Composite Structures 93 (2010) 216–224

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Effects of exposure to saline humidity on bond between GFRP and concrete

Manuel A.G. Silva *, Hugo C. Biscaia
UNIC, Faculdade de Ciências e Tecnologia, Universidade Nova de Lisboa, 2829-516 Caparica, Portugal

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

Article history: Based on extensive experimental program on effects of artificial environmental aging, its effects on
Available online 31 May 2010 strength and bond between outer GFRP reinforcement of RC beams and concrete are described and inter-
preted. Artificial aging of beams consisted of saline water immersion, salt fogging and cyclic tidal-like
Keywords: action causing degradation on mechanical properties that are reported and examined. Computational
Bond-slip modeling of these effects is also preliminarily described, considering post-aged constitutive properties
Artificial aging of the component materials and relevant non-linear material properties, to study bond-slip and beam
GFRP strengthening
response. Tests to develop a shear Mohr–Coulomb envelope as a rupture criterion on the layers adjacent
to the interface are also presented.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction
There are relatively few results available on environmental deg-
radation of the capacity of structural members strengthened with
FRP, namely due to changes on bond. The failure mechanisms of
beams with outer FRP reinforcement has however been extensively
studied [1]. Flexural and shear deformations endured by reinforced
concrete (RC) members determine the plate debonding mecha-
nisms. Shear failure and critical diagonal cracking are not consid-
ered in this study and attention concentrates on intermediate
crack mechanisms. When a small crack reaches a plate, the result-
ing stress concentration leads to cracking (IC) in the concrete adja-
cent to the interface. Shear can still be transferred across this crack
through aggregate interlock associated with localized forces or
stresses normal to the interface. Failure at the interface may also
relate to lack of ‘aggregate-interlock’ that can be measured from
pull-tests. Material shear slip can be measured in these tests. Oeh-
lers showed that, when debonding started at the plate end, large
stresses normal to the plate/concrete interface develop in the
vicinity and are resisted by strict tensile strength of the concrete,
instead of ‘aggregate-interlock’ shear resistance, although interface
shear stresses also develop [1].
A meso-scale finite element model for the simulation of interfa-
cial debonding failures in a pull test is described in [2]. Bond-slip
laws used to approach the FRP/concrete interface behavior, very of-
ten assume bi-linear or exponential laws. Typical bond-slip curves
consist of an ascending branch with continuous stiffness degrada-
tion till a peak bond stress is reached, followed by a descending
branch linear or non-linear until a zero bond stress is attained
* Corresponding author. Tel.: +351 214424589.
E-mail address: mgs@fct.unl.pt (M.A.G. Silva).
for a finite value of slip, as illustrated in Fig. 1. Bi-linear laws have
the advantage of providing an analytical solution of the problem
whereas exponential laws, with non-linear post-peak behavior,
approximate with good accuracy the concrete non-linear post-
peak behavior. A review of bond strength and bond-slip models
is offered in [3]. Non-linear finite element analyses (FEA) of RC
beams strengthened in flexure and in shear, simulating various
failure modes and including FRP debonding either at the plate
end or at intermediate cracks are found in [4]. A finite element
(FE) study shows that stresses vary strongly across the adhesive
layer [5]. The stresses calculated along the adhesive-to-concrete
interface were very different from those along the plate to adhesive
interface: near the end of the plate, the interfacial normal stress
was confirmed to be tensile along the adhesive–concrete interface,
but compressive along the plate–adhesive interface.
Preliminary work was also made in the computational modeling
of the GFRP/concrete interface behavior in reference specimens.
The computational analysis was based on 2D and 3D modeling
and both models led to results with no significant differences. Both
models predicted the maximum load and the maximum tensile
strength in the GFRP with fairly good accuracy. Maximum bond
stress and maximum slip showed larger differences between both
models. In any case, the 3D model is closer to the experimental re-
sults with relative errors under 18% whereas 37% were found with
the 2D model.
Knowledge of environmental physical degradation of strength
characteristics is required and has been the object of few publica-
tions. Some relevant results are briefly mentioned next. GFRP lam-
inates of epoxy matrix are vulnerable to moisture diffusion [6]
because most epoxies absorb between 1% and 7% moisture by
weight. Sorption or mass uptake may be due to (i) absorption, asso-
ciated with capillarity, or (ii) adsorption, a surface phenomenon

0263-8223/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruct.2010.05.018
 M.A.G. Silva, H.C. Biscaia / Composite Structures 93 (2010) 216–224 217

