively defines three critical
lengths for the debonding crack. The first critical
length (a1cr=65 mm) defines the buckling point under
the prescribed external load. The second critical length
(a2cr=105 mm) defines the point beyond which the crack
growth shifts from a spontaneous and an unstable
process into a stable one. The third critical length
(a3cr=185 mm) designates the point beyond which the
crack may be arrested. A qualitatively similar
phenomenon of transition from an unstable phase to a
stable one in the post buckling stage of a horizontally
cracked isotropic beam under axial compression was
analytically observed by Sheinman et al. (1998). The
results of this relatively simple case support and confirm
the unique trends and phenomena observed here. The
characterization of these trends and the ability of the
analysis to quantify and describe the unique behavior of
the strengthened beam are essential in order to asses the
reliability and the performance of the strengthened
member under its entire loading envelope.
A nonlinear analytical approach for the debonding
failure mechanism in RC beams strengthened with
externally bonded composite materials has been
presented. Emphasis has been placed on the cases where
various loading scenarios yield compressive stresses in
the bonded strip and critically accelerate the debonding
growth process. Two analytical methodologies have
been combined to yield the unified geometrically
nonlinear fracture mechanics modeling approach. First,
the overall geometrically nonlinear response of the
strengthened beam, the localized buckling of the
debonded FRP strip, and the stress and deformation
fields near the tips of the debonded region have been
analyzed through a geometrically nonlinear high order
model. Next, the fracture mechanics concept of the
energy release rate has been adopted for the assessment
of the tendency of the interfacial debonding crack to
burst and for the evaluation of the stability of the
debonding growth process.
The large deformations, moderate rotations and small
strains kinematic relations have been adopted in order to
account for the geometrically nonlinear effects in the
response of the strengthened beam. The variational
principle of virtual work, the lamination and beam
theories, the 2D theory of elasticity, and the high order
closed form solutions of the stress and displacement
776
fields in the adhesive layer have been used for the

derivation of the analytical nonlinear model. A
distinction between un-strengthened regions, strengthened
regions, and debonded regions with or without interfacial
contact have been made in order to account for the
unique topology of the delaminated strengthened beam.
A solution procedure that is based on the multiple
shooting method has been adopted for the numerical
solution of the resulting set of nonlinear differential
equations. Two approaches for the evaluation of the
fracture mechanics criterion of the energy release rate
have been discussed. The ERR has been determined
using the derivative of the total potential energy with
respect to the length of the debonded region.
The proposed analytical model has been used for the
study of a reinforced concrete beam strengthened with
an externally bonded FRP strip and subjected to a
uniformly distributed uplifting load. The investigation
has focused on the overall and the localized nonlinear
response of the beam, on the tendency of the debonded
region to grow, and on the stability of the debonding
growth process. The results have quantitatively revealed
the localized buckling and wrinkling of the un-bonded
FRP strip and have quantified the significant increase in
the tendency on the debonded region to grow in the post
buckling stage. Tracking down the interfacial crack
growth process has revealed that the ERR is
considerably increased once the critical debonding
length is reached. However, after it attains its maximal
value, the ERR tends to decrease and it reduces to
values that may go down below the critical fracture
energy. Thus, stabilization of the interfacial debonding
process and crack arrest are possible.
The analytical approach and the numerical results
presented here emphasize the importance of the localized
effects in the behavior of structural elements strengthened
with externally bonded composite material systems. Due
to the unique characteristics of the localized buckling and
wrinkling effects and, especially, due to the tendency of
the debonded region to grow, these effects may have a
critical influence on the performance, functionality, and
reliability of the entire strengthened element. The ability
of the analysis to quantify and to describe this type of
behavior is essential for the design of the strengthening
system, for the assessment of the ability of the element to
resist load, and, above all, for the safe and reliable use of
the strengthened member.
REFERENCES
ACI Committee 440.2R-02 (2002). Guide for the Design and
Construction of Externally Bonded FRP Systems for Strengthening
Concrete Structures, American Concrete Institute, Framington
Hills MI.
ASE 9-6-04_Frostig
4/12/06
5:10 pm
Page 777
Bakis, C.E., Bank, L.C., Brown, V.L., Cosenza, E., Davalos, J.F.,
Lesko, J.J., Machida, A., Rizkalla, S.H. and Triantafillou, T.C.
(2002). Fiber-reinforced polymer composites for construction
state-of-the-art review, Journal of Composites for Construction,
ASCE, Vol. 6, No. 2, pp. 7387.
Broek, D. (1974). Elementary Engineering Fracture Mechanics,
Nordhoff Int., Leyden, The Netherlands.
Buyukozturk, O. and Hearing, B. (1998). Failure behavior of precracked concrete beams retrofitted with FRP, Journal of
Buyukozturk, O., Gunes, O. and Karaca, E. (2004). Progress on
understanding debonding problems in reinforced concrete and
steel members strengthened with FRP composites, Construction
CEB-FIP (2001). Externally Bonded FRP Reinforcement for RC
Structures, CEB-FIP Bulletin No. 14, Fdration internationale
du bton (fib), Lausanne, Switzerland.
Deuring, M. (1993). Verstrken von Stahlbeton mit Gespannten
Faserverbundwerstoffen, EMPA Research report No. 224, Swiss
Federal Laboratories for Material Testing and Research,
Dbendorf, Switzerland. (in German).
Grace, N.F., Soliman, A.K., Abdel-Sayed, G. and Saleh, K.R.
(1999). Strengthening of continuous beams using fiber
reinforced polymer laminates, Proceedings of the Fourth
International Symposium on Fibre Reinforced Polymer
Reinforcement for Concrete Structures, SP-188, ACI, Farmington
Hills, MI, PP. 647657.
Kardomateas, G.A. and Pelegri, A.A. (1994). The stability of
delamination growth in compressively loaded composite plates,
International Journal of Fracture, Vol. 65, No. 3, pp. 261276.
Lau, K.T., Dutta, P.K., Zhou, L.M., and Hui, D. (2001). Mechanics
of bonds in an FRP bonded concrete beam, Composites Part B:
Engineering, Vol. 32, No. 6, pp. 491502.
Malek, A.M., Saadatmanesh, H. and Ehsani, M.R. (1998).
Prediction of failure load of RC beams strengthened with FRP
plate due to stress concentration at the plate end, ACI Structural
Journal, Vol. 95, No. 1, pp. 14252.
Mohamed Ali, A.S., Oehlers, D.J. and Bradford, M.A. (2002).
Interaction between flexure and shear on the debonding of RC
beams retrofitted with compression face plates, Advances is
Structural Engineering, Vol. 5, No. 4, pp. 223229.
Neale, K.W. and Laboissiere, P. (1997). State-of-the-art report on
retrofitting and strengthening by continuous fibre in Canada,
Non-metallic (FRP) Reinforcement For Concrete Structures, Vol. 1,
Japan Concrete Institute, pp. 2540.
Mukhopadhyaya, P. and Swamy, N. (2001). Interface shear stress:
a new design criterion for plate debonding, Journal of
Oehlers, D.J. (2005). Draft - Design Guidelines for RC Structures
Retrofitted with FRP and Metal Plates: Beams and Slabs, Centre
for Infrastructure Diagnosis Assessment and Rehabilitation
(CIDAR) The University of Adelaide (http://www.civeng.
adelaide.edu.au/index.html).
Rabinovitch, O. (2004a). Fracture-mechanics failure criteria for RC

beams strengthened with FRP strips - a simplified approach,
Composite Structures, Vol. 64, No. 34, pp. 479492.
Rabinovitch, O. (2004b). Nonlinear (buckling) effects in RC beams
strengthened with composite materials subjected to compression,
International Journal of Solids and Structures, Vol. 41, No. 20,
pp. 56775695.
Rabinovitch, O. and Frostig, Y. (2000). Closed-form high-order
analysis of RC Beams strengthened with FRP strips, Journal
of Composites for Construction, ASCE, Vol. 4, No. 2,
pp. 6574.
Rabinovitch, O. and Frostig, Y. (2001a). Delamination failure of
RC beams strengthened with FRP strips- a closed-form highorder and fracture mechanics approach, Journal of Engineering
Mechanics, ASCE, Vol. 127, No. 8, pp. 852861.
Rabinovitch, O. and Frostig, Y. (2001b). Non-linear high-order
analysis of cracked RC beams externally strengthened with FRP
strips, Journal of Structural Engineering, ASCE, Vol. 127,
No. 4, pp. 381389.
Rabinovitch, O. and Frostig, Y. (2003). Experiments and analytical
comparison of RC beams strengthened with externally bonded
CFRP strips, Composites Part B - Engineering, Vol. 34, No. 8,
pp. 663667.
Rice, J.R. (1968). A path independent integral and the
approximated analysis of strain concentration by notches and
cracks, Journal of Applied Mechanics, ASME, Vol. 35, No. 6,
pp. 379386.
Sheinman, I., Kardomateas, G.A. and Pelegri, A.A. (1998).
Delamination growth during pre - and post-buckling phases of
delaminated composite laminates, International Journal of
Solids and Structures, Vol. 35, No. 12, pp. 1931.
Smith, S.T. and Teng J.G. (2001). Interfacial stresses in plated
beams, Engineering Structures, Vol. 23, No. 7, pp. 857871.
Smith, S.T. and Teng, J.G. (2002a). FRP-strengthened RC beams.
I: Review of debonding strength models, Engineering
Smith, S.T. and Teng, J.G. (2002b). FRP-strengthened RC beams.
II: Assessment of debonding strength models, Engineering
Stoer, J. and Bulirsch, R. (1993). Introduction to Numerical
Analysis, Springer-Verlag, New York, N.Y.
Teng, J.G., Zhang, J.W. and Smith, S.T. (2002). Interfacial stresses
in reinforced concrete beams bonded with a soffit plate: a finite
element study, Construction and Building materials, Vol. 16,
No. 1, pp. 114.
Uy, B., Aw, C.S. and Yang, T. (2002). Local buckling of steel and
carbon fibre reinforced plastic plates restrained by concrete,
Proceedings of the 1st International Conference on High
Performance Structures and Composites, C.A Barebbia and W.P.
De Wilde, Editors, Seville, Spain, Wit press.
Varastehpour, H. and Hamelin, P. (1997). Strengthening of concrete
beams using fiber-reinforced plastics, Materials and Structures,
Vol. 30, No. 197, pp. 160166.
777
ASE 9-6-04_Frostig
4/12/06
5:10 pm
Page 778
ASE 9-6-05_Harries
4/12/06
5:10 pm
Page 779
Deterioration of FRP-to-Concrete Bond

Under Failure Loading
Kent A. Harries1,*, John Aidoo2, Andrew Zorn3 and Joshua Quattlebaum4
1University
of Pittsburgh, Pittsburgh PA
Institute of Technology, Terre Haute IN,
3DMJM Harris, Pittsburgh PA
4S&ME, Columbia SC
2Rose-Hulman
Abstract: Characteristics of FRP-to-concrete bond are observed to deteriorate under

cyclic loads. Deterioration has been observed under load conditions which result in a
stress range in the bonded FRP of only 4% of the material capacity. Current design
recommendations and accepted practice often neglect this detrimental effect. The
clear effects of bond deterioration under fatigue loading are demonstrated for
moderate-scale laboratory specimens and in full-scale specimens recovered from a
forty-year old interstate highway bridge. The results suggest additional limit states that
must be considered when even relatively small fatigue loads are present.
Key words: bond, CFRP, endurance limit, fatigue, flexural strengthening.
1. INTRODUCTION
Debonding of externally bonded fiber reinforced
polymer (FRP) composite materials used for repair of
reinforced concrete flexural elements is commonly
observed. The most common debonding mode reported
in earlier experimental literature is the plate end mode
(PE) which initiates at the curtailment of the FRP plate
and propagates toward the midspan of the beam
(Oehlers 2005). PE debonding may be effectively
mitigated by providing anchorage to resist the peel
forces that are generated normal to the FRP or by
simply extending the FRP plate across the entire shear
span. In beams having relatively long shear spans, or
where the PE mode has been effectively mitigated,
debonding initiates at flexural (FIC) and/or flexural/
shear (SIC) cracks near the region of maximum moment
(Oehlers 2005). Under loading, these cracks open and
induce high interfacial shear stress accompanied by a
small amount of peeling causing delamination, which
propagates across the shear span in the direction of
decreasing moment (Sebastian 2001; Oehlers 2005;
Quattlebaum et al. 2005; Aidoo et al. 2006). This mode
of debonding is properly termed intermediate [flexural
or shear] crack-induced debonding. Available guidance

for the design of externally bonded FRP systems often
takes the approach of limiting the stress (or strain) in the
FRP in order to mitigate debonding. The behaviour of
bonded FRP systems subject to fatigue loading,
however, is expected to degrade relative to their
statically loaded counterparts. Thus additional stress (or
strain) limits are recommended to mitigate bond failure
associated with deterioration due to fatigue loads.
2. FATIGUE BEHAVIOUR OF FRP
A substantial amount of data regarding the fatigue
behaviour and life prediction of FRP composites is
available, primarily through the aerospace industry
(National Research Council 1991). In conditions of
tension fatigue, where environmental effects are not
affecting behaviour, FRP composite behaviour is
dominated by the strain-limited creep-rupture process.
Plotted on a semi-log S-N curve (S the stress range
plotted on a linear scale; N is the number of cycles of
fatigue loading plotted on a log scale), carbon (CFRP)
and aramid (AFRP) composites exhibit strength
degradation due to tensile fatigue on the order of 5 to 8%
*Corresponding author. Email address: kharries@engr.pitt.edu; Fax: 412.624.0135; Tel: 412.624.9873.
779
ASE 9-6-05_Harries
4/12/06
5:10 pm
Page 780
Deterioration of FRP-to-Concrete Bond Under Failure Loading
per decade of logarithmic cycles (Curtis 1989; Roylance

and Roylance 1981). Glass (GFRP) composites exhibit
strength degradation on the order of 10% per decade
(Mandell 1982). By comparison, the stress range capacity
of mild steel degrades about 22% per decade (Tilly and
Moss 1982). Additionally, FRP composite materials do
not generally exhibit a clearly defined endurance limit
under conditions of tension fatigue.
The fatigue performance of FRP composite materials
depends on the matrix composition and, to some
extent, on the type of fiber (Curtis 1989). It has been
demonstrated that the matrix composition has a greater
effect on the fatigue performance than the type of fiber
used. No significant difference in fatigue performance of
FRP materials was observed when four different carbon
fiber types were placed in the same matrix, however,
using the same fiber in different epoxy resin matrices had
an effect on the slope of the resulting S-N curves (Curtis
1989). The individual fibers within unidirectional
composites have few defects and are consequently
resistant to crack initiation. Additionally, any crack that
does form travels through the matrix and is not
transmitted through adjacent fibers. It is these toughness
and crack arresting properties that contribute to the good
fatigue performance of FRP materials. Based on the
behaviour of FRP materials, it is not expected that the
fatigue behaviour of the FRP material itself will be a
design issue since the behaviour of the existing
reinforcing steel will dominate fatigue performance.
3. FATIGUE BEHAVIOUR OF EXTERNALLY
BONDED FRP REINFORCEMENT
In some of the previously reported studies of FRPretrofitted reinforced concrete flexural members
(Heffernan 1997; Masoud et al. 2001; Quattlebaum
et al. 2005; Aidoo et al. 2004, 2006), although the stress
range (S) in the primary longitudinal reinforcing steel is
reported to be reduced in the FRP-retrofitted beams,
there is no corresponding increase in the fatigue life (N).
In these cases, it would appear that the stress range in
the reinforcing steel, while initially reduced due to
the presence of the additional FRP reinforcement,
subsequently returned to the stress range corresponding
to the unretrofitted control specimens. Debonding of the
FRP retrofit material in the vicinity of the high flexural
stresses is believed to explain this behaviour.
Other studies (Barnes and Mays 1999; Shahawy and
Beitleman 1999) demonstrated both a significant
reduction in longitudinal steel stress and corresponding
increase in fatigue life with the application of FRP
reinforcing material. Meier et al. (1993) report that the
hybrid glass/carbon FRP material used in their study
was able to bridge cracks resulting from the fracture of
780
the primary longitudinal reinforcement in the beams

tested. In this case, fatigue loading was continued beyond
steel fracture until eventual fatigue failure of the hybrid
FRP reinforcing. These results are a clear indication of
sound bond between the concrete and FRP.
The different results obtained in different studies
appear to be associated with FRP-to-concrete bond and/or
the nature of the FRP-concrete interface. In all previously
reported studies, some extent of debonding of the FRP
material from the concrete substrate was observed.
Once debonding occurs, stress carried by the FRP is
redistributed back to the internal reinforcing steel in the
regions of debonding. Thus improvement in fatigue
performance is only affected as long as the FRP is
adequately bonded to the concrete.
Papakonstantinou et al. (2001) demonstrated that the
fatigue behaviour of the existing primary reinforcing
steel will govern the eventual fatigue performance of
FRP retrofitted concrete beams. In this study, the stress
level applied to the internal steel reinforcement was
maintained constant for both retrofitted and unretrofitted
specimens. No discernable difference in behaviour
between retrofitted and unretrofitted specimens having
the same reinforcing steel stress levels was observed. Of
course, the applied load to obtain the same steel stress in
the retrofitted specimens was increased. (Additionally in
this study of short beam specimens, FRP anchorage was
ensured by pinning the GFRP material beneath the
beam supports.)
Heffernan (1997), on the other hand, states that FRP
retrofitted beams should exhibit improved fatigue
characteristics over similar unretrofitted beams having
identical initial steel stresses. This conclusion stems
from the observations that concrete softens with cycling
(Neville 1975), requiring redistribution of applied
stresses to the tensile reinforcement. The additional
tensile reinforcement, in the form of the FRP retrofit
material, will clearly carry some of the redistributed
stresses thus relieving the steel reinforcement and
prolonging its fatigue life. Heffernan estimates that this
increase in fatigue life will be approximately 2%. Such
an increase in fatigue life is too small to capture with
certainty in a reasonably sized test program.
3.1. Observed Fatigue-Induced Failure Modes
The most common failure mode reported in the available
literature is the fatigue-induced fracture of the internal
steel reinforcement. The number of cycles to failure
when this failure mode controls the fatigue-life of the
specimens has not been found to be significantly different
from FRP-strengthened to unstrengthened reinforced
concrete beams. Nonetheless, both PE and FIC/SIC
debonding (Oehlers 2005) have been observed in fatigue
ASE 9-6-05_Harries
4/12/06
5:10 pm
Page 781
Kent A. Harries, John Aidoo, Andrew Zorn and Joshua Quattlebaum
Table 1. Summary of current guidelines for maximum allowable stress in bonded FRP subject
to fatigue loading
ACI 440.2R-02
Concrete Society TR55
JSCE Recommendations
CNR 200/2004
fib Bulletin 14
ffuf
ffu
ffud
Carbon FRP
Glass FRP
Aramid FRP
ffuf < 0.55ffu

ffuf < 0.80ffu
ffuf < 0.84ffud
ffuf < 0.50ffud
none
ffuf < 0.20ffu

ffuf < 0.30ffu
ffuf < 0.84ffud
ffuf < 0.50ffud
none
ffuf < 0.30ffu

ffuf < 0.70ffu
ffuf < 0.84ffud
ffuf < 0.50ffud
none
Stress limit in FRP under fatigue load conditions

Ultimate capacity of FRP
Stress limit in FRP to mitigate debonding under monotonic load conditions
tests. At high stress ranges, debonding failure has been

observed prior to reinforcing bar rupture, although
debonding progression has been observed at lower stress
ranges more representative of service conditions.
4. BEST PRACTICE FOR FATIGUE
LOADING CONDITIONS
Different approaches are taken by different entities in
addressing the behaviour of bonded FRP retrofit
measures subject to fatigue loads: the American Concrete
Institute (ACI 2002) and the UK Concrete Society (2004)
direct the designer to verify that the expected FRP
stress/strain is lower than an absolute limit associated
with the expected FRP material performance and not
specifically associated with debonding behaviour. The
Japan Society of Civil Engineers (JSCE 2000) and Italian
(National Research Council (Italy) 2004) approaches, on
the other hand, provide an additional reduction factor, to
be applied to debonding stress/strain limits determined
for static loading conditions. This factor is applied in
addition to others accounting for environmental exposure
or other effects. The fib (2001) does not quantitatively
address fatigue loading although it restricts the stress
range in the reinforcing steel to that present in an
unstrengthened beam. This approach implies that
accounting for the FRP reducing the stress in the existing
steel can be used to the designers advantage. Table 1
summarizes these approaches. Although not explicitly
stated (except in the Italian document), the stress in the
reinforcing steel of an FRP-repaired section must also be
verified for its fatigue performance. As discussed, this
latter limit will typically control fatigue behaviour.
The ACI and Concrete Society approaches recommend
limits on the permissible stress range in the FRP
for fatigue applications. In a realistic plate bonding
application with CFRP, such as this study, a stress range
of 0.55ffu to 0.80ffu (Table 1) would be essentially
impossible to achieve. Either delamination of the CFRP
through the concrete substrate prior to 0.55ffu being
attained or reinforcing bar failure rupture by CFRP
delamination would occur. For example, in work

conducted in relation to the present study, Reeve (2005)
reports the maximum stress developed in the CFRP in a
monotonic test was only 0.46ffu prior to the occurrence
of debonding.
Current conventional wisdom (described in Harries
2005) has the fatigue behaviour of FRP retrofitted
concrete sections controlled by the performance of the
internal reinforcing steel. This approach implies that
provided the endurance limit of the steel is not exceeded,
fatigue loading will not affect behaviour of the FRPretrofitted concrete section. The following experimental
results contradict this wisdom and furthermore suggest
that many best practice recommendations are nonconservative.
4.1. Why Consider Fatigue Loading
In most reinforced concrete practice, fatigue loading is
barely considered. The endurance limit of mild steel
reinforcement, approximately 125 MPa, is relatively
high compared to the magnitude of expected transient
stresses. In repair and retrofit applications, however,
transient loads may be effectively much larger due to a
variety of issues including: a) the retrofit is required for
increased transient loads (increased truck design loads,
for instance); b) the structure is badly deteriorated
resulting greater stresses being carried by the reinforcing
steel than originally intended; and c) construction errors
or omissions. An example of the latter is reported by
Aidoo et al. (2006) where existing steel was found to be
a lower grade (240 MPa rather than 400 MPa) than was
shown on as-built drawings of an interstate bridge.
5. EXPERIMENTAL PROGRAM
The present work synthesizes the results of a number of
fatigue tests (conducted under the direction of the
primary author) of CFRP-retrofitted concrete beams
reported by Zorn (2006), Quattlebaum et al. (2005) and
Aidoo et al. (2006). The reported specimens are
designated as the Z-series, Q-series and A-series,
781
ASE 9-6-05_Harries
4/12/06
5:10 pm
Page 782
Table 2. Material properties (CFRP and adhesive data reported by manufacturer)
Concrete strength, fc
Reinforcing steel yield strength, fy
Reinforcing steel tensile strength, fu
CFRP strength, ffu
CFRP tension modulus, Ef
CFRP rupture strain, fu
CFRP thickness, tf
Adhesive type
Adhesive strength, fau
Adhesive tension modulus, Ea
Adhesive rupture strain, au
Adhesive thickness, ta
1Zorn
Z-series1
Z-series1
Q-series2
A-series3
23.3 MPa
429 MPa
667 MPa
2800 MPa
155 GPa
0.018
1.4 mm
L
14 MPa
2227 MPa
0.063
1.6 mm
23.3 MPa
429 MPa
667 MPa
2800 MPa
155 GPa
0.018
1.4 mm
H
25 MPa
4482 MPa
0.010
1.6 mm
29.6 MPa
446 MPa
735 MPa
2800 MPa
155 GPa
0.018
1.4 mm
T
47 MPa
2069 MPa
0.033
1.6 mm
35 MPa
364 MPa
530 MPa
2800 MPa
155 GPa
0.018
1.4 mm
T
47 MPa
2069 MPa
0.033
1.6 mm
(2006); 2Quattlebaum et al. (2005); 3Aidoo et al. (2006)
152 mm
254 mm
2 - #3
Z-H1
Z-L1
Z-H2
Z-L2
Q-CL
Q-CH
Z-H2x1
Z-L2x1
Z-H4
Z-L4
4 - #4
25
51
25 51 25
102
Tested over simple span of 4570 mm

