Mechanical Behaviour and Durability of FRP To Steel Adhesively Bonded Joints

MechanicalBehaviourandDurabilityof
FRPtosteelAdhesivelybondedJoints
XuJiang
MechanicalBehaviourandDurabilityof
FRPtosteelAdhesivelybondedjoints
Proefschrift
ter verkrijging van de graad van doctor
aan de Technische Universiteit Delft,
op gezag van de Rector Magnificus prof. ir. K.C.A.M. Luyben,
voorzitter van het College voor Promoties,
in het openbaar te verdedigen
op woensdag 23 oktober 2013 om 12:30 uur
door
XuJIANG
MasterofScienceinStructuralEngineering,TongjiUniversity,China
geborenteShenyang,China
Dit proefschrift is goedgekeurd door de promotoren:

Prof. ir. F.S.K. Bijlaard
Copromotor:
Dr. M.H. Kolstein
Samenstelling promotiecommissie:
Rector Magnificus
Prof. ir. F.S.K. Bijlaard
Dr. M.H. Kolstein
Prof. dr. Airong Chen
Prof. dr.ir. P. van Bogaert
Prof. dr.ir. D.A. Hordijk
Prof. ir. A.C.W.M. Vrouwenvelder
Prof. dr. I.M. Richardson
voorzitter
Technische Universiteit Delft, promotor
Technische Universiteit Delft, copromotor
Tongji University
University of Gent
Technische Universiteit Delft
ISBN 978-90-5335-747-7
Copyright 2013 by Xu Jiang
All rights reserved. No part of the material protected by this copyright notice may
be reproduced or utilized in any form or by any means, electronica or mechanical,
including photocopying, recording or by any information storage and retrieval
system, without the prior permission of the author.
Printed by Ridderprint in the Netherlands
For my parents
Summary
During the last two decades, fiber-reinforced polymer (FRP) bridge decks
have been increasingly used as a competitive alternative for wood, concrete
and orthotropic steel decks, due to their various advantages: light-weight,
good corrosion resistance, low maintenance cost and rapid installation for
minimizing the traffic disturbing time. These advantages meet critical needs
for rehabilitation and new construction of pedestrian and highway bridges.
To be cost effective, FRP decks are usually supported by steel girders. For
the connection between FRP decks and steel girders, adhesive bonding
technique is usually considered as a preferable connecting method, which
can reduce construction time, save weight by eliminating fasteners, allow
more uniform load transfer, achieve better adaption to the brittle and
anisotropic nature of FRP materials and provide higher joint efficiency.
Despite the fact that FRP bridge decks and adhesive joints are already in
service in many FRP-steel composite bridges, mechanical behaviour and
long-term performance are still not clearly understood, which results in more
conservative designs of the FRP-steel composite bridges. To compensate
this lack, the overall aim of this project is to investigate mechanical
behaviours (in terms of strength and stiffness) of adhesively-bonded joints
between FRP bridge decks and steel girders, as well as durability of these
adhesively-bonded joints. As to the first aspect, considering the distribution
of traffic loads in the longitudinal and transverse directions of bridges, the
adhesive-bonded joints have been experimentally studied under six loading
conditions, including tensile loading, shear loading and four combining ratios
of tensile and shear loading. A specific tensile-shear loading device was
designed and then employed to offer six different angle loading conditions.
Different surface pretreatment methods (acetone (AC), sand paper (SP) and
sand blasting (SB)) were compared with regard to influences on the
stiffness, load-bearing capacity, failure mode and interfacial bonding quality
of adhesive joints. A Finite Element (FE) model was developed to simulate
the stress distribution throughout the adhesive joints under different loading
conditions, which proved that the failure of joints was induced by
combination of both tensile and shear stress peaks. The edge zone
i
(approximately 10mm from the ends of the adhesive layer) was the most
sensitive area to initiate the failure, where both the shear stress peak and
the tensile stress singularity were located.
Another critical aspect of this research is to characterize the durability of
FRP-to-steel adhesively-bonded joints under both temperature and moisture
effects. The influence of hydrothermal environmental aging on the
mechanical behaviours of adhesive joints has been studied and compared
with the un-aged adhesive joints. The shear-tensile failure criterions of
hydrothermal aged and un-aged adhesive joints were addressed. To better
understand the moisture effects, the moisture diffusion process in FRP
composite materials was characterized. Subsequently, the hydrothermal
degradation on the flexural and interlaminar properties of FRP laminates
was addressed. A coupled hygro-mechanical FE model was developed to
analyse the enviroment-dependent mechanical behaviours of FRP
lanimates. This FE model was first validated by test results of flexural tests
and subsequently employed in an inverse parameter identification method to
determine the elastic interlaminar shear modulus of FRP laminates.
Predictive equations for environment-dependent mechanical properties
(flexural and interlaminar) of FRP laminates were sustained by using the
least square method for the curve fitting.
Results of this research can contribute to the development of a design code
of FRP-steel composite bridges. They can also be used as a reference
information for understanding mechanical behaviours and durability of FRPto-steel adhesively bonded joints for other applications in civil engineering
field, such as strengthening of steel structures using FRP composite
materials.
ii
Contents
Summary.......................................................................................................................i
List of notations...........................................................................................................v
List of abbreviations.................................................................................................vi
Chapter 1 Introduction ............................................................................................. 1
1.1 Background .............................................................................................................. 1
1.2 Aim of the research .................................................................................................. 2
1.3 Outline...................................................................................................................... 3
Chapter 2 Literature review.....................................................................................7
2.1 FRP bridge decks ..................................................................................................... 7
2.2 FRP-steel composite bridge deck system ................................................................ 9
2.3 Adhesively-bonded joint ........................................................................................ 13
2.4 Environmental effects ............................................................................................ 16
References .................................................................................................................... 21
PartIMateriallevelresearch
Chapter 3 Moisture diffusion characteristics of FRP composites ......25
3.1 Introduction ............................................................................................................ 25
3.2 Moisture diffusion theory ...................................................................................... 25
3.3 Experiments ........................................................................................................... 29
3.3.1 Specimen preparation...................................................................................... 29
3.3.2 Gravimetric test process ..................................................................................31
3.4 Experimental results and discussion ...................................................................... 32
3.4.1 Pultruded FRP composites ..............................................................................32
3.4.2 Resin-infusion FRP laminates.........................................................................39
3.5 Conclusion ............................................................................................................. 45
References .................................................................................................................... 47
Chapter 4 Mechanical degradation of FRP laminates under hot/wet
environment ............................................................................................................. 49
4.1 Introduction ............................................................................................................ 49
4.2 Flexural property .................................................................................................... 50
4.2.1 Experiment ...................................................................................................... 50
4.2.2 Experimental results and discussion ............................................................... 54
4.2.3 Coupled hygro-mechanical FE analysis..........................................................63
4.3 Interlaminar shear property .................................................................................... 66
4.3.1 Experiment ...................................................................................................... 66
4.3.2 Experimental results and discussion ............................................................... 69
4.3.3 Determining the interlaminar shear modulus of FRP laminates .....................73
4.4 Conclusion ............................................................................................................. 85
References .................................................................................................................... 87
iii
PartIIJointlevelresearch
Chapter 5 Mechanical behaviour of FRPtosteel adhesivelybonded joints
before and after hydrothermal aging...................................................................89
5.1 Introduction ............................................................................................................ 89
5.2 Experiment ............................................................................................................. 91
5.2.1 Tensile-shear loading device........................................................................... 91
5.2.2 Test specimen.................................................................................................. 93
5.2.2.1 Un-aged specimen .................................................................................... 93
5.2.2.2 Hydrothermal aged specimen ................................................................... 98
5.2.3 Experimental procedure ................................................................................100
5.3 Mechanical behaviour of adhesively-bonded joints before hydrothermal aging . 106
5.3.1 Shear loading ................................................................................................ 106
5.3.1.1 Experimental results and discussion....................................................... 106
5.3.1.2 FE analysis ............................................................................................. 115
5.3.2 Tensile loading .............................................................................................. 125
5.3.2.2 FE analysis ............................................................................................. 132
5.3.3 Combination of shear and tension loading ....................................................141
5.3.3.2 FE analysis ............................................................................................. 147
5.4 Mechanical behaviour of adhesively-bonded joints after hydrothermal aging .... 155
5.4.1 Shear loading ................................................................................................ 155
5.4.2 Tensile loading .............................................................................................. 159
5.4.3 Combination loading of shear and tension.................................................... 162
5.5 Comparison of mechanical behaviours of FRP-to-steel adhesively-bonded joints
before and after hydrothermal aging .......................................................................... 167
5.5.1 Ultimate failure load and shear-tensile failure criterion ............................... 167
5.5.2 Failure mode ................................................................................................. 170
5.5.3 Stiffness......................................................................................................... 173
5.6 Conclusions......................................................................................................177
References .................................................................................................................. 181
Chapter 6 Conclusions and recommendations.................................................183
6.1 Conclusions .......................................................................................................... 183
6.1.1 Part I: Material level research ....................................................................... 183
6.1.2 Part II: Joint level research ........................................................................... 184
6.2 Recommendations for future research work ........................................................ 186
Acknowledgement..................................................................................................189
Curriculum vitae.....................................................................................................191
List of Publications .................................................................................................193

iv
List of notations
A
b
c
c
D
Dmax
D1
D2
D3
d
E
e
Fsbs
Fshear
Ftension
G13 (G23)
h
Kn
L
l
Mt
M
P
Pm
R
R2
S
t
w0
wt
x
Z
average
average
adhesive-bonding area
plate (specimen) width
moisture concentration
maximum equilibrium moisture concentration
equivalent moisture diffusion coefficient
maximum deflection of the center of the beam
moisture diffusion coefficients along the direction of length
moisture diffusion coefficients along the direction of width
moisture diffusion coefficients along the direction of thickness
depth of FRP beam tested
modulous
plate thickness
short-beam strength
shear load applied on the adhesive joint
tensile load applied on the adhesive joint
interlaminar shear modulus
specimen thickness
characteristic fractile factor
support span
plate length
moisture absorption content at the time t
equilibrium amount of absorption
load at the midspan on the load-deflection curve
maximum load observed during the short-beam test
rate of crosshead motion
R-square value
strength
time
specimens weight before exposure
specimens weight after exposure
space coordinate measured parallel to the diffusion
rate of straining of the outer fiber
stress in the outer fibers at midpoint
Von Mises stress
strain in the outer fibers at midpoint
average shear stress
average tensile stress
List of abbreviations

1D
2D
3D
AC
ASTM
DS
FE
FEA
FRP
GFRP
ISO
LVDT
PC
RC
RH
SB
SP
Tg
one dimensional
two dimensional
three dimensional
acetone
American Society for Testing and Materials
displacement sensor
finite element
finite element analysis
fiber reinforced polymer
glass fibre-reinforced polymer
International Organization for Standardization
linear variable differential transformer
personal computer
reinforced concrete
relative humidity
sand blasting
sand paper
glass transition temperature
vi
Chapter 1
Introduction
1.1 Background
The deteriorating state of the bridge infrastructure in many countries is well
documented all over the world. Conventional concrete decks, timber decks
as well as orthotropic steel decks are usually the major cause of structurally
deficient bridges. To address this issue, Fiber Reinforced Polymer (FRP)
bridge decks were developed to be a light-weight and durable alternative as
a means of deck replacement for older and deteriorated bridges, particularly
for bridges with steel girders. Due to the light-weight of FRP decks, the
reduced deck load may permit increased traffic loading without altering the
original state of the bridge. Moreover, the rapid installation of an FRP deck
also reduces bridge closure time for a rehabilitation project and minimizes
inconvenience to the daily traffic. Also, for new bridges, steel girder with an
FRP deck is a realistic option.
For the connection between the FRP decks and the steel girders, the
adhesive bonding technique is usually considered as a preferable
connecting method, which can reduce construction time, save weight by
eliminating fasteners, allow more uniform load transfer, achieve better
adaption to the brittle and anisotropic nature of FRP materials and provide
higher joint efficiency. Despite the fact that adhesive joints are already in
service in many FRP-steel composite bridges, the long-term performance,
gluing technique and design method are still not clearly understood.
Literature review shows that the adhesive joints have been intensively
investigated over the past 70 years. However, most researches are related
to the applications in aerospace and aircraft engineering. Only during the
last two decades, the adhesive joints have been increasingly used in civil
infrastructures, especially applied together with FRP composite materials.
But the adhesively-bonded joints utilized in civil infrastructures show
essential differences, including bond geometries (adhesive and adherent
thicknesses), fabrication processes, loading, curing conditions and service
1
Chapter1
environments. The FRP composite profiles and adhesive layers for

aerospace/aircraft structures are usually thin (0.11 mm), while in bridge
and building structures adherents and adhesive layers are comparatively
much thicker (220 mm). Furthermore, the design requirements and service
conditions of adhesive joints for infrastructures differ from those employed in
aerospace/aircraft structures. For instance, the service life of a bridge or
building is much longer than that of an aircraft. In many countries, 70 years
or more is expected to be the service life of bridges. Furthermore, curing
conditions are also different. Adhesives used in the civil engineering industry
are usually cured in ambient environments, but for aerospace applications
the temperatures of curing conditions are usually over 100C, which leads to
higher glass transition temperatures and the joints tend to be more durable.
Thus, researches conducted in aerospace engineering can only be applied
to the civil engineering field with limitations.
1.2 Aim of the research