Nomenclature

r compressive stress of the interface fc concrete strength

rult ultimate strength of the GFRP laminate fcm average concrete strength
rx longitudinal stress fctm average concrete tensile
s shear stress FGFRP tensile force on the GFRP
s(s) bond-slip law GIf fracture energy of mode I
sb bond stress GIIf fracture energy of mode II
sm maximum bond stress nP Popovic’s constant
De strain averaged between consecutive gauges P external load
DL length between consecutive gauges P0 maximum load
Dmax maximum displacement measured at maximum load s slip between concrete and the GFRP
eult ultimate strain of the GFRP laminate s0 maximum slip
er ultimate strain of the resin sm slip at maximum bond stress
/ internal friction angle at the interface tf GFRP thickness
c interface cohesion Tg vitreous transition temperature
E Young Modulus u displacement perpendicular to the interface bonded
Ecm average Young Modulus of concrete surface
Ef Young Modulus of the GFRP v displacement parallel to the interface bonded surface
Er Young Modulus of the resin

that generates heat and causes swelling. It induces plasticization of
the laminates, lowering of the glass transition temperature Tg, and
hydrothermal aging associated with chemical phenomena [7].
Moisture sorption affects the mechanical strength of laminates
with epoxy matrix [8] and contributes for loss of capacity of RC
beams externally strengthened with GFRP. Associated phenomena
and increased migration of water into flaws created inside the lam-
inates, especially in the interface with fibers, explain the behavior
of glass fiber-reinforced epoxy [8,9].
It has also been shown that ‘‘leaching” associated with the dif-
fusion of the alkali ions out of the glass structure into porous water
may occur and erode the GFRP reinforcement, a type of degrada-
tion more serious for reinforcement rods than external strips [10].
Lowering the vitreous transition temperature of the adhesive,
Tg, the bond strength deteriorates, independently of the concrete
substrate. Saline tests may, thus, impose a synergetic combined ef-
fect on the mechanical properties, depending on the temperature
of the water moisture.
Flexural strength of glass composites declines when immersed
in seawater for long periods and water absorption has been cited
as affecting the Mode I stability of delamination cracks in the
glass/polyester composite [11], a result to be confirmed for epoxy
matrices.
Tests simulating tidal effects on eight beams strengthened with
FRP submitted to dry/wet cycles, using saline water (15% concen-
tration of NaCl), with half of the specimens pre-cracked showed
much earlier degradation of the latter as also found by Spainhour
and Thompson [12]. Beaudoin et al. observed negligible effects of
tidal cycles on CFRP [13].
Fava et al. imposed artificial aging on beams with external FRP
plates and concluded that freeze–thaw cycles reduced bond shear
strength, while the beneficial effects of high humidity level charac-
terizing salt spray fog overcame eventual damage due to the chlo-
ride solution [14]. Higher deformability of the interface, i.e. high
peak slip, appeared to be the major effect found. Other authors
[15,16] had already mentioned similar effects.
Tests on the response of beams strengthened with external
GFRP laminates under monotonic static load, including some envi-
ronmental degradation, were also described by the authors [17]
with freeze thaw cycles revealing to be more aggressive to the
GFRP/concrete interface than salt fog cycles and immersion in 5%
of salt water.
Further experimental work and computational modeling are
needed and a contribution is described in the sequel. The objec-
tives of this study are twofold: (i) reporting experimental results
on effects of saline moisture on GFRP external reinforcement of
beams, mainly on bond between the composite and concrete and
(ii) presenting numerical models that lead to interpretation and
generalization of the experiments.

2. Experimental program

The effects on the mechanical properties of the ‘‘materials”, i.e.

GFRP laminates and concrete, and on the load capacity of structural
beams were considered separately. Specific aspects on bond-slip
data are treated separately.
The modification of the tensile strength of GFRP laminates was
quantified by standard tensile tests in a Zwick universal machine,
described below.