927
165
#4
A-C
Tested over
simple span
of 8025 mm
660
2 #8
95
3 #10
95
3 #11
76
204
343
Figure 1. Specimen geometry and retrofit details
respectively, based on their source documents. All

specimens used the same preformed CFRP strip
material. The primary objective of the Z-series study
was to address the effects of adhesive modulus and
CFRP reinforcement ratio and arrangement; thus the
parameters noted in Table 2 and Figure 1. The Q- and
A-series studies primary objectives were to address
CFRP retrofit geometry and included external
adhesive-applied, near-surface mounted (NSM), and
mechanically attached CFRP systems. The results of
these studies are reported in the source documents
cited above.
782
5.1. Z-Series and Q-Series Specimens

Details and specimen designations of the moderatescale Z- and Q-series specimens are shown in Figure 1.
Material properties of the concrete, internal reinforcing
steel, CFRP strips and adhesive system used are given
in Table 2. The equivalent longitudinal steel reinforcing
ratio (CFRP transformed to equivalent reinforcing
steel) of these specimens ranges from 0.0107 (Z-L1
and Z-H1) to 0.0128% (Z-L4 and Z-H4), where the
reinforcing ratio for the unretrofitted beam is 0.010.
These 4750 mm long, lab-built specimens, were
retrofitted with externally bonded CFRP and tested
over a clear span of 4570 mm resulting in a shear span
of 2285 mm (shear span-to-beam depth ratio, a/h =
9.0). All specimens reported were tested under
midspan fatigue loading at loading rates between 1.2
and 1.7 Hz. The test setup is shown in Figure 2(a).
All Z-series specimens were cycled between applied
loads of 4.4 kN and 22.2 kN. The resulting stress
ranges on both the internal reinforcing steel and the
CFRP are given in Table 3. Four specimens (Z-L1,
Z-L2x1, Z-H1 and Z-H2) exhibited fatigue-induced
reinforcing bar ruptures (FIRR) at fatigue lives (Nf)
indicated in Table 3. Four specimens (Z-L2, Z-L4,
Z-H2x1 and Z-H4) underwent 2,000,000 cycles
of fatigue loading without failure. These specimens
were designated fatigue-conditioned specimens
and subsequently loaded monotonically to failure.
An identical series of specimens were loaded
monotonically to failure without previous fatigue
conditioning (Reeve 2005) and thus serve as
companion control specimens for the latter fatigueconditioned tests. The monotonic control specimens
are designated M in Table 4 while those undergoing
fatigue conditioning (described in Table 3) are
designated F in Table 4.
ASE 9-6-05_Harries
4/12/06
5:10 pm
Page 783
Strain gauges on reinforcing steel

and CFRP spaced at 305 mm
222 kN actuator
Load cell
Midspan deflection
L = 4570 mm
(a) Z- and Q-series test set up (Z-series shown).
Load cell
490 kN actuator (fatigue)

1335 kN hydraulic cylinder (monotonic)
Strain gauges on reinforcing steel

and CFRP spaced at 305 mm
Midspan deflection
8025 mm simple span
(b) A-series test set-up.
Figure 2. Test set-up and instrumentation

Table 3. S-N results for fatigue specimens
Rebar stress
range (MPa)
Cycles to
CFRP stress range (MPa)
Specimen
failure, Nf
N=1
N = NF
N=1
N = NF
Z-H1
Z-L1
Z-H2
Z-L2
Z-H2x1
Z-L2x1
Z-H4
Z-L4
Q-CL
424,422
400,892
1,128,006
2,000,0001
2,000,0001
447,695
2,000,0001
2,000,0001
2,000,000
587,089
523,000
2,000,0001
239
248
199
207
220
207
175
174
150
250
230
217
241
250
230
196
201
204
203 (0.072ffu)
190 (0.068ffu)
188 (0.067ffu)
170 (0.061ffu)
199 (0.071ffu)
177 (0.063ffu)
143 (0.051ffu)
144 (0.051ffu)
232 (0.080ffu)
219 (0.078ffu)
205 (0.073ffu)
205 (0.073ffu)
190 (0.068ffu)
223 (0.080ffu)
188 (0.067ffu)
168 (0.060ffu)
159 (0.057ffu)
341 (0.120ffu)
320
993
400
993
341 (0.120ffu)
85 (0.030ffu)
387 (0.140ffu)
96 (0.034ffu)
Q-CH
A-C
CFRP debonding
observed?
no
no4
no4
no
yes
no4
yes
yes
no
yes
yes (Figure 4)
no
2,000,000 fatigue cycles followed by monotonic load to failure (see Table 4).
No instruments capable of assessing rebar strain operational following 2,000,000 cycles.
3 A-series specimens were recovered from a decommissioned bridge, thus reinforcing bar stresses are calculated rather than observed.
4 CFRP debonding observed only following fatigue-induced reinforcing bar rupture.
2
The Q-CH specimen was tested under a high stress

fatigue loading to FIRR. Specimen Q-CL was subject to
2,000,000 cycles of a low stress fatigue loading
followed by continued fatigue testing to FIRR at the
high stress fatigue load level as indicated in Table 3.
5.2. A-series Specimens

The 9150 mm long A-series specimens were recovered
from a 40-year old decommissioned interstate bridge
(commissioned in 1962) and retrofitted with externally
bonded CFRP strips as shown in Figure 1. The CFRP, in
783
ASE 9-6-05_Harries
4/12/06
5:10 pm
Page 784
Table 4. Summary of monotonic test results

Specimen
Z-L2
Z-H2x1
test designation1
Maximum load (kN)

Ultimate deflection (mm)
Displacement ductility
CFRP strain at
debonding ()
Maximum CFRP
strain ()
44.3
56.9
1.70
6688
45.5
67.3
2.05
4300
6688
7444
Z-H4
Z-L4
45.2
56.1
1.70
3200
45.3
51.3
1.45
3909
49.2
47.4
1.33
2850
53.8
52.6
1.30
3259
51.8
72.2
1.91
4540
6863
6970
4813
5860
6595
A-C
F
51.3
60.5
1.67
3854
823
96.0
5.20
4500
809
58.4
2.50
4000
5807
9500
6900
= monotonic control specimen;

F = monotonic test following 2,000,000 cycles fatigue loading
1M
this case, only increased the longitudinal reinforcing

ratio from 0.0230 for the unretrofitted beam to 0.0238
for the retrofitted beam. One fatigue test (reported
in Table 3 and designated F in Table 4) and one
monotonic test (M in Table 4) are reported here. The
specimens were tested over a simple span of 8025 mm,
resulting in a shear span of 4012 mm (a/h = 4.9). All
specimens were in excellent structural condition having
a sound concrete substrate on which to apply the CFRP
strips. Specimen A-C was subject to 2 million cycles of
fatigue loading (at 1.2 Hz) resulting in an applied
moment cycling between that corresponding to the
application of full in situ dead load (DL) to that
corresponding to the application of dead load plus an
AASHTO (2004) HS25 service load plus impact factor
(DL+(LL+IM)). As indicated in Table 3, this loading
protocol resulted in a stress range in the lower-most
reinforcing steel (Figure 1) of approximately 99 MPa.
Following 2,000,000 cycles of fatigue conditioning, this
specimen was tested monotonically to failure.
6. FATIGUE BEHAVIOUR
6.1. Z-Series and Q-Series Specimens
As shown in Table 3, all moderate-scale specimens were
tested at loading levels which subject the bonded CFRP to
initial stress levels ranging from 0.05ffu to 0.12fu (140 to
340 MPa). With cycling, these stress levels increased
to 0.06ffu to 0.14fu (170 to 390 MPa) but remained
significantly below established recommendations for
fatigue stresses on CFRP (Table 1). The reinforcing bar
stress ranges (Table 3), in all cases, exceeded the generally
accepted endurance limit of 125 MPa; thus fatigue-induced
reinforcing bar ruptures (FIRR) would be expected.
FIRR were observed in half of the specimens
reported. In these cases, with the exception of the high
stress range specimen Q-CH, debonding of the CFRP
was only observed in the vicinity of the bar rupture and
was attributed to the associated energy release in each
784
case. CFRP debonding was clearly observed in those

specimens (except Z-L2) which achieved 2,000,000
cycles of loading. In these cases, the debonding was
localized and associated with flexure-shear cracks along
the beam. In these cases, debonding was established using
a simple tapping test listening for a notable difference
in sound when the CFRP is tapped with a small metal
peening hammer. Figure 3 shows the observed debonding
in Specimen Z-L4 following 2,000,000 cycles of fatigue
conditioning. Debonding is always associated with the
toe of a flexure-shear crack and is observed to extend
across the entire cracked shear span.
Extensive debonding observed during the high stress
range cycling of Specimen Q-CH is shown in Figure 4(a).
Despite the extent of debonding, the high stress range on
the reinforcing steel resulted in a FIRR mode of failure
prior to complete debonding failure for this specimen.
Figure 4(b), taken at the same location as Figure 4(a)
following testing and removal of cover concrete, shows
a fractured reinforcing bar near an inclined flexure-shear
crack and the associated CFRP debonding progressing
from the toe of this crack.
Figure 5 shows the transient strains recorded for the
reinforcing bars and CFRP for Q-CL and Q-CH. In this
figure all strains are determined from strain gages
located at the same section. As expected, CFRP strains
are greater than reinforcing bar strains and both
generally increase with cycling. Evidence of debonding
is apparent in both cases. In Specimen Q-CH, the
reinforcing bar strains are clearly increasing at a greater
rate than the relatively constant CFRP strains indicating
some degree of strain discontinuity. In Specimen Q-CL,
there is a jump in CFRP strain at about 1,300,000
cycles. It is believed that this is evidence of the initiation
of debonding in the vicinity of the strain gage reported.
Such a jump was not evident at other gage locations.
As observed in Table 3, the stress ranges recorded in
both the steel and CFRP drift upwards with cycling.
ASE 9-6-05_Harries
4/12/06
5:10 pm
Page 785
Debonding
27.6 kNm
Cracking moment = 5.9 kNm
Bending moment at maximum applied load during fatigue cycling (22.2 kN)
Figure 3. Observed debonding of specimen Z-L4 (beam drawn to scale)
Inclined
inclined crack
crack
Debonding
debonding
Debonding crack running

along line of CFRP strip
(a) During fatigue cycling
Fatigue
fracture
fatigue fracture
of 13 mm bar
bar
(b) Post-test forensic image
Figure 4. Debonding observed during testing of Q-CH
4000
Q-CH
Strain, microstrain
3500
CFRP
Fatigue failure
3000
Rebar
2500
Rebar yield
2000
CFRP
1500
1000
Rebar
500
Q-CL
Data for final 600,000
cycles of Q-CL is not
available
1000000
2000000
Number of cycles
3000000
Figure 5. Transient strains in Q-CL and Q-CH
This increase indicates fatigue-induced damage to the

beams and is also reflected in observed increases in
deflections and/or decreases in apparent beam stiffness
(Zorn 2005). The increase in stress range with cycling may
be attributed to a) softening of concrete under repeated
compression loads (Neville 1975); and b) degradation of
the CFRP-to-concrete bond resulting in proportionally
more stress being carried by the reinforcing steel. Both
effects are in evidence in the present study.
6.2. Full-scale Specimen A-C
As indicated in Table 3, Specimen A-C was subjected to
2 million fatigue load cycles which subjected the bonded
CFRP to a stress range of approximately 0.03ffu. The

corresponding stress range in the lowermost internal
reinforcing steel was calculated to be approximately
99 MPa, below the expected endurance limit. The stress
range in the reinforcing steel of an unretrofitted
companion specimen (Aidoo et al. 2005) was 115 MPa.
Therefore the presence of the CFRP was effective in
relieving the reinforcing steel stress to some degree.
In this case, the stress ranges showed only a marginal
increase through the fatigue load history. During the
fatigue conditioning, little significant degradation was
visually observed in the specimen. The transient midspan
deflection increased from 8.1 mm in the first cycle (N = 1)
to 10.8 mm at N = 2,000,000. Additionally, there was a
1.5 mm residual deflection after N = 2,000,000. The
flexural stiffness of the beam degraded from 44.3 kN/mm
at N = 1 to 33 kN/mm at N = 2,000,000.
Damage accumulation curves, measured as
corresponding reinforcing bar and CFRP strains are
shown in Figure 6. The reinforcing bar strains are
calculated from linear instruments located on the side of
the concrete at the level of the bar; these instruments
have a gage length of 305 mm and thus capture average
strains. CFRP strains are determined from strain gages
located on the CFRP at the middle of the reinforcing bar
gage lengths (Aidoo et al. 2005). As can be seen from
Figure 6, apart from initial accumulation, there was little
damage accumulation through the 2,000,000 fatigue cycles.
785
ASE 9-6-05_Harries
4/12/06
5:10 pm
Page 786
1000
60
Rebar
800
Z-L4 monotonic
50
CFRP
600
400
0
500000
1000000
1500000
Number of cycles
2000000
Figure 6. Transient strains in specimen A-C during fatigue conditioning
The presence of more rapid initial damage accumulation

is an indication that the load level at which the girders
were cycled (DL+(LL+IM)) was rarely seen in the
42-year life of the bridge structure. This observation
was also supported by other tests conducted on
additional specimens from the same decommissioned
bridge (Aidoo et al. 2005).
Measured strains within the midspan region were used
to monitor the bond degradation in Specimen A-C. The
strains reported in Figure 6 correspond to the peak values
at the specified number of cycles. In Specimen A-C the
reinforcing bar strains are higher than the CFRP strains.
Since the CFRP is farther from the neutral axis than the
reinforcing steel, measured strains on the CFRP would
be expected to be greater than those at the level of the
reinforcing steel. The apparently lower CFRP strains, in
this case, resulted from the presence of cracks within the
gage length over which the reinforcing bar strains were
determined. These cracks increased in width as testing
progressed and resulted in local debonding of the CFRP,
thus recorded strains (from electrical resistance strain
gages on the CFRP) will be lower since they do not
correspond exactly to the location of the CFRP bridging
a crack. Similarly, CFRP strain gage measurements are
very local and will be lower if the gage does not
correspond to the exact location of the crack.
Although debonding was not visually observed
during the fatigue load cycles, relative slip or debonding
of the CFRP is evident as indicated by the increase in
strain accumulation being greater for the reinforcing
steel than for the CFRP. As shown in Figure 6, the
CFRP strains are relatively constant or decrease through
the load history while the reinforcing bar strains
increase to some degree. This behaviour is an indication
of incomplete stress transfer between the CFRP and
concrete substrate (and ultimately the reinforcing steel).
Thus relative slip is occurring.
7. EFFECTS OF FATIGUE CONDITIONING
Five of the fatigue specimens discussed above, having
achieved 2,000,000 cycles of fatigue loading, were
tested monotonically to failure (designated F in Table 4).
A summary of these tests and companion monotonic
786
Applied load at midspan (kN)
Strain, microstrain
Z-L4 fatigue
40
Unretrofit control specimen

30
20
Z-L4 fatigue conditioning
10
N = 1 to 2,000,000
0
0
20
40
60
Midspan deflection (mm)
80
100
Figure 7. Load-deflection behaviour of Z-L4-series specimens
tests having no fatigue conditioning (M in Table 4) is

reported in Table 4. (It is noted that the A-series
specimens, having come from a decommissioned bridge,
had some unknown fatigue history prior to testing. For
the sake of this work, all girders were assumed to have
the same history, thus all loading described here is in
addition to these extant conditions.)
For the relatively low CFRP stress ranges considered,
there was no notable effect on the stiffness or capacity
of the retrofitted beams following fatigue conditioning.
Although deflection results vary, the displacement
ductility (measured as the ratio of the peak displacement
to that at yield) was observed to decrease marginally due
to fatigue loading. This effect is primarily a reflection
of the permanent deformation following fatigue
conditioning (all displacement values are measured
from their pre-fatigue-conditioned baseline). The
representative load-deflection behaviour of Specimen
Z-L4 is shown in Figure 7.
Following fatigue conditioning, Specimen A-C was
tested monotonically to failure. The ultimate capacity
and stiffness of the member was essentially unchanged
from that of a monotonically tested companion. The
ultimate displacement, however, was only 56% of that
observed in the companion specimen. The complete
load-deflection responses of the A-C specimens are
shown in Figure 8.
7.1. Effectiveness of CFRP Retrofits measures
As a point of comparison, Figures 7 and 8 also show
load deflection behaviour of companion unretrofitted
control specimens. All Z- and Q-series specimens
demonstrated a marked improvement in load carrying
capacity accompanied by a reduction in displacement
capacity as represented by the Z-L4 specimens shown
ASE 9-6-05_Harries
4/12/06
5:10 pm
Page 787
900
A-C monotonic
Applied load at midspan (kN)
800
700
A-C fatigue
600
500
Unretrofit control specimen

400
A-C fatigue conditioning
300
N = 1 to 2,000,000
200
100
0
0
25
50
75
100
125
150
175
200
Midspan deflection (mm)
Figure 8. Load-deflection behaviour of A-series specimens
in Figure 7. This behaviour is typically of lightly

reinforced members (longitudinal steel reinforcement
ratio = 0.010) having a proportionally large increase in
reinforcement ratio resulting from the retrofit; in the
case of Z-L4, a 28% increase (to 0.0128).
The improvement in load carrying capacity of the
A-specimen was less significant while the reduction in
displacement capacity was quite significant (Figure 8).
Nonetheless, the improved performance, in this case is at
load levels immediately following yield of the specimen
where load carrying capacity is significantly increased.
This behaviour is expected in cases where the
longitudinal reinforcing ratio is high (0.023, in this case).
In the A-specimens, the addition of CFRP only increased
the longitudinal reinforcing ratio 3% to 0.0238.
7.2. Debonding of CFRP
All specimens tested monotonically to failure exhibited
debonding. The CFRP strains corresponding to
debonding are given in Table 4. It is noted that in all but
specimen Z-L2, some degree of debonding was evident
following the fatigue conditioning portion of the
loading. The results shown in Table 4 must be
interpreted carefully. The high modulus adhesive H
(Table 2) used in Specimens Z-H2x1 and Z-H4 resulted
in very low debonding strains for both the monotonic
and fatigue conditioned specimens. In these cases, there
was no apparent deleterious effect from the fatigue
cycling. In the other fatigue conditioned tests, however,
the lower modulus adhesives used (L and T in Table 2)
did exhibit marked reductions in the CFRP strain at
debonding as compared to their non-fatigue conditioned
companion specimens. Indeed, any improvement in
bond behaviour realised by the use of the softer
adhesives (Reeve 2005) was counteracted by the effects
of fatigue loading. The stiffer adhesive demonstrated no
such apparent degradation. It is noted, nonetheless, that

the lower modulus adhesives, L and T, regardless of the
effects of fatigue, outperformed the stiffer adhesive H;
requiring higher CFRP strains to initiate debonding.
8. FATIGUE GUIDELINES
Throughout this study debonding was observed at CFRP
strains notably lower than the American Concrete Institute
(2002) or Concrete Society (2004) recommended strain
limits (Table 1) even in the monotonic tests. These
limits appear to be non-conservative and clearly do not
account for issues of bond. In all cases reported,
including the monotonically loaded specimens, the ACI
(2002) stress level recommended for the mitigation of
fatigue failure of the CFRP of 0.55ffu was never
achieved.
The reduction factor approach proposed to account
for fatigue loading recommended by JSCE (2001) and
the Italian (National Research Council (Italy) 2004)
document appear to be more appropriate provided that
the debonding stress/strain determined for monotonic
loading is adequately determined. Nonetheless, the
results of the present work suggest that any reduction
factor associated with fatigue loading conditions is
dependent on properties of the adhesive system. Further
study is required to address this issue.
9. CONCLUSIONS
The limited study reported here indicates that behaviour
of externally bonded CFRP systems deteriorates when
subject to fatigue loading. The fatigue load levels
provided resulted in relatively large stress ranges on the
internal reinforcing steel and low stress ranges on the
significantly stronger (and somewhat less stiff) CFRP
retrofit materials.
In Specimens Q-CL and Q-CH, the fatigue stress
787
ASE 9-6-05_Harries
4/12/06
5:10 pm
Page 788
ranges in the CFRP did not exceed 0.14ffu although

clearly resulted in a significant accumulation of damage
affecting the behaviour of the member. In Specimen
A-C, a fatigue stress range in the CFRP as low as 0.04ffu
appears to have also affected the ultimate performance
of the member. These observations demonstrate that
under the fatigue loading conditions considered, the
durability of the bond and concrete substrate is critical
and the excellent fatigue performance of the CFRP itself
cannot likely be developed. Additionally, even very low
stress ranges result in degradation suggesting that there
is no (or at least a very low) endurance limit below
which fatigue-induced degradation is no longer a
concern. For the system investigated, fatigue induced
deterioration was measured as a reduction in CFRP
strain at the onset of debonding. Significantly more
study is required to adequately quantify the effects
described in this work.
ACKNOWLEDGEMENTS
Some experimental tests reported herein were carried
out with support from the South Carolina Department of
Transportation and the United States Federal Highway
Administration. Other tests were supported by the
University of Pittsburgh. The contributions of Fyfe
Company and Sika Corporation are also gratefully
acknowledged.
REFERENCES
3rd
AASHTO (2004). LRFD Bridge Design Specifications,

edition,
American Association of State Highway and Transportation
Officials.
ACI Committee 440 (2002). ACI 440.2R-02 Guide for the Design and
Construction of Externally Bonded FRP Systems for Strengthening
Concrete Structures, American Concrete Institute, 45 pp.
Aidoo, J., Harries, K.A. and Petrou, M.F. (2006). Full-scale
experimental investigation of repair of reinforced concrete
interstate bridge using CFRP materials, Journal of Bridge
Engineering, ASCE, Vol. 11, No. 3 (in press).
Aidoo, J., Harries, K.A. and Petrou, M.F. (2004). Fatigue behaviour of
CFRP strengthened reinforced concrete bridge girders, Journal of
Composites in Construction, ASCE ,Vol. 7, No. 6, pp 501518.
Barnes, R.A. and Mays, G.C. (1999). Fatigue performance of
concrete beams strengthened with CFRP plates, Journal of
Composites for Construction, ASCE ,Vol. 3, No. 2, pp. 6372.
Concrete Society (2004). Design Guidance for Strengthening
Concrete Structures Using Fibre Composite Materials, Technical
Report 55. Camberly, UK.
Curtis, P.T. (1989). The fatigue behaviour of fibrous composite
materials, Journal of Strain Analysis, Vol. 24, No. 4, pp. 235244.
Federation Internationale du Beton (fib) (2001). fib Bulletin 14:
Externally Bonded FRP Reinforcement for RC Structures.
788
Harries, K. (2005). Fatigue behaviour of bonded FRP used for

flexural retrofit, Proceedings of International Symposium on
Bond Behaviour of FRP in Structures (BBFS 2005), Hong Kong,
December, pp. 555560.
Heffernan, J. P. (1997). Fatigue Behaviour of Reinforced Concrete
Beams Strengthened with CFRP Laminates, Ph.D. Dissertation,
Department of Civil Engineering, Royal Military College of
Canada, Kingston, Ontario, May, 157 pp.
Japan Society of Civil Engineers (JSCE) (2001). Recommendations
for the Upgrading of Concrete Structures with Use of Continuous
Fiber Sheets, Concrete Engineering Series 41, Tokyo, 250 pp.
(available in English on CD).
Mandell, J.F. (1982). Fatigue behaviour of fiber resin composites,
Developments in Reinforced Plastics, Vol. 2, pp. 67107.
Masoud, S., Soudki, K. and Topper, T. (2001). CFRP-strengthened and
corroded RC beams under monotonic and fatigue Loads, Journal of
Meier, U., Deuring, M., Meier, H. and Schwegler, G. (1993).
Strengthening of structures with advanced composites, Alternate
Materials for the Reinforcement and Prestressing of Concrete,
J.L. Clarke, editor, Blackie Academic and Professional, Glasgow.
National Research Council (Italy) (2004). CNR-DT 200/2004 Guidelines
for Design, Execution and Control of Strengthening Interventions by
Means of Fibre-reinforced Composites, Advisory Committee on
Technical Regulations for Construction, Rome, 204 pp.
National Research Council (1991). Life Prediction Methodologies
for Composite Materials, Committee on Life Prediction
Methodologies for Composites NMAB 460, National Materials
Advisory Board, Washington DC, 66 pp.
Neville, A.M. (1975). Properties of Concrete, 2nd edition, Pitman.
Oehlers, D.J. (2005). Generic debonding mechanisms in FRP plated
beams and slabs, Proceedings of the International Symposium
on Bond Behaviour of FRP in Structures, December 7-9, 2005,
Hong Kong. pp. 5344.
Papakonstantinou, C.G., Petrou M.F. and Harries, K.A. (2001).
Fatigue of reinforced concrete beams strengthened with GFRP
sheets, Journal of Composites for Construction, ASCE, Vol. 5,
No. 4, pp. 246253.
Quattlebaum, J., Harries, K.A. and Petrou, M.F. (2005).
Comparison of three CFRP flexural retrofit systems under
monotonic and fatigue loads, Journal of Bridge Engineering,
ASCE, Vol. 10, No. 6, pp. 731740.
Reeve, B. (2005). Effect of Adhesive Stiffness and CFRP Geometry
on the Behaviour of Externally Bonded CFRP Retrofit Measures
Subject to Monotonic Loads, M.Sc. Thesis, University of
Pittsburgh. December.
Roylance, M. and Roylance, O. (1981). Effect of moisture on the
fatigue resistance of aramid-epoxy composite, Organic Coatings
and Plastics Chemistry, Vol. 45, pp. 784788.
Sebastian, M.W. (2001). Significance of midspan debonding failure
in FRP-plated concrete beams, Journal of Structural Engineering,
ASCE, Vol. 127, No. 7, pp. 792798.
Shahawy, M. and Beitelman, E.T. (1999). Static and fatigue
ASE 9-6-05_Harries
4/12/06
5:10 pm
Page 789
performance of RC beams strengthened with CFRP laminates,