The overall aim of this project is to investigate the mechanical behaviour (in
terms of strength and stiffness) of adhesively-bonded joints between FRP
bridge decks and steel girders, as well as the durability of these adhesivelybonded joints. As to the first aspect, considering the distribution of traffic
loads in the longitudinal and transverse directions of bridges, the adhesivebonded joints have been experimentally studied under six loading
conditions, including tensile loading, shear loading and four combining ratios
of tensile and shear loading. Shear stress occurring in adhesive joints is due
to the composite action between FRP decks and steel girders in the
longitudinal direction of the bridge. The deck and steel girder tend to bend
together to carry the traffic load. Thus, the adhesive joint is in the shear
stress state to transfer the loading from FRP deck to steel beam, see Fig.
1.1 a). As shown in Fig. 1.1 b), in the transverse direction of the bridge,
loading on left traffic lanes causes up-lift forces locally on the adhesive joint
at the right side of the bridge, which results in the through-thickness tensile
stress in the adhesive joints between FRP decks and steel girders. The
above two phenomenon can also take place simultaneously resulting in a
combination of shear and tensile stress in the adhesively-bonded joint.
Chapter1Introduction
a) Shear stress in the longitudinal direction
b) Tensile stress in the transverse direction

Fig. 1.1. Typical stress states in an adhesively-bonded joint of an FRP-steel composite
bridge
Different surface pretreatment methods are compared with regard to the

influence on the stiffness, load-bearing capacity, failure mode and interfacial
bonding quality of adhesive joints. A Finite Element (FE) model is developed
to simulate the stress distribution throughout the adhesive joints under
different loading conditions. Another critical aspect of this research is to
characterize the durability of FRP-to-steel adhesively-bonded joints under
temperature and moisture effects. The influence of hydrothermal
environmental aging on the mechanical behaviours of adhesive joints has
been studied. To better understand the moisture effects, the moisture
diffusion process in FRP composite materials is characterized.
Subsequently, the hydrothermal degradation on the flexural and interlaminar
properties of FRP laminates is addressed. Results of this research can
contribute to the development of a design code on FRP composite materials
and structures for application in the civil engineering field, particularly for the
FRP-steel composite bridge.
1.3 Outline
This thesis contains six chapters, as illustrated in Fig. 1.2. After the first
introductory chapter, a literature review is provided in Chapter 2, which
Chapter1
gives a brief state-of-the-art in FRP bridge decks, the FRP-steel composite

bridge deck system, adhesively-bonded joints and environmental effects.
The research work mainly consists of two parts: Part I - material level
research (Chapter 3 and Chapter 4) and Part II - joint level research
(Chapter 5).
In Chapter 3, the moisture diffusion characteristic of two types of FRP
composite materials (pultrusion and resin-infusion) are studied by
gravimetric experiments in four environmental conditions. Based on the
moisture diffusion theory, the moisture diffusion coefficients are determined.
Subsequently, the FE model for simulating the moisture diffusion process in
the FRP composites is developed and validated by the experimental results,
which provides a numerical technical basis for coupling the moisture
diffusion and mechanical analysis of FRP composites. Chapter 4
investigates the influence of moisture and temperature on the mechanical
properties (flexural and interlaminar shear) of FRP laminates by employing
the three-point bending tests. One cycle of the moisture absorptiondesorption process is considered. The environment-dependent degradation
of flexural modulus and strength as well as shear strength is experimentally
addressed. Furthermore, the coupled hydro-mechanical FE model is
developed and employed to determine the environment-dependent
interlaminar shear modulus by an inverse parameter identification approach.
Finally, the predictive equations of mechanical degradation of FRP
composite properties (flexural and interlaminar shear) are presented.
In Chapter 5, firstly, the mechanical behaviour (in terms of load-deformation
and stress-deformation) of FRP-to-steel adhesively-bonded joints is
experimentally and numerically studied. A specific tensile-shear loading
device is designed and then employed to offer six different angle loading
conditions, including the pure tensile, 18, 36, 54, 72 and pure shear
loading. The 18, 36, 54, 72 angle loading conditions are considered as
the combination of tensile and shear loads in four different ratios. The
influence of different surface pretreatment methods on the mechancial
performance of adhesive joints is also investigated under tensile and shear
loading. By experimental investigations, a tensile/shear failure criterion of
4
Chapter1Introduction
adhesively-bonded joints is addressed. Furthermore, a three-dimensional

FE model of the adhesive joint is developed and validated by experimental
results. Subsequently, linear elastic simulations are performed to
characterize the stress distribution in the adhesive joint under six different
loading conditions. The mesh-dependency of the FE model is also
addressed in the longitudinal, transverse and through-thickness direction of
the adhesive joint. Subsequently, further research is focused on mechanical
behaviours of the FRP-steel adhesively-bonded joint after hydrothermal
aging. The previous tensile/shear loading device is employed again to offer
the six different loading angles for testing the four-month aged adhesive
joints. A tensile/shear failure criterion of four-month aged adhesively-bonded
joints is obtained. To better understand the influence of hydrothermal aging
on mechanical behaviours of the adhesively-bonded joint, a comparison
between the experimental results of un-aged joints and hydrothermal aged
joints is conducted with regard to the ultimate failure load, failure criterion,
failure mode and stiffness. Furthermore, the post curing mechanism induced
by the elevated temperature of hydrothermal aging environment is
discussed.
Finally, Chapter 6 gives the overall conclusions of this research, as well as
recommendations for future research work.
Chapter1
Chaptter 1
Introdu
uction
Chapter 2
Liteerature review
PartI: Matteriallevelrresearch
Chapter
C
4
Mechanical
M
degradationn of FRP
laminates un
nder hot/weet
environment
e
t
C
Chapter 3
M
Moisture difffusion charaacteristics
off FRP compposites
PartII: Joi ntlevelressearch

Chapter 5
Mechaniical behavio
ours of FRP
P-to-steel ad
dhesively-bo
onded jointss before
hydrotherm
mal aging
Mechaniical behavio
ours of FRP
P-to-steel ad
dhesively-bonded jointts after
hydrotherm
mal aging
Compariison on mecchanical behhaviours of FRP-to-steel adhesivellybonded joiints before and
a after hyydrothermal aging
Chapterr 6 Concluusions and reecommendaations

Fig. 1.2. O
Outline of the
t thesis
Chapter 2
Literature review
2.1 FRP bridge decks
Fibre reinforced polymer (FRP) materials, especially glass-fibre-reinforced
polymer (GFRP) composites are being increasingly applied in civil
engineering as a competitive alternative to traditional materials, such as
concrete, timber and steel [1-5]. From a review of FRP composites for
construction [1], FRP composites were firstly commercialized to meet the
higher performance challenges of space exploration and air travel in the
1960s and 1970s. Thanks to cost reduction of the continued growth of the
FRP industry, FRP composites finally found their acceptance in the
conservative infrastructure construction industry during the late 1980s and
throughout the 1990s. For the design of FRP structures, optimization design
methods were introduced and compared in literature [2], which
compensated for the lack of design standards for FRP infrastructure.
Throughout the past two decades, one conspicuous application of FRP
materials in civil engineering is FRP bridge decks for rehabilitation of old
bridges and the construction of new bridges [3-8]. Current commercially
available FRP decks can be classified into two categories according to the
types of assembly and construction [9]: sandwich panels (Fig. 2.1 a) and b))
and multi-cellular type panels (Fig. 2.1 c) and d)). The sandwich decks are
mainly manufactured by the resin vacuum infusion technique, and the
cellular decks are made by the pultrusion technique.
Chapter2
(a) ECOSAFE (Infra Composites, Netherlands) (b) Honeycomb Deck (Kansas, USA)
(c) ASSET (Fiberline, Denmark)
(d) DuraSpan (Martin Marietta Composite, USA)
Fig. 2.1. FRP bridge decks
The growing acceptance of FRP bridge decks can be attributed to their

pronounced advantages [10]:
Light weight. FRP bridge decks weigh about 10-20% of a reinforced
concrete deck. Consequently, using an FRP deck to replace a concrete
deck reduces the dead load significantly. A lighter dead load can be
translated into savings throughout the structure and the foundations are
reduced for new structures.
Corrosion resistance. Corrosion of the reinforcing steel is the main cause
of premature deterioration of RC bridge decks. The use of road de-icing
salts accelerates this corrosion. FRP composites possess a higher tolerance
for frost and de-icing salts.
Rapid installation with minimum traffic disruption and factory making.
Factory made FRP deck panels offer several advantages over cast-in-place
concrete decks. These are:
1) Quality of the product can be closely monitored in the controlled factory
environment.
2) During manufacturing the potential for inclement weather is eliminated.
3) Once the superstructure is prepared, the fabricated deck structure can
be installed quickly with light lifting cranes.
8
Chapter2Literaturereview
High strength to weight ratio.

Longer service life and lower maintenance cost. Life cycle cost savings
have been shown to more than offset the relatively high initial cost of the
FRP materials compared to conventional materials. The service life of the
FRP deck can be about three times larger than concrete decks. However,
few public agencies select materials based on projected life-cycle costs,
most materials are chosen on the experience and judgement of the engineer,
agency preferences and industry standard practice, generally with a strong
bias towards minimizing initial construction costs.
2.2 FRP-steel composite bridge deck system

To be cost effective, the FRP decks are usually supported by steel girders,
as shown in Fig. 2.2. Steel girders enhance the ductility of this composite
bridge system after failure loading achieved, which compensates for the
brittle characteristics of FRP composites. Between the FRP decks and steel
girders, the adhesive bonding technique is usually employed as a preferable
connection method. In recent years, this kind of FRP composite girder
system was utilized [3, 4, 11]. In Knipperss research [3], it was employed as
a flyover across the federal road B3 in Germany. The high durability of FRP
composites and the fast assembly of the bridge were decisive factors for this
application. Through Cassity et al.s experimental investigation [11], the
degree of composite action between cellular FRP decks and steel girders
was studied and subsequently adopted in a rehabilitation project of an old
and deteriorated bridge. Through these projects, valuable experience was
gathered concerning in-situ constructions, and the good performance of
FRP-steel composite deck system was confirmed.
Chapter2
Fig. 2.2. FRP-to-Steel composite girder system [12]
In most cases, FRP decks have to compete with concrete decks. For the
widely used concrete-steel composite bridge, concrete decks are usually
designed to behave as the top chord of the composite girder in the
longitudinal direction of the bridge, and the stiffness and load-bearing
capacity of the bridge can be significantly increased. To be competitive,
FRP decks also need to be capable to contribute as part of the longitudinal
top chord for maintaining the full composite action. To achieve this, the
adhesively-bonded joint between FRP decks and steel girders must obtain
the full loading transfer capacity, which implies a linear strain distribution
through the depth of the hybrid cross-section. In literature, some researches
[7, 11-15] highlighted that the composite action between FRP decks and
steel girders is of great importance. In the research of Keller and Gurtler [14],
a 30% decrease in deflection and a 56% increase in load-bearing capacity
were experimentally identified, both relative to a single steel I-shape beam
acting alone, for a 7.5m span simple-supported composite specimen with
the FRP deck adhesively bonded to the steel beam. The full load transfer
capacity was evident through the adhesive joint between the FRP deck and
the steel beam, as shown in Fig. 2.3.
10
Fig. 2.3. Axial strain distribution in the mid-span cross-section of the Asset and DuraSpan
hybrid Girders [16]
The plate-bending behaviour of a pultruded GFRP bridge deck system was

investigated by full-scale experiments and numerical modeling [12], as
shown in Fig. 2.4. Particularly, the through-thickness performance of
adhesive joints between FRP bridge decks and steel girders was studied.
Tensile stress distribution in the adhesive joint is non-uniform with high
stress concentrations underneath the FRP webs of the cellular deck,
attaining stress concentration factors higher than 12, see Fig. 2.5 and 2.6.
Fig. 2.4. FE model of test set-up
11
Chapter2
Fig. 2.4. FE model of test set-up
Fig. 2.5. Normalized through-thickness tensile stress distribution in adhesive layer/bridge

deck interface (x-direction) over center of steel girder
Fig. 2.6. Normalized through-thickness tensile stress distribution in adhesive layer/bridge

deck interface (y-direction) below vertical web at x=608 mm
12
2.3 Adhesively-bonded joint

As mentioned in reference [9], the load carrying connections in an FRP
composite bridge include componentcomponent connections to form
modular FRP bridge deck panels (henceforth referred as component level
connection), panelpanel connections to form FRP bridge deck systems
(henceforth referred as panel level connection), and FRP deck-to-support
connections to form bridge superstructures (including deck-girder, deckabutment and deck-barrier connections, etc., henceforth referred as system
level connection). For these connections, the adhesive bonding technique is
usually considered to be an excellent alternative. As compared to bolted or
stud connections, adhesively-bonded connections can reduce construction
time, save weight by eliminating fasteners, introduce more uniform load
transferring and provide better long-term performance. Bolted connections
usually result in much higher stress concentrations where cracks occur.
Furthermore cutouts of FRP plates can provide path for moisture
penetration. Adhesive connections, however, are more material-adapted,
since larger surfaces can be glued together and no holes are made, thus
reducing concentrated stresses.
As mentioned above, lots of studies on FRP composite adhesive joints were
conducted in the aerospace engineering field [17-20]. But these experiences
and knowledge cannot be directly applied to civil engineering applications
because of essential differences in geometries, types of fibres and matrix,
fabrication methods, curing processes and service environmental conditions.
Recently, some researches [21-23] were conducted for civil engineering
applications, focusing on the mechanical performance of adhesively-bonded
single-lap joints and double-lap joints. These adhesive joints were
composed of pultruded GFRP composite profiles glued by epoxy adhesives,
as illustrated in Fig. 2.7.
13
Chapter2
Fig. 2.7. Geometry of lap joint specimens (not to scale)
Parametric studies were conducted experimentally and numerically on the

overlap length, the adhesive layer thickness, the adherent thickness and the
degree of chamfering of the adherents. The results (Fig. 2.8) indicated that
the combination of local through-thickness tensile (peeling) and shear
stresses was the most severe stress-state and usually initiated the failures
in the adhesive fillet and in the outer fibre-mat layers of the adherents below
the joint edges. Further researches [24, 25] offered a probabilistic strength
prediction method on the adhesive joints under quasi-static axial tensile
loading.
14
a) Shear stress distribution
b) Through-thickness stress distribution