2.1. Tensile tests of GFRP laminates

The coupons for tensile tests were made of unidirectional GFRP

Fig. 1. Bond-slip laws commonly used (based in [3]). laminates 1.3 mm thick, 250 mm long and 25 mm wide (SEH51/
218 M.A.G. Silva, H.C. Biscaia / Composite Structures 93 (2010) 216–224
Tyfo fabric) cut from 2-ply GFRP laminates and tested in accor-
dance with ASTM D3039/D 3039M. Average results of the mechan-
ical tests made at the onset of artificial aging were rult = 496 MPa,
eult = 2.42% and E = 20.5GPa, values designated hereafter as refer-
ence or control values.
Moisture cycles with distilled water, subjecting standard flat
coupons to 12 h at 20% relative humidity and 12 h at 90%, keeping
temperature at 40 °C, led to gradual decay of the average tensile
ultimate stress, a decrease that reached 11% at 10,000 h.
Tensile tests, along the direction of alignment of the glass fibers,
were also made after salt fog cycles (8 h salt fog at 35 °C and 98%
humidity, followed by 16 h drying at room temperature) at differ-
ent times up to 10,000 h. The tensile strength showed some initial
fluctuation, and decreased more than 12.0% at 10,000 h.
DMTA tests showed, for salt fog cycles, a decrease of 4 °C on the
initial average glass transition temperature of Tg = 66 °C, a fact
coherent with the recorded ultimate strength decrease.
Effects of immersion in distilled water at 22 °C showed no sig-
nificant degradation of the GFRP laminates [8]. Strength and strain
reductions of up to 35% are mentioned by the supplier for water
immersion at 38 °C, effect probably linked also to the higher tem-
perature of the water.
2.2. Bending tests and bond degradation
The flexural failures were imposed on specimens composed of
two concrete blocks joined by a metallic hinge, in the upper side,
and connected by two layers of GFRP on the under side, Fig. 2.
The composite plate was bonded by the wet lay up technique for
a length of 240 mm on each block, length known to be longer than
the effective length.
Steel reinforcement consisted of four /3 mm steel (S500) rods.
Stirrups /3 mm are spaced 5 cm apart. All specimens were pre-
treated with sand blasting to improve the surface roughness.
The beams were instrumented with strain gauges bonded to the
GFRP strip, spaced approximately 40 mm, to measure the strain
distribution along the GFRP laminate for different loading levels.
Two pressure cells MT-KCM/300 controlled the applied loading.
The initial tests performed at the time that artificial aging
started are designated as reference tests.
The laminates were manufactured with unidirectional glass
fiber Tyfo SH 51. Results showed a linear constitutive law, and an
average tensile failure stress of 500.2 MPa and an ultimate strain
2.26% i.e. an average E = 20.39 GPa. The same Tyfo S Epoxy that
impregnated the glass fibers was used as adhesive. The average
mechanical properties of the resin were rr = 72.4 MPa, Er = 3.18
GPa and er = 5.0%.
Fig. 2. Post-failure surface of laminate with chunks of concrete.
Concrete, in the first batch of specimens, led to an average cubic
compression strength at 28 days of 43.6 MPa, fcm = 34.87 MPa,
fctm = 2.69 MPa, Ecm = 32.0 GPa.
Values at later stages were extrapolated based on results from
compressive tests of concrete at different times. The differences
found were irrelevant for this study.
Epoxy S strength is higher than the tensile strength of concrete
(2.7 MPa), an essential requirement to transfer the loads through
bond. Typical post-failure configurations of the bending tests are
shown in Fig. 2. As the load increased, the bond stresses migrated
to the extreme side of the composite reinforcement. Initially, con-
crete develops low tensile stresses, but, under increasing load, the
stresses rise, concrete cracks and stresses on GFRP ‘‘propagate” to-
wards the extreme sides of the strip. Where concrete cracks, the
load ceases to be carried by the corresponding layer of concrete
and is carried further out by the GFRP laminate. When stresses in
the GFRP are constant in consecutive gauges, the average bond
stress vanishes, meaning that debonding took place.
Separation of the GFRP laminate from concrete, in general, fol-
lowed concrete tensile failure and thick layers of concrete were
found adherent to the GFRP strip. Average results from bending
reference tests indicate a load capacity of 28.7 kN corresponding
to a maximum tensile force of 25.4 kN on the GFRP strip, calculated
from Fig. 3,
P
F GFRP ¼  ðb  aÞ ð1Þ
2h
where FGFRP = force on the GFRP strip, P = external load , h = distance
from the center of the metallic hinge to the GFRP strip, a, b = dis-
tances from the center of the beam to the point of load application
and support, respectively.
The assumption of linear behavior of the laminate leads to an
average bond stress:
De  E f  t f
sb ¼ ð2Þ
DL
where DL is the length between consecutive strain measurements
De, Ef is the GFRP Young Modulus and tf is the GFRP thickness.
The tensile stress on the GFRP laminate and the stepwise average
bond stress for a reference specimen are shown in Fig. 4, for some
values of the external load and up to its maximum value.
The average maximum bond stress and effective length for the
reference specimens was, respectively, 3.71 MPa and 150 mm.
Immersion in salt water increased the tensile strength of con-
crete, leading to a gain of 21% on the capacity of beams at
10,000 h. Failure surfaces are shown in Fig. 5, for tests made at
1000 h, 5000 h and 10,000 h of immersion; patterns are similar,
 M.A.G. Silva, H.C. Biscaia / Composite Structures 93 (2010) 216–224
Fig. 3. Scheme of external loading.
Fig. 4. GFRP and bond stresses at different stages of loading (Specimen REF-2).
Fig. 5. Water immersion 1000, 5000 and 10,000 h – failure surfaces.
219
whereas failure had taken place at increasing levels of tensile
stress.
Salt fog spray submitted the beams for 8 h followed by 16 h dry-
ing at 35 °C and these cycles were repeated till a time limit of
10,000 h was reached.
The failure surfaces, for both salt fog and immersion in salt
water, occurred at the interface of the GFRP and concrete, and
the load capacity was only slightly affected when compared to
the reference values.
The cycles used to simulate the effects of wet–dry cycles of salt
water (5% weight of NaCl) considered total immersion for 3 days
followed by total out of water positioning of the specimens for
4 days.
The surfaces of failure from the mechanical tests, at the inter-
face GFRP–concrete, give clues to the relative effects of the aging
processes. At 1000 h, Fig. 6 shows more pronounced concrete fail-
ure for salt water immersion, and greater deterioration of bond for
salt fogging. Loss of strength confirmed the visual inspection of the
failure surfaces as displayed in Fig. 7 where the evolution of the
capacity of the beams along time is shown for each type of
exposure.
The average maximum bond stress and effective length for the
reference specimens was, respectively, 3.71 MPa and 150 mm. Ta-
ble 1 shows the ratio of the average maximum bond stresses and
effective bond length after 10,000 h of aging by the corresponding
average values for the reference beams. The beneficial effects of the
salt water on concrete and the fact that salt water diffusion had not
been enough to erode the fibers explain the larger than one stress
ratio.
For tide like cycles and especially for salt fog, at 10,000 h, the
length of bonded strip beyond which no additional contribution
could be obtained for traction on the composite plate decreased.
Immersed specimens, after 10,000 h, gave the same results as at
reference time. These results are consistent with the distribution of
the stresses on GFRP along the bond interface as e.g. shown in
Fig. 8 where the length of the interface that is mobilized after aging
is smaller than on the reference specimens. However, for the im-
mersed specimens, more interface length was mobilized and a
higher rupture load was found.
The bending tests for coupons submitted to 5% of salt water
(immersion, fog and tide like cycles) revealed an initial improve-
ment of the interface behavior leading to higher maximum loads
when compared to the reference specimens. The aged tests also
revealed that the effective length tends to decrease with the
time exposure. This effect was more evident at higher stages
of ageing, excepting the case of total immersion in salt water.
220 M.A.G. Silva, H.C. Biscaia / Composite Structures 93 (2010) 216–224