Journal of Structural Engineering, ASCE, Vol. 125, No. 6,
pp. 613621.
Tilly, G.P. and Moss, D.S. (1982). Long Endurance Fatigue of Steel
Reinforcement, IABSE Reports, International Association for
Bridge and Structural Engineering, 37, Zurich, Switzerland,
pp. 229238.
Zorn, A. (2006). Effect of Adhesive Stiffness and CFRP Geometry on the
Behaviour of Externally Bonded CFRP Retrofit Measures Subject to
Fatigue Loads, M.Sc. Thesis, University of Pittsburgh. April.
APPENDIX: NOTATION
a
shear span length
DL dead load (AASHTO 2004)
Ea tensile modulus of adhesive
Ef tensile modulus of FRP
fau strength of adhesive
fc concrete strength
ffu ultimate capacity of FRP
ffud stress limit in FRP to mitigate debonding under
monotonic load conditions
ffuf stress limit in FRP under fatigue load conditions
fu reinforcing tensile strength
fy reinforcing yield strength
h
overall height (depth) of beam
IM impact load (AASHTO 2004)
LL live load (AASHTO 2004)
N number of cycles of fatigue loading
Nf number of fatigue cycles to failure
S
stress range
ta
thickness of adhesive
tf
thickness of FRP
au rupture strain of adhesive
fu rupture strain of FRP
789
ASE 9-6-05_Harries
4/12/06
5:10 pm
Page 790
ASE 9-6-06_Zhang
4/12/06
5:09 pm
Page 791
Advances in Adhesive Joining of Carbon

Fibre/Polymer Composites to Steel Members
for Repair and Rehabilitation of Bridge Structures
L. C. Hollaway1,*, L. Zhang2, N. K. Photiou1, J. G. Teng2 and S. S. Zhang2
1School
2Department
of Engineering, University of Surrey, Guildford, Surrey, UK

of Civil and Structural Engineering, The Hong Kong Polytechnic University, Hong Kong, China
Abstract: Generally, it is relatively easy to obtain a high joint strength with most
modern bonding systems if these are cured under ideal factory conditions. However,
civil engineering construction joints are cured on site and are required to last the
lifetime of the structural member in harsh environmental conditions. This implies that
site joints might not be constructed and cured as well as those fabricated in the factory.
The paper compares two possible methods for bonding an FRP composite
patch/plate to a steel adherend suitable for construction sites. The first method utilises
the accepted technique of bonding two dissimilar materials using a two-part cold cure
adhesive. The second method employs a pre-impregnated FRP composite plate in
conjunction with a compatible film adhesive; these comparisons are undertaken by
examining the results of double-strap butt joint tests. A possible site technique using
the pre-impregnated FRP composite to upgrade a steel beam whilst it is under a low
frequency vibration load is investigated; this represents the repair of a steel bridge
constantly under traversing traffic.
The butt joint test results show that on average, the pre-impregnated composite in
conjunction with the film adhesive leads to a failure load which is 15% higher than that
of the cold setting adhesive technique. The results of the rehabilitated beam tests show
that the bonded joint between the pre-impregnated CFRP composite and the steel
adherend did not suffer any significant damage from the low frequency vibrations
imposed upon the steel beam during the cure period.
Key words: Polymer composites, pre-impregnated composites, carbon fibres, glass fibres, film adhesive, cold
cure adhesives.
1. INTRODUCTION
Currently, one of the major uses of fibre reinforced
polymer (FRP) composites in the civil infrastructure is
in the repair and rehabilitation of highway and railway
bridge structures constructed from concrete and steel
materials. It has been estimated that in the USA, more
than one third of all bridges are substandard and of these
more than 43% are made from steel (Tavakkolizadeh
and Saadatmanesh 2003). The latest National Bridge

Inventory (NBI) states that steel bridges rank high in the
group recommended for improvement; a similar
situation exists in Europe. In the UK, the majority of
steel and composite bridges on the National Trunk Road
Network, managed by the Highways Agency, are
relatively new, but a large number of old (those over
100 years old) wrought iron and early steel structures
*Corresponding author. Email address: L.Hollaway@surrey.ac.uk; Fax: +44-1483-450984; Tel: +44-1483-689280.
791
ASE 9-6-06_Zhang
4/12/06
5:09 pm
Page 792
Advances in Adhesive Joining of Carbon Fibre/Polymer Composites to Steel Members for Repair and Rehabilitation of Bridge Structures
are to be found on the railway and canal network. The

major degradation problem with steel bridges is
corrosion, lack of proper maintenance and fatigue due to
traffic vibrations on the bridge. As the cost associated
with repair and rehabilitation is far less than that of
replacement, it is clear that the former solution, which,
in addition to being more cost-effective, takes much less
time to complete and causes less disruption to the
traffic, should be investigated before a more drastic
decision is taken to replace these bridges.
The superior mechanical, fatigue and life-time
properties of carbon fibre reinforced polymer (CFRP)
composites, of both high modulus (H-M) and ultra-high
modulus (UH-M) fibres, make them ideal candidates
for the repairing and strengthening of steel bridge and
building structures. However, to date, only a limited
amount of research appears to have been undertaken
(Mertz and Gillespie 1996; Mosallam and Chakrabarti
1997; Liu et al. 2001), although there is clear evidence
that such studies are on the increase (Hollaway et al.
2004; Cadei et al. 2004; Schnerch et al. 2004; Schnerch
et al. 2006). These rigid composite plates, which would
generally be bonded to the soffit of a structural steel
member using a two-part cold cure adhesive, are
manufactured in a factory by the pultrusion or the
pre-impregnation (prepreg) techniques. The correct
installation of these composite plates is a critical factor
in the long term performance of the upgraded beam.
All adhesive joints used in the civil engineering
industry must be durable and capable of maintaining
effective performance throughout their service life,
which will extend over many years. Consequently,
careful selection of the materials to be used and the
bonding techniques to be employed must be made.
Knowledge of the environment where the polymer joint
will be used must also be available. Therefore, the five
important areas to be considered at the design stage of
the repair procedure for metallic structures, especially
for the flexural (and shear) upgrading of metallic
railway bridge structures, using FRP composites are:
(i) the most relevant adhesive polymer for the
fabrication technique and a knowledge of its
mechanical and in-service properties extending
over a number of years;
(ii) the relevant FRP composite plate for the repair
method and its mechanical and in-service
properties;
(iii) the pre-treatment technology for the substrate;
(iv) the fabrication method used for the manufacture
of the FRP plate and the method of bonding it to
the substrate; and
(v) the possible vibrational forces which metallic
bridge structures may be subjected to.
792
The aim of this paper is to examine the five items above,

particularly in relation to a new composite plate and
bonding system (i.e. the prepreg/film adhesive system)
to be introduced into the construction industry. The
characteristics of this system are compared with those of
the more conventional technique of bonding a factory
made pre-cast FRP plate with a two-part adhesive (i.e.
the rigid plate/two-part adhesive system), by examining
the test results of:
(i) double-strap butt joints composed of a
discontinuous steel member and two FRP cover
plates; and
(ii) steel flexural members bonded with composite
prepregs.
2. ADHESIVE POLYMERS
Adhesive polymers are two-part systems, namely the
resin and the polymerising agent; adhesives used on site
are generally cold cure systems. These adhesives take
several days to reach about 90% polymerisation and
should, ideally, be post-cured at a higher temperature;
the length of time of this cure depends upon the value of
the post-cure temperature. The curing process occurs
within the bulk of the adhesive, although adhesion
occurs at the interface of the two adherends to be joined.
Van der Waal's forces also contribute to the adhesion, as
these are the normal attraction between atoms and
molecules. However, these forces are in addition to
chemical bonding, mechanical interlocking, diffusion
and electrostatic attraction, which also play a part in
forming the joint of the adherends. Ideally, for
construction sites, the cold cure epoxies should be post
cured to a temperature of at least 50C for complete
polymerisation and should also be toughened to
avoid brittle behaviour. There are two important
considerations which must be addressed when the
technique of adhesive bonding is being considered.
These are: (a) the mechanical properties of the adhesive
polymer, particularly the glass transition temperature
(Tg), and (b) the proposed environment and projected
life-span of the adhesive polymer.
All amorphous polymers, (epoxy systems normally
fall within this category), have a Tg value. Below this
temperature they are relatively hard and inflexible and
are normally described as 'glassy', but above it they tend
to be soft and flexible and are described as being
rubbery. Clearly, all adhesives should be maintained at
a temperature below their Tg and should not be exposed
to an environment which fluctuates above and below
this critical temperature. Most epoxy adhesive polymers
used in civil engineering will have a Tg value of about
15C to 20C above their cure temperature; if these
polymers are then post cured say to 50C their Tg value
ASE 9-6-06_Zhang
4/12/06
5:09 pm
Page 793
L.C. Hollaway, L. Zhang, N.K. Photiou, J.G. Teng and S.S. Zhang
will rise to a maximum value of 70C. However, there

is a limiting value to which the Tg can rise irrespective
of the post cure temperature; this will depend upon the
adhesive used. Consequently, it is important to take into
account the glass transition temperature property of any
adhesive used in construction.
The main area that should be considered when
repairing (rehabilitating) steel bridges is the local
environment into which the structure is located. The
ingress of moisture, particularly at the edges of the joint,
can cause serious loss of strength across the entire joint.
Analytically, it can be shown that the thicker the
adhesive the better the load transfer is under shear,
although overall joint stiffness is decreased. However,
as the adhesive thickness increases so does the likely
occurrence of bond-line porosity; this decreases the
shear and peel strengths markedly over the life-span of
the joint. The use of silanes to increase the durability of
the steel/epoxy bond does not affect the initial bond
strength. However, the compatibility between the
adhesive polymer and the silane is an important
property to be considered as was shown by the early
work of Gettings and Linlock (1972), who found that
premixing with -glycidoxpropyltrimethoxysilane (-GPS)
considerably improved the durability of grit blasted
steel joints, whereas two other silanes did not.
3. THE FRP COMPOSITE PLATES USED
CURRENTLY IN REPAIR
The two techniques generally used in construction are:
(i) the wet lay-up system where the adhesive polymer is
also the matrix material for the composite, and (ii) a
rigid factory-made FRP plate, either a pultruded or a
pre-impregnated (prepreg) plate; both plates are cured at
an elevated temperature of between 120C and 135C
and, in addition, the pultrusion would have a high level
of consolidation in the die and the prepreg plates would
be cured under vacuum. These two FRP plate systems
would be bonded in place using a two-part cold cure
adhesive; the temperature of cure would be equal to the
ambient temperature of the proposed site. The wet layup system would use a cold cure polymer (the combined
adhesive and matrix).
For the upgrading of structural steel members, the fibre
component of the FRP material, used for the manufacture
of the upgrading plates, would be either the H-M or the
UH-M carbon fibre; the polymer would be either an
epoxy or, perhaps, a vinylester material.
4. PRE-TREATMENT OF THE SUBSTRATE
To achieve reliable intermolecular forces between the
substrate and the adhesive, it is necessary to pre-treat the
substrate; this operation would require different
treatments for adherends of different materials. For

metals including aluminium, the pre-treatment would
range from a mechanical/physical to a chemical method.
The treatment for concrete, which requires careful
considerations depending upon the state of the concrete,
usually consists of a mechanical grit-blasting method.
Fibre reinforced polymer (FRP) composites are highly
polar and are, therefore, very receptive to adhesive
bonding. Pre-formed FRP composite plates generally
have a peel-ply covering the surface to be bonded,
which is removed immediately before bonding; this
will reveal a clean and rough composite surface. The
surfaces of the adherend must be checked every time
before bonding.
Problems can be experienced when bonding
dissimilar materials to metals, particularly in the civil
engineering industry, where the adhesive joint is made
in service, possibly in a hostile environment, and
required to be functional for many years. They have to
be pre-treated carefully, i.e. usually solvent degreasing,
abrasion and grit blasting, immediately followed by a
primer solution in the form of a thin organic coating,
which is painted onto the substrate to prevent oxidation
of the cleaned surface. In addition, as the primer has a
much lower viscosity than the adhesive, it can achieve
greater contact with the substrate. The pre-treatment of
aluminium includes etching and anodising procedures
and, therefore, by definition, causes chemical
modification to the surfaces involved. A chemical pretreatment could be applied to steels; preliminary tests
have indicated that it might be a more appropriate
treatment than grit-blasting the substrate, especially
when strengthening using the pre-impregnation method,
which is described in Section 5 of this paper. The
procedure involves degreasing the substrate using a
suitable solvent (e.g. MEK Ensolv or white spirit),
which is applied by hand wipes or by spray, but only if
the local environment is suitable. The solvent would
then be removed by wiping with clean cloth, with the
final traces of solvent then allowed to evaporate.
Alternatively, the solvent could be removed by the
application of warm air. The surface could then be
treated with an industrial methylated spirit, or with a gel
containing phosphoric acid and a thixotropic agent.
The gel is then applied to the substrate using a
polythene, polypropylene or fluorocarbon bristle
brush and allowed to set for 10 minutes at ambient
temperature. After this time, the substrate is then
cleaned with clean tap water, (not de-ionised water).
After this operation, it is imperative that all water
particles are removed from troughs and depressed areas
of the steel. This treatment not only cleans the treated
surface, but builds a thin, passive oxide onto the metal
793
ASE 9-6-06_Zhang
4/12/06
5:09 pm
Page 794
surface. Often, on construction sites, chemical treatments

are discouraged because of the safety regulations related
to site operations, more so when using acids. The
chemicals described above are contained within
applicators and the treatment is, therefore, no more
dangerous than any other site operations provided all
necessary precautions are taken regarding hazardous
liquids/vapours and a risk assessment is made. However,
this form of treatment is only undertaken in very
special circumstances.
5. THE PROPOSED PREPREG/FILM
ADHESIVE SYSTEM
The laboratory tests used in the investigation to compare
the curing of the prepreg/film adhesive system with that
of the rigid plate/two-part adhesive system in upgrading
a structural steel member, included (i) double-strap butt
joint tests, and (ii) steel beam tests.
5.1. The FRP Prepreg Composite Material
The Advanced Composite Group Ltd. (ACG)
Derbyshire, UK, has developed a prepreg carbon fibre
composite for the aerospace, automotive and marine
industries. ACG is now introducing these systems into
the construction industry and has developed two
matrices and a compatible film adhesive, particularly
suited for use in civil engineering. In conjunction with
these two matrices, the high modulus (H-M) and ultrahigh modulus (UH-M) carbon fibre composites are
produced. These material systems present potential new
repair techniques for construction and could have
considerable advantages over the rigid plate/two-part
adhesive system.
This FRP composite is a hot-melt factory-made
prepreg; the compatible film adhesive is also factory
made. The cure procedure when the two components are
brought together on site. Two types of fibre are used in
the prepreg composite, namely a high modulus (H-M)
carbon fibre and an ultra-high- modulus (UH-M) carbon
fibre in an epoxy matrix. In this paper, the bonding
characteristics of double-strap (H-M) CFRP prepreg
composite butt joints, fabricated using the in-situ cure
technique for the prepreg/film adhesive system, are first
compared with that of the rigid plate/the two-part
adhesive system. Factory-made (H-M) CFRP prepreg
composite was used for the FRP cover plates and these
were bonded onto the discontinuous steel substrate with
Sikadur 31. Following this, comparisons of the strengths
of double-strap butt joints constructed from the H-M
CFRP and the UH-M CFRP prepreg composites using
the prepreg/film adhesive method are also made. Two
cure procedures are considered in these comparisons,
one is that of 65C for 16 hours and the other is that of
794
80C for 4 hours.

5.2. The Component Parts and Fabrication
of the Prepreg System
The fabrication technique for the ACG prepreg
composite is similar to that employed in the aerospace
industry for patch repair of aerospace systems. The
matrix for the present system, however, has been
specifically developed to take account of the civil
engineering requirements. Furthermore, when the ultra
high-modulus (UH-M) carbon fibre is used in a civil
engineering application, it is manufactured from the
pitch precursor; the aerospace industry uses the
polycarbonitrile (PAN) fibre precursor. The pitch fibre is
a comparatively low-cost fibre, but the non-uniformity of
the manufactured fibre, which occurs between batches, is
a problem in the aerospace industry, but it is acceptable
in the construction industry. The prepreg/film adhesive
methodology consists of a factory-made hot melt prepreg
composite and a compatible film adhesive. Both
components are then transported to site and fabricated
onto the structural member and cured under an elevated
temperature and a vacuum assisted pressure of 1 bar.
This procedure enables the composite and the film
adhesive to be compacted and cured in one operation.
Furthermore, the vacuum applied to the system extracts
the air and reduces the volume of voids in the
composite to a minimum. A further advantage of this
system is that the polymer materials are cured at an
elevated temperature, one which ensures that the Tg
is as high as it is practically possible to achieve on
site. The value of the Tg is dependent upon the cure
temperature of this system.
During the normal repair of a civil engineering
structure with the ACG prepreg system, this temperature
would be 65C for 16 hours, which is acceptable to the
civil engineering industry. If it is convenient and
possible for an on-site operation, this composite
material can be cured at 80C for 4 hours. The 4-hour
cure procedure is a major advantage for, say, a railway
bridge structure that is exposed to constant vibrations
during normal working conditions, but which can be
closed to train movement for four hours, perhaps during
the night; longer closure periods may incur heavy
financial penalties to the contractor. However, Zhang
et al. (2006) have shown that the 16-hour cure procedure
is also relevant for this loading condition. The latter
process may even be more advantageous because the
composite/adhesive system is cured under vacuum and
any vibrations set up by the train motion may assist in
the extraction of any air from voids. Furthermore, the
cold cure adhesives currently used in construction take
5 to 7 days, depending upon the site temperature, to
ASE 9-6-06_Zhang
4/12/06
5:09 pm
Page 795
reach 90% polymerisation. After this time, if the system

is not post-cured, polymerization will continue, but over
a very long time and will probably never reach 100%;
the vibrations will impair the polymerisation mechanism
of the cold cure adhesive polymer.
5.3. Materials Used in the Double-Strap
Butt Joint Tests
5.3.1. Prepreg composite materials
and film adhesive
The composite materials used in this investigation were
manufactured from an epoxy polymer and unidirectionally
aligned (UH-M) carbon fibres, with a modulus of elasticity
of 640 GPa at 600 g/m2 or unidirectionally aligned
(H-M) carbon fibres, with a modulus of elasticity of
230 GPa at 600 g/m2. The materials used have the
following specifications:
The UH-M CFRP
-V T M 2 6 4 F R B / K 6 3 7
12-600-35%RW
The H-M CFRP
-VTM264FRB/PANEX3550K-600-35%RW
The 45 biaxial GFRP
-VTM264FRB/EBX60232%RW
Adhesive film
-VTA260/PK13-313g(compatible film adhesive
with the composite matrix
VTM264FRB)
The geometrical and mechanical properties of the
composite materials and the film adhesive are given in
Table 1. Advanced Composite Group Ltd. (ACG) at
Derbyshire, UK manufactured all prepregs and the film
adhesive used in this work.
The fibre volume fraction for the CFRP composites
was 56%. This is a lower percentage compared with that