Fig. 2.8. Stress distribution at 100kN for DN 100.1/5 specimen (100mm overlap, inner
profile)
The stress analysis on the adhesively-bonded joints is usually approached

by a closed-form analytical model or a finite element analysis (FEA). For a
fast and easy answer, a closed-form analysis is preferable. A review [17] on
analytical models of adhesively-bonded joints shows that almost all
analytical models for adhesively-bonded lap joints are two-dimensional,
which are generally sufficient because the variation of stresses in the width
direction are significantly lower than that in the direction of the loading. The
15
Chapter2
linear elastic analysis is supposed to be more appropriate for both

adherents and adhesive, because the inclusion of material non-linearity
renders the solution too complex. However, for the complex geometries and
elaborate material models, an FEA is more suitable, which allows many
tests to be simulated that would take too long to perform or be too
expensive in practice, such as a geometrical parametric study or selection of
appropriate material properties. In recent years, by improving the
computational power of present machines, three-dimensional FE models [19,
26, 27] are more preferable for the stress-strain analysis on the adhesive
joint than a two-dimensional FE model. By employing the three-dimensional
FE models, the behaviour out of the plane can be addressed in a more
trustworthy way.
2.4 Environmental effects

Although FRP composites are increasingly being used in civil infrastructure
applications, their durability and long-term performance are still not
comprehensively understood. More comprehensive understanding and
accurate experimental data are required, since the service life of
infrastructures is generally expected to be more than 50 years. With a view
to the fact that most FRP infrastructures were constructed in the last two
decades, the research correlating to the effects of environmental
degradation on the mechanical behaviour and long-term performance of
FRP structures are very limited in literature. In the natural life of such
applications, the FRP composites are usually exposed to harsh and
changing environments, involving large variations in temperature and
humidity. In Ashcroft et al.s research [28], three typical outdoor
environments were selected and considered as hot/wet environment
(Innisfail, Australia), hot/dry environment (Cloncurry, Australia) and
temperate environment (Farnborough, UK). Average temperature and
relative humidity of these three places are listed and shown in Table 2.1 and
Fig. 2.9 respectively.
16
Table 2.1. Average climatic conditions at outdoor environments
Environment
Hot/wet
Hot/dry
Temperate
Temperature
23C
25C
10C
Relative humidity
83%
55%
78%
Monthly rainfall
297mm
39mm
49mm
(a) Temperature
(b) Relative humidity

Fig. 2.9. Average monthly temperature and relative humidity in outdoor environments
The hot/wet exposure is supposed to be the severest environmental

condition to degrade the performance of FRP composites [29-37], which will
17
Chapter2
decrease the service life of FRP composite structures. The absorbed

moisture will cause plasticization, saponification or hydrolysis that will
induce both reversible and irreversible changes in the micro structure of the
polymer, which will lead to degradation in their mechanical, chemical and
thermo-physical characteristics [38-40]. Degradation due to moisture
absorption may significantly reduce the service life of FRP structures. The
elevated temperature usually accelerates this process. Thus, moisture
diffusion in FRP composites is considered to be one of the major reliability
concerns for the long-term performance of FRP structures.
In the aerospace engineering field, the influence of moisture absorption on
mechanical properties of FRP composites and adhesive material is well
documented in literature [29-34, 41-53]. Absorbed moisture can cause
pronounced changes in modulus, strength and ultimate strain [50]. Moisture
content of submerged FRP composites increases by diffusion. The research
of Garcia et al. [47] indicates that the absorbed moisture can cause matrix
cracking, fibre-matrix debonding, and corrosion of glass fibres, which results
in a reduction of strength and stiffness of the FRP composite. Phifer [49]
recorded that tensile strength and stiffness reductions are 60 percent and 10
percent, respectively, for E-glass/vinyl ester composites submerged in fresh
water for a period of about 2 years. Doxsee et al. [51] correlated the
interlaminar shear strength with the moisture concentration at the plane of
failure in aramid/epoxy composites. In the research of Hu et al. [53] on
fibre/polylactide composites, after 24 hour aging under 70C in saturated
water vapour condition, the tensile strength of uncoated fibre/polylactide
composite specimen was 85.4% of the specimens without aging. After 72
hour aging, the tensile strength has badly deteriorated both for coated and
uncoated specimens, less than 30% of unaged specimens. Interlaminar
shear and flexural properties of FRP composites are generally more
sensitive to moisture effects than tensile properties, since tensile properties
are dominated by the fibres. But the glass and carbon fibre reinforcement
does not absorb moisture. Only the fibre-matrix interface offers a
preferential pathway for moisture ingression [45, 46]. For FRP bridge decks,
the mechanical degradation of FRP composites will result in reduction of
effective deck widths and the degree of composite action between decks
and supporting girders, as well as overall stiffness of FRP bridges.
18
Consequently, the residual strength and service life of bridges will be less
than expected.
Given the strong correlation between the rate of mechanical degradation
and moisture absorption, it is of immense importance to understand the
moisture diffusion process in polymers. The research [33] suggested that
moisture diffuses into composite materials by three different mechanisms: (1)
diffusion of water molecules inside the micro-gaps between polymer chains,
(2) capillary transport into the gaps and flaws at the interfaces between
fibres and polymer, and (3) transport by micro-cracks in the matrix, formed
during the compounding process. Understanding the whole diffusion
process by which moisture enters an FRP composite is critical to identify the
location of damage, analyse the mechanical degradation as well as predict
the residual strength and service life of FRP structures. Hence, it is
important to know that moisture concentration distribution throughout
sections of FRP composites as a function of time. As it is difficult to measure
moisture concentration distribution throughout structural sections by
experimental methods, gravimetric experiments [54-61] on the thin sections
of FRP material are usually employed to obtain the moisture diffusion
coefficients by recording the weight of absorbed water in a specimen as a
function of aging time. Post et al. [57] obtained the higher diffusion rates and
maximum moisture uptakes of a pultruded polyester/E-glass profile than
typically reported for this class of material, which may be attributable to a
larger microscopic void volume in the matrix resulting from rapid cure during
pultrusion. Pierron [54] proposed a novel method for the identification of 3D
moisture diffusion parameters on an epoxy resin reinforced by a glass fibre
cloth. An optimization solution was employed to get the moisture saturation
level only based on the slope of the initial linear part of the gravimetric curve,
since the saturation of thick FRP specimens usually lead to very long
conditioning times. For the case of Kevlar epoxy composite [60], the
moisture diffusion was two orders of magnitude more rapid in the composite
than in the base resin and the solubility was three to four times more than
that accommodated by the base resin. It was speculated that the rapid
diffusion in the composite was due to preferential diffusion of moisture along
the matrix/fibre interface. This conclusion was confirmed by the research of
Leman [61] on sugar palm fibre reinforced epoxy composites.
19
Chapter2
Generally, the moisture diffusion experiments for FRP composites are