2.3. Pull-off tests

Pull-off tests were also performed. A direct tensile force was ap-
plied to a partial core, cut through the GFRP material into the
underlying concrete slabs, as prescribed by EN 1542 [18]. The envi-
ronmental conditions were the same as described for the speci-
mens tested by bending. The average result for reference pull-off
stress was 4.09 MPa. The results normalized by the reference value
after 10,000 h revealed the highest pull-stress reduction (0.824) for
immersion in salt water. In the case of salt fog cycles and wet–dry
tests, the normalized results were 0.875 and 0.829, respectively.
There was a decrease of strength varying from 12.5% to 17.6%, with
salt fog being least detrimental. Results confirm the difficulty to
extrapolate from pull-off tests to beam flexural capacity, due to
the intrusive nature of the preparation for testing and to bond fric-
tion caused by swelling [19]. Except for one test out of a total of 42
Fig. 6. Tide simulation, salt fogging and immersion at 1000 h – failure surfaces.
tests [20], failure was due to tensile rupture in the concrete slab,
Fig. 9.

2.4. Shear tests for Mohr–Coulomb envelope

The interface finite element available in the computer code li-

brary is based on fracture mechanics and simulates the contact be-
tween GFRP and concrete. These interface elements introduce the
Mohr–Coulomb failure criterion that requires the knowledge of
the interface cohesion, c, and friction angle, /. The authors de-
signed experimental shear tests to determine those parameters
and rupture law (s = c + rtan /) [21] and applied them in the pres-
ent study.
The set-up for the tests conceived to obtain the envelope is
shown in Fig. 10 Unidirectional GFRP laminate 2.5 mm thick,
520 mm long and 80 mm wide (SEH51/Tyfo fabric) and Tyfo S
Epoxy were used. Standard tests of concrete cubes led to average
compressive strength 47.0 MPa and standard deviation 2.4 MPa.
Typical aspect of the surface of GFRP laminate after failure is
Fig. 7. Evolution of the bending load capacity under artificial aging.
also depicted in Fig. 10 and average results are tabulated in Table 2.
The interface cohesion was obtained from test mc-1. The fric-
tion angle / was obtained from the definition of the Mohr–Cou-
Table 1 lomb rupture envelope law and was found by applying different
Ratios of maximum bond stress and effective bond length at 10,000 h by values from values of compressive stress to the GFRP/concrete interface (mc-
reference specimens. 2 to mc-4). The maximum load found for pulling under zero nor-
Coupons Stress ratio Leffective ratio mal pressure (r = 0) was 55.1 kN (27.55 kN for each GFRP/concrete
interface). The tests, in general, revealed an increase of compres-
Salt fog cycles 1.10 0.75
Tide-like cycles 1.12 0.88 sion of 0.5 MPa that caused an increase of the maximum pulling
Immersion 1.16 1.00 strength of 15 kN which reveals the linearity of the Mohr–Coulomb
rupture envelope.

Fig. 8. Evolution of GFRP stresses along the interface length – 10,000 h of aging and
25 kN external load.

As a consequence, the slip distribution along the bonded area

was inferior to that observed in the reference specimens. Fig. 9. Aspect of fractured surfaces after pull off test.
 M.A.G. Silva, H.C. Biscaia / Composite Structures 93 (2010) 216–224 221

sm s
sðsÞ ¼  s  esm þ1 ð5Þ
sm
Bond-slip laws were established from measurements made
with strain gauges, at each aging stage and condition. Available
data show that the bond strength FGFRP is directly proportional to
the square root of the fracture energy at the interface Gf [22]
approximated by
Z sm
Gf ¼ sðsÞds ð6Þ
o

for each environmental aging, with different cases shown in Fig. 11

for 10,000 h exposure, as well as at the onset of artificial ageing.

