The Tenth East Asia-Pacific Conference on Structural Engineering and Construction

August 3-5, 2006, Bangkok, Thailand

A FRACTURE-MECHANICS MODEL FOR DEBONDING OF EXTERNAL FIBRE

Paththini M.M. ACHINTHA1 and Chris J. BURGOYNE2

ABSTRACT: A Fracture-mechanics model for debonding of external fibre reinforced polymer plates
on reinforced concrete beams is presented. The conventional methods of concrete-FRP interface
analysis use finite element models, which require details of unknown and unknowable interface
characteristics. The present model assumes flaws in the vicinity of the interface and assesses whether
sufficient energy can be released to cause these flaws to propagate. Energy released by an extension of
an existing flaw depends on the change of recoverable energy stored in the system. This paper
concentrates on the moment-curvature model for a cracked reinforced-concrete beam under a prestress
caused by the force in a FRP plate. The use of the proposed model to determine the energy released
from the system with the extension of an existing flaw is also presented. The energy required to create
the associated new surfaces depends on interface fracture energy which is first reviewed and methods
to determine is also discussed.

KEYWORDS: concrete beams, flexural-strengthening, fracture-energy, interface-debonding,

1. INTRODUCTION

Reinforced Concrete (RC) beams flexurally strengthened with externally bonded Fibre Reinforced
Polymer (FRP) plates often fail by plate debonding, but a rational analytical approach to these failures
has yet to be developed. Due to the typical premature and brittle nature of these debonding failures,
inadequately designed strengthening applications may become ineffective and reduce the level of
safety. Therefore, proper understanding of the concrete-FRP interface debonding is required for safe
and reliable application of externally bonded FRP systems for strengthening of RC elements.

Two modes of debonding failures of external FRP plates on RC beams have been observed by the
researchers (Figure 1) [1]. The first initiates somewhere in the middle of the beam at an intermediate
crack and propagates outwards. Existing models to analyse this mode of debonding are generally
based on the results from shear lap tests of FRP plates on small-scale concrete specimens [2], but it is
uncertain whether these shear lap tests actually simulate the debonding mechanism of real beams. The
other mode of debonding initiates at the plate ends and propagates inwards. Interfacial stress-based
models for plate end debonding can be found in the literature but these models are based on linear-
elastic theories. Significant cracking is expected at the onset of debonding and hence the applicability
of these models is questionable. Finite Element (FE) models have been used for concrete-FRP
interface analysis [1] but they require details of interface characteristics unknown and unknowable,

1
PhD Student, Department of Engineering, University of Cambridge, UK
2
Reader in Concrete Structures, Department of Engineering, University of Cambridge, UK
such as the shape of the adhesive layer, especially at the layer ends; interface texture; shape, size and
location of voids in the adhesive layer.

In addition to these strength-based models for debonding of external FRP plates on concrete, some
empirical models are also available in the literature. ACI Committee 440 [3] suggests a limit on the
FRP plate strain to prevent debonding, based on a database of test results, but the theoretical
justification is unclear: these strain limits depend only on the stiffness properties of the FRP plates and
do not take account of the concrete-FRP interface properties, width of the FRP plate or the plate
curtailment location, all of which should be expected to affect debonding.

Fracture-mechanics models assume that, since flaws are inevitable in the interface, what matters is
whether these flaws can propagate. The energy needed to form new surfaces depends on the interface
fracture energy and must be compared with the energy released by the system when the flaw extends,
which in turn depends on the change of strain energy stored in the system. This paper is primarily
concerned with the determination of that strain energy.

It is necessary to determine the strain energy stored in a reinforced concrete (RC) beam, to which is
attached a FRP plate that may be debonded over part of its length; that is not a trivial problem. It is
well-known that when a RC beam cracks its stiffness does not immediately change to that of a section
where the tension concrete can be fully disregarded. Various empirical models, such as Branson’s Ieff
concept [4], have been used to model this behaviour, primarily with a view to being able to predict the
deflections of RC beams to check their compliance with code limits. Such models normally work in
moment-curvature space (M-κ) rather than stress-strain space (σ-ε) and thus do not attempt to model
individual cracks. Two versions are normally presented; one designed to model the overall beam
stiffness and the other to determine the local curvature.