limited to a fairly short time, normally no more than 5 years. But the
expected service life of infrastructures such as bridges is more than 50
years. Thus, the short-term experimental investigations are not sufficient to
estimate the long-term performance of FRP structures. To achieve this aim,
some accelerated experimental methods were developed by the researches
[30, 37, 62, 63], in which the temperature or atmospheric pressure of the
environmental aging conditions were raised beyond the normal service
conditions to accelerate the moisture diffusion and degradation process.
These accelerating experimental methods were confirmed to be useful and
time-effective to investigate the durability of FRP composites and adhesive
materials. But some researches indicated that the high aging temperatures
approaching the glass transition temperature of specimens would improve
the mechanical performance by post-cure or deteriorate the materials by
inducing thermal cracks, which do not occur in the real utilisation of FRP
composite structures. Another method for studying the long-term moisture
diffusion behaviour in FRP structures is the Finite Element (FE) Method.
Using the moisture diffusion coefficients determined by short-term
gravimetric experiments, the moisture diffusion process in real FRP
structures can be predicted by FE analysis [64]. The research [65]
investigated the moisture diffusion in an adhesively-bonded composite
connection under two environmental conditions (45C-85%RH and 90C97%RH) by FE analysis, parametrically studying the unidirectional and
multidirectional composites as well as two different fillet shapes, i.e., a
rectangular and a triangular fillet.
20
References:
[1] Bakis CE, Bank LC, Brown VL, Cosenza E, Davalos JF, Lesko JJ, et al. Fiberreinforced polymer composites for construction-state-of-the-art review. J Compos Constr.
2002;6(2):73-87.
[2] Awad ZK, Aravinthan T, Yan ZG, Gonzalez F. A review of optimization techniques
used in the design of fibre composite structures for civil engineering applications. Mater
Design. 2012;33:534-44.
[3] Knippers J, Pelke E, Gabler M, Berger D. Bridges with Glass Fibre-Reinforced
Polymer Decks: The Road Bridge in Friedberg, Germany Structural Engineering
International. 2010;20(4):400-4.
[4] Luke S, Canning L, Collins S, Knudsen E, Brown P, Taljsten B, et al. Advanced
composite bridge decking systemproject ASSET. Structural Engineering International.
2002;12(2):76-9.
[5] Reising RMW, Shahrooz BM, Hunt VJ, Neumann AR, Helmicki AJ, Hastak M. Close
look at construction issues and performance of four fiber-reinforced polymer composite
bridge decks. J Compos Constr. 2004;8(1):33-42.
[6] Hejll A, Taljsten B, Motavalli M. Large scale hybrid FRP composite girders for use in
bridge structures - theory, test and field application. Compos Part B-Eng. 2005;36(8):57385.
[7] Moses JP, Harries KA, Earls CJ, Yulismana W. Evaluation of effective width and
distribution factors for GFRP bridge decks supported on steel girdersl. J Bridge Eng.
2006;11(4):401-9.
[8] King L, Toutanji H, Vuddandam R. Load and resistance factor design of fiber
reinforced polymer composite bridge deck. Compos Part B-Eng. 2012;43(2):673-80.
[9] Zhou A, Keller T. Joining techniques for fiber reinforced polymer composite bridge
deck systems. Compos Struct. 2005;69(3):336-45.
[10] Hollaway LC. A review of the present and future utilisation of FRP composites in the
civil infrastructure with reference to their important in-service properties. Constr Build
Mater. 2010;24(12):2419-45.
[11] Cassity P, Richards D, Gillespie J. Compositely Acting FRP Deck and Girder System.
Structural Engineering International. 2002;2:71-5.
[12] Schollmayer M. Through-Thickness Performance of Adhesive Connections Between
FRP Bridge Decks and Steel Main Girders. Lausanne: Ecole Polytechnique Federale de
Lausanne; 2009.
[13] Park KT, Kim SH, Lee YH, Hwang YK. Degree of composite action verification of
bolted GFRP bridge deck-to-girder connection system. Compos Struct. 2006;72(3):393400.
[14] Keller T, Gurtler H. Composite action and adhesive bond between fiber-reinforced
polymer bridge decks and main girders. J Compos Constr. 2005;9(4):360-8.
[15] Keelor DC, Luo Y, Earls CJ, Yulismana W. Service load effective compression flange
width in fiber reinforced polymer deck systems acting compositely with steel stringers. J
Compos Constr. 2004;8(4):289-97.
[16] Gurtler HW. Composite action of FRP bridge decks adhesively bonded to steel main
girders Lausanne: Ecole Polytechnique Federale de Lausanne; 2004.
[17] da Silva LFM, das Neves PJC, Adams RD, Spelt JK. Analytical models of adhesively
bonded joints-Part I: Literature survey. Int J Adhes Adhes. 2009;29(3):319-30.
[18] da Silva LFM, das Neves PC, Adams RD, Wang A, Spelt JK. Analytical models of
adhesively bonded joints-Part II: Comparative study. Int J Adhes Adhes. 2009;29(3):33141.
21
Chapter2
[19] Diaz J, Romera L, Hernandez S, Baldomir A. Benchmarking of three-dimensional
finite element models of CFRP single-lap bonded joints. Int J Adhes Adhes.
2010;30(3):178-89.
[20] Wahab MMA, Crocombe AD, Beevers A, Ebtehaj K. Coupled stress-diffusion
analysis for durability study in adhesively bonded joints. Int J Adhes Adhes.
2002;22(1):61-73.
[21] Keller T, Vallee T. Adhesively bonded lap joints from pultruded GFRP profiles. Part I:
stress-strain analysis and failure modes. Compos Part B-Eng. 2005;36(4):331-40.
[22] Keller T, Vallee T. Adhesively bonded lap joints from pultruded GFRP profiles. Part
II: joint strength prediction. Compos Part B-Eng. 2005;36(4):341-50.
[23] Vallee T, Keller T. Adhesively bonded lap joints from pultruded GFRP profiles. Part
III: Effects of chamfers. Compos Part B-Eng. 2006;37(4-5):328-36.
[24] Vallee T, Correia JR, Keller T. Probabilistic strength prediction for double lap joints
composed of pultruded GFRP profiles part I: Experimental and numerical investigations.
Compos Sci Technol. 2006;66(13):1903-14.
composed of pultruded GFRP profiles - Part II: Strength prediction. Compos Sci Technol.
2006;66(13):1915-30.
[26] Ichikawa K, Shin Y, Sawa T. A three-dimensional finite-element stress analysis and
strength evaluation of stepped-lap adhesive joints subjected to static tensile loadings. Int J
Adhes Adhes. 2008;28(8):464-70.
[27] Goncalves JPM, de Moura MFSF, de Castro PMST. A three-dimensional finite
element model for stress analysis of adhesive joints. Int J Adhes Adhes. 2002;22(5):357-65.
[28] Ashcroft IA, Digby RP, Shaw SJ. A comparison of laboratory-conditioned and
naturally-weathered bonded joints. J Adhesion. 2001;75(2):175-201.
[29] Shao YX, Kouadi S. Durability of fiberglass composite sheet piles in water. J Compos
Constr. 2002;6(4):280-7.
[30] Bank LC, Gentry TR, Barkatt A. Accelerated Test Methods to Determine the LongTerm Behavior of Frp Composite Structures - Environmental-Effects. J Reinf Plast Comp.
1995;14(6):559-87.
[31] Nogueira P, Ramirez C, Torres A, Abad MJ, Cano J, Lopez J, et al. Effect of water
sorption on the structure and mechanical properties of an epoxy resin system. J Appl
Polym Sci. 2001;80(1):71-80.
[32] Earl JS, Shenoi RA. Hygrothermal ageing effects on FRP laminate and structural
foam materials. Compos Part a-Appl S. 2004;35(11):1237-47.
[33] Robert M, Roy R, Benmokrane B. Environmental Effects on Glass Fiber Reinforced
Polypropylene Thermoplastic Composite Laminate for Structural Applications. Polym
Composite. 2010;31(4):604-11.
[34] Karbhari VM. E-glass/vinylester composites in aqueous environments: Effects on
short-beam shear strength. J Compos Constr. 2004;8(2):148-56.
[35] Daly HB, Brahim HB, Hfaied N, Harchay M, Boukhili R. Investigation of water
absorption in pultruded composites containing fillers and low profile additives. Polym
Composite. 2007;28(3):355-64.
[36] Karbhari VM, Xian GJ. Hygrothermal effects on high V(F) pultruded unidirectional
carbon/epoxy composites: Moisture uptake. Compos Part B-Eng. 2009;40(1):41-9.
[37] Muliana A, Nair A, Khan KA, Wagner S. Characterization of thermo-mechanical and
long-term behaviors of multi-layered composite materials. Compos Sci Technol.
2006;66(15):2907-24.
[38] Levy RL, Fanter DL, Summers CJ. Spectroscopic Evidence for Mechanochemical
Effects of Moisture in Epoxy-Resins. J Appl Polym Sci. 1979;24(7):1643-64.
22
[39] Birger S, Moshonov A, Kenig S. The Effects of Thermal and Hygrothermal Aging on
the Failure Mechanisms of Graphite-Fabric Epoxy Composites Subjected to Flexural
Loading. Composites. 1989;20(4):341-8.
[40] Mikols WJ, Seferis JC, Apicella A, Nicolais L. Evaluation of Structural-Changes in
Epoxy Systems by Moisture Sorption-Desorption and Dynamic Mechanical Studies. Polym
Composite. 1982;3(3):118-24.
[41] Springer GS. Enivormental effects on composite materials, vol.1. Lancaster, PA:
Technomic Pubnishing company; 1981.
[44] Loh WK, Crocombe AD, Wahab MMA, Ashcroft IA. Modelling anomalous moisture
uptake, swelling and thermal characteristics of a rubber toughened epoxy adhesive. Int J
Adhes Adhes. 2005;25(1):1-12.
[45] Haque A, Mahmood S, Walker L, Jeelani S. Moisture and Temperature Induced
Degradation in Tensile Properties of Kevlar Graphite Epoxy Hybrid Composites. J Reinf
Plast Comp. 1991;10(2):132-45.
[46] Akay M, Mun SKA, Stanley A. Influence of moisture on the thermal and mechanical
properties of autoclaved and oven-cured Kevlar-49/epoxy laminates. Compos Sci Technol.
1997;57(5):565-71.
[47] Garcia K, Hayes MD, Verghese N, Lesko JJ. The effects of cycling moisture aging on
glass/vinyl ester composite system. Proceedings of 3rd International Conference on
Progress in Durability Analysis of Composite Systems1998. p. 173-9.
[48] Phifer SP. Hygrothermal evaluation of pultruded polymer composite laminatesExperimentation, analysis, and prediction. Blacksburg, VA: Virginia Tech; 2003.
[49] Phifer SP. Hygrothermal evaluation of pultruded polymer composite laminatesExperimentation, analysis, and prediction. Blacksburg, VA: Virginia Tech; 2003.
[50] Shen CH, Springer GS. Moisture absorption and desorption of composite materials,
environmental effects on composite materials: Technomic Publishing Company, Westport,
CT; 1981.
[51] Doxsee LE, Janssens W, Verpoest I, Demeester P. Strength of Aramid-Epoxy
Composites during Moisture Absorption. J Reinf Plast Comp. 1991;10(6):645-55.
[52] Arun KV, Basavarajappa S, Sherigara BS. Damage characterisation of glass/textile
fabric polymer hybrid composites in sea water environment. Mater Design.
2010;31(2):930-9.
[53] Hu RH, Sun MY, Lim JK. Moisture absorption, tensile strength and microstructure
evolution of short jute fiber/polylactide composite in hygrothermal environment. Mater
Design. 2010;31(7):3167-73.
[54] Pierron F, Poirette Y, Vautrin A. A novel procedure for identification of 3D moisture
diffusion parameters on thick composites: Theory, validation and experimental results. J
Compos Mater. 2002;36(19):2219-43.
[55] Bao LR, Yee AF. Moisture diffusion and hygrothermal aging in bismaleimide matrix
carbon fiber composites - Part I: uni-weave composites. Compos Sci Technol.
2002;62(16):2099-110.
carbon fiber composites: part II - woven and hybrid composites. Compos Sci Technol.
2002;62(16):2111-9.
23
Chapter2
[57] Post NL, Riebel F, Zhou A, Keller T, Case SW, Lesko JJ. Investigation of 3D
Moisture Diffusion Coefficients and Damage in a Pultruded E-glass/Polyester Structural
Composite. J Compos Mater. 2009;43(1):75-96.
[58] Barjastech E, Nutt SR. Moisture absorption of unidirectional hybrid composites
Compos Part a-Appl S. 2012;43:158-64.
[59] Katzman HA, Castaneda RM, Lee HS. Moisture diffusion in composite sandwich
structures. Compos Part a-Appl S. 2008;39(5):887-92.
[60] Aronhime MT, Neumann S, Marom G. The Anisotropic Diffusion of Water in Kevlar
Epoxy Composites. J Mater Sci. 1987;22(7):2435-46.
[61] Leman Z, Sapuan SM, Saifol AM, Maleque MA, Ahmad MMHM. Moisture
absorption behavior of sugar palm fiber reinforced epoxy composites. Mater Design.
2008;29(8):1666-70.
[62] Davies P, Evrard G. Accelerated ageing of polyurethanes for marine applications.
Polym Degrad Stabil. 2007;92(8):1455-64.
[63] Moulzakis DE, Zoga H, Galiotis C. Accelerated environmental ageing study of
polyester/glass fiber reinforced composites (GFRPCs). Compos Part B-Eng.
2008;39(3):467-75.
[64] Jiang X, Kolstein H, Bijlaard FSK. Moisture diffusion in glass-fiber-reinforced
polymer composite bridge under hot/wet environment. Compos Part B-Eng.
2013;45(1):407-16.
[65] Abdelwahab MM, Ashcroft IA, Crocombe AD, Shaw SJ. Diffusion of moisture in
adhesively bonded joints. J Adhesion. 2001;77(1):43-80.
24
Chapter 3
Moisture diffusion characteristics
of FRP composites*
3.1 Introduction
The subject of this chapter is to study the moisture diffusion process in FRP
composites (pultrusion and resin-infusion). By gravimetric experiments, the
moisture diffusion in FRP composites has been characterized under four
environmental conditions. Based on the analytical solution from the onedimensional and the three-dimensional moisture diffusion theory, the
diffusion coefficients were determined by the least-square curve fitting to the
experimental data. The FE models with the same dimensions as test
specimens were developed and validated against the experimental results.
By employing the FE model, the three dimensional diffusion coefficients
were also validated. This research provides a numerical technical basis for
coupling the moisture diffusion and mechanical analysis of FRP composites
to predict the residual strength of FRP structures exposed to hot/wet
environments.
3.2 Moisture diffusion theory

Most of the studies on moisture diffusion in FRP composites rely on the onedimensional Fickian process, the equation of which is expressed as [3, 4]:
c
2c
D 2
t
x
(3.1)
where c represents the moisture concentration, x the space coordinate

measured parallel to the diffusion, and D the moisture diffusion coefficient in
The content of this chapter is partially published in [1] Jiang X, Kolstein H, Bijlaard FSK. Moisture
diffusion in glass-fiber-reinforced polymer composite bridge under hot/wet environment. Compos
Part B-Eng. 2013;45(1):407-16. and [2] Jiang X, Kolstein H, Bijlaard FSK, Qiang X. Effects of
hygrothermal aging on glass-fibre reinforced polymer laminates and adhesive of FRP composite
bridge: Moisture diffusion characteristics Composites Part A: Applied Science and Manufacturing.
2014, 57: 4958.
25
Chapter3
the x direction. D is supposed to be independent of the spatial and temporal

coordinates. For a plate of infinite dimensions, the boundary conditions are:
c ci 0 x e
c c x 0; x e
t0
(3.2)
t0
where e is the plate thickness. The analytical solution [3], giving the
moisture concentration c at time t, is expressed as:
c
4 (1) n
Dt
1
exp 2 2 (2n 1) 2
c
n 0 (2n 1)
e
(3.3)
where c is the maximum equilibrium moisture concentration. The

expression is integrated, giving the moisture absorption content Mt as a
function of time, in Equation (3.4):
8
M t M 1 2

(2n 1)
n0
Dt
exp 2 2 (2n 1) 2
e
(3.4)
where M is the equilibrium amount of absorption. The theoretical Fickian

diffusion process is shown in Fig. 3.1.
Fig. 3.1. Fickian diffusion process.
For the initial linear part of the Fickian diffusion curve ( Dt / e 0.28 ), the
identification of the moisture diffusion coefficient D is performed as follows.
The total gravimetric curve (Fig. 3.1) is used to determine M. Then, D is
calculated from the linear part using two points at times t1 and t2.
e
D
4M
M1 M 2
t t
2
1
26
(3.5)
Chapter3MoisturediffusioncharacteristicsofFRPcomposites
In practice, this simple procedure is mostly used for the thin-section of an

FRP composite plate in the through-thickness direction, although the
hypothesis on which it relies, i.e., the fact that the plate is infinite in the
plane directions, is incorrect for the generally small specimens used in the
climate chambers.
For a thick plate of FRP composites, to satisfy the above assumption
infinite plate, the use of very large specimens is unavoidable to achieve
the required surface-to-thickness ratio. But these large specimens are
inconvenient to store in environmental chambers and weight measuring.
Thus, other approaches need to be developed to analytically depict threedimensional moisture diffusion process in the thick plate of FRP composites.
Firstly, due to the simplicity and mathematical tractability, the equivalent
diffusion coefficient method is usually used [3]. It assumes that the total
mass of moisture absorbed is equal to the total amount of moisture
absorbed from the six surfaces independently. According to this assumption,
the moisture uptake can be expressed as:
M t 4c
eb
D1 el D2 bl D3
(3.6)
where l and b are the length and width of the plate and D1, D2 and D3 are the
moisture diffusion coefficients along the direction of length, width and
thickness, respectively.
As c = mlbe, then Eq. 3.6 comes to be:
Mt
4m
e
t e
e
D1
D2 D3
l
b
(3.7)
By analogy to the infinite plate solution, for the three dimensional moisture
diffusion it is possible to define an equivalent moisture diffusion coefficient D
by:
e D1 e
D D3
l
D
b
3
D2
1
D3
(3.8)
Employing the gravimetric curve (Fig. 3.1) obtained from experiments, the
equivalent D can be addressed by Eq.3.5. Then three groups of FRP
27
Chapter3
specimens with different aspect ratios fully enable the identification of the
three dimensional diffusion coefficients: D1, D2 and D3, from Eq.3.8.
However, this approach has limits, which will be discussed hereafter in
further detail.
The other method to obtain the D1, D2 and D3 is depending on the full threedimensional moisture diffusion theory [3]. The three-dimensional Fickian
differential equation is written as:
c
2c
2c
2c
D1 2 D2
D3 2
t
x1
x2 2
x3
(3.9)
With the boundary conditions as:

c ci 0 x1 l ;0 x2 b ;0 x3 e
c c x1 0, x2 0, x3 0; x1 l , x2 b, x3 e
t0
t0
(3.10)
A closed-form solution to the above is given as:

ct ci
1 1 1
64
(2m 1) x1
1 3
exp(Qt ) cos
c ci
m 0 n 0 p 0 2m 1 2n 1 2 p 1
l
m
(2 p 1) x3
(2n 1) x2
cos
cos
b
e
(3.11)
With
2
2
2m 1 2
2n 1
2 p 1
Q D1
D2
D3

b
e
l
2
(3.12)
By integrating on the space variables, the moisture uptake content can be

expressed as:
8 3
M t M 1 2
Qt
exp(
)
2
2
2
m 0 n 0 p 0 (2m 1) (2n 1) (2 p 1)
(3.13)
It is noted that the three-dimensional moisture diffusion analytical

expression is rather complicated. There is no analogical equation like Eq.
3.5 to directly identify the moisture diffusion coefficients D1, D2 and D3 from
the experimental gravimetric curve (Fig. 3.1). A novel method for
identification of three-dimensional moisture diffusion coefficients on thick
28
FRP plates was developed by Pierron [5], which is based on an optimization

scheme. The idea is to build up an objective function q as:
q M t ti M i
(3.14)
where Mt(ti) is the moisture content calculated from Eq.3.13 at time ti and Mi
the moisture content experimentally obtained from the gravimetric curve at
time ti. Minimizing q with respect to D1, D2 and D3 enables the identification.
This process is realized by fitting the best Fickian least-square curve to the
experimental data points.
3.3 Experiments
3.3.1 Specimen preparation
Two typical FRP composites are selected for this research: pultrusion profile
and resin-infusion laminates. For the pultruded FRP composites, specimens
are cut from the ASSET FRP bridge deck element (Fig. 3.2) produced by
Fiberline Composites A/S [6].
a) Outer web
b) Inner web
c) Flange
Fig. 3.2. FRP ASSET bridge deck element and test specimens
These triangular shape profiles are manufactured by the pultrusion process,

and then bonded together to form the bridge deck. The lay-up consists of
longitudinal rovings, surrounded by a continuous strand mat and a surfacing
29
Chapter3
veil, as shown in Fig. 3.3. In the roving part, fibres are unidirectional oriented
in the direction of pultrusion. The surfacing veil is added outside the mat part
to protect against environmental attacks. The shapes consist of an average
62% E-glass fibres (volume fraction) embedded in an isophthalic polyester
matrix.
Fig. 3.3. Typical cross-section view of pultruded FRP composites [7]
It is obvious from Fig. 3.2 that specimens from different parts (inner web,
outer web and flange) of the ASSET bridge deck element are composed of
different volume fractions of roving and mat layer. It can be expected that
deviations of moisture diffusion properties can occur on different parts of the
cross-section. Thus, it is of great interest to conduct separate moisture
diffusion experiments on the inner web, outer web and flange parts. Nominal
dimensions of pultruded FRP specimens are listed in Table 3.1.
Table 3.1 Nominal dimension of FRP specimens
Pultrusion
profile
Innerweb
Outerweb
Flange
Resininfusion
Laminate
Square
Rectangular
Smallsquare
Length(mm)
Width(mm)
Thickness(mm)
100
100
100
9.80
7.80
15.6
2.80
2.80
2.80
Length(mm)
Width(mm)
Thickness(mm)
100
100
50
100
50
50
2.82
2.82
5.64
For resin-infusion FRP laminates, specimens are manufactured by resin

vacuum infusion (Infra Composite BV [8]) and then cut into specific
dimensions (see Fig. 3.4). The resin used is polyester. In order to obtain the
three-dimensional moisture diffusion coefficients D1, D2 and D3, specimens
30
were deliberately prepared with three different aspect ratios, as listed in