generally used for pultruded FRP composite plates for
rehabilitation.
5.3.2. Two-part cold cure adhesive
The two-part cold cure adhesive used in this
investigation was Sikadur 31, a product of Sika Ltd.,
Welwyn Garden City, Herts, UK. This is a typical
thixotropic epoxy resin adhesive, widely used in the
construction industry. The mechanical properties of this
adhesive are given in Table 2.
Initially, two adhesive thicknesses were investigated
using double strap joint tests, namely 1.0 mm, and
0.5 mm; the failure loads of the joints were very similar.
However, on civil engineering sites it is not recommended
to use a thickness lower than 1 mm for a two-part
adhesive due to the difficulty of forming a satisfactory
joint. For the controlled laboratory manufacture of
the joint and to provide a more rational comparison
with the 0.1 mm film adhesive, the decision was made
to use 0.5 mm thick adhesive joints. In addition,
the application of one or two layers of film was
investigated independently (Photiou et al. 2003), and it
was concluded that there is no benefit, when using the
cured-on-site prepregs for steel substrates, to have more
than one film layer.
5.3.3. Steel plates
Two types of thin steel plates, type A and type B, were
used in the experimental work on double-strap joints;
each had slightly different mechanical characteristics.
Type A, which was used at The University of Surrey,
UK, was manufactured in the UK. Type B, which was
Table 1. Mechanical properties of FRP materials and film adhesive
Material
UHM-CFRP (Unidirectional)
HM-CFRP (Unidirectional)
GFRP (45 to line of action
of load)
Film Adhesive
Thickness
per layer
(mm)
Tensile
strength
(MPa)
0.3
0.6
0.4
1120
2110
215
0.1
32
Elastic
modulus
(GPa)
270
135
16
3.7
Ultimate
strain
(%)
Poisson
ratio
0.4
1.6
1.7
0.32
0.28
0.15
0.9
0.37
Table 2. Properties of Sikadur 31*
Material
Tensile
strength
MPa
Modulus of
elasticity
GPa
Shear
strength
MPa
Temperature
range
C
Sikadur 31
14.8
6.87.3
21
530
*Abstracted from the manufacturer's data sheets.
795
ASE 9-6-06_Zhang
4/12/06
5:09 pm
Page 796
used at The Hong Kong Polytechnic University, China,

was manufactured in Hong Kong. It was difficult to
obtain an exact match for the two types. Type A steel
was characterised by a modulus of 205 GPa and a 0.2%
proof stress of 700 N/mm2 and a limit of proportionality
at approximately 550 N/mm2, whereas type B steel had
a modulus of 200 GPa, a yield stress of 430 MPa and an
ultimate strength of 660 MPa.
5.4. Surface Preparation of Specimen
5.4.1. Steel surface preparation for bonding
The preparation of the types A and B steel adherends
was undertaken as discussed in Section 4 for site
work. Initially, both types of adherends were solvent
degreased; then Type A was subjected to grit blasting
using angular chilled iron grit (Guyson 1989) of 0.18
nominal particle size and Type B was subjected to
sandblasting at a pressure of 0.6 MPa. Finally,
immediately before applying the adhesive, both types
were solvent degreased again. In this study, the surface
was not primed and due to laboratory restrictions the
bonding operation was not undertaken for two days after
the surface preparation; however, the surface was
protected from the laboratory environment by covering
it with polyethylene.
5.4.2. Preparation of the rigid FRP plates
for use with the two-part cold
cure adhesive
The composite prepregs were stored at 20C
immediately after its factory manufacture and were
maintained at this temperature until required for use.
Before cutting the prepreg samples to size, the material
was allowed to thaw for approximately 2 hours. In the
present study, only three double-strap joint tests of
identical geometry and materials were conducted on
the rigid plate/two-part adhesive system; the rigid
composite cover plates used in these tests were each
formed from one 0.6 mm thick H-M CFRP prepreg
layer. The first stage was to fabricate the plates by
covering the prepreg layers with a Halar film and a
breather blanker before the whole unit was placed in a
vacuum bag. The bag, under a vacuum-assisted pressure
of 1 bar, was then exposed to a temperature of 65C
for 16 hours. After the cure cycle was completed, the
temperature of the oven and the sample were reduced to
room temperature at a steady rate. The plates were then
cut to the required size in preparation for forming the
joint. The surfaces of the sides to be bonded were lightly
roughened with 'Emery' cloth and solvent washed in
preparation for the bonding operation. Small glass
spheres of nominal diameter of 0.5 mm were used to
obtain the required thickness of adhesive for each joint.
796
During the manufacture of the joint, a jig was used to

place the CFRP cover plates on both sides of the
discontinuous steel adherend. Weights were applied to
the top horizontal surface of the sample to stabilise the
joint during polymerisation of the adhesive.
5.4.3. Preparation of prepreg CFRP plates
for use with film adhesive
The composite plate samples and the film adhesive were
cured in one operation following the same preparation
procedure as described above, except that, before curing,
one H-M or two UH-M CFRP prepreg layers were
placed on each of the two sides of the discontinuous steel
adherend, with the adhesive film positioned between the
two composite plates and the discontinuous steel
adherend. The joints were then exposed to the same
pressure and cure temperatures as discussed above. In
some of the specimens, two GFRP layers, one between
the adhesive film and the steel adherend and another one
as the outermost layer, were included.
5.5. Failure Modes and Ultimate Loads
of the Double-Strap Butt Joint Tests
This section discusses the results of double-strap
butt joint tension tests, covering the two plate bonding
systems and the two different curing procedures for the
prepreg/film adhesive system:
(i) The rigid plate/two-part adhesive system
(Sikadur 31);
(ii) The prepreg/ film adhesive system, cured under
a vacuum assisted pressure of 1 bar, and with
the final stage cure being at 65C for 16
hours, for both H-M CFRP and UH-M CFRP
composites; and
(iii) The prepreg/ film adhesive system, cured under
a vacuum assisted pressure of 1 bar, and with the
final stage cure being at 80C for 4 hours, for
both H-M CFRP and UH-M CFRP composites.
Table 3 summarises the failure loads (peak loads
reached in the tests) of all 10 series of tests:
Series 1:
The rigid plate/two-part adhesive
system3 mm thick Type A steel plate
and H-M CFRP prepreg cover plates.
Series 2:
The prepreg/film adhesive systemfinal
stage curing at 65C/16 hours, 3 mm
thick Type A steel plate and H-M CFRP
prepreg cover plates.
Series 3:
thick Type B steel plate and H-M CFRP
Series 4:
B
C
A
B
C
A
B
C
A
B
C
A
B
C
A
B
C
A
B
C
A
B
C
H-M CFRP/GFRP
H-M CFRP/GFRP
UH-M CFRP/GFRP
UH-M CFRP/GFRP
UH-M CFRP
UH-M CFRP
H-M CFRP
H-M CFRP
Adhesive film
Adhesive film
Adhesive film
Adhesive film
Adhesive film
Adhesive film
Adhesive film
Adhesive film
Adhesive film
Sikadur 31
80C/4 hours
65C/16 hours
80C/4 hours
65C/16 hours
80C/4 hours
65C/16 hours
80C/4 hours
65C/16 hours
65C/16 hours
Ambient
temperature
* Abstracted from Photiou et al. (2006);

Low value due to experimental eccentricity of load; ignored in calculating the average failure load.
The failure modes have been estimated by visual observation of the failed specimens.
10
H-M CFRP
H-M CFRP
Adhesive
Type
Thickness
(mm)
37.02
36.57
42.0
46.0
32.0
25.4
29.5
31.8
43.93
32.89
43.46
39.4
35.4
34.4
44.5
49.0
47.0
41.16
38.18
40.71
43.46
37.71
43.23
32.2
29.5
29.1
36.4
33.1
36.4
34.26
Failure
load
(kN)
41.47
40.02
46.83
36.4
40.23
28.90
44.0
35.95
35.3
30.27
Average
Failure load
(kN)
Debonding between the CFRP

and the GFRP followed by
rupture of composite plates
Debonding between the CFRP
and the GFRP followed by
Rupture of composite plates
Interfacial debonding and

100% interfacial on both sides
80% cohesive and

20% interfacial on both sides
Failure
mode
5:09 pm
2*
A
B
C
A
B
C
A
Specimen
Curing
condition
4/12/06
1*
Series
Composite
plate
Steel
plate
Table 3. Failure loads and modes of double-strap joints
ASE 9-6-06_Zhang
Page 797
797
ASE 9-6-06_Zhang
4/12/06
5:09 pm
Page 798
thick Type B steel plate and H-M CFRP

Series 5:
thick Type A steel plate and UH-M
CFRP prepreg cover plates.
Series 6:
thick Type B steel plate and UH-M
Series 7:
thick Type A steel plate and UH-M
Series 8:
thick Type B steel plate and UH-M
Series 9:
thick Type B steel plate and H-M
CFRP/GFRP prepreg cover plates.
Series 10:
thick Type B steel plate and H-M
CFRP/GFRP prepreg cover plates.
It is noted that the GFRP laminates have fibres at +45
to the longitudinal direction of the specimens and
consequently in that direction the laminate is matrix
dominated and will have a limited effect on the
combined strength/stiffness of the uni-directional
CFRP/GFRP laminates.
The first double-strap joint tests used Type A steel
plates of 3 mm thickness to compare the two plate
bonding systems, namely, the rigid plate/two-part
adhesive system and the prepreg/film adhesive system.
After performing these tests it was felt necessary to
increase the thickness of the discontinuous steel
adherends to 6 mm to prevent any likelihood of yielding
of the steel. Two further tests were then undertaken to
ascertain whether there was any significant difference
between the failure modes and ultimate loads of the
double-strap joints of the 6 mm thick Type A steel and
those of the 6 mm thick Type B steel; the tests were
undertaken on double-strap joints using H-M CFRP
composite adherends bonded with film adhesive and
cured at 65C for 16 hours.
It can be seen from Table 3 that joints of the rigid
plate/two-part adhesive system (series 1) have a lower
average failure load (30.27 kN) than joints of the
prepreg/film adhesive system cured at 65C for 16 hours,
which have an average failure load of 35.3 kN. The
failure modes for the joints of the former system varied
798
between cohesive and interfacial, whereas, for the joints

of the latter system, all specimens failed by the
interfacial mode on both sides of the steel adherend.
Ideally, the adhesive layer should not be the weak link
in the joint and wherever possible the joint should be
designed to ensure that the adherends fail before the
bond layer. However, in a steel/CFRP composite joint
the adhesive is much weaker than the CFRP composite
or the steel adherends, so bond failure occurs before
failure in the adherends. Consequently, in a well-bonded
steel/CFRP composite joint, failure should occur within
the adhesive (cohesive failure) or within the composite
adherend (CFRP inter-laminar failure). Failure at the
adherend-adhesive interface (interfacial or bond-line
failure) generally indicates that a stronger bond should
be achieved.
The test results of series 2 and 3 (Table 3) indicate
that the failure loads do not depend on the type of steel
and the grit material used. These two series of tests were
on joints with H-M CFRP composite prepreg cover
plates bonded using film adhesive, cured at 65C for
16 hours. Their average failure loads are 35.3 kN and
35.95 kN, respectively.
Results from series 3 and 4 (Table 3) show that the
cure temperature of 80C for 4 hours led to a higher
failure load (viz. 44 kN) than was achieved with a lower
cure temperature of 65C for 16 hours, where the
average failure load is 35.95 kN. This would seem to
indicate that it is possible to achieve a greater degree of
cure with a higher temperature. This observation is also
apparent from the results of series 5 and 6 on UH-M
CFRP double-strap joints cured at 65C and 80C
respectively, where the respective failure loads are
28.90 kN and 40.23 kN. Furthermore, a similar
observation can be made about the tests of series 7 and
8 on UH-M CFRP/GFRP double-strap joints cured at
65C and 80C respectively.
In a separate internal investigation, a finite element
analysis was undertaken to ascertain the effects of
internal thermal stresses on a double-strap joint under
a load of 30 kN when the temperature reduces from
80C to 20C. The interfacial shear stresses and the
longitudinal stresses under this condition were
compared with those for a joint under a load of 30 kN
and in steady thermal state. Compared to the value
of the steady thermal state condition, the high cure
temperature condition was found to lead to a decrease of
22% in the maximum value of interfacial shear stress at
the free end; this difference reduced to zero at a distance
of 15 mm from the free end. A decrease in the
longitudinal stress of 25% was observed over the central
portion of the joint. This favourable result is due to the
much smaller value of the coefficient of thermal
ASE 9-6-06_Zhang
4/12/06
5:09 pm
Page 799
expansion of the carbon fibre compared to that of steel.

Due to the more favourable loading characteristics of
the film adhesive compared with those of the widely
used two-part standard civil engineering adhesive (20%
improvement, illustrated in Table 3), tests on joints with
the UH-M CFRP/GFRP or UH-M CFRP composite
bonded with the Sikadur 31 two-part adhesive to steel
adherends were not undertaken. Consequently, the
investigations focused on the prepreg/film adhesive
system and the two cure temperatures, namely: 65C for
16 hours and 80C for 4 hours.
The double-strap joints of series 5-8 all failed by the
tensile rupture of the UH-M CFRP cover plates. A
comparison of series 5 and 7 indicates that under the
same curing condition at 65C, the double-strap joints of
series 7, manufactured from the UH-M CFRP/GFRP
laminates, carried 26% greater load on average than that
carried by those of series 5 manufactured from the
UH-M CFRP laminates. The results of series 6 and 8, in
which specimens were cured at 80C, lead to a similar
observation, where an average extra load of 16% was
achieved with the inclusion of the GFRP layers. Further
tests for the latter condition were undertaken, but which
are not reported here, included measurements of FRP
strains at the middle section. These strain readings
showed that the maximum strains reached a value
of about 3400 , which is comparable to the tensile
rupture strain of the UH-M CFRP composite. Furthermore,
referring to series 9 and 3, which were cured at 65C,
the average failure load of the specimens of the H-M
CFRP/GFRP laminates is 12% greater than that carried
by those with the H-M CFRP laminates. However, the
results of series 10 and 4, the same observation cannot
be made. It may be noted that the joints of series 3, 9
and 10 experienced some debonding before tensile
rupture of the FRP laminates, but final failure was still
due to rupture of the FRP laminates. It thus appears
that the inclusion of GFRP layers is likely to lead to a
higher failure load, and two reasons are possible: (a)
the direct contribution of the GFRP to the tensile
resistance of the cover plates and a more gradual
transfer of stresses from the steel adherend to the cover
plates. However, since there is a significant scatter in
the test results and the benefit of the GFRP layers was
not observed in some of the tests, this conclusion needs
further validation.
6. THE UPGRADED BEAMS
Photiou et al. (2006) tested artificially degraded steel
beams strengthened with either the H-M CFRP/GFRP
composite or the UH-M CFRP/GFRP composite cured
at 65C for 16 hours. If this curing procedure is adopted,
then one significant issue is the effect of traffic-induced
vibrations during the curing of the prepreg/film

adhesive system, which was not examined by Photiou
et al. (2006). Alternative, the curing procedure of 80C
for 4 hours may be used, for which traffic induced
vibration is not longer an issue.
Based on the above considerations and the results
given in the preceding section, further work was
conducted on the two of the possible combinations of
the composite laminates and the curing procedure:
(i) the H-M CFRP/GFRP prepreg composite cured at
65C for 16 hours, and (ii) the UH-M CFRP/GFRP
prepreg composite cured at 80C for 4 hours. It may be
noted that some other combinations appear equally
promising. The effectiveness of these two methods, both
based on the prepreg/film adhesive system, in upgrading
steel beams were examined by testing: (a) a steel beam
upgraded by the first method exposed to an external
2 point loading system varying between a total upper
load limit of 40 kN and a total lower load limit of 5 kN
under a frequency of 1 Hz throughout the 16 hours
cure period, and (b) a steel beam upgraded by the
second method under a steady state condition during
polymerisation. These two loading systems represent
the condition when the bridge is open to traffic during
polymerisation of the polymer and the condition when
the bridge is closed to traffic for 4 hours in a 24 hour
cycle, respectively (Figure 1).
6.1. Materials Used
The rectangular tubular steel beams had external crosssectional dimensions of 120 mm 80 mm, a wall
thickness of 6 mm, and a total length of 2000 mm
(Figures 2 and 3). A 6 mm thick flat plate of dimensions
56 mm (width) 2000 mm (length) was welded onto the
compression flange of each beam. The exposed tension
flanges of the beams were grit blasted to the Swedish
Code SA 21/2 Grade (BS 7079). The steel from which
the beams were manufactured was characterised by a
modulus of elasticity of 209 GPa, a yield strength of 305
MPa and an ultimate stress of 400 MPa. The positions of
the longitudinal strain gauges are shown in Figure 2.
The upgrading material systems were made from
hybrids of H-M or UH-M unidirectional CFRP and 45
GFRP composites. Both the UH-M and the H-M CFRP
composites comprised two double-ply laminates, and
each ply thickness was 0.3 mm and 0.6 mm, respectively;
the carbon fibres were aligned with the longitudinal
direction of the beam. Three single plies of GFRP were
used with glass fibres positioned at 45 to the
longitudinal direction of the steel beam. The matrix
resin system and the film adhesive, which had a typical
thickness of about 110-120 microns, had the same
specification as those used for the double-strap joints
799
ASE 9-6-06_Zhang
4/12/06
5:09 pm
Page 800
300 mm
300 mm
150 mm
150 mm
CFRP
Adhesive
(or CFRP/GFRP composite)
(Two-part adhesive or film adhesive)
Steel adherend
Figure 1. Schematic representation of the double-strap Joint
80
56
6.2. Test Procedure

The two beams were tested statically to failure under
the action of two point loads. The span for the test was
1700 mm and the two external loads were positioned at
200 mm on either side of the mid-span of the beam
(spaced at a 400 mm distance). Three displacement
transducers were placed under the beam to monitor its
deflections: one at mid-span, and one under each load.
In addition, the movements of beam supports were
measured by placing one displacement transducer at
each support. The test set-up for the two rehabilitated
steel beams is shown in Figure 3.
50
15
35
120
Strain
gauges
The adhesive film was cut to size and laid onto the
beam, as were the various laminates in the stacking
sequence, as described above. The FRP composites on
the beam were covered by a Halar film and breather
blanket, and the whole was then covered with a heating
blanket before placing it in a vacuum bag. The bag
under a vacuum-assisted pressure of 1 bar was exposed
to the relevant temperature (either 65C for 16 hours
under the vibrating load condition or 80C for 4 hours
without imposed loads). After curing, the temperature of
the heating system was reduced to room temperature at
a steady rate.
60
Figure 2. Beam cross-section and positions of strain gauges
P
Distribution beam
Steel beam
400 mm
FRP composite
1600 mm
50 mm
50 mm
6.3. Test Results

Figure 4 shows the total load against the central
deflection for: (i) the H-M CFRP/GFRP rehabilitated
beam cured at 65C under a vibration load varying
between an upper limit of 40 kN and a lower limit of
5 kN with a frequency of 1 Hz throughout the 16 hours
cure period, (ii) the UH-M CFRP/GFRP rehabilitated
beam cured at 80C and (iii) the control beam. Figures 5
and 6 present the strain distributions in the pure bending
regions of these two beams, respectively.
Figure 3. Set-up for beam test

200
800
UH-M CFRP/GFRP 80oC

160
Load (kN)
and were made from low temperature curing resin

systems (cure temperature being 65C or 80C), suitable
for civil engineering applications.
The various layers of the UH-M and the H-M prepreg
composite material were stacked as follows: firstly, a
single 45 GFRP ply, followed by one double-ply
CFRP laminate, followed by a single 45 GFRP ply,
followed by one double-ply CFRP laminate and, finally,
followed by one single 45 GFRP ply. The first GFRP
ply was adjacent to the film adhesive and, therefore,
separated the CFRP composite from the steel beam.
It may be noted that in upgrading the steel beams, the
H-M CFRP/FRP laminate used had almost the same
axial stiffness as that of the UH-M CFRP/FRP laminate.
H-M CFRP/GFRP 65oC
120
Control beam
80
40
10
20
30
40
Deflection at mid-span (mm)
50
60
Figure 4. Total load against mid-span deflection
ASE 9-6-06_Zhang
4/12/06
5:09 pm
Page 801
subjected to vibration load limits of 5 kN and 40 kN, and

with the UH-M CFRP/GFRP composite plate cured at
80C, are illustrated in Figures 5 and 6, respectively.
Figure 5 clearly shows the shift in the neutral axis
position as yielding develops, but no such shift is seen
in the beam shown in Figure 6, as the strain values here
are much lower.
Distance from the bottom (mm)
140
120
40 kN
100 kN
140 kN
170 kN
180 kN
100
80
60
40
20
0
100007500 5000 2500 0
2500 5000 7500 10000
Microstrain
Figure 5. Distributions of strains for beam upgraded with H-M

CFRP/GFRP at a cure temperature of 65C and under a vibration
load of (5-40 kN)
Distance from the bottom (mm)
140
40 kN
70 kN
90 kN
120 kN
135 kN
120
100
80
60
40
20
0
3000
2000
1000
1000
2000
3000
Microstrain
Figure 6. Distributions of strains for beam upgraded with UH-M

CFRP/GFRP at a cure temperature of 80C
It is seen from Figure 4 that the beam with the UH-M

CFRP/GFRP failed at 140 kN with a central deflection
of 13.5 mm, (span/126). Likewise, the beam with the
H-M CFRP/GFRP failed at 180 kN with a central
deflection of 38 mm, (span/44.7); both failures were due
to the tensile rupture of the composite plate. Figure 5
shows that the ultimate strain of the H-M CFRP/GFRP
composite observed in the beam test was 1%, which is
lower than the ultimate axial tensile strain of about 1.6%
from tensile coupon tests. Likewise, the ultimate strain
of the UH-M CFRP/GFRP composite observed in the
beam test was 0.28% (Figure 6), which is lower than the
ultimate axial tensile strain of about 0.4% from tensile
coupon tests.
The strain distributions within the cross-sections
of the steel beams, which were upgraded with the
H-M CFRP/GFRP composite plate cured at 65C and
7. DISCUSSION AND OBSERVATIONS

This experimental research has examined and compared
two bonding systems which are available to the civil
engineering industry for the purpose of joining FRP
composites in the form of patch/plate components to
steel structural members; these are:
(i) The rigid plate/two-part adhesive system, where
the thickness of the adhesive was 0.5 mm;
currently, this is the preferred method for the
civil engineering industry.
(ii) The prepreg/film adhesive system, where the
thickness of the adhesive film was 0.1 mm. Both
components were cured simultaneously on the
structure under a vacuum assisted pressure of 1
bar and at low elevated temperatures of 65C for
16 hours and 80C for 4 hours, respectively.
Double-strap joint tests were conducted and these tests
indicated that the rigid plate/two-part adhesive system
led to lower failure loads than the prepreg/film
adhesive system. The investigation then focussed the
use of the more robust system, namely the prepreg/film
adhesive system, to upgrade a steel beam. This
upgrading was performed whilst the beam was under a
low frequency vibration load. This loading represents
the upgrading of a steel bridge whilst it is being
subjected to continuous traffic.
Both a H-M and an UH-M CFRP composite were
used in the investigation. The former had an elastic
modulus of about 60% of that of steel, whereas the latter
had an elastic modulus higher than steel, but a low
ultimate strain. From the results of the double-strap joint
tests, it is shown that the UH-M CFRP/GFRP cured at
80C for 4 hours and the H-M CFRP/GFRP cured under
the same condition are generally the most efficient, with
the former of which giving the greater failure load. The
H-M CFRP/GFRP composite cured at 65C for 16 hours
is slightly less efficient than the same composite cured
at 80C for 4 hours. Initially due to the vastly different
values of the coefficients of thermal expansion of the
FRP composite and the steel, it was not clear whether
the thermal stresses set up as the temperature was
reduced to ambient from 80C might affect the degree of
bond between the two components; it appears that it has
no detrimental effect.
Depending upon the ambient environmental temperature,
801
ASE 9-6-06_Zhang
4/12/06
5:09 pm
Page 802
the rigid plate/two-part adhesive system takes between