3.1. Effects of environmental exposure

Under saline moisture ageing the fracture energy reached more

than twice the value on the reference specimens, allowing the pre-
diction of a gain of the maximum force anchored in the GFRP of
around 50% due to the effects of salt water. These are attributed
to increases of the concrete strength capacity during the aging pro-
cess in salt water. It is noticed that the increase of the maximum
load P0, read from the tests, was smaller than predicted by the pro-
portionality to the square root of Gf.
Fig. 10. Shear tests for characterization of Mohr–Coulomb. An increment of fracture energy has also been reported in the
literature for CFRP plates bonded to concrete blocks and submitted
to salt fog spray [24]. The results revealed an increase of 60% of
Table 2 fracture energy, after one month, and the bond-slip laws deter-
Shear tests GFRP–concrete.
mined by the authors followed the same pattern as those in Fig. 11.
Test r (MPa) F (kN) s (MPa) / Dmax (mm) It is remarked that in the fib bulletin 14 [25] the maximum force
mc-1 0 38.8 2.80 0 2.25 that can be transmitted to the FRP does not take into account the
mc-2 2.01 77.6 5.20 1.28 6.11 degradation of the bond due to aggressive environmental agents
mc-3 0.99 70.0 4.69 2.04 4.49 although, as mentioned, this issue can play an important role.
mc-4 0.50 55.2 3.69 2.24 3.56
Fig. 12 shows representative curves of the slip along the length
of the bonded strip, at 1000 h, 5000 h and 10,000 h, for each type of
3. Bond-slip chosen environmental action, when specimens were loaded with
25 kN.
Cracking of concrete at the interface originates large bond stres- The first 1000 h of salt fog did not significantly change the
ses in the vicinity of cracks and the GFRP plate slips relatively to material, whereas immersion and wet–dry cycles caused signifi-
the concrete surface, as confirmed from data generated from strain cant degradation.
gauges readings in the tests. Slip is generally denoted by ‘‘s” and, At 10,000 h, comparing the average results of the slip distribu-
when originated from a single flexural crack, it decreases until tion for the three environmental types of actions, it can be seen
the strains in the composite plate and the adjacent concrete are that the interface was strongly affected, with minor slips between
identical and s = 0. The slip strain between the composite and the GFRP and concrete.
adjacent concrete is given by ds/dx. Denoting by rx the tensile
stress on the laminate of thickness tf, and s(s) the relationship be- 3.2. Numerical estimates
tween the bond stress and s, from equilibrium, is governed by
The study of the slip distribution along the interface follows
@ rx sðsÞ work available in the literature [23,26]. The main difference be-
 ¼0 ð3Þ
@x tf tween both studies is that in [23], Eq. (4) was solved assuming a
bi-linear bond-slip law whereas in [26] Popovic’s non-linear
The strain on the GFRP may be obtained from ef = ds/dx, where ef
bond-slip law was used:
is the tensile strain in the FRP. From Hooke’s law, with Ef as Young’s
modulus of the laminate, the equation of equilibrium is obtained

@2s sðsÞ
 ¼0 ð4Þ
@x2 Ef  tf
A shooting method to solve the problem and relevant physical
considerations can be found on [22].
The relationship s–s has non-linear nature. The average bond
stress can be calculated by (2). If bi-linear bond-slip law is as-
sumed, as in Fig. 1, three parameters, maximum bond stress (Eq.
(2)), sm, ‘‘elastic” slip, sm, and maximum slip, s0 (both, obtained
by integration of the strains along the GFRP) are required to estab-
lish a solution to (4). However, exponential bond-slip laws can also
be used that approximate better the after-peak behavior of the
bond and require only two parameters (sm and sm), Fig. 11. Bond-slip curves after 10,000 h of degradation.
222 M.A.G. Silva, H.C. Biscaia / Composite Structures 93 (2010) 216–224

Tide simulation FDT Reference

Reference Stage 1: 1,000h Stage 2: 5,000h Stage 3: 10,000h 0.8
0.8 0.7
0.7 0.6

Slip (mm)
0.6 0.5
Slip (mm)

0.5
0.4
0.4
0.3
0.3
0.2 0.2
0.1 0.1
0 0.0
50 100 150 200 250 300 0 100 200 300
Strain Gauge distance from plate end (mm) Strain Gauge distance from plate end (mm)

Salt Fog Cycles Fig. 13. Comparison of slip distribution along the length of beam – reference tests
Reference Stage 1: 1,000h Stage 2: 5,000h Stage 3: 10,000h
and numerical treatment (P = Pmax).