Table 1. The 2.82mm thickness specimens are made up of three plies of
standard 0.94mm EQX1200, which is a glass-fibre reinforced polymer
composites (54% glass content by weight). Layup configuration of EQX1200
is listed in Table 3.2. The 5.64mm thickness specimen is made up of six
plies of standard 0.94mm EQX1200.
a) Square
b) Rectangular
c) Small square
Fig. 3.4. FRP laminate specimens

Table 3.2. Properties of FRP laminates (supplied by manufacturer [9])
Product
name
EQX1200
Total
weight
(g/m2)
1193
Weightuniformity(g/m2)
Yarnroving
Knit
yarn
0o
+45o
90o
45o
283 300 300 300
10
3.3.2 Gravimetric test process

Generally, the moisture diffusion process in FRP composite materials is
investigated by gravimetric tests [10-15]. The whole experiment process
follows the test code ASTM D5229/D5229M-92 [16]. For both pultruded FRP
composites and resin-infusion FRP laminates, four replicates are tested in
each specific aging condition. Four environmental aging conditions are
selected with regard to normal service environments of bridge decks, which
are 20C-50% RH (relative humidity), 20C-water, 40C-96% RH and 40Cwater. The 40C-96% RH and 40C-water conditions are considered as
typical hot/wet environments for the application of an FRP composite bridge.
The temperature and relative humidity of each aging condition is kept
constant during the whole process of testing. The 20C-50% RH condition is
obtained by putting specimens in a climate room, with constant temperature
31
Chapter3
and relative humidity at 20C and 50% RH. The 20C-water condition is
obtained by putting specimens in a water filled glass container, which is also
kept in the same climate room. The 40C-96% RH is provided in a climate
chamber, of which the temperature and relative humidity are controlled to be
40C and 96% RH. 40C-water condition is obtained by putting specimens
in a water filled glass container, which is kept in the same chamber at stable
temperature 40C. Prior to putting specimens into the environmental
conditions, all the specimens are dried in an oven at 40C and the weight of
specimens is periodically checked until no changes in weight occur. This
status is assumed to be the original stage of the whole moisture diffusion
process. For tracking the change of weight, after each specific time interval,
each specimen is removed from the environmental conditioning chamber,
weighed quickly using a precise balance with the accuracy of 0.00001g and
then returned to the chamber. For the water immersed aging conditions,
before measuring the self-weight, the residual liquid water trapped on the
surface of the specimens must be wiped away. The procedure is repeated
until the samples reach a saturation level. The moisture uptake content (Mt)
absorbed by each specimen is calculated according to its weight before
exposure (w0) and after exposure (wt) as follows:
Mt
wt w0
w0
(3.15)
3.4 Experimental results and discussion

3.4.1 Pultruded FRP composites
After a 24-day aging period in water and vapour environmental conditions,
the moisture uptake curves of the pultruded FRP specimens are presented
in Fig. 3.5, where Mt is plotted vs. t to show the initial linear diffusion curve.
All the detailed test data can be found in the Stevin II lab report.
32
3.5
Mt (%)
3.0
2.5
2.0
20 , 50% RH
40 , 96% RH
20 , water
40 , water
1.5
1.0
0.5
0.0
0
200
400
600
800
1/2
1000
1200
1400
1600
1/2
t (s )
a) Inner web
3.5
Mt (%)
3.0
2.5
2.0
20 , 50% RH
40 , 96% RH
20 , water
40 , water
1.5
1.0
0.5
0.0
0
200
400
600
800
1/2
1000
1200
1400
1600
1/2
t (s )
b) Outer web
3.0
Mt (%)
2.5
2.0
1.5
20 , 50% RH
40 , 96% RH
20 , water
40 , water
1.0
0.5
0.0
0
200
400
600
800
1/2
1000
1200
1400
1600
1/2
t (s )
c) Flange
Fig. 3.5. Moisture absorption process of pultruded FRP specimens from different parts
33
Chapter3
For the initial part of all curves, moisture uptakes of the pultruded FRP
specimens increase abruptly. All the FRP specimens aged in the specific
environmental condition reached the moisture saturation level within two
days. To analytically model this moisture diffusion process, the onedimensional Fickian model is employed due to its simplicity and
mathematical tractability [3, 4]. As aforementioned, it is assumed that the
FRP plate is infinite and the moisture only diffuses in the through-thickness
direction. But for this study, only small scale specimens could be prepared
due to the limited geometry of the ASSET bridge deck element. It is
generally noted that in continuous fibre composites, the bulk diffusion
properties are orthotropic due to the material heterogeneity difference along
and transverse to the fibre direction. Results of Aronhime et al.s
experiments [17] showed that the moisture diffusion rates along the fibre
direction were much higher than those transverse to the fibre direction, and
they were in a different order of magnitude. In this study, the pultruded FRP
specimens are cut in the vertical direction of pultrusion, which means the
cutting surfaces (XY plane in Fig. 3.6) of FRP specimens are vertical to the
fibre direction of the roving part. Thus, the amount of moisture content
diffused through the cutting surfaces of FRP specimens is much larger than
that through the edge surfaces of FRP specimens. Moreover, the edge
surfaces of FRP specimens are protected from environmental attacks by the
surfacing veil. So, the moisture absorption from the edge surfaces of
specimens can be neglected. Therefore, in this study, the moisture diffusion
process in FRP specimens can be assumed to be essentially onedimensional through the thickness direction (Z direction in Fig. 3.6) of FRP
specimens.
Z
X
Y
Fig. 3.6. Illustration of coordinates in the pultruded FRP specimen
34
For the pultruded FRP composites, the initial moisture diffusion process is
too fast to get enough data points to comply the assumption ( Dz t / e 0.28 )
of the Fickian diffusion theory. In order to get the accurate value of Fickian
diffusion coefficient D, the analytical model (Eq. 3.4) is employed to fit the
experimental data points by the least-square method. After the analytical
curve fitting, the Fickian diffusion coefficients and maximum moisture
contents at equilibrium are obtained and summarized in Table 3.3, and
analytical curves are plotted in Fig. 3.5.
Table 3.3 Moisture diffusion coefficients of pultruded FRP composites
Specimen
Aging condition
D (10-5 mm2/s)
M (%)
Inner web
20C , 50% RH
40C , 96% RH
20C , water
40C , water
3.164
6.105
4.717
6.022
0.141
0.403
3.306
3.272
Outer web
20C , 50% RH
40C , 96% RH
20C , water
40C , water
3.804
7.365
9.278
24.92
0.149
0.380
3.461
3.489
Flange
20C , 50% RH
40C , 96% RH
20C , water
40C , water
4.425
5.750
8.847
29.38
0.137
0.361
2.915
2.751
Comparing with the gravimetric experiments in the vapour aging conditions,

experiments conducted in water result in higher moisture saturation levels,
around 3% of the initial self-weight of FRP materials. Specimens immersed
in water at temperatures of 20C and 40C develop almost identical
moisture uptakes to each other. Only for the initial linear phase, the moisture
diffusion into FRP specimens under 40C is faster than that of 20C. The
moisture saturation levels for both aging temperatures are close to each
other. Results (Table 3.3) show that the rates of moisture uptake are more
sensitive to temperature and the moisture saturation levels of FRP material
are only dominated by humidity of aging conditions. For the vapour
environmental conditions, the comparison of the mass variation exhibited by
both FRP profiles depicts significant differences, with the moisture uptake
for 20C -50% RH aging condition being considerably slower and lower in
moisture equilibrium than that exhibited by the 40C -96% RH aging
35
Chapter3
condition, which is reasonable according to the aforementioned conclusion.

Comparing the results of different part specimens, it is found that there is no
considerable variability in the value of maximum equilibrium moisture
contents, where the flange absorbs about 2.8% in water and the inner and
outer webs absorb about 3.3% in water. Only the moisture diffusion rates of
inner web specimens in water conditions are considerably lower than the
other two parts.
It is noted that analytical solutions for moisture diffusion prediction are
limited to simple geometry. It is not feasible to predict the moisture diffusion
process in the complex shaped FRP composite profiles, such as the ASSET
bridge element (Fig. 3.2). Thus, it is of great interest to develop the Finite
Element (FE) method to simulate the moisture diffusion process in FRP
composites. As we know, the moisture diffusion process is a transient
phenomenon, which relates to the moisture concentration distribution
throughout the profiles as a function of exposure time. So the transient field
analysis is the most appropriate form for FE based diffusion analysis. The
commercial FE analysis package ABAQUS 6.8 is employed for the moisture
diffusion analysis. In ABAQUS 6.8, the governing equations for mass
diffusion are an extension of Fickian equations: they allow for non-uniform
solubility of the diffusing substance in the base material and for mass
diffusion driven by gradients of temperature and pressure. The basic
solution variable (used as the degree of freedom at the nodes of the mesh)
is the normalized concentration, which is the mass concentration of the
diffusing material divided by its solubility in the base material. Therefore,
when the mesh includes dissimilar materials that share nodes, the
normalized concentration is continuous across the interface between the
different materials [18]. The mass diffusion element DC3D8 is used in FE
analysis, which is a 8-node linear brick element. The FE model is built with
the same dimensions as the test specimens (see Fig. 3.7).
36
a) Inner web
b) Outer web
c) Flange
Fig. 3.7. FE models of pultruded FRP specimens
Herein, only the moisture diffusion in the 40C-water aging condition is

simulated by FE analysis, depending on which the coupled hygro-thermal
mechanical analysis can be conducted. The moisture diffusion coefficients
obtained from tests (see Table 3.3) are used as input parameters for
material properties in the FE model. The typical Fickian diffusion is
considered for the pultruded FRP composites. All of these parameters are
assumed to be constant during the whole diffusion analysis. Small
integration time steps are inevitably needed to achieve the convergence and
enhance the accuracy, especially at the initial stage of FE diffusion analysis.
Integration time steps are automatically controlled. The moisture
concentration of six surfaces of FE models are set to be fully saturated as
the boundary conditions to active the kinetics of the diffusion process.
Because the specimens are fully immersed in water, the variation in the
moisture boundary concentration with time is assumed to be negligible. The
results of mass concentration at integration points (CONC) are output to
a .fil result file. The specific post-processing subroutine is developed and
performed by FORTRAN to get the moisture absorption curves as a function
of time. The subroutine realizes: 1) calculating the moisture content of each
element (at the time t) by using the mass concentration (conc) of the
element multiplied by this element volume; 2) summarising the moisture
content (Mt) of each element to obtain the amount of moisture content in the
FE model.
37
Chapter3
Fig. 3.8 shows the comparison of analytical solutions and FE analysis

results together with experimental data for moisture diffusion in pultruded
FRP composite specimens under the 40C -water condition.
3.5
3.0
1.5
Mt (%)
2.5
1.0
2.0
1.5
0.5
1.0
test data
FE analysis
analytical model
0.5
0.0
0
0.0
0
200
400
600
800
1/2
1000
1200
1400
1/2
t (s )
Fig. 3.8. a) inner web FRP composites

4.0
3.5
Mt (%)
3.0
2.5
2.0
1.5
1.0
test data
FE analysis
analytical model
0.5
0.0
0
200
400
600
800
1/2
1000
1200
1/2
t (s )
Fig. 3.8. b) outer web FRP composites
38
1400
3.0
2.5
Mt (%)
2.0
1.5
1.0
test data
FE analysis
analytical model
0.5
0.0
0
200
400
600
800
1/2
1000
1200
1400
1/2
t (s )
c) flange FRP composites

Fig. 3.8. Comparison of FE results, analytical models and experimental data
It is manifest from Fig. 3.8 a) - c) that analytical curves and FE results agree
very well with each other. Only for the initial diffusion stage, deviation from
two curves takes place, which is due to boundary conditions of the FE
model. At time=0, the moisture concentration of edge surfaces of FE models
is fully saturated, which means the total moisture content before running the
FE analysis is not zero. As the diffusion analysis is going on, the FE results
gradually approach analytical curves and finally they superpose each other.
Thus, the FE model is validated by analytical curves and experimental data.
It can be used in future work to predict the moisture diffusion in the
pultruded FRP composites with complex geometry and longer aging times.
3.4.2 Resin-infusion FRP laminates