5 and 7 days for the adhesive to achieve about 90%
polymerisation. This process continues after this period,
but it is unlikely that it will ever reach 100%. Moreover,
the present tests showed that this system led to lower
failure loads than the prepreg/film adhesive system. It is
concluded, therefore, that the rigid plate/two-part
adhesive system is not desirable for bridge structures
which are subjected to vibrations due to the passage of
traffic over the bridge. Such loadings can damage the
polymerisation process.
As a result of the considerations discussed above,
attention was then directed to the following two specific
FRP upgrading methods, both based on the prepreg/film
adhesive system:
(i) the UH-M CFRP/GFRP cured at 80C for 4
hours; with this method, it was assumed that the
bridge would be closed for 4 hours in a 24 hour
period; and
(ii) the H-M CFRP/GFRP cured at 65C for 16 hours;
with this method, it was assumed that the bridge
would remain open for the full period of cure.
It should be noted that: (i) the joint thicknesses (i.e.
adhesive thicknesses) are different for the film adhesive
and the two-part adhesive, and (ii) only one layer of
adhesive film was used with the prepreg composite. A
linear elastic finite element analysis was undertaken in
another study (Photiou 2005) and one of the parameters
that this analysis yielded was the longitudinal shear
stresses at the interface between the steel and the FRP
composite at its free end and at the discontinuity of the
steel substrate. The peak shear stresses at the free end of
a 0.1 mm thick adhesive joint, i.e. one film adhesive
thick, were 57% greater in value compared with those in
the 0.5 mm thick joint. Likewise, at the position of the
discontinuity of the steel adherend, the corresponding
value was 21%. The higher values of the peak shear
stresses, as the adhesive thickness decreases, have been
noticed by other researchers. However, one advantage
of using a pre-preg composite and a compatible
adhesive film is that the whole upgrading system is
fabricated and bonded onto the steel substrate in one
operation; this involves compaction under pressure and
curing at an elevated temperature. This allows an
intermixing of the adhesive and the matrix of the
composite to take place during curing, thereby
producing a high degree of molecular interlocking and
an excellent degree of compaction of the two component
parts. The void ratio of the cured prepreg composite was
measured as about 2%.
Results from previous finite element analysis
(Photiou 2005) showed that the inclusion of a low
modulus GFRP composite layer between the steel and
802
the CFRP composite leads to a more gradual transfer of

stresses between the steel plate and the CFRP
composite, which is believed to be beneficial to the
bond capacity. However, as noted in Section 5.5, this
benefit was not observed in some of the present doublestrap joint tests; this may be due to the fact that most of
these specimens failed by the tensile rupture of
composite plate. Another possible advantage of a GFRP
layer between the steel and the CFRP composite is that
the effect of thermal stresses between the two materials
is expected to be reduced.
The beam which used the UH-M CFRP/GFRP
composite as the upgrading material, failed as a result of
carbon fibre breakage. This breakage occurred at an
ultimate strain of 0.28% and a total load of 140 kN with
a central deflection of 13.5 mm (Span/126). The H-M
CFRP composite failed at a higher load of 180 kN and
an ultimate strain of 1.0%; the failure was also due to
carbon fibre breakage. The beam with the H-M carbon
fibre composite failed after substantial inelastic
deformation of the steel, whereas the beam upgraded
with the UH-M modulus carbon fibre composite showed
minimal inelastic deformation. The ultimate load in the
former beam is 33% higher than that achieved in the
latter one.
It is noticed that the failure loads of the UH-M
CFRP/GFRP composite double-strap butt joints cured at
temperatures of 65C and 80C are 12% and 16% higher
than those of the corresponding H-M CFRP composite
double-strap butt joints respectively. The reverse trend
is true for the upgraded beams. This is because the steel
beam upgraded with the UH-M CFRP/GFRP composite
failed by the rupture of the FRP composite at a tensile
strain of 0.28, when the resistance of the steel beam had
not been fully mobilised. Therefore, for the upgrading of
a steel beam, a more ductile material, such as the H-M
CFRP/GFRP composite examined in the present study,
appears to be more suitable.
8. CONCLUSIONS
The general conclusion from the present work is that an
attractive system for upgrading a steel bridge, which is
being continuously traversed by traffic, is to use a
suitable prepreg composite and a compatible film
adhesive, both being cured simultaneously. The cure
temperature of 65C and the vacuum assisted pressure
of 1 bar, which were applied in the present experiments,
developed a satisfactory bond and a good compaction
for the prepreg/film adhesive system under a vibration
load. The experiments also showed that a cure
temperature of 80C for 4 hours is at least as effective as
the lower temperature of 65C for 16 hours in upgrading
structural steel members. For the upgrading of steel
ASE 9-6-06_Zhang
4/12/06
5:09 pm
Page 803
beams, a more ductile material rather than a brittle

material (e.g. the H-M CFRP/GFRP composite rather
than the UH-M CFRP/GFRP composite examined in the
present study), is needed to more fully mobilise the
resistance of the steel beam.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the financial support
provided through a Kan Tong Po Visiting Professorship,
jointly funded by a Kan Tong Po grant, the Royal
Society, The Hong Kong Polytechnic University; and
through grants from The Hong Kong Polytechnic
University (Project codes: BBZH and G-YX47).
The authors also gratefully acknowledge the support
received from Mr. E. Shahidi and the staff of
Advanced Composite Group Ltd., Heanor, Derbyshire,
UK, especially for the helpful discussions throughout
these investigations.
REFERENCES
BS 7079 (1989). Preparation of Steel Substrates before Application
of Paints and Related Products.
(2004). Strengthening Metallic Structures Using ExternallyBonded Fibre-Reinforced-Polymers. CIRIA Report C595, London.
Gettings, M. and Kinlock, A.J. (1977). Surface analysis of
polysiloxane/metal oxide interfaces, Journal of Materials
Science, Vol. 12, pp. 25112518.
Hollaway, L.C., Chryssanthopoulos, M.K. and Moy, S.S.J. (Eds).
(2004). Advanced Polymer Composites for Structural
Applications in Construction, Proceedings of ACIC 2004,
Woodhead Publishing, Cambridge.
Liu, X., Silva, P.R. and Nanni, A. (2001). "Rehabilitation of steel
bridge members with FRP composite materials", Composites in
Construction, Proceedings of CCC 2001, Porto, Portugal, J.
Figueiras, L. Juvandes and R Furia (Eds), pp. 613617.

Mertz, D. and Gillespie, J. (1996). "Rehabilitation of steel bridge girders
through the application of advanced composite material", NCHRP
93-ID11, Transportation Research Board, Washington, D.C.
Mosallam, A.S. and Chakrabarti, P.R. (1997). "Making connection",
Civil Engineering, ASCE, pp. 5659.
Photiou, N.K., Hollaway, L.C. and Chryssanthopoulos, M.K. (2003).
"Characterisation of adhesively bonded plates for upgrading
structural steelwork", Proceedings (CD-ROM), 10th International
Conference on Structural Faults and Repair, 13 July, London, UK.
Photiou, N.K. (2005). Rehabilitation of Steel Members Utilising
Hybrid FRP Composite Material Systems, PhD Thesis, University
of Surrey, UK.
Photiou, N.K., Hollaway, L.C. and Chryssanthopoulos, M.K. (2006).
"Strengthening of an artificially degraded steel beam utilising a
carbon/glass composite system", Construction and Building
Schnerch, D., Dawood, M., Rizkalla, S. and Sumner, E. (2006).
"Proposed design guidelines for strengthening of steel bridges
with FRP materials", Construction and Building Materials,
in press.
Schnerch, D., Stanford, K., Sumner, E. and Rizkalla, S. (2004).
"Strengthening steel structures and bridges with high modulus
carbon fiber reinforced polymer, resin selections and scaled
monopole behavior", Transportation Research Record, No. 1892,
pp. 237245.
Tavakkolizadeh, M. and Saadatmanesh, H. (2003). "Strengthening
of steel-concrete composite girders using carbon fibre reinforced
polymer sheets", Journal of Structural Engineering, ASCE, Vol.
129, No. 1, pp. 3040.
Zhang, L., Hollaway, L.C., Teng, J.G. and Zhang, S.S. (2006).
'Strengthening of steel bridges under low frequency vibrations",
Proceedings, Third International Conference on FRP Composites
in Civil Engineering (CICE 2006), Miami, USA.
803
ASE 9-6-06_Zhang
4/12/06
5:09 pm
Page 804
ASE 9-6-07_Schnerch
4/12/06
5:09 pm
Page 805

D. Schnerch 1,*, M. Dawood 2, S. Rizkalla 2, E. Sumner 2 and K. Stanford 3
1
Wiss, Janney, Elstner Associates, Inc., 245 First Street, Suite 1200, Cambridge, Massachusetts, USA
Department of Civil, Construction, and Environmental Engineering, North Carolina State University, USA
3 Sutton-Kennerly and Associates, Greensboro, North Carolina, USA
Abstract: Recent research has focused on rehabilitation and strengthening of steel

structures and bridges using fiber reinforced polymer (FRP) materials. The bond
behavior of FRP materials to steel structures is quite different from that of concrete
structures. Preliminary test results showed the occurrence of very high bond stresses
for most strengthening applications due to the amount of strengthening required for
steel structures and bridges. In this paper, surface preparation methods and means of
preventing galvanic corrosion are discussed. The results of an experimental program
for selection of suitable adhesives through determination of the development length is
discussed as well as preliminary testing showing the importance of proper detailing of
the ends of the FRP strips. The shear stress distribution determined in the experimental
program is compared to analytical models using a stress-based approach. The
remainder of the paper focuses on the current methods for determining bond stresses
and their use for the design of FRP strengthening system for steel structures.
Key words: high modulus CFRP, steel bridge, strengthening, rehabilitation, bond stresses.
1. INTRODUCTION
Innovative methods are required for the strengthening
and rehabilitation of steel structures that are deficient
due to the demand to increase the specified load and/or
deterioration as a result of corrosion. A considerable
body of research has established the successful use of
carbon fiber reinforced polymer (CFRP) materials for
strengthening concrete structures. With the introduction
of new high modulus CFRP materials, the possibility for
providing a solution to the ongoing problem of
infrastructure deterioration may be extended to steel
structures as well. Bond stresses may be much more
critical for steel structures than for concrete structures
since more strengthening material is needed for steel
structures to achieve a similar increase in strength due to
the inherent high strength of steel and also since the
debonding failure does not occur in the substrate as in
concrete structures. Further complications may arise due
to the potential for galvanic corrosion between the
carbon and steel materials. Despite these challenges,
since many structures built in the post-World War II era

are already past their design life, the inventory of
deteriorated steel structures and bridges in need of
rehabilitation are extremely significant (Tang and Hooks
2001). The aim of this paper is to examine the proper
techniques for the bond of CFRP to steel structures and
compare the results of the experimental program to
analytical methods of determining bond stresses. The
determination of these bond stresses is necessary for the
design of the CFRP strengthening for steel structures.
2. BONDING OF HIGH MODULUS CFRP
STRIPS TO STEEL STRUCTURES
Proper installation of high modulus CFRP strips is
essential in ensuring both the long-term performance of
the system and that the behavior of the system matches
the intentions of the designer. A certain level of care
and expertise is required to ensure that these goals are
met. Research into the nature of bonding between FRP
materials and metallic structures was first investigated
*Corresponding author. Email address: dschnerch@wje.com; Fax: +1-617-225-9081; Tel: +1-617-225-9080.
805
ASE 9-6-07_Schnerch
4/12/06
5:09 pm
Page 806
by the aerospace and naval industries. Later adoption of

the technique for civil engineering applications has
typically used CFRP instead of GFRP due to the
more reasonable necessary thickness of the applied
strengthening material. The axial stiffness required for
strengthening steel structures is much higher than for
concrete or masonry structures that have limited tensile
strength, justifying the use of much higher modulus
strengthening materials.
2.1. Surface Preparation
Bonded joints are often the most effective way to join
two different adherends, since the resulting stress
concentrations at the joint are much lower than for
bolted connections. Furthermore, the anisotropic nature
of most CFRP materials would preclude bolting as a
connection method. To ensure full utilization of the
applied CFRP material, surface preparation of the steel
must be undertaken to enhance the formation of
chemical bonds between the adherend and the adhesive.
This requires a chemically active surface that is free
from contaminants. Most surface treatment involves
cleaning, followed by removal of weak layers and then
re-cleaning (Mays and Hutchinson 1992). Degreasing is
a necessary first step in preparing most metals to
remove, oils and other potential contaminates. Brushing,
ultrasonic or vapor degreasing systems are claimed to be
most efficient in removing this surface contamination,
especially when sufficient amounts of solvent are used
(Hashim 1999). Contamination may then be removed
with the excess solvent, rather than simply redeposited
on the surface as the solvent evaporates.
The most effective means of achieving a highenergy steel surface is by grit blasting (Sykes 1982;
Hutchinson 1987; Hollaway and Cadei 2002). Parker
(1994) found that for composite joints, those that were
grit blasted had higher peel strengths than those that
were hand abraded. Grits are found to have a clean
cutting action that can expose a clean surface, unlike
wire brushing. Grit blasting procedures using angular
grit, removes the inactive oxide and hydroxide layers
by cutting and deformation of the base material. The
size of the grit will also affect the surface profile of the
steel. Harris and Beevers (1999) confirmed that finer
grit particles produced smoother surfaces than coarser
particles in an investigation using three-dimensional
profilometry measurements. For two of the three grits
studied, smoother surfaces exhibited higher surface
energy readings as determined from static contact
angle measurements. However, the initial joint
strengths were independent of the coarseness of the
grit. Furthermore, the long-term durability was not
affected by the surface profile.
806
Following grit blasting, the surface may be contaminated

with fine abrasive dust. It has is generally been agreed
that abrasive dust should be removed prior to bonding.
Hollaway and Cadei (2002) state that the dust should be
removed by dry wipe, or by a vacuum head with brushes
and that solvent cleaning should be avoided. This is due
to the assumption that solvent wiping only partially
removes the dust, and redistributes the remaining dust on
the entire surface. However, several different studies have
shown that solvents may be used to clean the surface after
grit blasting without resulting in poor bond performance
(El Damatty et al. 2003; Photiou et al. 2004). If solvents
are used, it may be beneficial that they be applied in
excess so that any debris removed by the solvent is
removed from the surface and is not redeposited after the
solvent evaporates.
2.2. Durability of CFRP-Steel Bonded Joints
FRP materials typically have excellent resistance to
corrosion and chemical attacks, resulting in the
expectation of a long life of the repair with little or no
maintenance required. However, the adhesive and steel
may be affected by long term exposure to moisture,
especially in conjunction with salts resulting from
deicing of roadways or from ocean spray. The effects of
moisture or temperature that is acting in conjunction
with an applied stress, may influence the behavior of the
joint due to stiffness change of the resin resulting from
the exposure (Karbhari and Shulley 1995). In general,
adhesive joints subjected to high humidity, saturation
with water or extreme temperatures, will result in a
reduction of the joint strength. Despite the change in the
mechanical properties of the adhesive, the primary
mechanism for strength reduction in bonded steel joints
in wet environments is the influence of interfacial
attack in displacing the adhesive from the adherend
(Hutchinson 1987; Hashim 1999). Moisture diffusing
through the adhesive layer is energetically attracted to
high-energy substrate surfaces, resulting in adsorption
of water molecules, thereby displacing secondary bonds
between the adhesive and substrate. Compounding this
effect is that moisture ingress occurs at the edges of a
joint, where the bond stresses may be the highest.
Adhesion promoters, such as silanes, have been
shown to increase the durability of steel-epoxy bonds
without affecting the initial bond strength (McKnight
et al. 1994). Similar findings have been reported for grit
blasted aluminum surfaces (Allan et al. 1988). For
aluminum naval structures, silane incorporated into the
adhesives themselves was shown to be less effective
than providing a separate silane layer. Silane adhesion
promoters are also noted to greatly reduce the
variability of bond performance, while protecting the
ASE 9-6-07_Schnerch
4/12/06
5:09 pm
Page 807
D. Schnerch, M. Dawood, S. Rizkalla, E. Sumner and K. Stanford
freshly prepared surface from damage, exposure to

environmental conditions and contamination prior to
bonding the FRP material. Gettings and Kinloch (1977)
found that durability was improved only when there was
evidence of primary bonding between the polysiloxane
primer and the steel surface. Due to the promising
results associated with the use of silanes, they have been
used in field applications such as the strengthening of
bridge 1-704 that carries southbound traffic on Interstate
95 in Delaware (Miller et al. 2001).
Prevention of galvanic corrosion is necessary for the
long-term durability of any CFRP strengthening applied
to a metallic structure. In general, the requirements for
galvanic corrosion are that the two metals must be in
direct electrical contact, the metals must have sufficient
potential difference, they must be bridged by an
electrolytic solution and a cathodic reaction must be
sustained on the noble metal (Francis 2000). This
electrolytic solution may be generated by the presence
of water with a salt, fertilizer, acid or a combustion
product. If all of these conditions are met, current will
flow through the electrolyte from the anodic metal to the
cathodic metal. The cathodic metal is then protected
from corrosion, but the anodic metal may suffer even
greater corrosion. The reactions that occur due to
galvanic corrosion, are similar to those that would occur
on a single metal, however the rate of attack is increased
for the anode. Carbon is a very noble cathodic material
that can drive the corrosion of many different metals
galvanically coupled to it. Steel and aluminum have
similar positions in the galvanic series, and behave
anodically relative to the carbon. The composite itself
may be degraded by the galvanic process (Miriyala et al.
1992). In this study, the polymer material was found to
be degraded on the cathodic surface, although it was not
known whether this was due to direct involvement in the
cathodic reaction or due to chemical attack of the
polymer by some product of the cathodic reaction.
2.3. Prevention of Galvanic Corrosion
Considerable research has been focused on the
prevention of galvanic corrosion. In general, to prevent
against galvanic corrosion the flow of corrosion currents
must be prevented. This may be achieved by insulating
the dissimilar metals from one another or by preventing
a continuous bridge of electrolytic solution between the
two by coating with a water resistant sealant (Evans and
Rance 1958). If the two metals are not in contact,
galvanic corrosion cannot occur. Brown (1974) studied
the corrosion of different aircraft metals connected
to CFRP by adhesive bonding or bolting. For the
specimens connected by adhesive bonding there was no
accelerated corrosion attack. This behavior was claimed
to be due to the insulating behavior of most structural

adhesives in not allowing electrical contact between the
two materials. Tavakkolizadeh and Saadatmanesh
(2001) completed an experimental study to determine the
CFRP/steel corrosion rate when subjected to seawater
and deicing salt solutions. The effect of different epoxy
thicknesses and the removal of fiber sizing agents with
different solvents were also examined. The effect of a
thin coating of epoxy (0.25 mm) was found to be
significant as was the sizing applied to the fibers. In
general, thicker epoxy films between the steel and
CFRP surfaces were shown to significantly slow the
corrosion rate of the steel.
To reduce the possibility of galvanic corrosion, a
non-conductive layer, such as an epoxy film or GFRP
sheet, can be used to isolate the carbon fiber from the
steel. In the current study, it was found that a uniform,
1 mm adhesive thickness could be achieved between the
steel surface and the CFRP strip by mixing a small
amount of glass beads into the adhesive prior to
application (Dawood 2005). West (2001) concluded that
either an adhesive layer or a GFRP layer effectively
isolated the two components and protected against
galvanic corrosion. Although accelerated tests have
been developed to determine the performance of lab
scale specimens, there is little correlation between these
tests and typical environmental exposure. This is an area
were further research needs to be directed.
A water resistant sealant on the surface can be used to
prevent ingress of any electrolytic solution, and preventing
one of the necessary conditions for galvanic corrosion to
occur. Brown and De Luccia (1977) noted that, for
aluminum to carbon fiber samples, the use of a water
resistant sealant or the use of a GFRP barrier performed
equally in a corrosive salt-spray environment, but that a
combination of a nonconductive barrier plus a sealant was
the most promising approach to control corrosion. This
was similar to the technique that was later used by Allen
et al. (1982) for protecting aluminum aircraft structures
strengthened with CFRP material. A moisture barrier of
aluminum foil was bonded over the strengthened area and
extended past this area on all sides. The aluminum patch
in turn, was protected by a chopped glass strand mat
finished with additional epoxy resin. This ensured that the
strengthened region would remain free from moisture.
Considerable attention has been focused on the use of a
GFRP insulation layer, rather than relying on the
insulating properties of the adhesive on its own. However,
the introduction of GFRP material may be less durable
than the adhesive. There are two possible reasons for this.
First, moisture intake may be accelerated due to water
traveling more quickly along the glass fiber-resin interface
than through the bulk adhesive itself (Choqueuse et al.
807
ASE 9-6-07_Schnerch
4/12/06
5:09 pm
Page 808
1997). The second reason is that salts can leach out of the
glass fibers themselves. This causes a concentration
gradient that can draw more water into the interface or into
voids within the joint. The pressure generated by this
process can cause the voids to blister, resulting in damage
to the surrounding material (Frieze and Barnes 1996).
Tucker and Brown (1989) have found that glass fibers
placed within a carbon fiber composite result in the
blistering of the composite by creating conditions
favorable for the development of a strong osmotic
pressure within the composite. Clearly, water being drawn
within the bond line by osmotic pressure is not favorable
for maintaining a durable bond. Other materials may be
more suitable for this purpose. Hollaway and Cadei (2002)
reported that a polyester drape veil was installed to
provide insulation between the carbon fiber and the cast
iron to prevent direct contact between the CFRP and the
steel, although no durability information was given for this
combination of materials. Finally, although fiber-glass or
epoxy films can be used to provide effective insulation,
Sloan and Talbot (1992) note that few materials retain
their insulating properties for more than a few years due
to wear, chemical breakdown or electrolyte absorption. A
monitoring program could also be initiated to identify
cathodic sites so that galvanic corrosion damage could be
stopped or mitigated.
2.4. CFRP Detailing
For ease of shipping and handling CFRP strips are
typically manufactured in finite lengths that are suitable
for strengthening most typical short span girders. To
facilitate the implementation of the strengthening
system to longer span girders, it is necessary to develop
an effective technique to splice adjacent lengths of the
CFRP materials. The use of a bonded splice cover plate
is a promising technique to ensure continuous transfer
of forces across splice joints. However, the use of this
technique requires a careful consideration of localized
bond stress concentrations that may occur at or near
splice locations. Other researchers have demonstrated
that careful detailing of the ends of a CFRP strip can
significantly reduce the bond stress concentrations that
typically occur at or near the strip ends. Tapering of the
CFRP strips at their edge avoids imposing a local stress
concentration at the boundary of the joint. For lapped
joints it was recommended that peel stresses should be
designed out of the joint by tapering the ends of the
overlap (Hart-Smith 1980). In the case of joints made to
FRP adherends, it was also noted that this would also
reduce the possibility of an interlaminar failure within
the FRP. Allan et al. (1988) recommended finishing
steel to CFRP joints with a 10:1 taper (5.7 degrees) at
their ends to reduce stress concentrations.
808
An alternate method of reducing the stress