0.8
0.7
0.6 4.1. Model to study failure of beam specimens
Slip (mm)

0.5
0.4 Preliminary analyses were carried out using a finite element
0.3
code suitable for RC modeling, developed at the University of Pra-
0.2
0.1 gue [21]. The code allowed the option of the smeared crack concept
0 assuming isotropic behavior until concrete cracks. With further
0 50 100 150 200 250 300 loading the crack direction is assumed to remain approximately
Strain Gauge distance from plate end (mm) fixed. The shear stiffness in the cracked concrete, and the shear
resistance decrease with crack opening [21]. Softening laws are
Immersion
used to simulate crack formation in the crack process zone (CPZ)
Reference Stage 1: 1,000h Stage 2: 5,000h Stage 3: 10,000h
where the stresses decrease on the cracked face. Afterwards, crack-
0.7
0.6
ing is governed by non-linear fracture mechanics. A crack band
length is introduced to control the localization of deformations at
Slip (mm)

0.5
0.4 failure. The decrease of compressive strength of cracked concrete
0.3 is also considered and the hardening/softening plasticity model
0.2 of the concrete compressive state is based on the failure surface
0.1 of Menétrey–William in the case of the 3D analysis [21].
0 It is remarked that preliminary studies made in a 2D model
0 50 100 150 200 250 300
where the external GFRP strip was crudely modeled by an ‘‘equiv-
Strain Gauge distance from plate end (mm)
alent external rod” [15] was later improved introducing interface
Fig. 12. Slip distribution along the length of beam at 1000 h, 5000 h and 10,000 h,
finite elements that follow the Mohr–Coulomb rupture criterion
for each type of environmental aging. with tension cut off [28]. The criterion was parametrized using re-
sults of shear tests (Section 2.4) that led to cohesion 2.80 MPa and
friction angle equal to 1.88.
The interface between concrete and steel bars followed MC90
sb s nP [29] and steel reinforcement obeys an elasto-plastic law with
¼  ð7Þ
sb;max sm ðnP  1Þ þ ðssm ÞnP Young Modulus 195 GPa and yielding at 323 MPa. The tensile crack
model for Mode I (opening) behavior and the sliding crack model
where sb,max is the bond stress of the FRP/concrete interface; nP is
for Model II (sliding) behavior are shown in Fig. 14.
the Popovic constant made equal to 3.0; sm is the slip at sb,max;
and sb;max ¼ 3:5fc0:19 , where fc is the concrete strength.
4.2. Finite element models
Other complex functions that pretend to reproduce with good
accuracy the non-linear GFRP/concrete post-peak behavior can be
Two dimensional models essentially based on quadrilateral ele-
used. For instance, differential Eq. (4) can be also solved introduc-
ments of 5 mm side with four nodes, and three dimensional solid
ing the exponential bond-slip law defined by Eq. (5). Eq. (7) is
brick elements with eight nodes, 10 mm side, were developed,
assumed in the following analysis. Popovic’s constant can be ad-
with finer meshes near the singularities. The 2D models had 265
justed to experimental data. The post-peak branch of the bond-slip
law decreases very quickly when nP decreases [26]. From the
experiments, it was possible to introduce y0 (0) = ef = 6.4  103
(maximum measured strain at central section of the reference
specimen) and y0 (L) = 0 (L = 240 mm i.e. zero strain at the end of
GFRP plate). The results obtained using 10 equal intervals are
shown in Fig. 13.
A detailed description of alternative models for bond-slip can be
found in [27].

4. Numerical modeling

Numerical techniques were used to solve the problems formu-

lated above and are briefly described next. Fig. 14. Bi-linear softening laws.
 M.A.G. Silva, H.C. Biscaia / Composite Structures 93 (2010) 216–224 223

elements and 3312 nodes. The 3D-GFRP elements were also mod-
P=10kN, FEA P=20kN, Experiment
eled with brick elements with 1 mm side. A total of 8775 elements
P=10kN, Experiment Pmax, FEA
were generated in the 3D model, corresponding to a total of 13,282
P=20kN, FEA Pmax, Experiment
nodes. In a 2.33 GHz Dual Core computer with 1 GB of RAM mem-
1
ory, the 3D analysis took about 16,571 s, more than three times the 0.9
507 s taken by the 2D analysis. Fig. 15 shows both models used in 0.8
this work. 0.7

Slip (mm)
0.6
4.3. Comparison between experimental and finite element analysis 0.5
results 0.4
0.3
0.2
The differences of the average experimental values to those ob-
0.1
tained by computations are fairly small as shown in Table 3 and
0
below 11% except for the maximum bond stress and the maximum 0 50 100 150 200 250 300
slip, perhaps because averaging of experimental results led to Strain Gauge distance from plate end (mm)
poorer representation of extreme values. Microcracks have direct
influence on the results and numerical accuracy of their location Fig. 16. Evolution of the slip on GFRP throughout the bond length: 3D numerical
-4.000E-05 results and experimental values.