There are two complications that have been addressed in the present work. If the FRP plate is bonded
to the concrete it behaves as reinforcement, but the plate will not be expected to affect cracking in the
same way as internal reinforcement. If unbonded, the FRP plate will act as a prestressing element,
inducing both force and moment in the beam (Figure 2). Most existing models do not cope with either
of these effects, but if the model is to determine the strain energy in the system it will have to handle
both. The model is explained below and its accuracy verified by comparisons with experimental data
available in the literature.
Centroidal axis
Applied loads
M eff M app

F p
External
Midspan debonding Effective actions moment
RC beam
propagation on RC section
F p

Plate-end debonding
propagation
Intermediate crack FRP Plate FRP plate

Figure 1. Debonding modes of external FRP plates Figure 2. Actions on RC section

2. OVERVIEW OF THE PROPOSED FRACTURE-MECHANICS MODEL

In the proposed model it is assumed that an existing flaw extends by small amount; the energy needed
to form the new surfaces (EF) is compared with the energy released from the system (ER), which can
be determined from energy conservation before and after the flaw extension. The energy needed to
form new surfaces can be calculated when the interface fracture energy and the extension of the
fracture surface are known. When a beam bends, energy is put into the structure by the loads, some of
which is dissipated, whilst some is stored as strain energy, shown schematically in Figure 3. When the
flaw extends the energy distribution changes and some of the strain energy can be released. Thus, the
elastic strain energy stored in the system before and after the flaw extension is required to determine
the corresponding energy released from the system. If E R ≥ E F the flaw will extend; if not there is
insufficient energy for the flaw to propagate.

3. ELASTIC STRAIN ENERGY IN A FRP STRENGTHENED RC BEAM SECTION

The Elastic Strain Energy (ESE) stored in the system before and after the after the flaw extension can
be determined if the moment and curvature in the RC beam are known, as well as the strain and force
in the FRP plate. However, a rational analytical approach to determine ESE in a FRP plate
strengthened RC beam section has yet to be developed. Branson’s effective second moment of area
(Ieff ) expression (Equation (1)) indirectly accounts for tension stiffening effects of cracked concrete
and successfully predicts deflections and moment-curvature relationships of RC beams [5]. The
stiffness is interpolated between the corresponding full- (uncracked Iun) and zero- (fully-cracked Ifc)
tension stiffening states. The interpolation coefficient represents the extent of cracking of the section.
Mcr and Mapp in Equation (1) are the moments causing first cracking and the externally applied
moment on the section. Ieff is taken as Iun when Mapp ≤ Mcr. Ieff is the effective second moment of
area of the equivalent transformed concrete section of modulus Ec so the effective curvature of the
section ( κ eff ) can be determined from Equation (2).

{
I eff = (M cr M app ) I un + 1 − (M cr M app )
4 4
}I fc ⇒ if M app > M cr (1)

κ eff (
= M app E c I eff ) (2)

As described, this model determines the local curvature of the section. A modified form is used when
an average Ieff of the beam is needed to determine deflection. In this case the exponent in Equation (1)
is reduced from 4 to 3. This method has been widely verified for conventional RC beams and will now
be extended to deal with the more complex problem of beams with external FRP plates.

4. PROPOSED MOMENT-CURVATURE MODEL FOR FRP STRENGTHENED RC BEAMS

Branson’s Ieff expression addresses the behaviours of RC beams under nominal service moments
where tension steel remains elastic and concrete shows virtually a linear-elastic behaviour in
compression. At higher compressive strains concrete is nonlinear. FRP-strengthened RC beams
sections are subjected to loads higher than those for which they were originally designed, but it is still
possible to define equivalent elastic stiffness (EIeq) of such a section (Equation (3)). κ and M in
Equation (3) are the curvature and the effective moment on the RC section about the section’s
centroidal axis. EIeq represents the equivalent elastic stiffness of the section. The proposed model
(Equations (4) and (5)) follows Branson’s Ieff expression but is based on this EIeq of the sections.

EI eq = M κ (3)

In normal RC beams the axial force is zero so the centroidal location does not matter, but because of
the prestressing effect of the FRP plate the location of the centroid is important. Iun and Ifc in
Equation (1) should be calculated relative to the centroidal axes of the uncracked and fully-cracked
sections respectively and Ieff is related to the centroidal axis of the imaginary effective section [6].
Sakai and Kakuta [5] presented an empirical expression for the effective centroidal axis depth (αeff) of
a partially-cracked RC beam section subjected to an axial force as a transition between uncracked and
fully-cracked sections. The applied moments should also be calculated about the relevant centroidal
axes to define the corresponding equivalent stiffness from Equation (3).
When the amount of cracking of a RC section increases, the tension-stiffening eventually becomes
ineffective. In Branson’s original model the stiffness becomes asymptotic to the fully-cracked state.
The present model (Equations (4) and (5)) assumes that the beam is fully-cracked at the moment
causing first yielding of tension steel and uses a slightly modified form of the interpolation used in
Equation (1).
EI eq −eff = K EI eq −un + (1 − K ) EI eq − fc (4)
m
[ m
]
K = (M cr − c M app − c ) 1 − {(M app − c − M cr − c ) (M y − c − M cr − c )} ⇒ if M cr − c < M app − c < M y − c (5)