concentration at the end of the joint is to locally increase
the adhesive thickness. Wright et al. (2000) found that
increasing the thickness of the bond line resulted in a
reduction in the stiffness of the adhesive layer, reducing
the stress concentration at the ends of the joint and
thereby increasing its overall strength. Slight edge
preparation of the steel was shown to further reduce the
stress concentration. Earlier work had shown that
increasing the adhesive thickness was more effective
than tapering the adherend (Price and Moulds 1991).
It may be possible to combine the effects previously
discussed to further reduce the stress concentration by
producing a reverse-tapered joint. For this type of joint
the adherend is tapered, while the adhesive thickness is
also increased as a result of the taper. Price and Moulds
(1991) found this type of joint to be superior for loads
that are applied statically or cyclically. For steel plates
bonded to aluminum the use of a reverse-taper improved
the fatigue life by about a factor of four Allan et al.
(1988). For repairs consisting of multiple plys of a
thinner CFRP material, like sheets, the interlaminar shear
and peel stresses can be reduced by reverse tapering of
subsequent plys of material. This was accomplished by
increasing the length of each ply so that the shortest layer
is on the inside and progressively longer layers are
towards the outside (Ong and Shen 1992). Analytical
modeling of single lap joints has also shown that reverse
tapering is a highly efficient technique in reducing the
stress peaks in both the adherend and in the adhesive,
thereby improving joint strength (Hildebrand 1994).
More recently, finite element analysis has demonstrated
that the use of a reverse taper and a spew fillet can
significantly reduce the peak shear and peeling stresses
which occur near the end of a bonded joint, with the
majority of the advantage being achieved for taper and
filet angles of 45 (Belingardi et al. 2002). While the
effectiveness of implementing a reverse taper has been
established analytically, there are relatively few
experimental studies that investigate the suitability of
this type of detailing reported in the literature. Typically,
in civil engineering applications, thick and stiff
composite plates are used for strengthening as compared
to the relatively thin and flexible composite materials
used in aerospace applications. Consequently, the bond
stresses can be more critical and proper detailing of the
joint ends is essential.
3. EXPERIMENTAL PROGRAM
3.1. Introduction
There have been only limited published studies on the
bond length of CFRP strips applied to steel flexural
members. Nozaka et al. (2005) studied the bond
ASE 9-6-07_Schnerch
4/12/06
5:09 pm
Page 809
behavior of cracked steel girders. For the strips and

adhesives studied, the failure was always by debonding.
It was noted that the shear ductility at failure seemed to
be the most important parameter in insuring a high CFRP
strain at failure, since the adhesive would rapidly yield as
the CFRP strip was loaded. This results in the CFRP strip
having a high strain at failure since the ductility of the
adhesive reduces the stress concentration at the end of
the strip that can result in premature debonding.
Debonding from the ends of the CFRP strip will occur if
the adhesive stresses at the end of the strip are exceeded
or if these stresses exceed the bond strength to the CFRP
or steel. The development length found for the adhesives
studied was found to be less than 203 mm. In another
study, a total of seven steel beams were strengthened
with standard modulus CFRP strips (Lenwari et al.
2006). The beams that were strengthened with the
shorter CFRP strips failed due to debonding of the CFRP
while the beams strengthened with 1200 mm long strips
failed by rupture of the CFRP. The study indicated
that the critical stress intensity factor near the end of the
strengthening plate can be used to calculate the
debonding strength of the strengthened member based on
a fracture mechanics approach.
Only a limited number of studies have been reported
that investigate the suitability of splicing FRP strips for
civil infrastructure applications. Stallings and Porter
(2003) recommend that splices should be located such
that the maximum strain at the end of the splice on the
main CFRP strip does not exceed a limiting value. Yang
and Nanni (2002) recommend a minimum splice length
to ensure rupture of CFRP strips, but used relatively thin
strips in their study. Neither of the reported studies
indicates that the CFRP splices were detailed in any way
to reduce bond stress concentrations at the CFRP ends.
In the present study, the bond behavior of uncracked
steel flexural members, typical of most steel structures,
strengthened with high modulus CFRP strips was
examined. These strips had a tensile elastic modulus of
338 GPa and an ultimate elongation of 3.32 millistrain,
with an essentially linear stress-strain behavior until
rupture. Additional specimens were tested to determine
the performance of typical splice connections,
considering the effect of different joint details.
3.2. Flexural Test Specimens for Bond
A flexural type of test specimen was used to study the
bond performance for the adhesive bonding of high
modulus CFRP strips. This type of specimen was used
due to the expectation that the CFRP strips would
generate significant normal, or peel stresses, and have
similarly proportioned shear and normal stresses to the
larger structures they represent. The test specimens
consisted of a wide flange steel member, typically

designated SLB 100 x 4.8. This designation represents
the nominal depth in millimeters and the mass in
kilograms per meter. An additional grade A36 steel plate
was stitch welded to the compression flange to simulate
the strain profile of a bridge girder that acts compositely
with a concrete deck. Welding was completed using E70
grade weld material and 4.8 mm fillets on either side of
the steel plate.
Strengthening of each specimen was completed by
bonding the high modulus CFRP strip to the bottom of
the tension flange. Each of the strips was cut to a width
of 36 mm and the thickness of each strip was 1.45 mm.
Surface preparation of the steel was completed by
sandblasting, followed by cleaning with liberal amounts
of acetone to remove any dust. Since the CFRP strips
used did not have a peel ply, the strips were lightly
sanded and wiped with methanol prior to bonding. The
length of the bonded CFRP strip used was varied from
50-200 mm. Figure 1 shows the specimen dimensions.
Material properties for the adhesives used were typically
not available from the manufacturers. The properties for
the SP Spabond adhesive were determined in accordance
with the ASTM D638-03 standard test method. The
behavior of the adhesive in tension was approximately
linear until failure with a tensile strength of 37.1 MPa,
initial tensile modulus of 3007 MPa and a Poissons ratio
of 0.38. More details of the material properties for the
adhesive as well as material properties for the steel beam
and steel cover plate are listed in Schnerch (2005).
A four-point bending test was used to load the beams,
as shown in Figure 2, with the development length
defined as the distance from one of the load points to the
end of the CFRP strip, in a region of constant shear force
and decreasing bending moment towards the end of the
strip. Lateral bracing of the top flange was provided
at the supports. Load was applied at a constant
displacement rate of 0.75 mm/minute. In addition to
measurement of the deflection using linear voltage
displacement transducers, longitudinal strain was
measured using electrical resistance strain gauges, with
a 6 mm gauge length, bonded to the inside of the
compression flange, inside the tension flange and on the
CFRP strip. Strain gauges were positioned at midspan,
at the load points, and every 25.4 mm from the load
point to 6.4 mm from the end of the CFRP strip.
3.3. Tension Test Specimens for Spliced
Connection
Three double-lap type shear coupons were fabricated and
tested to investigate the bond behavior between CFRP
strips. These specimens were tested to simulate the
behavior of bonded cover-plate splice joints. Three
809
ASE 9-6-07_Schnerch
4/12/06
5:09 pm
Page 810
38 mm
102 mm
6.5 mm
2.5 mm
1.45 36 mm CFRP strip
3.3 mm
57 mm
Figure 1. Cross-section dimensions of typical flexural test

specimen for bond
different configurations of bonded joints were considered

as shown schematically in Figure 3. The typical test
specimen consisted of two 8 mm thick main plates that
were fabricated by bonding together two 4 mm thick
CFRP strips. The main plates were connected by two 4
mm thick splice plates as shown in Figure 3. Joint
configuration A was fabricated with all strip ends square
with no spew fillets. Joint configurations B and C
incorporated reverse tapers and spew fillets, with angles
of approximately 20 degrees, to help reduce the localized
stress concentrations at various locations on the test
specimens as shown. The width of the test coupons was
35 mm. The splice plates had an overlap length of 200
mm on either side of the joint and a 2 mm gap was
maintained between the main plates at the splice location
that was subsequently filled with adhesive. All of the
coupons were instrumented with electrical resistance
strain gauges, at the locations shown in Figure 3(a), to
measure the distribution of the stresses in the splice

plates throughout the bonded joint. An additional two
strain gauges were installed away from the splice
location to measure the strain in the main plate. The
coupons were loaded in axial tension with a constant
displacement rate of 0.5 mm/min. The CFRP strips used
in this study had a higher modulus of elasticity than in
the previous study, since the strips were pultruded with
a higher fiber volume fraction of the same high
modulus carbon fiber. The tensile modulus of elasticity
was determined to be 460 GPa and the SP Spabond
345 adhesive was used to bond the joints. A uniform
adhesive thickness of approximately 1 mm was
maintained by mixing a small amount of glass spacer
beads into the adhesive prior to fabrication of the joints.
4. RESULTS OF FLEXURAL TEST
SPECIMENTS FOR BOND
Flexural specimens were loaded until a steel tension
flange strain of 8 millistrain was reached. Prior to this
level of strain being reached, the CFRP strip either
ruptured, near its ultimate elongation for beams where
sufficient development length was provided, or
debonded from the steel for beams with insufficient
development length. In determining the most suitable
adhesives for bonding the CFRP strips to steel,
the CFRP strip strain at failure in conjunction with
observation of the failure mode provided the best
indication of which adhesives were able to fully utilize
the CFRP material at the shortest development lengths.
Table 1 summarizes the results of the adhesive selection
phase in order from the adhesive with the shortest
development length to the adhesive with the longest.
Typically, high CFRP strain at failure and rupture of the
CFRP indicate complete utilization of CFRP materials.
Two adhesives, Weld-On SS620 and SP Spabond 345,
were found to have the shortest development lengths of
76-102 mm. The remaining adhesives had development
102 mm
A36 steel plate

SLB 100 4.8
Bonded CFRP strip
Development length
Length of CFRP strip
813 mm
864 mm
Figure 2. Loading configuration of typical flexural test specimen for bond
810
ASE 9-6-07_Schnerch
4/12/06
5:09 pm
Page 811
200 mm
6 mm
50 mm
1 mm thick adhesive layer
100 mm
8 mm thick CFRP main plate
50 mm
4 mm thick CFRP splice plate
100 mm
(a) Configuration A (square ends, no fillet)
Reverse taper, adhesive fillet
Square ends, no fillet

(b) Configuration B

(c) Configuration C
Strain gauge
Figure 3. Detail of splice joint configurations for tension test specimens (side view)
Table 1. CFRP strip strain at rupture/debonding for tested adhesives and development lengths
Development Length
Adhesive
203.2 mm
152.4 mm
127 mm
101.6 mm
76.2 mm
50.8 mm
Weld-On SS620
SP Spabond 345
Vantico Araldite 2015
Jeffco 121
Fyfe Tyfo MB2
Sika Sikadur 30
3.077 rupture
2.878 rupture
3.094 rupture
2.981 rupture
3.470 rupture
2.814 debond
2.964 rupture
2.943 rupture
2.980 rupture
3.276 rupture
3.060 debond
2.662 rupture*
3.161 rupture
3.111 rupture
2.820 rupture
2.438 debond
2.096 debond
2.903 rupture
2.433 debond
2.772 debond
2.589 debond
1.833 debond
*Underlined values are the average of two test results.
lengths as follows: the Vantico Araldite 2015 and Jeffco

121 adhesives had a development length of 102-127 mm,
Fyfe Tyfo MB had a development length of 152 mm and
Sika Sikadur 30 had a development length of more than
203 mm. In all cases, debonding progressed extremely
rapidly starting from one end of the CFRP strip to the
other side of the strip. The failure surface typically was
mixed with some adhesive remaining adhered to the steel
and some adhesive adhered to the CFRP strip.
Six of the specimens tested were instrumented with
strain gauges positioned along the development length of
the bonded CFRP strip on one side of the beam. These
strain measurements were recorded at discrete locations.
The difference in tensile strain between two gauge
locations must be balanced by the shear force acting
between the CFRP strip and the steel substrate, as noted
by Garden et al. (1998). The average shear stress could
then determined between the two gauge locations as,
av = E frp t frp
2 1
x2 x1
(1)
where 2 1 is the difference in strain between two

adjacent gauges and x2 x1 is the distance between the
gauges. The longitudinal strain at the tip of the CFRP
sheet was taken to be zero in order to calculate the shear
stress between the end of the strip and the location of the
first strain gauge. As shown in Table 2, the specimen
using the SP Spabond 345 adhesive had the highest
shear stresses of the tests with one ply of CFRP strips. It
is possible that some of the other adhesives could have
developed higher shear stresses, had the CFRP strips not
ruptured first.
4.1. Comparison to Predicted Behavior
The analysis of bonded joints in general has been
investigated using analytical and finite element
techniques. The advantage of analytical bond modeling
is that since stress singularities at the material interfaces
are avoided, consistent results can be achieved quickly
(Xiong and Raizenne, 1996). Besides the need for
significant computing time, which makes parametric
studies tedious, difficulties can arise in modeling the
811
ASE 9-6-07_Schnerch
4/12/06
5:09 pm
Page 812
Table 2. Maximum shear stress (MPa) and failure mode for beams strengthened by adhesive bonding
of CFRP strips using different development lengths
Development Length
Resin
Weld-On SS620
SP Spabond 345
Jeffco 121
plys
254 mm
203.2 mm
127 mm
101.6 mm
1
1
2
1
2
49.9 rupture
61.8 rupture
21.3 rupture
17.7 rupture
36.7 rupture
13.3* debond
*This average shear stress was determined over the last 25.4 mm of the CFRP strip, unlike the remaining values that were determined over the last 6.4 mm.
adhesive since the elements within the adhesive tend to

have high aspect ratios, and the results may vary
significantly depending on the mesh used. Analysis
methods have been completed to determine the
critical shear and normal adhesive stresses based on
compatibility of deformations among the beam being
strengthened, the adhesive and the FRP strip. The
solutions are valid in the linear-elastic range of the
materials. Due to the large difference in flexural
stiffness between the beam being strengthened and the
FRP material acting alone, simplifications can be made
in the derivation of the adhesive stresses. The
assumption of constant shear and normal stresses in the
adhesive across the thickness of the adhesive layer leads
to the result that the approximate solutions do not satisfy
the zero boundary condition at the ends of the adhesive
layer (Buyukozturk et al. 2004). One such method
developed by Smith and Teng (2001) was used to
compare the experimental interfacial stress values to
those predicted by the analytical procedure, as shown in
Figure 4. This test used the 101.6 mm development
length and the SP Spabond 345 adhesive. Higher-order
analysis, which accounts for the distribution of adhesive
stresses through the adhesive thickness, may provide a
more accurate solution at the very ends of the adhesive
layer, however considering a safety factor of up to 17
may be necessary for the design of adhesive joints, this
level of accuracy may not be justifiable (Institution of
Structural Engineers 1999).
4.2. Results of Tension Test Specimens for
Spliced Connection
The load-strain behavior at the center of the splice joint
is shown for the three joint configurations in Figure 5.
The initial stiffness of all three joint configurations was
similar up to a load level of 40 kN. At the 40 kN load
level, joint configurations A and B exhibited a sudden
increase of the measured strain which was possibly due
to cracking of the adhesive within the center of the joint.
None of the other strain gauges along the splice plates
812
exhibited a similar increase of strain, which suggests

that the cracking occurred at the joint center. Cracking
of the adhesive resulted in a corresponding loss of
stiffness of the joint as can be seen in the Figure 5. Joint
configuration C did not exhibit a similar increase of
strain suggesting that cracking did not occur and that the
reverse taper was effective in reducing the stress
concentration at the center of the joint. Joint
configuration A failed suddenly due to debonding of the
CFRP splice plates at a load level of 90 kN while joint
configuration B failed by debonding of the plates at a
load level of 160 kN with additional cracking occurring
within the joint at a load level of 144 kN. Debonding of
the splice plates of joint configuration C occurred at a
load level of 190 kN. The test results indicate that for the
current system under investigation, implementing a
reverse taper and a spew fillet significantly increased the
capacity of the splice joint.
4.3. Comparison to Predicted Behavior
Based on the measured strains at the surface of the
splice plates, the corresponding axial stresses at various
locations along the splice plate were calculated. The
measured stress distributions are shown in Figure 6(a)
and (b), for joint configurations A and C, respectively at
a load level of 80 kN. Albat and Romilly (1999) present
an analytical model that can be used to calculate the
distribution of axial stress in the cover plates of doublelap type shear coupons with square plate ends. The
calculated stress distribution is also presented in the
figure for reference purposes. Due to the presence of a
localized shear stress concentration near the square plate
ends at the center of the joint, the calculated stress
distribution exhibits a sharp peak at this location as
shown in Figure 6. Inspection of Figure 6 (a) indicates
that the calculated stress distribution closely matches
the measured trend of the stresses for joint configuration
A, which was fabricated with square plate ends.
Inspection of Figure 6 (b), however, indicates that for
joint configuration C, which was fabricated with
ASE 9-6-07_Schnerch
4/12/06
5:09 pm
Page 813
406.4 mm
50.8 mm
A36 steel plate welded to compression flange
SLB 100 4.8

C
L
Bonded CFRP strip
152.4 mm
60
50
40
30
20
Predicted from bond model
10
From strain distribution of test beam

0
0.0
50.8
101.6
152.4
203.2
254.0
304.8
355.6
406.4
Distance from centre of beam (mm)
Figure 4. Comparison of predicted shear stress distribution and shear stress distribution determined from testing
of beam using 101.6 mm development length
reverse-tapered plate ends within the center of the joint,

the measured stress at the center of the splice was
considerably lower than that predicted by the model.
This further indicates that the presence of the reverse
taper can help to reduce the localized shear stress
concentration near the plate ends.
The average shear stress between the strain gauges
was determined using Equation (1) above. The average
strain was taken to occur at the midpoint between
adjacent strain gauges. The experimentally determined
strain distributions for the three tested joints are shown
in Figure 7. The calculated stress distribution,
determined using the analytical model is also shown,
whereby the experimentally determined shear stress
distributions for all three joint configurations closely
match the calculated shear stress distribution. This
indicates that the analytical model can be used to
accurately predict the shear stresses for the bonded

joints. It should be noted that the measured shear
stresses are sensitive to the spacing between the strain
gauges. Since the strain gauges in this study were placed
with a relatively large spacing, the measured strain
distributions did not accurately capture the reduction of
the shear stress concentrations near the plate ends.
5. DESIGN FOR BOND
An elastic stress-based analysis is recommended for the
design. Although, an elastic approach may neglect
reserve capacity in the adhesive after yielding of the
adhesive, it is desirable to have yielding of the section
occur before the adhesive becomes inelastic since
yielding of the member is more visually apparent then
any non-linear behavior of the adhesive and steps can be
taken to address the overloading of the structure, if any.
813
ASE 9-6-07_Schnerch
4/12/06
5:09 pm
Page 814
250
Strain guage
16
200
12
Configuration C
Load (kN)
150
10
Configuration B
100
6
Configuration A
Average shear stress (MPa)
14
50
2
0
0.0000
0.0005
0.0010
Strain (mm/mm)
0
0.0020
0.0015
400
400
350
350
Splice plate axial stress (MPa)
Splice plate axial stress (MPa)
Figure 5. Load strain behavior at the center of the splice joints
300
250
200
Calculated
Measured
150
100
50
300
250
Calculated
200
Measured
150
100
50
0
50
50
150
250
350
(a) Joint configuration A
450
0
50
50
150
250
350
(b) Joint configuration C
450
Figure 6. Comparison of measured and calculated axial stress distribution in the splice cover plate
It has further been recommended that the maximum

bond stresses in an adhesive joint should not exceed
2030 percent of the ultimate strength of the adhesive
under repeated fatigue loading conditions (Cadei et al.
2004). Designing a joint beyond its elastic strength may
also result in poor creep performance.
An analytical procedure was developed for
determining the bond stresses. This procedure allows
the analysis of beams with bonded FRP strips to the
tension side of the beam. The analysis includes the
814
effect of the applied loading, the thermal effects

resulting from differing coefficients of thermal
expansion, as well as any prestressing applied to the
FRP strip before bonding. Thermal effects should be
considered for any structure that is subjected to thermal
changes. More details regarding the analysis, in addition
to techniques to reduce the stresses at the ends of the
CFRP strips may be found in Schnerch (2005).
As indicated by Cadei et al. (2004), the strength of
the bond must be determined empirically since this
ASE 9-6-07_Schnerch
4/12/06
5:09 pm
Page 815
40
30
Adhesive shear stress (MPa)
At 80 kN
Load level
20
10
10
20
30
50.0
50.0
Joint A
150.0
Joint B
250.0
Joint C
350.0
450.0
Albat & Romily
Figure 7. Experimental and analytical shear stress distributions
strength depends not only upon the properties of the

substrates and adhesives, but also upon the degree of
surface preparation that is expected. The characteristic
strength of an adhesive system can be established using
small-scale single or double lap shear coupon tests.
The test coupons should be prepared using the same
materials, surface preparation techniques and application
techniques as will be used for the strengthening project
to ensure that the coupon test results are representative
of the expected behavior. The maximum shear stress, ,
and normal stress, , can be determined from these
preliminary tests using well established bond models
such as those outlined for the case of double-lap shear
specimens (Hart-Smith 1980). Typically, the maximum
shear and normal stresses are approximately coincident
along the length of the adhesive joint. These stresses can
then be used to determine the maximum principal stress,
of the adhesive or interface. A design value can then be
determined using an appropriate factor of safety
considering the sudden debonding failure mode that
does not give any warning prior to failure.
6. CONCLUSIONS
Surface preparation and detailing are essential to ensure
satisfactory performance of bonded joints between steel
and FRP materials, which are capable of sustaining the
high interfacial stresses necessary to realize the full
strength of these materials. The primary challenge
towards the successful implementation of FRP materials
for steel strengthening is the performance of the bond. It
is not only necessary to consider the short term bond
performance, but new research should attempt to