P=10kN, FEA P=20kN, Experiment

P=10kN, Experiment Pmax, FEA
P=20kN, FEA Pmax, Experiment
160
GFRP stress (MPa)
140
120
100
Y

80
60
X 40
20
0
0 80 160 240 320
Strain Gauge distance from plate end (mm)

Fig. 17. Evolution of the stresses on GFRP along the bond length: 3D numerical
results and experimental values.

and formation is not very high, with the 3D model being more
effective than the 2D model. In general, for low level of load, the
numerical bond stress distribution is close to the experimental.
Given the proximity of the results obtained with the 2D and 3D
models, as per Table 3, only the computational values from the 3D
model are shown in Fig. 16, versus the experimental values.
Fig. 17 compares the evolution of the stresses on GFRP along the
Fig. 15. 2D and 3D mesh discretization showing crack propagation at maximum bond length, both calculated in the 3D model and obtained from
load capacity. the tests on reference coupons. It is seen that, for a same fixed load,
the 3D model underestimated the GFRP stresses. However, at

Table 3
Comparison between experimental average values and results of 2D/3D numerical
1.4
GR-REF-2
modeling. 1.2 GR-REF-1
Experimental ATENA Relative ATENA Relative 1
GR-REF-4
2D error (%) 3D error (%)
0.8
P/P0