After about 250-day aging in water and vapour environmental conditions,
the average moisture uptake curves of FRP laminate specimens are
presented in Fig. 3.9. All the detailed test data can be found in the Stevin II
lab report.
39
Chapter3
1.0
0.8
Mt (%)
0.6
20 C, 50% RH
o
40 C, 96% RH
o
20 C, water
o
40 C, water
0.4
0.2
0.0
0
1000
2000
3000
1/2
4000
5000
6000
1/2
t (S )
a) 100mm100mm2.82mm
1.0
Mt (%)
0.8
0.6
0.4
20 C, 50% RH
o
40 C, 96% RH
o
20 C, water
o
40 C, water
0.2
0.0
0
1000
2000
3000
1/2
4000
5000
6000
1/2
t (S )
b) 100mm50mm2.82mm
1.0
0.8
Mt (%)
0.6
0.4
o
20 C, 50% RH
o
40 C, 96% RH
o
20 C, water
o
40 C, water
0.2
0.0
0
1000
2000
3000
1/2
4000
5000
6000
1/2
t (S )
c) 50mm50mm5.64mm
Fig. 3.9. Moisture absorption process of FRP laminate specimens
40
Similar to the pultruded FRP composites, in the initial part of all curves,
moisture uptakes increase linearly and then regularly slow down until
reaching the moisture saturation level. Moisture absorption content varies
significantly in four aging conditions.
Table 3.4 Moisture diffusion coefficients of FRP laminates based on one dimensional
diffusion theory
Specimen
Square
Rectangular
Small square
D(10-7mm2/s)
1.347
1.963
1.438
3.040
1.542
2.380
1.676
3.602
2.830
5.198
3.101
7.330
Aging condition
20C, 50%RH
40C, 96%RH
20C, water
40C, water
20C, 50%RH
40C, 96%RH
20C, water
40C, water
20C, 50%RH
40C, 96%RH
20C, water
40C, water
M (%)
0.148
0.694
0.535
0.890
0.140
0.706
0.519
0.889
0.087
0.632
0.471
0.754
As listed in Table 3.4, the specimens aged in the 40C-water condition

result in the highest moisture saturation level (M), which is more than six
times of that in the 20C-50% aging condition. It is confirmed once more that
the hot/wet environment can accelerate the moisture-induced deterioration
process of FRP materials. As compared to the maximum moisture uptakes
of pultruded FRP composites addressed above, the M of FRP laminates
are considerably lower. Moreover, for the last part of the moisture diffusion
process of the Square and Rectangular specimens under the 40C-water
aging condition, the curves drop from the moisture equilibrium content,
which means the mechanism of mass loss took place. The same
phenomenon is also found in other researches [19, 20], when the
environmental temperature approach the glass transition temperature (Tg)
of FRP composites. It indicates that the material has experienced some
form of physical and/or chemical degradation: hydrolysis of the matrix, chain
breakage, creation of small molecules and extraction (leaching) out of these
molecules from the composite. From Table 3.4, it can be also found that the
moisture saturation levels of Square and Rectangular specimens are much
close to each other for the four aging conditions, but deviate a bit from that
of Small square specimens. This can be attributed to the manufacture and
curing process of FRP laminates, which could influence the microstructure
41
Chapter3
of FRP composite materials. In this research, the Square and Rectangular

specimens are cut from the same FRP plate, while the Small square
specimens are cut from another thicker plate. This could be the reason to
explain the deviation of moisture saturation levels. Moreover, mass loss can
also influence the saturation level, since they could occur to different extents
in the different sample types.
Firstly, for the purpose of convenience, the experimental results are fitted to
one dimensional moisture diffusion theory (Eq.3.4) by using the least-square
method. The equivalent diffusion rates D and maximum moisture contents
(M) at equilibrium of different aspect ratio specimens in four aging
conditions are obtained. For comparison, the analytical curves are drawn in
Fig. 3.9. Good agreement is achieved between experimental results and
analytical fitting curves. Subsequently, by employing Eq.3.8, three
dimensional Fickian diffusion coefficients D1, D2 and D3 are obtained and
summarized in Table 3.5.
Table 3.5 Moisture diffusion coefficients of FRP laminates in three directions from Eq. 3.8
Aging condition
20C 50%RH
40C 96%RH
20C water
40C water
D1 (10-6mm2/s)
1.121
2.829
1.071
3.501
D2 (10-6mm2/s)
0.786
2.547
1.132
3.021
D3 (10-6mm2/s)
0.097
0.123
0.102
0.202
To validate the above three dimensional Fickian diffusion coefficients, FE

analysis are used herein. The whole FE modeling process is the same as
mentioned above for the pultruded FRP composites. The typical threedimensional Fickian diffusion analysis is considered for resin-infusion FRP
laminates. The D1, D2 and D3 (from Table 3.5) are used as input parameters
for the FE model. Subsequently, the FE results are obtained and shown in
Fig. 3.10 (indicated as 1D FE diffusion curve), together with the analytical
model and experimental data for FRP laminates with three aspect ratios
under the 40Cwater environmental condition. It is manifest that the
moisture uptake of 1D FE analysis is relatively slower than that of
experimental results. It can be explained that, due to the assumption of the
equivalent diffusion coefficient method (Eq. 3.6), the amount of moisture
absorption is equal to the one dimensional moisture uptake from six
surfaces independently. However, in tests, specimens absorb moisture from
42
six surfaces homogeneously, which means in the corner between two

adjacent surfaces, moisture uptake from one direction can interact with that
from the other direction. This interaction mechanism can disturb the typical
one-dimensional moisture diffusion process and correspondingly decrease
the diffusion rate. Eq. 3.6 does not take this mechanism into account, so the
three directional moisture diffusion coefficients D1, D2 and D3 derived from
Eq. 3.8 are lower than the real values. Thus, the conclusion can be drawn
that only at the initial moisture diffusion stage, the assumption of Eq. 3.6 can
be satisfied by neglecting the interaction mechanism on moisture absorption
at the corner area between adjacent surfaces.
1.0
0.9
0.8
0.7
Mt (%)
0.6
0.5
0.4
3D FE
1D FE
test data
1D analytical model
3D analytical model
0.3
0.2
0.1
0.0
0
1000
2000
3000
4000
1/2
5000
6000
1/2
t (S )
Fig. 3.10. a) Square

1.0
0.9
0.8
0.7
Mt (%)
0.6
0.5
0.4
0.3
3D FE
1D FE
test data
1D analytical model
3D analytical model
0.2
0.1
0.0
0
1000
2000
3000
1/2
4000
1/2
t (S )
Fig. 3.10. b) Rectangular

43
5000
6000
Chapter3
0.8
0.7
Mt (%)
0.6
0.5
0.4
0.3
3D FE
1D FE
test data
1D analytical model
3D analytical model
0.2
0.1
0.0
0
1000
2000
3000
4000
1/2
5000
6000
7000
800
1/2
t (S )
c) Small square
Fig. 3.10. Comparison of FE results, analytical models and experimental data of FRP
laminate specimens aged at the 40Cwater condition
In order to obtain the accurate three directional moisture diffusion

coefficients, the full three-dimensional diffusion theory needs to be
employed. Eq. 3.13 indicates the moisture uptake as the function of time
based on three dimensional coordinates. As mentioned before, an
optimization routine is developed to minimize the least squared error,
between test data and the equation prediction for all the specimens with
different aspect ratios. As the results of this optimized solution, D1, D2 and
D3 of FRP laminates in four aging conditions are listed in Table 3.6, of which
the values differ significantly from those in Table 3.5. Fig. 3.10 shows the FE
diffusion curves based on the three-dimensional diffusion theory, indicated
as 3D FE. Perfect agreement is achieved between FE results, analytical
solution and test data. Thus, the FE analysis suggests that Eq. 3.8 is not
reliable for determination of moisture diffusion coefficients of anisotropic
polymeric materials. The three-dimensional moisture diffusion theory is
inevitably needed. Whats more, development of FE model is confirmed to
be useful to simulate and validate the moisture diffusion process. With these
moisture diffusion coefficients (D1, D2 , D3 and M), the moisture diffusion
process in FRP laminates with other complex geometries can be easily
modeled by FE analysis, even for thicker sections or much longer
environmental aging durations.
44
Table 3.6 Moisture diffusion coefficients of FRP laminates in three directions from Eq.
3.13
Aging condition
20C 50%RH
40C 96%RH
20C water
40C water
D1 (10-6mm2/s)
0.869
9.243
0.862
9.607
D2 (10-6mm2/s)
1.081
9.403
0.927
9.631
D3 (10-6mm2/s)
0.125
0.187
0.182
0.318
From Table 3.6 it can be found that, for each aging condition, the moisture
diffusion rate through the thickness of FRP laminates (D3) is lower than D1
and D2, especially for the aging conditions of 40C temperature. Meanwhile,
the D1 and D2 are close to each other. This implies that the moisture is
inclined to diffuse along the fibre direction, and with a view to the layup
configuration of FRP laminates (Table 3.2), it is reasonable that D1 and D2
achieve approximately the identical value. Furthermore, comparing the
moisture diffusion coefficients between different aging conditions, the
maximum values of D1 and D2 are obtained almost simultaneously in both
40C-96%RH and 40C-water conditions, which indicates that the
environmental temperature dominates the moisture diffusion rates along the
fibre direction. For the moisture diffusion transverse to the fibre direction, the
influence of temperature is not that significant, since the absolute value of
D3 is too low to be sensitive to the various aging conditions. However, no
matter moisture diffusion rates (D1, D2 and D3) or saturated content (M), the
40C-water condition always leads to the highest values, in comparison with
the other three aging conditions. It confirms that the hot/wet condition should
be considered to be a hostile service environment for the utilization of FRP
structures.
3.5 Conclusions
By gravimetric experiments, the moisture diffusion characteristics of
pultruded FRP composites and resin-infusion FRP laminates under four
environments were studied. For the pultruded FRP specimens, the moisture
saturation levels were reached under four aging conditions within only two
days. The one-dimensional Fickian diffusion analytical model was employed
to fit the experimental data by the least-square method. Subsequently, the
moisture diffusion coefficients were determined. In comparison with the
vapour aging conditions, FRP composites immersed in water result in much
45
Chapter3
higher moisture uptake levels, around 3% of the initial self-weight. The