correlate the performance of accelerated tests to longterm field performance.
Small-scale flexural tests were conducted to
investigate adhesive bond performance. Considerable
variation in the development length and maximum
CFRP strain at failure was found among the adhesives
studied. Several of the adhesives were able to achieve
rupture of the CFRP strip at relatively short
development lengths. The current study indicates that
the development length is proportional to the number of
plys of CFRP strips. Existing analytical techniques are
the most suitable tools that are currently available to
designers in describing the bond behavior of steel to
CFRP bonds. These models predict a shear stress
concentration at the end of the CFRP, which was
confirmed in the current experimental program.
The preliminary experimental results also confirm the
findings of previous finite element studies that the
stress concentration at the end of the strip
can be significantly reduced by implementing a reverse
taper and spew fillet. Adhesive selection and careful
detailing are critical to the satisfactory performance of
steel members strengthened with externally bonded
CFRP strips.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the financial
support provided by the National Science Foundation
Industry/University Cooperative Research Center and
Mitsubishi Chemical FP America, Inc.
815
ASE 9-6-07_Schnerch
4/12/06
5:09 pm
Page 816
REFERENCES
Albat, A.M. and Romilly, D.P. (1999). A direct linear-elastic
analysis of double symmetric bonded joints and reinforcements,
Composites Science and Technology, Vol. 59, pp. 11271137
Allan, R.C., Bird, J. and Clarke, J.D. (1988). Use of adhesives in
repair of cracks in ship structures, Materials Science and
Technology, Vol. 4, pp. 853859.
Allen, K.W., Chan, S.Y.T. and Armstrong, K.B. (1982). Coldsetting adhesives for repair purposes using various surface
preparation methods, International Journal of Adhesion and
Adhesives, Vol. 2, No. 4, pp. 239247.
Belingardi, G., Goglio, L. and Tarditi, A. (2002). Investigating the
effect of spew and chamfer size on the stresses in metal/plastics
adhesive joints, International Journal of Adhesion and
Adhesives, Vol. 22, pp. 273282.
Brown, A.R.G. (1974). Corrosion of CFRP to metal couples in
saline environments, Proceedings of the 2nd International
Conference on Carbon Fibres, London, England, February, Paper
No. 35, pp. 230241.
Brown, S.R. and De Luccia, J.J. (1977). Corrosion characteristics of
naval aircraft metals and alloys in contact with graphite-epoxy
composites, Proceedings of the Environmental Degradation of
Engineering Materials Conference, Virginia Polytechnic Institute
and State University, Blacksburg, Virginia, October, pp. 277288.
Buyukozturk, O., Gunes, O. and Karaca, E. (2004). Progress on
understanding debonding problems in reinforced concrete and
steel members strengthened using FRP composites, Construction
(2004). Strengthening Metallic Structures Using Externally
Bonded Fibre-Reinforced Polymers. Publication C595,
Construction Industry Research and Information Association
(CIRIA), London, UK, 234 pp.
Choqueuse, D., Davies, P., Mazeas, F. and Baizeau, R. (1997).
Aging of composites in water: Comparison of five materials in
terms of absorption kinetics and evolution of mechanical
properties, High Temperature and Environmental Effects on
Polymeric Composites: 2nd Volume, ASTM Special Technical
Publication 1302, Thomas S. Gates and Abdul-Hamid Zureick,
eds., American Society for Testing and Materials, pp. 7396.
Dawood, M. (2005). Fundamental Behavior of Steel-Concrete
Composite Beams Strengthened with High Modulus Carbon
Fiber Reinforced Polymer Materials, Masters Thesis, North
Carolina State University, 213 pp.
El Damatty, A., Abushagur, M. and Youssef, M.A. (2003).
Experimental and analytical investigation of steel beams
rehabilitated using GFRP Sheets, Steel and Composite
Evans, U.R. and Vera, E.R. (1958). Corrosion and its Prevention at
Bimetallic Contacts, Her Majestys Stationery Office, London,
England.
Francis, R. (2000). Bimetallic Corrosion: Guides to Good Practice
in Corrosion Control, National Physical Laboratory, Teddington,
816
Middlesex, 15 pp.
Frieze, P.A. and Barnes, F.J. (1996). Composite materials for
offshore application new data and practice, Proceedings of the
28th Annual Offshore Technology Conference, Construction and
Installation, Houston, Texas, May, Vol. 3, pp. 247253.
Garden, H.N., Quantrill, R.J., Hollaway, L.C., Thorne, A.M. and
Parke, G.A.R. (1998). An experimental study on the anchorage
length of carbon fibre composite plates used to strengthen
reinforced concrete beams, Construction and Building
Gettings, M. and Kinloch, A.J. (1977). Surface analysis of
polysiloxane/metal oxide interfaces, Journal of Materials
Science, Vol. 12, No. 12, pp. 25112518
Harris, A.F. and Beevers, A. (1999). The effects of grit blasting on
surface properties for adhesion, International Journal of
Adhesion and Adhesives, Vol. 19, No. 6, pp. 445452.
Hart-Smith, L.J. (1980). Further developments in the design and
analysis of adhesive-bonded structural joints, ASTM Special
Technical Publication 749, Joining of Composite Materials: A
Symposium, K.T. Kedward, Ed., April, pp. 331.
Hashim, S.A. (1999). Adhesive bonding of thick steel adherends for
marine structures, Marine Structures, Vol. 12, No. 6, pp. 405423.
Hildebrand, M. (1994). Non-linear analysis and optimization of
adhesively bonded single lap joints between fibre-reinforced
plastics and metals, International Journal of Adhesion and
Adhesives, Vol. 14, No. 4, pp. 261267.
Hollaway, L.C. and Cadei, J. (2002). Progress in the technique
of upgrading metallic structures with advanced polymer
composites, Progress in Structural Engineering Materials, Vol.
4, No. 2, pp. 131148.
Hutchinson, A.R. (1987). Surface pretreatment the key to durability,
Proceedings of the International Conference on Structural Faults &
Repair, University of London, July, pp. 235244.
Institution of Structural Engineers (1999). A Guide to the Structural
Use of Adhesives, The Institution of Structural Engineers,
London, UK, 51 pp.
Karbhari, V.M. and Shulley, S.B. (1995). Use of composites for
rehabilitation of steel structures - determination of bond
durability, Journal of Materials in Civil Engineering, Vol. 7,
No. 4, pp. 239245
Lenwari, A., Thepchatri, T. and Albrecht, P. (2006). Debonding
strength of steel beams strengthened with CFRP plates, Journal
of Composites for Construction, ASCE, Vol. 10, No. 1, pp. 6978.
Mays, G.C. and Hutchinson, A.R. (1992). Adhesives in Civil
Engineering, Cambridge University Press, New York, New York,
333 pp.
McKnight, S.H., Pierre, E.B., John, W., Gillespie, Jr. and Karbhari,
V.M. (1994). Surface preparation of steel for surface bonding
applications, Infrastructure: New Materials and Methods of
Repair, Proceedings of the 3rd Materials Engineering
Conference, ASCE, Kim D. Basham, Ed., Nov., San Diego,
California, pp. 11481155.
Miller, T.C., Michael, J.C, Dennis, R.M. and Jason, N.H. (2001).
ASE 9-6-07_Schnerch
4/12/06
5:09 pm
Page 817
Strengthening of a steel bridge girder using CFRP plates, Journal

of Bridge Engineering, ASCE, Vol. 6, No. 6, pp. 514522.
Miriyala, S.K., Tucker, W.C., Rockett, T.J. and Brown, R. (1992).
Degradation of carbon reinforced polymer composites under
galvanic coupling conditions, Proceedings of the 33rd
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics,
and Materials Conference, Dallas, Texas, April, pp. 30363045.
Nozaka, K., Carol, K.S. and Jerome, F.H. (2005). Effective bond
length of carbon-fiber-reinforced polymer strips bonded to
fatigued steel bridge I-girders, Journal of Bridge Engineering,
ASCE, Vol. 10, No. 2, pp. 195205.
Ong, C.L. and Shen, S.B. (1992). The reinforcing effect of
composite patch repairs on metallic aircraft structures,
International Journal of Adhesion and Adhesives, Vol. 12, No. 1,
pp. 1926.
Parker, B.M. (1994). Adhesive bonding of fibre-reinforced
composites, International Journal of Adhesion and Adhesives,
Vol. 14, No. 2, pp. 137143.
Photiou, N.K., Hollaway, L.C. and Chryssanthopoulos, M.K.
(2004). An ultra-high modulus carbon/glass fibre composite
system for structural upgrading of steel members, Proceedings
of the 2nd International Conference on FRP Composites in Civil
Engineering CICE 2004, Seracino, R., ed., December,
pp. 741748.
Price, A. and Moulds, R.J. (1991). Repair and strengthening of
structures using plate bonding, Construction and Building
Schnerch, D.A. (2005). Strengthening of Steel Structures with High
Modulus Carbon Fiber Reinforced Polymer (CFRP) Materials,
Ph.D. Dissertation, North Carolina State University, 265 pp.
Sloan, F.E. and Talbot, J. B. (1992). Corrosion of graphite-fiberreinforced composites I - galvanic coupling damage, Corrosion,
Vol. 48, No. 10, pp. 830838.
Smith, S.T. and Teng, J.G. (2001). Interfacial stresses in plated
beams, Engineering Structures, Vol. 23, No. 7, pp. 857871.
Stallings, M. and Porter, N.M. (2003). Experimental investigation
of lap splices in externally bonded carbon fiber-reinforced plastic
plates, ACI Structural Journal, Vol. 100, No. 1, pp. 310.

Sykes, J.M. (1982). Surface treatments for steel, Surface Analysis
and Pretreatment of Plastics and Metals, D.M. Brewis, Ed.,
Applied Science Publishers, Ltd., Essex, England, pp. 153174.
Tang, B.M. and John, M.H. (2001). FRP Composites industry is
changing the American bridge building industry, Proceedings of
the International Conference on FRP Composites in Civil
Engineering, , J.G. Teng, ed., Hong Kong, December, Vol. 2,
pp. 16571663.
Tavakkolizadeh, M. and Hamid, S. (2001). Galvanic corrosion of
carbon and steel in aggressive environments, Journal of
Composites for Construction, ASCE, Vol . 5, No. 3, pp. 200210.
Tucker, W.C. and Richard, B.(1989). Blister formation on
graphite/polymer composites galvanically coupled with steel in
seawater, Journal of Composite Materials, Vol. 23, No. 4,
pp. 389395.
West, T.D. (2001). Enhancements to the Bond between Advanced
Composite Materials and Steel for Bridge Rehabilitation,
Masters Thesis, University of Delaware, 207 pp.
Wright, P.N.H., Wu, Y. and Gibson, A.G. (2000). Fibre
reinforced composite-steel connections for transverse ship
bulkheads, Plastics, Rubber and Composites, Vol. 29, No. 10,
pp. 549557.
Xiong, Y. and Raizenne, D. (1996). Stress and failure analysis of
bonded composite-to-metal joints, AGARD Specialist Meeting on
Bolted/bonded Joints in Polymeric Composites, AGARD conference
proceedings 590, Florence, Italy, September, pp. 9.19.11.
Yang, X. and Nanni, A. (2002). Lap splice length and fatigue
performance of fiber-reinforced polymer laminates, ACI
Materials Journal, Vol. 99, No. 4, pp. 386392.
APPENDIX: NOTATION
Efrp Modulus of elasticity of FRP
tfrp
thickness of FRP strip
shear stress
av
average shear stress
normal or peel stress
817
ASE 9-6-07_Schnerch
4/12/06
5:09 pm
Page 818
ASE 9-6-08_Kliger
4/12/06
5:08 pm
Page 819
Experimental and Numerical Investigation

of the Behaviour and Strength of Composite
Steel-CFRP Members
Mohammad Al-Emrani* and Robert Kliger
Department of Civil and Environmental Engineering, Steel and Timber Structures, Chalmers University of Technology, Gteborg, Sweden
Abstract: The use of advanced composite materials to strengthen and repair existing
structures is increasing rapidly. One specific area in which the technique has recently
been introduced is the strengthening of metallic structures with bonded carbon-fibre
laminates. In this paper, the behaviour of composite steel-CFRP members is studied
experimentally and using FE analysis. A new type of test specimen has been developed
to examine the effect of various material parameters on the behaviour and strength of
bonded steel-CFRP elements. Finite element analysis of the tested elements was also
employed to examine the type and magnitude of interfacial stresses in the bond line of
these elements. Moreover, different types of fracture mode could be examined by
testing composite elements with different combinations of CFRP-laminates and
adhesives. The tested composite elements displayed different behaviour and a large
difference in strength and ductility could be observed.
Key words: bonding, strengthening, steel, CFRP, adhesive, interfacial, shear stresses.
1. INTRODUCTION
The use of composite materials for the strengthening and
repair of existing structures has increased significantly
during the last decade. The high durability and fatigue
resistance of these materials and their superior strengthto-weight ratio enable them to compete easily with other
traditional building materials in this field. After having
initially been used for concrete structures, the use of
composite materials for strengthening and repair work
has been extended to include timber and masonry and
more recently also metallic structures. Several field
applications have been reported in which steel and
wrought-iron structures (mostly bridges) have been
strengthened or repaired using adhesively bonded
carbon fibre laminates (CFRP) (Luke 2001; Miller et al.
2001). In some cases, pre-stressing of the laminates was
employed to obtain a more effective strengthening
schedule and higher utilisation of the high tensile

strength of the composite material (Bassetti 2001).
Several experimental studies have also been conducted
in this area. The results of laboratory tests on steel by
Bassetti (2001), Mertz et al. (1996), Jiao et al. (2004) and
composite steel-concrete beams (Al-Saidy et al. 2004;
Sen et al. 2001; Tavakkolizadeh and Saadatmanesh
2003) strengthened with adhesively bonded CFRP have
been reported. The vast majority of these tests were
devoted to beams in bending and they cover a wide
range of application areas, including the upgrading of
bending capacity, increasing bending stiffness, repairing
fatigue-cracked elements and increasing the fatigue
strength of steel beams. In general, the main focal point
of the reported experimental studies has been the degree
of strengthening that can be obtained by using this
technique. In almost all cases, large-scale beams were
used for this purpose. More detailed studies focusing on
*Corresponding author. Email address: Mohammad.Al-Emrani@chalmers.se; Tel: +46-31-772-2258.
819
ASE 9-6-08_Kliger
4/12/06
5:08 pm
Page 820
Experimental and Numerical Investigation of the Behaviour and Strength of Composite Steel-CFRP Members
the behaviour and strength of composite steel-CFRP

elements are somewhat scarce. Furthermore, analytical
work conducted in this field mainly emanates from the
work performed on concrete beams and is limited to
linear elastic analysis. Consequently, knowledge relating
to the behaviour and strength of composite steel-CFRP
elements is still lacking. In particular, the behaviour,
strength and possible fracture modes of these composite
members in the post-yielding stage need to be studied
more closely. Standard double-lap tests are apparently
not suitable for this purpose, as yielding takes place in
the unstrengthened part of the test specimen (Colombi
et al. 2006). Knowledge about the appropriate selection
of materials (carbon fibre, adhesives) for the different
applications (increased strength in the ULS, increased
stiffness, fatigue and so on) is also lacking. For
example, the general opinion that high-modulus
laminates are more effective in steel applications does
not always hold true.
Research related to the strengthening and repair of
metallic structures using bonded CFRP has been
ongoing at Chalmers University of Technology in
Sweden since 2003. Various types of laboratory test and
analytical and numerical work have been performed in
this field. In this paper, the results of an experimental
study of the behaviour and strength of composite steelCFRP elements are presented. Finite-element analysis
of the tested specimens has also been conducted and the
results are presented and discussed. The main objectives
of the study were as follows.
1. To study the behaviour of composite steel-CFRP
elements, in particular in the post-yielding stage.
2. To examine possible fracture modes that can be
obtained for steel elements strengthened by

bonded CFRP.
3. To evaluate the effect of various material
properties on the strength and ductility of the
composite elements, as well as on the type of
fracture mode obtained for these elements.
As standardised tests that can be used for these purposes
are lacking, a new type of test specimen was developed
for the tests. The shape and dimensions of the new test
specimen were established on the basis of the results
obtained from extensive finite-element analysis.
2. MATERIAL TESTS
A wide range of different carbon-fibre-reinforced
composites, with a variety of strength and stiffness
classes, is currently available in the market. Typically,
laminates with an ultimate tensile strength of 1300 to
3500 MPa, with axial stiffness ranging from 450 to 150
GPa, are available. Two adhesive-laminate systems,
supplied by two different manufacturers, were used in
the experimental study reported in this paper. These
systems are designated A and B. The relevant material
properties for the laminates and adhesives first had to be
established. Tension tests were performed on dog-bone
adhesive specimens at four different curing times and
four different temperatures. Typical stress-strain curves
for the two adhesives used in the tests are given in
Figure 1. Similar tension tests were also carried out on
the CFR laminates. Table 1 shows the relevant material
properties for the two types of adhesive used in the tests,
while the strength and stiffness of the laminates are
shown in Table 2. For the laminates, the measured
values differed somewhat from the values provided by
40
Adhesive B
Adhesive A
35
Stress (MPa)
30
25
20
15
10
5
0
0
0.002
0.004
0.006
0.008
0.01
Strain
Figure 1. Stress-strain curves obtained at room temperature for the adhesives used in the tests
820
ASE 9-6-08_Kliger
4/12/06
5:08 pm
Page 821
Mohammad Al-Emrani and Robert Kliger
Table 1. Material properties of the two types

of adhesive used in the tests
(average values at room temperature after seven
days of curing)
Adhesive
A
B
E [GPa]
ult [MPa]
ult [x103]
14
6.5
32
24
3
7
0.27
0.35
Table 2. Measured strength and stiffness

(parallel to the fibres) of the four types
of laminate used in the tests
CFRP type
A12
B12
B17
B40
Thickness
1.2
1.43
1.95
4.4
E//
//,ult
155
174
383
362
1932
1855
1252
(1252)*
*No reliable value for the ultimate tensile strength of this type of laminate
could be obtained. As it was made of the same material as B17, the same
tensile strength was assumed for both laminates.
the manufacturer. This measured value was 80% of the

value provided by the manufacturer when it came to the
longitudinal stiffness of the two high-modulus laminates
B17 and B40.
3. DEVELOPMENT OF TEST SPECIMEN
AND FINITE-ELEMENT MODELLING
Figure 2 shows schematically the load effects along the
bond line in a steel beam strengthened with adhesively
bonded CFRP and loaded in bending. Four possible
failure modes can be identified.
1. Rupture of the laminate when the maximum axial

stress in the laminate reaches its ultimate strength
2. Debonding failure at the end of the laminate
3. Debonding failure in the middle of the laminate
4. Interlaminar failure (delamination) at the end of
the laminate
Debonding failure at the ends or in the middle of the
laminate can be governed by either the cohesive strength
of the adhesive (failure within the adhesive) or the
interfacial strength of one of the adhesive-adherend
interfaces. The governing failure mode in a specific
beam will primarily depend on: the properties of the
materials used (such as laminate and adhesive stiffness
and strength), geometrical properties (such as laminate
and adhesive thickness and width and laminate end
preparation) and the mechanical properties (like the
characteristics of the bonding surfaces and possible
defects in the bond line).
Figure 3 shows the final test specimen, which was
developed to realise the objectives of the tests. Initially,
the shape and dimensions of the specimen were to be
chosen so that:
successive yielding of the steel adherent could be
obtained, starting from the middle of the
specimen,
the fracture modes expected to be obtained in
steel beams with bonded CFRP (Figure 2) could
be reproduced.
In the first step, an FE model of the test specimen was
created. The model included one-eighth of the test
specimen, taking advantage of the symmetry. For this
model, brick elements with 20 nodes were used for all
parts (i.e. the steel plate, adhesive and laminate). The
shape and dimensions of the steel plate were modified
Portion of the beam undergoing

plastic deformation
y
x
yx
yy
Before
yielding
yx
yy
After
yielding
Figure 2. Schematic illustration of the principal load effects in a steel beam strengthened with bonded CFRP
821
ASE 9-6-08_Kliger
4/12/06
5:08 pm
Page 822
2 mm adhesive with varying laminate thickness
900
150
100
200
T = 10
25
90
36
R = 400
R = 400
Figure 3. The final shape and dimensions of the test specimen developed to study the behaviour and strength of composite steel-CFRP
Steel
CFRP
Adhesive
(a)
Steel
CFRP
Adhesive
(b)
Figure 4. The two finite-element models used in the development of the test specimen: (a) Model I with 20-node brick elements; (b) Model
II with 8-node brick elements
several times. In each case, the analysis was conducted

with four different laminate-adhesive combinations, one
from system A and three from system B, see Table 1.
For each analysis, the maximum longitudinal strain in
the laminate and the interfacial shear stresses in the
bond line at the end and middle of the laminate were
examined. The final shape and dimensions of the test
specimen were chosen so that the different fracture
modes expected for steel beams would be reproduced in
the tests.
After testing, the overall performance of the FE
model was validated using strain readings from several
strain gauges installed at different locations on the
laminates and the steel plate in each specimen. In order
to examine the accuracy of the first model, a second FE
model with much finer mesh was also constructed and
analysed. The main parameter, which was investigated
by comparing the results obtained from the two models,
is the effect of element size, particularly near the end of
the laminate, on the interfacial normal and shear stresses
in this location. Figure 4 shows the two FE models
examined in the study and the main features of the two
822
models are given in Table 3.

Both models included a concave fillet of adhesive
at the end of the laminate. A fillet of this kind is often
produced in practice from the excess adhesive
material when making the joint. The fillet at the end of
the joint also helps to avoid the numerical singularity
problems that are encountered when modelling joints
with dissimilar materials (Teng et al. 2002). Nonlinear material models were used for the steel material,
as well as for the adhesives. The models were derived
from standard tension tests on dog-bone adhesive and
steel specimens. The laminates were modelled as
linearly elastic material.
Both FE models were capable of reproducing the
behaviour of the tested specimens. The overall stiffness
of the composite elements, as well as the distribution of
strain at various locations along the laminates and the
steel plate, was in good agreement with the measured
experimental values, both in the elastic and in the postyielding stage, cf. Figure 5. Further, the interfacial
normal and shear stresses along the bond line were
examined at three different sections through the
ASE 9-6-08_Kliger
4/12/06
5:08 pm
Page 823
Table 3. The main features of the two models examined for FE analysis
of the test specimens
Model
No. of nodes
20-node brick
8-node brick
Stress on the laminate surface (MPa)
I
II
Element type
No. of elements through

adhesive thickness
No. of elements
17207
73935
3324
65380
2
8
1400
FEM, P = 63,2 kN
FEM, P = 215 kN
Measured P = 63 kN
Measured P = 215 kN
1200
1000
800
600
400
200
0
50
100
150
200
250
300
Distance from specimen centre line (mm)
Figure 5. Measured and calculated longitudinal strain (presented as stresses) in the laminate for specimen B17 (The values correspond to
two load levels before and after the yielding of the steel plate, namely 63 and 215 kN respectively)
adhesive thickness: the steel-adhesive interface, the

middle of the adhesive layer and the laminate-adhesive
interface. The results produced by both models were
almost identical. However, both models are unable to
reproduce the behaviour of the strengthened beam after
debonding at the mid-span, when vertical cracks in the
adhesive layer occur after the yielding of the steel
beam. Based on the results of this initial investigation,
the first model (model I) was chosen to analyse the
tested specimens.
4. PRODUCTION AND PREPARATION
OF TEST SPECIMENS
AND TESTING PROCEDURE
Four composite steel-CFRP elements and one reference
element (with steel only) were manufactured and tested.
One of the composite elements was made using system
A (i.e. adhesive A and laminate A12), while system B
was used in the remaining three elements. In order to
avoid any residual stress effects that might result from
flame cutting and the mechanical preparation of the steel
plate, the latter was cut to the final shape using a water
jet. The laminates were also cut to the desired length and
width using the same technique. Prior to bonding, the
opposite surfaces of the steel plate where the laminates
were to be bonded were sandblasted. Immediately
thereafter, the steel surfaces were cleaned with alcohol

and coated with a thin layer of primer to prevent
oxidation and improve bonding. Following the curing of
the primer, the laminates and the bonding areas of the
steel plate were covered with adhesive and a roller was
used to press the laminate onto the steel surface. Small
spacers made of timber were used to ensure a uniform
adhesive layer thickness of 2 mm. The specimens were
then left to cure for seven days. The adhesive layer was
given a concave fillet at the ends of the laminates. Each
test specimen was equipped with a number of strain
gauges (2 mm long measurement) to monitor the
longitudinal strain at different locations in the laminate
and the steel plate. A universal uniaxial testing machine
was used to load each test specimen in tension at a
displacement rate of 0.25 mm/min.
5. LOAD EFFECTS IN THE COMPOSITE
SPECIMENS
When the composite specimen is loaded, the steel plate
and the laminates bonded on each side carry a percentage
of the load according to their relative axial stiffness.
Stiffer laminates (with a larger E-modulus and/or higher
thickness) will attract a larger percentage of the total
applied axial load. Owing to the non-uniform shape of
the steel plate (and thereby also its axial stiffness), the
823
ASE 9-6-08_Kliger
4/12/06
5:08 pm
Page 824