The stiffnesses in Equation (4) are calculated at the relevant centroid, but it is found desirable to use a
fixed reference axis to calculate K, to avoid complications caused by moving the prestressing force.
The moments in Equation (5) are calculated relative to the beam’s centreline. The exponent m in
Equation (5) is taken as 4, but for the equivalent expression used to calculate the movement of the
centroid (A in Equation (6)) a value of 3.5 is found to give a better fit to the data. When the effective
centroidal axis depth (αeff) is known the effective moment on the section can be calculated and then
the effective curvature of the section from Equation (3).
α eff = Aα un + (1 − A ) α fc (6)
The complete derivation of this proposed moment-curvature model and comparison between the
model predictions against experimental data available in the literature will be presented elsewhere.
From the comparisons shown there (an example is given in Figure 4) it is possible to conclude that the
proposed moment-curvature model is accurate and reliable and thus can be used to calculate strain
energies in FRP strengthened RC beams.

Moment 70
Energy lost in steel yielding
60

B A = First cracking
Energy lost 50
Applied moment (kNm)

in cracking B = First yielding of tension steel

30

A 20
Recoverable strain energy
10
Proposed model
Curvature 0
Experimental (Beam A3.1 Reference [7])
0 0.5 1 1.5 2 2.5 3 3.5
Curvature -5
x 10

Figure 3. Energy in flexure Figure 4. Moment-curvature comparison

5. ENERGY RELEASED FROM THE SYSTEM WITH THE EXTENSION OF FLAW

The Strain Energy (SE) in a FRP strengthened RC beam section consists of three components which
are summed in the proposed model. 1) SE in the RC beam due to flexure; 2) SE in the RC beam due to
axial force and 3) SE in the FRP plate. Energy released from all of these components with the
extension of flaw should be considered; the beam is divided into short segments and the energy
releases in each segment are calculated. Since the strain distributions of the segments outside the
fractured zone cannot be altered by the extension of the flaw, unless the load changes, it is assumed
that only the segments within the fractured zone contribute for the energy released from the system.

It is now possible to consider the energy state of the beam before and after the flaw extension. The
elastic strain energy is given by the assumption that the beam would unload elastically from any
loaded state, using the effective stiffnesses calculated as in Section 4. There would be some permanent
deformation. By accounting for the energy that is dissipated in the concrete, either in flexural-tension
cracking, or nonlinear elasticity, and by yielding of the tensions reinforcement, it is possible to
determine the elastic strain energy that can be used to extend the fracture. Space does not permit a full
description of that process.

Account should also be taken of the strain energy stored in the FRP plate, which can be assumed to
remain linearly elastic, and also of the effect of the axial force induced in the RC beam which will
have associated strain energy. The full model also accounts for beams which have been unloaded and
reloaded.

6. CONCRETE-FRP INTERFACE FRACTURE ENERGY

The energy released by the system when the flaw extends is used to create new surfaces and the
propagation path of the existing interface flaw takes the path that requires least energy. Experimental
evidence confirms that debonding fractures of external FRP plates on RC beams generally propagate
through the concrete just above the interface [8]. However, it is also reported that poor surface
preparation prior to the FRP plate bonding or low-strength adhesives can lead failure in the interface
or within the adhesive itself [9]. However, with the availability of high-strength adhesives and well-
established surface preparation techniques, the proposed model assumes debonding fractures
propagate through the concrete. Thus, the interface fracture energy (GF) should be that of concrete.

The opening mode (Mode I) is the most common fracturing behaviour in concrete. It is reported that
“It would seem impossible to propagate cracks in concrete under pure in-plane shear mode (Mode II)
conditions, because of frictional forces. Under mixed mode loading, it appears that concrete fractures
locally take place in pure Mode I, although the directions of propagation depend on the mixity
conditions” [10] and in any event overall compatibility conditions should preclude other modes of
crack opening. However, with complex stress conditions expected at the interface flaw tips there are
consensus among researchers that the fracture of concrete in FRP strengthened RC beams propagates
under mixed mode loading (combination of Modes I and II loadings) [8].

Elices et al. [11] have performed extensive FE modelling to predict the fracture paths and load versus
crack-mouth-opening-displacement variations in notched concrete specimens under mixed mode
loadings. Their numerical simulations were based on Mode I fracture properties of concrete and they
reported excellent correlation with the experimental results, which agrees with RILEM Report 5 [12]
that says that concrete fractures propagate under pure Mode I conditions even under mixed-mode
loading. The present model therefore assumes that debonding fractures of external FRP plates on RC
I
beams propagate as pure Mode I fractures of concrete and uses the Mode I fracture energy ( G FC ).