experimental data and fitted analytical curves confirm that elevated
temperatures could accelerate the moisture diffusion process, and the
maximum moisture equilibrium contents are dominated by the humidity of
aging environments. No significant variations on moisture diffusion
characteristics of pultruded FRP specimens from different parts of the FRP
deck profile exist. The FE modelling for simulating the moisture diffusion
process in pultruded FRP composites were developed, and validated
against the gravimetric experimental data and analytical solution.
For resin-infusion FRP laminate specimens, more aging time was needed
for specimens to achieve the moisture saturated levels. In comparison with
the pultruded FRP composites, the FRP laminates obtained much lower
values for both moisture diffusion rate and saturated content, which
indicates good corrosion resistance to environmental effects. Consequently,
the surfacing veil is not necessary to be added to lay-ups of FRP laminates
for protection against environmental attacks. The experimental results
confirmed once more that the hot/wet environment can accelerate the
moisture-induced deterioration process of FRP materials. Furthermore, the
mass loss mechanism was observed for Square and Rectangular
specimens under 40C-water aging condition. For identification of threedimensional moisture diffusion coefficients of anisotropic FRP materials, the
equivalent diffusion coefficient method has proven to be unreliable by
employing the three-dimensional FE diffusion analysis. The threedimensional moisture diffusion theory is inevitably needed for developing the
optimization scheme. The research described in this chapter can be
attributable for the coupled hygro-thermal stress-strain analysis on FRP
structures, which develops a step towards the prediction of long-term
performance and life-time estimation of FRP-steel composite bridges.
46
References:
[1] Jiang X, Kolstein H, Bijlaard FSK. Moisture diffusion in glass-fiber-reinforced
polymer composite bridge under hot/wet environment. Compos Part B-Eng.
2013;45(1):407-16.
[2] Jiang X, Kolstein H, Bijlaard FSK, Qiang X. Effects of hygrothermal aging on glassfibre reinforced polymer laminates and adhesive of FRP composite bridge: Moisture
diffusion characteristics Composites Part A: Applied Science and Manufacturing. 2014, 57:
49-58.
[4] Crank J. The mathematics of diffusion. 2nd editon. Oxford: Clarendon Press; 1975.
[5] Pierron F, Poirette Y, Vautrin A. A novel procedure for identification of 3D moisture
diffusion parameters on thick composites: Theory, validation and experimental results. J
Compos Mater. 2002;36(19):2219-43.
[6] http://www.fiberline.dk/.
[7] Zhou A, Keller T. Joining techniques for fiber reinforced polymer composite bridge
deck systems. Compos Struct. 2005;69(3):336-45.
[8] http://www.infracomposites.com.
[9]http://pdf.nauticexpo.com/pdf/vetrotex-ocv-reinforcements/quadraxial-fabrics/2789113674-_2.html.
[10] Shao YX, Kouadi S. Durability of fiberglass composite sheet piles in water. J Compos
Constr. 2002;6(4):280-7.
[11] Post NL, Riebel F, Zhou A, Keller T, Case SW, Lesko JJ. Investigation of 3D
Moisture Diffusion Coefficients and Damage in a Pultruded E-glass/Polyester Structural
Composite. J Compos Mater. 2009;43(1):75-96.
[12] Robert M, Roy R, Benmokrane B. Environmental Effects on Glass Fiber Reinforced
Polypropylene Thermoplastic Composite Laminate for Structural Applications. Polym
Composite. 2010;31(4):604-11.
[13] Karbhari VM, Xian GJ. Hygrothermal effects on high V(F) pultruded unidirectional
carbon/epoxy composites: Moisture uptake. Compos Part B-Eng. 2009;40(1):41-9.
carbon fiber composites: part II - woven and hybrid composites. Compos Sci Technol.
2002;62(16):2111-9.
carbon fiber composites - Part I: uni-weave composites. Compos Sci Technol.
2002;62(16):2099-110.
[16] D5229/D5229M-92 A. Standard Test Method for Moisture Absorption Properties and
Equilibrium Conditioning of Polymer Matrix Composite Materials. PA, United States:
ASTM International; 1992.
[17] Aronhime MT, Neumann S, Marom G. The Anisotropic Diffusion of Water in Kevlar
Epoxy Composites. J Mater Sci. 1987;22(7):2435-46.
[18] Abaqus Analysis User's Manual, 6.8 version.
[19] Daly HB, Brahim HB, Hfaied N, Harchay M, Boukhili R. Investigation of water
absorption in pultruded composites containing fillers and low profile additives. Polym
Composite. 2007;28(3):355-64.
[20] Liao K, Schultheisz CR, Hunston DL, Brinson LC. Long-term durability of fiberreinforced polymer-matrix composite materials for infrastructure applications: A review. J
Adv Mater-Covina. 1998;30(4):3-40.
47
Chapter3
48
Chapter 4
Mechanical degradation of FRP
laminates under hot/wet
environment
4.1 Introduction
In the present chapter, the influence of moisture and temperature on the
mechanical properties of Glass-Fiber-Reinforced polymer (GFRP) laminates
is investigated. In the use of GFRP bridge decks, the GFRP laminates are
mainly loaded by the wheel load in the through-thickness direction.
Therefore, it is of great importance that the flexural and interlaminar shear
properties are studied. For this study, three point bending tests were
performed with different support spans to study the flexural and shear
properties of FRP laminates by varying the thickness to support span ratio.
The mechanical properties of dry (0%Mt/M), moisture unsaturated
(30%Mt/M and 50%Mt/M) and moisture saturated specimens (100%Mt/M)
under both 20C and 40C test temperatures are compared. One cycle of
moisture absorption-desorption process is also included in this study to
investigate how the residual damage induced by the moisture diffusion
degrades the mechanical properties of FRP laminates. Furthermore, to
better understand the environment-dependent mechanical performance of
FRP laminates, a coupled hygro-mechanical FE model was developed by
writing a specific postprogressing subroutine to work together with the FEM
software ABAQUS, and subsequently validated by the flexural test results.
Based on this coupled hygro-mechanical FE model, an inverse parameter
identification approach to short-beam shear tests was developed and then
employed to determine the environment-dependent interlaminar shear
modulus of FRP laminates by minimizing the difference between the
numerically predicted material response and experimental measured data.
49
Chapter4
4.2 F
Flexural property
4.2.1 Experim
ment
To in
nvestigate the flexu
ural prope
erty of FR
RP laminates, three
e point be
ending
tests are emplo
oyed. The
e whole te
est procedure follow
ws the stan
ndard testt code
ASTM
M D790-10 [1]. FRP
P laminate
d in this chapter
c
arre the sam
me as
es studied
those
e used in Chapter 3,
3 which a
are manuffactured by
b resin va
acuum inffusion
(Infra
a Composite BV) an
nd then cu
ut into spe
ecific dime
ensions w ith the principal
axis parallel to
o the warrp directio
on of the woven ro
oving ( ass shown in Fig.
4mm thick
k specime
en is com
mposed of six laye
ers of standard
4.1). The 5.64
mm EQX1
1200. The
e layup co
onfiguratio
on of eac
ch piece o
of the standard
0.94m
0.94m
mm EQX1
1200 is illustrated i n Table 4.1.
4 The mechanica
m
al propertties of
FRP laminatess supplied
d by the m
manufactu
urer are shown
s
in Table 4.2
2. The
nal length
h and widtth of the specimen
ns are selected to b
be 150mm
m and
nomin
20mm
m respectively. Acc
cording to the stand
dard test code
c
AST
TM D790-1
10 [1],
the sp
pecimen length
l
sha
all be suffficient to allow
a
for overhangin
ng on eac
ch end
of at least 10%
% of the support
s
sp
pan. The specimen
n width sh
hall not ex
xceed
ourth of th
he supporrt span forr specimens greater than 3.2
2mm in de
epth.
one fo
w
weft
warp
5.64mm
150m
mm
Fig. 4.1. FRP laminnate specim

men for flexu
ural tests
50
Chapter4MechanicaldegradationofFRPlaminatesunderhot/wetenvironment
Table 4.1. FRP laminate properties of EQX1200 (supplied by manufacturer)
Product
name
EQX1200
Total
weight
(g/m2)
1193
0o(warp)
283
Yarnroving
+45o
90o(Weft)
45o
300
300
300
Knityarn
10
Table 4.2. Mechanical properties of FRP laminates (supplied by manufacturer)

Property
Tensile(ISO5274)
Compression(ISO8515)
Flexural(ISO14.125)
Meanvalue
Warp
Weft
Warp
Weft
Warp
Weft
Strength
331MPa
314MPa
220MPa
200MPa
473MPa
433MPa
Modulus
18GPa
17GPa
14GPa
14GPa
13GPa
11GPa
The numbering of the specimens is given in Table 4.3, with regard to the
moisture uptake content, test temperature, absorption/desorption process
and replicated number. Two test temperatures 20C and 40C are selected
for the three point bending tests, which are controlled by the climate
chamber with the tolerance of 2C, as shown in Fig. 4.2.
Fig. 4.2. Climate chamber

51
Chapter4
Table 4.3. FRP laminate specimens for flexural tests
Specimenidentification
Set1
Set2
Set3
Set4
Set5
Set6
Set7
F0%20C
F0%40C
F30%20C
F30%40C
F50%20C
F50%40C
F100%20C
F100%40C
F50%20CD
F50%40CD
F30%20CD
F30%40CD
F0%20CD
F0%40CD
Test
Mt/M
flexural
flexural
flexural
flexural
flexural
flexural
flexural
flexural
flexural
flexural
flexural
flexural
flexural
flexural
0
0
30%
30%
50%
50%
100%
100%
50%
50%
30%
30%
0
0
Test
After
Numderof
temperature desorption specimens
20C
no
5
40C
no
5
20C
no
5
40C
no
5
20C
no
5
40C
no
5
20C
no
5
40C
no
5
20C
yes
5
40C
yes
5
20C
yes
5
40C
yes
5
20C
yes
5
40C
yes
5
As listed in Table 4.3, the test in each condition is repeated five times to
investigate the spread of test results. The hydrothermal aging condition is
40C-water, which is supposed to be a severe hot/wet condition for FRP
laminates, as discussed in Chapter 3. In total, 70 pieces of specimens are
prepared. During the hydrothermal aging process, all the specimens are
immersed in the water at the temperature of 40C, except for the F-0-20C
and F-0-40C specimens, which are the as-received reference specimens
(Set-1 in Table 4.3). The as-received specimens are stored in the laboratory
environment. The moisture content of them is very low, and thus can be
ignored. As illustrated in Table 4.3, the Set-2 specimens (F-30%-20C and
F-30%-40C) are tested at 30% relative moisture uptake content. The Set-3
specimens (F-50%-20C and F-50%-40C) are tested at 50% relative
moisture uptake content. The Set-4 specimens (F-50%-20C and F-50%40C) are tested at the moisture saturation level (100% relative moisture
uptake content). Until this point in time, the above test process is considered
as the moisture absorption process. Then, the remaining specimens are all
taken out of the hydrothermal aging environment, and put into an oven at a
temperature of 42C to dry them, which is considered as the moisture
desorption process. In this way, the Set-5 specimens (F-50%-20C-D and F50%-40C-D) are tested at 50% relative moisture uptake content after a
certain time of moisture desorption. Subsequently, the Set-6 specimens (F30%-20C-D and F-30%-40C-D) are tested at 30% relative moisture uptake
52
content after the moisture desorption. The final Set-7 specimens (F-0-20CD and F-0-40C-D) are the fully dry specimens after one cycle of moisture
absorption-desorption process. Herein, the symbol D indicates the
moisture desorption.
As shown in Fig. 4.3, the whole three point bending test set-up is put into
the chamber. According to ASTM D790-10 [1], the support span-to-depth
ratio is 16:1. Thus, the support span is proposed to be 90.24mm, but it
varies between different groups of specimens, since the value of the support
span is exactly calculated based on the average real thickness of each
group of specimens. The radii of the loading nose and supports are 5.0
0.1mm. The whole loading process employs the strain rate of
0.01mm/mm/min. Correspondingly, the rate of crosshead motion is 2.4
mm/min, which is calculated as follows [1]:
R ZL2 / 6d
(4.1)
R = rate of crosshead motion, mm/min,

L = support span, mm,
d = depth of FRP beam tested, mm,
Z = rate of straining of the outer fiber, mm/mm/min.
The specimen is deflected until the load drops to 30% of the maximum load
or until a maximum displacement of mid-span reaches 10mm, whichever
occurs first (as illustrated in Fig. 4.4). The experimental data is recorded per
second. To track the moisture absorption process in FRP laminate
specimens, gravimetric tests are also conducted, the test procedure of
which can be found in section 3.3.2.
53
Chapter4
Fig. 4.3. Flexural test device

Load
Load
FMax
FMax
70%FMax
10mm
Displacement
a) Drop to 30% of the maximum load
Displacement
b) Maximum displacement of 10mm
Fig. 4.4. Termination rule of the flexural test
4.2.2 Experimental results and discussion

Fig. 4.5 shows the moisture absorption process of FRP laminate flexural test
specimens immersed in water of 40C, compared with the FE moisture
diffusion analysis. Moisture content (calculated from Eq. 3.15) is established
as the function of the square root of time. It can be found that the moisture
saturation level is about 0.77% (M), which is in line with the gravimetric
experimental results obtained in Chapter 3. The specimens developed a
54
similar moisture diffusion curve as that simulated by the FE analysis, which

confirms the accuracy of the FE moisture diffusion model.
1 month
1 day
3 months
1 year
6 months
110
0.8
100
0.7
90
Mt (%)
70
0.5
60
0.4
50
0.3
40
Mt/M (%)
80
0.6
30
0.2
FE analysis
test data
0.1
20
10
0
0.0
0
1000
2000
3000
1/2
4000
5000
6000
1/2
t (S )
Fig. 4.5. Comparison of moisture uptake curve between test results and FE analysis on
FRP specimens for flexural tests
The typical failure mode of specimens under flexural tests is shown in Fig.
4.6, where rupture occurs in the outer surface of the test specimen. All the
detailed test data can be found in the Stevin II lab report. In order to obtain
the E-modulus of FRP laminates, the stress and strain at the midspan of
FRP specimens are calculated as follows.
outer surface failure
Fig. 4.6. Failure mode of the flexural test specimen
55
Chapter4
According to ASTM D790-10 [1], the flexural stress in the outer surface of
the specimen at midpoint is calculated by means of the following equation:
3PL / 2bd 2
(4.2)
where:
= stress in the outer fibers at midpoint, MPa,
P = load at the midspan on the load-deflection curve, N,
b = width of FRP beam tested, mm.
The flexural strain, that nominal fractional change in the length of an
element of the outer surface of the test specimen at midspan, is calculated
for any deflection using Eq. 4.3:
6 Dmax d / L2
(4.3)
where:
= strain in the outer fibers at midpoint, MPa,
Dmax = maximum deflection of the center of the beam, mm
The stress-strain curves are presented in Fig. 4.7. To make the comparison
more clear, the curve of only one specimen of each test condition is
presented, which is selected visually as the average curve of the five
specimens.
56
stress [MPa]
400
F-0%-20C
F-0%-40C
F-30%-20C
F-30%-40C
F-50%-20C
F-50%-40C
F-100%-20C
F-100%-40C
F-50%-20C-D
F-50%-40C-D
F-30%-20C-D
F-30%-40C-D
F-0%-20C-D
F-0%-40C-D
300
200
100
0
0.000
0.005
0.010
0.015
0.020
0.025
0.030
0.035
0.040
strain [mm/mm]
Fig. 4.7. Stress-strain curves of FRP specimens under flexural tests
The E-modulus is represented by the chord modulus, which is calculated

from two discrete points on the load-deflection curve, using Eq. 4.4 [1]:
E ( 2 1 ) /( 2 1 )
(4.4)
1 , 1 and 2 , 2 are the flexural stress and strain selected at two points of
stress-strain curves (Fig. 4.7) in the linear and stable range.