50
100
150
200
300
250
Shear stress (MPa)
1
2
3
4
250
260
270
280
290
300
3
Analytical
5
5
6
Mid-adhesive
Steel-adhesive
Laminate-adhesive
Figure 6. The variation in shear stress along the bond line in specimen B12 analytical and FE results
distribution of axial force in the laminate will also be

non-uniform. Shear stresses will therefore be present
along the whole bond line of the specimen, but the
magnitude of these stresses will be highest at the end of
the laminate. Figure 5 shows an example of the
variation in axial stress in the laminate along half its
length (from the middle of the laminate to its end)
obtained from the FE model for two different values of
applied axial load. The corresponding values obtained
from strain measurements for the same loads are also
shown. The percentage of the axial load carried by the
laminates increases substantially after the yielding of
the steel plate.
An example of the variation in shear stress along the
bond line of specimen B12 is shown in Figure 6 at the
three different sections through the adhesive thickness.
The shear stress derived analytically for a steel plate with
a constant width is also shown. The analytical solution,
which is based on the assumption of constant shear stress
through the adhesive thickness (Eqn 1), gives valid
results for the distribution of shear stress near the end of
the laminate, where the width of the specimen is
constant. Eqn 1 was derived for a steel plate with a
constant width and two laminates bonded on each side of
the plate. The derivation is not included in this paper;
however, it can be found in Al-Emrani et al. (2005).
EL tL
sinh( x )
( x ) = P
cosh(L)
Es ts + 2 EL tL
(1)
In Eqn 1, EL and tL are the E modulus and the thickness

of the laminate, Es and ts are the corresponding values
for the steel plate, x is defined in Figure 11 and L is the
laminate length. Ga and ta are the shear modulus and the
824
thickness of the adhesive and the parameter is defined

in Eqn 2.
=
Ga 1
2
+
E s t s
ta EL tL
(2)
As can be seen in Figure 6, the shear stress near the end

of the laminate is not constant through the adhesive
thickness. This observation was also made in a previous
FE analysis of adhesive joints (Teng et al. 2005) and in
higher order analytical solutions of the problem (HuiShen et al. 2001). Due to the presence of the adhesive
fillet at the end of the laminate, the shear stress at the
steel-adhesive interface is reduced. At the laminateadhesive interface, where the shear stress has its
maximum value, the analytical solution provides a very
good approximation.
The same observations could be made for the other
three specimens. Table 4 lists the values of maximum
shear and normal stresses at the three different sections
through the adhesive thickness for the four tested
specimens at the same applied load (P = 63 kN). The
behaviour of all the specimens is totally elastic at this
load level. Specimens B40 and B17, which comprise
laminates with an ultra-high E modulus, experience
higher shear stresses, with B40 as the highest of the two
due to larger laminate thickness. When it comes to
specimens A12 and B12, the latter had a thicker
laminate with a higher E modulus. Nevertheless, the
magnitude of maximum shear stress at the ends of the
laminates was comparable to that obtained for specimen
A12. This can be explained by the fact that, in specimen
B12, an adhesive with a lower shear modulus was used,
which contributed to a less steep distribution of shear
ASE 9-6-08_Kliger
4/12/06
5:08 pm
Page 825
Table 4. Maximum interfacial normal and shear stresses obtained

from the FE analysis of the tested specimens (P = 63 kN)
Steel-adhesive
A12
B12
B17
B40
Mid-adhesive
Laminate-adhesive
max
max
max
max
max
max
4.4
4.3
7
8.8
6.8
5.8
8.8
11.0
5.9
5.3
8.2
9.6
6.3
5.4
8.1
10.4
6.6
6.4
10
12.0
4.1
3.8
6.0
8.3
6
Laminate-adhesive
Mid-adhesive
Normal stress (MPa)
Steel-adhesive
0
275
280
285
290
295
300
305
2
Figure 7. An example of the variation in interfacial normal stress near the ends of specimen B12 (P = 63 kN) as obtained
from the FE analysis
stress near the end of the laminate and thereby to a lower

maximum shear stress.
In addition to the interfacial shear stresses, normal
stresses (perpendicular to the plane of the laminate) also
exist in the adhesive joint near the ends of the laminate.
These stresses are the result of the eccentricity in load
transfer from the steel plate to the laminate. As such,
they are expected to be higher in specimens with thicker
(and stiffer) laminates. Figure 7 shows an example of
the variation in normal stresses in the adhesive joint.
The maximum normal stress values for all the tested
specimens are given in Table 4. In all the specimens, the
maximum normal stress value appears at the end of the
laminate on the steel-adhesive interface. These stresses
converge to zero at the end of the adhesive fillet. At the
end of the laminate-adhesive interface, normal stresses
are compressive in a very localised area, which is also in
agreement with the findings presented in previous FE
studies of bonded joints (Teng et al. 2002).
Following the yielding of the steel plate in the middle
of the specimen, the plate itself is unable to carry any
additional load and any increase in the applied load has to

be transferred to the laminate. This can easily be seen
considering the variation in axial stress in the middle of
the laminate with an applied load (cf. Figure 13). This
transition of axial force from the steel plate to the
laminate is associated with interfacial shear stresses in the
bond line, which might be high enough to cause
debonding. Another way to look at the problem is that
these interfacial shear stresses are caused by the large
difference in longitudinal strain between the steel plate
(which is now yielding and losing its initial stiffness) and
the laminate, which is still elastic and stiff. This
difference in strain has to be accommodated by the
adhesive, which (if relatively stiff) might crack or debond
from the steel surface. This type of failure cannot be
predicted by the type of FE model employed in this study.
Figure 8 shows the distribution of shear stresses
along the bond line, obtained from the FE analysis of
specimen A12 at three different stages of loading. The
yielding of the steel plate in the middle of the specimen
took place at a load of 120 kN. Prior to yielding, very
825
ASE 9-6-08_Kliger
4/12/06
5:08 pm
Page 826

0
2
50
100
150
200
250
300
Shear stress (MPa)
4
6
8
10
P = 63 kN
P = 150 kN
12
P = 184 kN
14
16
Figure 8. The variation in interfacial shear stress along the bond line in specimen A12 at three different levels of applied load (The vertical
dotted lines show the progress of yielding from the middle of the specimen (x = 0))
low shear stresses are present in the bond line near the
middle of the specimen due to the change in the steel
plate width, as has previously been discussed. With the
successive yielding of the steel plate, higher shear
stresses are created near the middle of the specimen.
These are increased as the yielding in the plate
progresses (due to increased plastic strain) and the
location of maximum shear stress shifts towards the end
of the laminate. At higher loads (P = 184 kN), the
magnitude of maximum shear stress in the middle of the
specimen is higher than that at the ends of the laminate.
Debonding (cohesive or interfacial) might therefore
begin in the middle of the specimen before failure at the
end of the laminate is obtained. This observation was
also made for specimen A12.
6. TEST RESULTS
Figure 9 shows the load-displacement curves obtained for
the four composite elements. The results obtained from
the reference specimen (steel only) are also shown for
comparison. The same results are listed in Table 5. The
following observations can be made.
1. As one might expect, the delay in the onset of
yielding in the composite specimens in
comparison to the reference specimen is highest
for the laminates with a high E modulus (the
difference between B17 and B40 being attributed
to the higher thickness in the latter), see Table 2.
2. The highest degree of strengthening (i.e.
increase in the ultimate load) was obtained for
specimen B12. This specimen contains the
laminates with the lowest stiffness and the
lowest thickness. The ductility was also highest
826
for this specimen. Specimen B17 with a stiffer

laminate produced a slightly lower strengthening
effect but far lower ductility.
The four composite elements displayed clearly different
behaviour. In addition to the observed difference in
ultimate load and ductility, the fracture modes displayed
by the specimens were also different. In fact, the tests
succeeded in reproducing the four different fracture
modes which are expected to be obtained in steel beams
strengthened with bonded CFRP, see Figure 1. The main
observations concerning the failure of the composite
elements are listed in Table 6.
6.1. Delamination or Debonding at the Ends
of the Laminates Due to Maximum
Shear Stress
Delamination and/or debonding were observed in three
of the four tested specimens. However, this was the final
mode of failure in two only, B40 and A12. The
behaviour of specimen B40 was fully elastic up to
failure (i.e. without the yielding of the steel plate). The
specimen failed in a very brittle manner due to high
interlaminar shear stress at the ends of the laminate. The
fracture surface was within the laminate itself, see
Figure 10. At failure, the maximum axial stress in the
laminate was only 25% of its actual tensile strength.
Figure 11 shows the variation in shear stress with an
applied load at some distance from the end of the
laminate. The shear stress values were derived from
readings from strain gauges that were placed near the
end of the laminate, also shown in Figure 8. These
values agreed very well with those obtained from the FE
analysis (cf. Figures 6 and 8). It should be noted that the
ASE 9-6-08_Kliger
4/12/06
5:08 pm
Page 827
Table 5. The strengthening effects of bonding CFRP to the steel plates in the tested specimen
Specimen
Steel
A12
B12
B17
B40
Yielding
load (kN)
Increase in
yielding load
(%)
Ultimate load
(kN)
Increase in
ultimate load
(%)
Deformation
at failure
(mm)
101
114
123
147
14
23
47
>57
140
186
251
247
157
33
80
76
12
12.51
3.34
7.51
2.10
0.95
Table 6. Failure modes observed from testing

First observed
failure mode
Specimen
A12
Load
(kN)
Debonding
(adhes./steel) in the
middle of the specimen;
extended towards the
ends
Debonding at the ends
of the laminates
extending inwards
B12
B17
B40
Final failure
mode
Load
(kN)
150
Debonding at the ends of the laminates
184
244
Tensile rupture of the laminate in the

middle of the specimen
230
210
Tensile rupture of the laminate in the

middle of the specimen
247
Delamination at the ends of the laminates

in a very brittle manner
157
300
B-17, Pmax = 247 kN
B-12, Pmax = 251 kN
B-40, Pmax = 157 kN
250
Load (kN)
200
150
A-12, Pmax = 186 kN

Reference specimen, Pmax = 140 kN
100
50
0
0
6
8
Displacement (mm)
10
12
14
Figure 9. Load-deformation curves obtained for the five tested specimens
827
ASE 9-6-08_Kliger
4/12/06
5:08 pm
Page 828
Figure 10. The fracture surfaces at one end of the laminate in specimen B40: (a) Laminate surface; (b) Steel plate surface where the carbon
fibres are still bonded to the adhesive
30
25
20
Shear stress
15
10
x 1
FL(X)
FP(X)
0
0
50
100
150
5
B-12, x = 2.5 mm
10
P1
250
1 =
1
.E .t
x 1 L L
350
B-17, x = 1.5 mm
B-40, x = 3.0 mm
A-12, x = 3.0 mm
15
Applied load (kN)
Figure 11. Shear stresses at the end of the laminates in the four tested specimens as derived from strain measurements (It should be noted
that these stresses are derived at slightly different distances (x) from the end of the laminates, as the distances from the end of the laminate
to the first strain gauge (x) were slightly different. Notations are in accordance with Eqns 1 and 2)
distance at which the shear stresses are derived varies

due to some discrepancy in the location of the strain
gauges, 3-6 mm from the end of the laminate.
The stiff and thick laminate in specimen B40
attracted a major percentage of the applied load and the
laminate experienced a high concentration of shear
stress at its end. This led to the sudden and premature
interlaminar shear failure (delamination) of the laminate
in this specimen. Unfortunately, no information is
available on the interlaminar strength of the laminates
used in the study. However, using the FE analysis of
specimen B40, a value of maximum principal stress in
the adhesive near the end of the laminate of 24.7 MPa
828
could be derived (at the laminate-adhesive interface).

This value is very close to the strength of adhesive
B used in this specimen (cf. Figure 1).
The second specimen which failed due to delamination
at the end of the laminate was specimen A12. Prior to
failure, this specimen experienced successive debonding
in the middle. This is discussed in more detail below.
Even specimen B12 showed some signs of debonding/
delamination at the ends of the laminates just before
failure. The loading of this specimen could, however, be
continued, with the debonding areas at the ends of the
laminates extending gradually. The failure of this
specimen was determined by the tensile rupture of the
ASE 9-6-08_Kliger
4/12/06
5:08 pm
Page 829
Final debonding
failure
Initial debonding due to steel yielding
Final debonding
failure
Figure 12. The failure surface in specimen A12
laminate in the middle of the specimen.

6.2. Debonding in the Middle of the Specimen
due to the Successive Yielding
of the Steel Plate
As has been mentioned before, the successive yielding of
the steel plate starting from the middle of the specimen is
associated with the gradually increasing interfacial shear
stresses in the bond line at these locations. These shear
stresses are primarily caused by the large difference
between the very high strain in the part of the steel plate
which has yielded and the stiff laminate which is still
behaving in an elastic manner. If the adhesive layer fails
to accommodate this large difference in deformation,
failure of the bond line might be obtained.
This was the case for specimen A12. In this specimen,
the adhesive used to bond the laminate to the steel plate
(adhesive A) was two times stiffer than the adhesive used
in the other three specimens (adhesive B). At some load
close to the ultimate load, debonding was observed
between the adhesive and the steel plate. The debonded
area then extended successively from the middle of the
specimen towards the ends of the laminates. This was
associated with a gradual increase in the shear stresses at
the ends of the laminate and failure was finally obtained
due to debonding at the ends, see Figure 12. The residual
bonding length at both ends of the laminates was around
120 mm. Specimen A12 was the only specimen in which
debonding in the middle of the specimen due to yielding
could be observed during the tests. FE analysis of the
tested specimens also shows that specimen A12
experienced high shear stresses in the middle. These
stresses were even higher that those present at the end of
the laminate (cf. Figure 8). It should be pointed out here
that the type of FE models employed in this study does not
make it possible to capture the behaviour/load effects in
areas where local debonding or delamination takes place.
Due to the more flexible epoxy used in specimen
B12, much larger plastic strain in the steel plate could

take place without initiating debonding in the middle of
the specimen. However, evident signs of high shear
stresses in the bond line in the middle of the specimen
due to the yielding of the steel plate could also be
observed for specimens B12 and B17 after the tests were
ended. The adhesive layers in this location showed very
clear signs of deformation through the thickness, with
an inclination of about 45 degrees.
6.3. Tensile Rupture of the Laminates
Two specimens (B12 and B17) failed as a result of the
tensile rupture of the laminate in the middle of the
specimen. The tensile strength of the laminates in these
two specimens could thus be fully utilised. In both
cases, the laminate started to carry an increased
percentage of the applied load due to the yielding of the
steel plate until the ultimate strength of the laminate was
reached, see Figure 13. Even though the final failure of
the laminate itself was brittle, the overall behaviour of
specimen B12 was very ductile, as, due to its low
stiffness, the laminate used in this specimen allowed for
the large plastic deformation of the specimen before
final failure. Specimen B17, however, displayed very
limited ductility before failure. This specimen contained
laminates with an ultra-high E modulus. The ultimate
load (and the degree of strengthening) was, however,
almost the same for both specimens.
Adhesively bonded composite materials are now
widely used for strengthening and repairing existing
structures. In later applications, the material, often in
the form of unidirectional laminates, has been
employed to strengthen metallic structures such as steel
and composite bridges. In this paper, the results of an
experimental and numerical study of the behaviour and
strength of steel elements strengthened with bonded
829
ASE 9-6-08_Kliger
4/12/06
5:09 pm
Page 830
300
250
Applied load (kN)
200
150
100
B-17
50
0
0.0
200.0
400.0
600.0
800.0
1000.0
B-12
B-40
Yielding of the steel plate

Rapture of the laminate
1200.0
1400.0
1600.0
A-12
1800.0
2000.0
Axial stress in the middle of the laminate (MPa)
Figure 13. The variation in maximum axial stress in the middle of the laminate in the four composite specimens
carbon-fibre laminates are presented. A new type of test

specimen has been developed for this purpose, based on
extensive FE analysis. Five specimens, one with just
steel and four with various types of adhesively bonded
carbon-fibre laminate, were tested. The behaviour
displayed by the tested specimens varied widely. The
degree of strengthening (i.e. the increase in the ultimate
load-carrying capacity) that could be achieved by
bonding the laminates varied between 12% and 80%.
The degree of ductility which was observed for the
tested specimens also varied, ranging from brittle to
very ductile failure. The tests succeeded in reproducing
the failure modes that can be obtained in steel elements
strengthened with bonded composite laminates. The
tensile rupture of the laminate, as well as debonding and
delamination failures in the middle and at the ends of the
laminates, could be observed. The stiffness of the
composite laminates and the adhesive used for bonding
both play a dominant role in the type of fracture mode
and thereby the ultimate load-carrying capacity that can
be achieved for steel elements strengthened with
adhesively bonded laminates.
REFERENCES
Al-Emrani, M. and Linghoff, D. (2005). Advanced Composite
Materials for the Strengthening and Repair of Metallic Structures,
Chalmers University of Technology, Civil and Environmental
Engineering, Steel and Timber Structures, Report 2005:7.
Al-Saidy, A.H. Klaiber, F.W. and Wipf, T.J. (2004). Repair of steel
composite beams with carbon fiber-reinforced polymer plates,
830
Journal for Composites for Construction, ASCE, Vol. 8, No. 2,

pp. 163173.
Bassetti, A. (2001). Lamelles prcontraintes en fibres de carbone
pour le renforcement de ponts rivets endommags par fatigue,
Thse N 2440, EPF Lausanne, Switzerland.
Colombi, P. and Poggi, C. (2006), Strengthening of tensile steel
members and bolted joints using adhesively bonded CFRP
plates, Construction and Building Materials, Vol. 20, Nos 12,
pp. 2233.
Jiao, H. and Zhao, X. L. (2004). CFRP strengthened butt-welded
very high strength (VHS) circular steel tubes, Thin-Walled
Luke, S. (2001). The use of carbon fibre plates for the strengthening
of two metallic bridges of an historic nature in the UK,
Proceedings, International Conference on FRP Composites in
Civil Engineering, Hong Kong.
Mertz, D.R. and Gillespie, J.W. (1996). Rehabilitation of steel
bridge girders through the application of advanced composite
materials, IDEA Program, Transportation Research Board
National Research Council (NCHRP), University of Delaware.
Miller, T.C., Chajes, M.J., Mertz, D.R. and Hastings, J.N. (2001).
Strengthening of a steel bridge girder using CFRP plates,
Journal of Bridge Engineering, ASCE, Vol. 6, No. 6, pp. 514522.
Sen, R., Liby, L. and Mullins, G. (2001). Strengthening steel bridge
sections using CFRP laminates, Composite Part B-Engineering,
Vol. 32, No. 4, pp. 309322.
Shen, H.S., Teng, J.G. and Yang, J. (2001). Interfacial stresses in
beams and slabs bonded with thin plate, Journal of Engineering
Mechanics, ASCE, Vol. 27, No. 4, pp. 399406.
Tavakkolizadeh, M. and Saadatmanesh, H. (2003). Strengthening
ASE 9-6-08_Kliger
4/12/06
5:09 pm
Page 831
of steel-concrete composite girders using carbon fiber reinforced

polymer sheets, Journal of Structural Engineering, ASCE, Vol.
129, No. 1, pp. 3040.
Teng, J.G. and Zhang, L. (2005). Finite element predictions of
interfacial stresses in beams bonded with a thin plate,
Proceedings, International Symposium on Bond Behaviour of
FRP in Structures (BBFS 2005), December 2005, Hong Kong,
China.
Teng, J.G., Zhang, J.W. and Smith, S.T. (2002). Interfacial stresses
in reinforced concrete beams bonded with a sofit plate: a finite
element study, Construction and Building Materials, Vol. 16,
No. 1, pp. 114.
APPENDIX: NOTATION
El E modulus of the laminate
Es E modulus of the steel
Ea E modulus of the adhesive
Ga Shear modulus of the adhesive
L
Laminate length
P
Applied load
ta
Thickness of the adhesive
tl
Thickness of the laminate
ts
Thickness of the steel plate
Strain
Constant
Shear stress
831
ASE 9-6-08_Kliger
4/12/06
5:09 pm
Page 832
ASE 9-6-09_best paper
4/12/06
5:08 pm
Page 833
Best Paper Award

The Best Paper Award of this journal was set up in 2003. The aim of this annual Award is to give special recognition
to authors of the most meritorious paper published by this journal in a given year as judged by its International
Editorial Board and International Advisory Board. The Award includes a certificate for each author of the winning
paper plus US$200 to be shared by the authors. Each author of the winning paper will also receive a free
subscription of the journal for two years. The selection process involves careful deliberations by the International
Editorial Board and the International Advisory Board, coordinated by the Editor-in-Chief and the three Associate
Editors.
It gives me great pleasure to announce that the following paper has been selected to receive the 2005 Best Paper
Award for papers published in the journal in 2005:
Behavior of the Tree Branches, Trunk, and Root Anchorage by Nonlinear Finite Element Analysis,
Vol. 8, No. 1, pp. 1-14, by Y.B. Yang, Y.T. Yang, H.H. Su
Please join me in congratulating the authors of the winning paper and in thanking the members of both the
International Editorial Board and the International Advisory Board for all their hard work.
Jin-Guang Teng
Editor-in-Chief
On behalf of the Local Editorial Team
833
ASE 9-6-09_best paper
4/12/06
5:08 pm
Page 834

Bond Behavior of FRP in Structures ASE 9-6

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Bond Behavior of FRP in Structures ASE 9-6

Hochgeladen von

Copyright:

Verfügbare Formate

ASE 9-6 front page

Volume 9 Number 6 December 2006

Construction site of a railway project in Hong Kong where GFRP

ASE 9-6 front page

ASE 9-6 contents

Advances in Adhesive Joining of Carbon Fibre/Polymer Composites to

Experimental and Numerical Investigation of the Behaviour and Strength

ASE 9-6 contents

promoting research collaboration on FRP-toconcrete bond behaviour

providing a forum for exchange, discussion and

establishing design models and

developing a state-of-the-art report on the

Advances in Structural Engineering Vol. 9 No. 6 2006

process by two expert reviewers selected from the

The Importance of Bonding A Historic Overview

Key words: adhesives, structural bonding, joints, construction industry, organic.

resins from tree sap. Archaeologists have also uncovered

Advances in Structural Engineering Vol. 9 No. 6 2006

The Importance of Bonding A Historic Overview and Future Possibilities

billiard balls, which had been made of ivory. The era of

Increased speed of production, in particular for

Wider selection of materials. It is possible to mix

Local stress concentrations can be avoided and

Compared to welded structures bonded structures

Advances in Structural Engineering Vol. 9 No. 6 2006

Table 1. Organic adhesive classification, after (Mays and Hutchinson 1992)

corn, rice, wheat, and potatoes are probably better

Advances in Structural Engineering Vol. 9 No. 6 2006

system. When a drop of cyanoacrylate adhesive is put

The Importance of Bonding- A Historic Overview and Future Possibilities

the carburettor of an automobile engine. Today,

although a degree of flexibility is required in the adhesive

Advances in Structural Engineering Vol. 9 No. 6 2006

Table 2. Typical material properties for epoxy adhesives

Cold-curing epoxy adhesive

Hot-curing epoxy adhesive

Figure 1. The importance to wet the surface to be bonded

bonding as a method of fastening is that the surfaces

Advances in Structural Engineering Vol. 9 No. 6 2006

Figure 2. Different forces acting on a bonded joint

versus shear, cleavage and compression, as shown

The Importance of Bonding- A Historic Overview and Future Possibilities

generally much lower than between metals and

Advances in Structural Engineering Vol. 9 No. 6 2006

Carbon-fiber composite (CFRP)

Shear strength (MPa)

Hang-ons for cars

Plastics for cars

Large body shells

Advances in Structural Engineering Vol. 9 No. 6 2006

no need to enhance the stiffness of the crash resistance

The Importance of Bonding- A Historic Overview and Future Possibilities

different bonded parts and materials for the Mercedes

have a high abrasion resistance and good adhesion at

Advances in Structural Engineering Vol. 9 No. 6 2006

(Kecsmar 2003). As a consequence, either the structural

Advances in Structural Engineering Vol. 9 No. 6 2006

structures. During the 1950s and 1960s an enormous

The Importance of Bonding- A Historic Overview and Future Possibilities

compression inflicting crushing of wood. In terms of

Advances in Structural Engineering Vol. 9 No. 6 2006

bridge replaced an existing bridge, the West Mill bridge,

Advances in Structural Engineering Vol. 9 No. 6 2006

complicated and increases the challenge to increase the