Rupture load 28.74 29.91 6.1 31.73 10.3 FEA - 3D

(kN)
0.6
GFRP tension 25.39 26.42 4.1 28.03 10.4 0.4 GR-REF-3
(kN)
Max. bond 3.71 2.71 37.2 3.15 17.8 0.2
stress (MPa) 0
Maximum 0.63 0.68 7.9 0.68 7.9 0 0.2 0.4 0.6 0.8 1
strain (%)
Δ /Δ max
Max. normal 132.4 138.5 4.4 138.7 4.6
stress (MPa)
Max. slip (mm) 0.71 0.93 23.8 0.86 17.8 Fig. 18. External load versus non-dimensional displacement: experimental versus
results from 3D models (FEA).
224 M.A.G. Silva, H.C. Biscaia / Composite Structures 93 (2010) 216–224
collapse, the maximum stress on the GFRP is very similar to that
obtained from the experiments.
Dimensionless force–displacement results obtained from the
models are compared with the experimental results in Fig. 18,
where Dmax is the displacement measured at the maximum load,
P0, obtained according to the fib bulletin 14 [25] and found to be
25.3 kN [15]. Some differences between the experimental results
and the finite element model are visible, but partially attributed
to some imperfect representation of material properties and
curing.
5. Conclusions
Artificially accelerated tests were made and results interpreted
in order to study the influence that diffusion of water, with and
without NaCl, has on degradation of strength of beams with outer
reinforcement of GFRP.
Immersion in salt water, dry–wet cycles simulating tidal sub-
mersion in saline water and salt fog spraying designed to especially
examine the effects on bond between concrete and the GFRP outer
strip led to the following conclusions:
 Slip was affected by artificial aging, decreasing with time for all
three conditions.
 Salt water immersion caused improvement of concrete proper-
ties, higher plasticization of interface and a more evenly distrib-
uted stresses, with failure resulting from tensile fracture of
adjacent layer of concrete.
 Bond was visibly more degraded by the salt fog spray tests.
 A gain of beam capacity was found after 10,000 h of immersion
in salt water, whereas it decreased for salt fogging and dry–wet
tests.
 Less slip in the GFRP/concrete interface is associated with
higher rupture load.
 The results obtained by a 2D and 3D analysis did not differ sig-
nificantly, except for the maximum bond stress and maximum
slip at the interface, where the 3D model gave better
representation.
 Tests to obtain parameters that characterize Mohr–Coulomb
failure criterion were defined and results preliminarily used
on 3D numerical models.
In general, a systematic, broader study of accelerated environ-
mental effects of moisture diffusion on FRP–concrete bond contin-
ues to be required, including the need to consider synergetic
effects that are present in actual structures.
Acknowledgement
The authors are grateful to Fundação para a Ciência e Tecnologia
for partial financing of the work under Project DUST – PTDC/ECM/
100538/2008.
References
[1] Oehlers DJ. FRP plates adhesively bonded to reinforced concrete beams:
generic debonding mechanisms. Adv Struct Eng 2006:737–50.
[2] Lu XZ, Ye LP, Teng JG, Jiang JJ. Meso-scale finite element model for FRP sheets/
plates bonded to concrete. Eng Struct 2005;27(4):564–75.
[3] Lu XZ, Ye LP, Teng JG, Jiang JJ. Bond slip models for FRP sheets/plates bonded to
concrete. Eng Struct 2005;27(6):920–37.
[4] Neale KW, Ebead UA, Abdel Baky HM, Elsayed WE, Godat A. Analysis of the
load–deformation behaviour and debonding for FRP-strengthened concrete
structures. Adv Struct Eng 2006;9(6):751–63.
[5] Teng JG, Zhang JW, Smith ST. Interfacial stresses in reinforced concrete beams
bonded with a soffit plate: a finite element study. Construct Build Mater
2002:1–14.
[6] Zhou J, Lucas JP. Hygrothermal effects of epoxy resin. Part I: the nature of water
in epoxy. Polymer 1999;40:5505–22.
[7] Gumen RV, Jones RF, Attwood D. Prediction of glass transition temperature and
key engineering properties of dry/wet epoxy based composite matrix. In:
Thirteenth international conference on composite materials (ICCM-13),
Beijing, ID 1178; 2001.
[8] Silva MAG. Aging of GFRP laminates and confinement of concrete columns.
Compos Struct 2007;79:97–106.
[9] Thomason JL. The interface region in glass fibre-reinforced epoxy resin
composites: 2. Water absorption, voids and the interface. Composites
1995;26(7):477–85.
[10] Chin JW, Aouadi K, Nguyen T. Effects of environmental exposure on fiber
reinforced plastic materials used in construction 1997. J Compos Technol Res
1997;9(4):205–13.
[11] Kootsookos A, Mouritz AP, St. John NA. Comparison of the seawater durability
of carbon- and glass-polymer composites. In: Thirteenth international
conference on composite materials (ICCM-13), Beijing, ID 1178; 2001.
[12] Spainhour LK, Thompson I. Effect of carbon fiber jackets on reinforced concrete
columns exposed to a simulated tidal zone. In: Fiber composite in
infrastructure, proceedings of the second international conference on fiber
composites in infrastructure ICCI’98, vol. 1. Tucson; 1998. p. 426–39.
[13] Beaudoin Y, Labossière P, Neale W. Wet–dry action on the bond between
composite materials and reinforced concrete beams. In: Proceedings from first
international conference on durability of fiber reinforced polymer (FRP)
composites for construction. Sherbrooke; 1998. p. 537–46.
[14] Fava G, Mazzotti C, Poggi C, Savoia M. Durability of FRP–concrete bonding
exposed to aggressive environment. FRPRCS-8, University of Patras, Patras,
Greece, July 16–18, 2007.
[15] Biscaia HEC. Debonding between external GFRP reinforcement and concrete
structural elements. MSc thesis, Instituto Superior Técnico, Universidade
Técnica de Lisboa; 2006 [in Portuguese].
[16] Marreiros R. Degradação da aderência entre CFRP e betão armado devida a
envelhecimento acelerado. MSc thesis, Instituto Superior Técnico,
Universidade Técnica de Lisboa; 2005 [in portuguese].
[17] Biscaia HEC, Silva MAG. Environmental effects on bond of GFRP external
reinforcement to RC beams. In: 8th International symposium on fiber
reinforced polymer reinforcement for concrete structures, FRPRCS-8, Patras,
Greece, July 16–18, 2007.
[18] BS EN 1542:1999. Products and systems for the protection and repair of
concrete structures – tests methods – measurements of bond strength by pull-
off.
[19] Sen R, Shahawy M, Sukumar S, Rosas J. Effects of tidal exposure on bond of
CFRP rods. In: Proceedings of the second international conference on fiber
composites in infrastructure. ICCI98, vol. 2. Tucson; 1998. p. 512–23.
[20] Silva MAG, Biscaia H. Degradation of bond between FRP and RC beams.
Compos Struct 2008;85(2):166–74.
[21] Cervenka V, Jendele L, Cervenka J. ATENA program documentation – part 1 –
theory. Prague: Cervenka Consulting; 2003.
[22] Wu ZS, Yuan H, Niu HD. Stress transfer and fracture propagation in different
kinds of adhesive joints. J Eng Mech, ASCE 2002;128(5):562–73.
[23] Liu IST, Oehlers DJ, Seracino R. Study of intermediate cracking debonding in
adhesively plated beams. J Compos Construct, ASCE 2007;11(2):175–83.
[24] Fava G, Mazzotti C, Poggi C, Savoia M. Durability of FRP-concrete bonding
exposed to aggressive environment. In: 8th International symposium on fiber
reinforced polymer reinforcement for concrete structures, FRPRCS-8, Patras,
Greece, July 16–18, 2007.
[25] fib bulletin 14. Externally bonded FRP reinforcement for RC structures.
Technical report; 2001.
[26] Ferracuti B. Strengthening of RC structures by FRP: experimental analyses and
numerical modelling. Dissertazione presentata per il conseguimento del titolo
di Dottore di Ricerca in Meccanica delle Structture, Università degli Studi di
Bologna.
[27] Lua XZ, Tengb JG, Yea LP, Jianga JJ. Bond-slip models for FRP sheets/plates
bonded to concrete. Eng Struct 2005;27:920–37.
[28] Biscaia HEC, Silva MAG, Chastre CC. Experimental characterization and
numerical modeling of the interface GFRP/concrete. In: 7th Cong
experimental mechanics. Vila Real, Portugal; 2008 [in Portuguese].
[29] Comité Euro-International du Betón. CEB-FIP model code; 1990.