Concrete fracture cannot be explained using Linear Elastic Fracture (LEF) theories due to the
existence of a significant nonlinear fracture process zone ahead of a crack tip, which is neglected in
I
the derivation of LEF theories. Therefore, Mode I fracture energy of concrete ( G FC ) should be
determined using available concrete fracture theories. RILEM Report 5 [12] recommends the use of a
I
two-parameter, effective-crack-size model and Bažant’s size effect law to determine G FC using
experimental results from central-notched three-point bending tests. Hillerborg’s fictitious crack
I
model can also be used to determine G FC but this requires a concrete softening curve which is
generally obtained by means of stable tensile tests on specimens. These tests are less accurate and less
reliable than notch tests because the location of the crack is not known a priori and on most occasions
multiple cracking occurs due to material heterogeneity [11]. The CEB-FIP Code [13] presents an
I
empirical expression for G FC , based on available experimental data, as a function of concrete
compressive strength and the maximum aggregate size.
 I
If this Mode I fracture energy of concrete ( G FC ) is known it can be compared with the energy release
rate (GR) of the system. If G R ≥ G FC
I
the flaw will extend and failure is to be expected.

7. CONCLUSIONS

The outline of a fracture-mechanics model for debonding of external fibre reinforced polymer plates
on reinforced concrete beams has been presented. The paper has shown that a modified version of
Branson’s effective stiffness model can be used to determine the elastic strain energy stored in the
beam. The energy released from the system with the extension of an existing interface-flaw can be
calculated and compared with the energy required to create new fracture surfaces. This will decide
whether the flaw will propagate. The energy required to create the associated new surfaces depends on
the Mode I concrete fracture energy, which can be determined from standard concrete fracture tests.

8. REFERENCES

1. Leung, C.K.Y., “Delamination Failure in Concrete Beams Retrofitted with a Bonded Plate”.
Journal of Materials in Civil Engineering, Vol. 13, No.2, 2001, pp.106-113.
2. Pham, H.B. and Al-Mahahaidi, R., “ Prediction Models for Debonding Failure Loads of Carbon
Fibre Reinforced Polymer Retrofitted Reinforced Concrete Beams”. Journal of Composites for
Construction, Vol. 10, No.1, 2006, pp.48-59.
3. ACI Committee 440. Guide for Design and Construction of Externally Bonded FRP Systems for
Strengthening Concrete Structures. American Concrete Institute, Farmington Hills, Mich, 2002.
4. Branson D.E., “Design Procedures for Computing Deflections”. ACI Journal Proceedings, 1968,
Vol. 65, No. 9, 1968, pp. 730 - 735.
5. Sakai K. and Kakuta, Y., “Moment-Curvature Relationships of Reinforced Concrete Members
Subjected to Combined Bending and Axial Force”. ACI Journal Proceedings, Vol. 77, No.3,
1980, pp.189-194.
6. Teng, S. and Branson, D.E., “Initial and Time-Dependent Deformation of Progressively Cracking
Nonprestressed and Partially Prestressed Concrete Beams”. ACI structural Journal, 1993, Vol. 90,
No.5, pp.480 - 488.
7. Spadea G., and Bencardino F. and Swamy R.N., “Structural Behaviour of Composite RC beams
with Externally Bonded CFRP”. Journal of Composites for Construction,1 998, Vol.2,No.3,
pp.132-137.
8. Gunes, O., “Debonding Problems in Beams Strengthened Using Externally Bonded FRP
Reinforcemnt”. Proceedings of COBRAE 2005, Conference on Bridge Engineering with Polymer
Composites, Zurich, Switzerland, March 30-April 1, 2005,
9. Buyukozturk, O., Gunes, O. and Karaca, E., “Progress on Understanding Debonding Problems in
Reinforced Concrete and Steel Members Strengthened using FRP Composites”. Construction and
Building Materials, Vol. 18, 2004, pp. 9-19.
10. Karihaloo, B., Fracture Mechanics and Structural Concrete. Addison Wesley Longman Limited,
Harlow, Essex CM20 2JE, UK, 1995.
11. Elices, M., Guinea, G.V., Gơmez, J. and Planas, J., “ The cohesive zone model: advantages,
limitations and challenges”. Engineering Fracture Mechanics, Vol. 69, 2002, pp.137-163.
12. Shah, S.P. and Carpinteri, A., (edited). RILEM Report 5, Fracture Mechanics Test Methods for
Concrete, RILEM, International Union of Testing and Research Laboratories for Materials and
Structures, London, UK, 1991.
13. Comite Euro-International du Beton, Federation Internationale de la Precontrainte, CEB-FIP
model code, Lausanne, Switzerland, 1991.