The flexural strength is the maximum flexural stress sustained by the test
specimen during the flexural test. The environment-dependent flexural
properties (including E-modulus and strength) are shown in Fig. 4.8 and
Table 4.4. The R-square value for each curve is also present in Fig. 4.8,
which indicates how well the curve fits test data points.
57
Chapter4
18000
2
R =0.6784
16000
E-modulus (MPa)
14000
R =0.6329
12000
10000
8000
6000
absorption
absorption-desorption
4000
2000
0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Mt/M
Fig. 4.8. a) E-modulus, 20C

450
400
R =0.8402
300
250
R =0.9570
200
150
absorption
100
50
0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
Mt/M
Flexural strength (MPa)
350
Fig. 4.8. b) Strength, 20C
58
0.7
0.8
0.9
1.0
18000
16000
2
R =0.6326
E-modulus (MPa)
14000
12000
R =0.0632
10000
8000
6000
absorption
4000
2000
0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Mt/M
c) E-modulus, 40C
450
400
Flexural strength (MPa)
350
300
2
R =0.8760
250
200
R =0.9403
150
absorption
100
50
0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Mt/M
d) Strength, 40C
Fig. 4.8 Environment-dependent flexural property degradation of FRP laminates
59
Chapter4
Table 4.4 Flexural property degradation of FRP laminates

Specimen
identification
F0%20C
F0%40C
F30%20C
F30%40C
F50%20C
F50%40C
F100%20C
F100%40C
F50%20CD
F50%40CD
F30%20CD
F30%40CD
F0%20CD
F0%40CD
Emodulus*
(MPa)
16609
15409
15873
13874
15038
13870
14022
12780
14408
13019
15059
12336
16333
13095
Standard
Flexural
Standard
Deviation(MPa) strength*(MPa) Deviation(MPa)
386
411
6.89
852
375
8.49
867
299
8.88
500
249
11.42
710
269
17.03
535
252
17.57
514
265.
13.74
538
214
6.08
551
260
15.46
832
239
17.85
740
324
37.04
166
277
9.85
204
410
10.80
698
314
13.92
* mean value of five specimens
To be simplified, the predictive equation for the E-modulus degradation as

the function of moisture content is curve fitted by the linear interpolant
function, while for the flexural strength the exponential function is used. All
the curve fitting processes are conducted by using the MATLAB R2011b
software, employing the least square method. The obtained predictive
equation is as follows:
E-modulus, 20C, absorption process:
E= -2725Mt/M+16609
(4.5)
E-modulus, 20C, absorption-desorption process:

E= -2428Mt/M+16333
(4.6)
Flexural strength, 20C, absorption process:

(
S 103
Mt
1.1)
M
257
(4.7)
Flexural strength, 20C, absorption-desorption process:
S 15.5
Mt
1.87)
M
60
242
(4.8)
E-modulus, 40C, absorption process:

E= -2795Mt/M+15409
(4.9)
E-modulus, 40C, absorption-desorption process:

E= -628Mt/M+13095
(4.10)
Flexural strength, 40C, absorption process:

(
S 123
Mt
1.04)
M
221
(4.11)
Flexural strength, 40C, absorption-desorption process:

(
S 3.23
Mt
4.25)
M
166
(4.12)
All the predictive curves are illustrated in Fig. 4.8 for comparison with the
experimental results.
Considering firstly the specimens tested at 20C (Fig. 4.8a)), the E-modulus
of FRP laminates is decreasing gradually as the moisture content increases
from fully dry to fully saturated. The E-modulus of the moisture saturated
specimen is 14022 MPa (as shown in Table 4.4), which is 15.6% lower than
that of the unconditioned dry specimen. For the specimens with the moisture
content of 30% and 50% of the saturation level, the loss of E-modulus is
4.4% and 9.5% respectively. With regard to specimens in the moisture
desorption process, the E-modulus does not decrease significantly as
compared to the specimens at the same moisture uptake level. Accordingly,
the slight loss of E-modulus is 1.7%, 5.1% and 4.2% at the moisture uptake
level of 0%, 30% and 50% respectively. In terms of flexural strength (see
Fig. 4.8b)), there is a general exacerbation of decreasing between the fully
dry specimens and 30% moisture content specimens, regardless of the
moisture absorption/desorption process. More than 20% loss of flexural
strength is evident. After this, as the moisture uptake content increases, the
flexural strength of specimens is slightly decreasing, until reaching 265MPa
of the moisture fully saturated specimens. In the end, the total drop of
flexural strength is 35.4% of the fully dry specimens. Similar to the E61
Chapter4
modulus degradation, the difference of flexural strength between absorption

process and desorption process is very limited.
With regard to the environment-dependent flexural properties of FRP
laminates at 40C, the E-modulus of FRP laminates is regularly decreasing
from 15409 MPa (dry) to 12780 MPa (moisture fully saturated), and then
slightly increases to 13095 MPa after fully drying in the moisture desorption
process. Different from that under 20C, there is a significant unrecoverable
loss (15.0%) of the E-modulus for the dry specimens tested at 40C.
Accordingly, the loss of 11.1% and 6.1% of the E-modulus is evident for the
specimens with 30% and 50% moisture content respectively in the
desorption process. As to the flexural strength of specimens tested under
40C, a rapid decrease is observed at a moisture content of about 30% of
moisture saturation level, and then the flexural strength slightly decreases to
214 MPa of the moisture fully saturated specimens, which is the lowest
value among the whole series of tests. It is 42.9% lower than that of the
unconditioned dry specimens (375 MPa, as listed in Table 4.4) tested at
40C, and 47.9% lower than that of the unconditioned dry specimens (410
MPa) tested at 20C. This most severe loss of flexural strength indicates
that the combination of moisture and temperature effects can significantly
influence the mechanical properties of FRP laminates. Comparing the
unconditioned dry specimens with the dry specimens after the moisture
desorption, a decrease of 16.4% of the flexural strength is observed for the
specimens tested at 40C. However, for the specimens with a moisture
content of 30% and 50%, the loss of flexural strength is not obvious. Even a
slight increase is observed for the specimens with 30% moisture content in
the moisture desorption process.
In conclusion, the hot and wet environment sincerely degrades the flexural
properties of FRP laminates, and in turn influences the durability of FRP
composite structures. As to the moisture effects, as stated in the researches
[2, 3], the degradation of the mechanical properties of FRP composite
materials was due to water plasticization during the moisture uptake process
and the disruption of hydrogen bonds between the molecular chains in the
polymer. Furthermore, the fibre/matrix debonding, matrix cracking due to the
moisture/thermal cycle and ultimately fiber breaking would also contribute.
62
With regard to the temperature effects, when the test temperature was
approaching the Tg (glass transition temperature) of FRP composite
materials, the mechanical performance such as E-modulus, strength and
fatigue resistance significantly decreased [4, 5]. The recommended working
temperature for FRP composite structures should be at least 20C lower
than the Tg of FRP composite materials. Furthermore, the researches [5, 6]
confirmed that a decrease of Tg was evident when the moisture uptake
content increased in the FRP composite materials. Meanwhile, as already
proven in Chapter 3, the high temperature can speed up the moisture
diffusion process. Thus, the interaction between moisture and temperature
effects accelerates the environmental degradation process on the FRP
composite material, which in turn explains why the combination of moisture
and temperature effects seriously deteriorates the mechanical properties of
FRP materials.
4.2.3 Coupled hygro-mechanical FE analysis

Generally, the environmental degradation experiments (mainly concerning
moisture and temperature effects) for FRP composite materials are limited
to a fairly short time, normally no more than 5 years. But the expected
service life of infrastructures such as bridges exceeds 50 years. Thus, the
short-term experimental investigations are not sufficient to estimate the
long-term performance of FRP structures. To achieve this aim, some
accelerated experimental methods were developed by the researches [710], in which the temperature or atmospheric pressure of the environmental
aging conditions were raised beyond the normal service conditions to
accelerate the moisture diffusion and degradation process. These
accelerating experimental methods were confirmed to be useful and timeeffective to investigate the durability of FRP composites and adhesive
materials. But researches indicated that the high aging temperatures
approaching glass transition temperature of specimens would improve the
mechanical performance by post-cure or deteriorate the materials by
inducing thermal cracks, which do not occur in the real utilization of FRP
composite structures. Another method for studying the environmental
degradation of mechanical behaviour of FRP materials and structures is the
predictive FE modeling [11]. The first step is modeling moisture transport
63
Chapter4
through FRP structures in order to determine the moisture concentration

distribution through the cross-sections as a function of time. The material
parameters required for the transient diffusion FE analysis are the diffusion
coefficient and the solubility coefficient, which can be obtained from the
short term gravimetric experiments, as stated in Chapter 3. And then, based
on the obtained moisture concentration distribution, the environmentdependent mechanical behaviour of FRP structures can be investigated
using the coupled hygro-mechanical FE analysis. The input moisturedependent material properties of FRP composites are obtained by the
material level tests (such as flexural test, tensile test and short-beam shear
test).
To develop the coupled hygro-mechanical FE modeling method, in this
section, the FE model of the FRP flexural test specimen is built (see Fig.
4.9). The moisture diffusion process in the FRP specimen is modeled by the
transient diffusion FE analysis (thoroughly described in Chapter 3) and
validated previously by comparing with the gravimetric experimental results,
as shown in Fig. 4.5. From the moisture diffusion analysis, the moisture
concentration distribution across the FRP specimen section is obtained as a
function of time, which can be read into the stress analysis as a predefined
field variable. The predictive equations (Eq. 4.5, 4.6, 4.9 and 4.10) of
environment-dependent material properties serve as the input of fielddependent material properties of FRP composites, which is obtained by the
flexural tests. Hence, the E-modulus of each element is determined by the
local moisture concentration distribution. Thus, the coupled hygromechanical FE modeling is realized on the flexural test specimen.
Subsequently, it can be employed to simulate other material level tests
(such as short-beam shear tests) and FRP structures.
3
1
2
Fig. 4.9. FE model of the flexural test specimen
To validate this coupled hygro-mechanical FE modeling method, two groups

of FRP specimens are selected: F-50%-20C and F-30%-40C-D. Firstly, the
64
moisture concentration distribution in these two specimens is shown in Fig.

4.10 and Fig. 4.11. Based on this two predefined fields, the coupled hygromechanical FE analysis is conducted. The FE analysis is linear-elastic. The
element used is C3D8R. In total 2000N load is applied at mid-span of the
specimen.
Fig. 4.10. Nominal moisture concentration distribution across the mid-plane of the FRP
specimen with 30% moisture uptake content (time= 24 hours)
Fig. 4.11. Nominal moisture concentration distribution across the mid-plane of the FRP
specimen with 50% moisture uptake content (time= 229 hours)
Comparison of experimental and FE results on the load-deflection curve of

F-50%-20C specimens and F-30%-40C-D specimens are illustrated in Fig.
4.12 and Fig. 4.13 respectively. Good agreement is obtained in the linear
stage, which means the environment-dependent stiffness of FRP specimens
with different moisture uptake contents can be predicted by the coupled
hygro-mechanical FE model. Hence, the accuracy of FE model is proven.
65
Chapter4
1400
1200
Load [N]
1000
800
600
F-50%-20C-01
F-50%-20C-02
F-50%-20C-03
F-50%-20C-04
FEM
400
200
0
0
Deflection [mm]
Fig. 4.12. Comparison of experimental and FE results on the load-deflection curve of F50%-20C specimens
1400
Load [N]
1200
1000
800
600
F-30%-40C-D-01
F-30%-40C-D-02
F-30%-40C-D-03
F-30%-40C-D-04
F-30%-40C-D-05
FEM
400
200
0
0
Deflection [mm]
Fig. 4.13. Comparison of experimental and FE results on the load-deflection curve of F30%-40C-D specimens
4.3 Interlaminar shear property

4.3.1 Experiment
To investigate the interlaminar shear property of FRP laminates, three point
bending tests of short-beam FRP specimens are employed. The whole test
66
Chapter4
4MechanicaldegradattionofFRPlaminatesunderhot//wetenviro
onment
proce
edure follo
ows the standard
s
test code
e ASTM D2344/D2
2344M-00
0 [12].
FRP laminatess studied in this se
ection are the same as thosse discuss
sed in
on 4.2.1. As show
wn in Fig
g. 4.14, th
he nominal thickne
ess, width
h and
Sectio
length
h of the FRP shorrt-beam sspecimen are 5.64mm, 12m
mm and 40mm.
4
Numb
bering of specimen
ns is show
wn in Table 4.5, with regard tto the mo
oisture
uptakke conten
nt, test temperatture, abs
sorption/de
esorption process
s and
repliccated num
mber. The numberin
ng method
d of short-beam sh
hear specimens
is sim
milar to that of the flexural te
est specim
mens. On
nly the firsst charactter F
(repre
esenting the flexurral tests) is chang
ged to S, which representts the
short--beam shear tests.. The who
ole aging condition process is the sam
me as
that o
of the flexural tests. During t he aging time, the moisture uptake co
ontent
of eacch specim
men is reco
orded by using the gravimetrric test me
ethod.
12mm
m
5.64mm
40mm
Fig. 4.14.
4
FRP laaminate short beam speecimen
Tablle 4.5. FRP laminate sppecimens fo
or the short-b
beam shearr tests
Sp
pecimeniden
ntification
Set1
Set2
Set3
Set4
Set5
Set6
Set7
S0
0%20C
S0
0%40C
30%20C
S3
S3
30%40C
S5
50%20C
S5
50%40C
S1
100%20C
S1
100%40C
S5
50%20CD
S5
50%40CD
S3
30%20CD
S3
30%40CD
S0
0%20CD
S0
0%40CD
Test
Mt/M
shear
shear
shear
shear
shear
shear
shear
shear
shear
shear
shear
shear
shear
shear
0
0
30%
30%
50%
50%
100%
100%
50%
50%
30%
30%
0
0
mderof
Test
Aftter
Num
temperatture desorrption speccimens
20C
n o
5
40C
n o
5
20C
n o
5
40C
n o
5
20C
n o
5
40C
n o
5
20C
n o
5
40C
n o
5
20C
yees
5
40C
yees
5
5
20C
yees
40C
yees
5
20C
yees
5
40C
yees
5
The sshort-beam
m shear te
est device
e is shown
n in Fig. 4.15.
4
Acco
ording to ASTM
A
D234
44/D2344M
M-00 [12],, the loadiing span length-to-s
specimen thickness
s ratio
is 4. Conseque
ently, the support sspan is prroposed to be 22.5
56mm. It varies
v
ng differen
nt groups of specim
mens, sinc
ce the valu
ue of the ssupport sp
pan is
amon
exacttly calcula
ated base
ed on the
e average real thickness of each gro
oup of
67
Chapter4
specimens. The diameter of the loading nose and supports is 6.00mm and
3.00mm, respectively. The speed of testing is set at a rate of crosshead
movement of 1.0mm/min. The specimen is deflected until the load drops to
30% of the maximum load or until a maximum displacement of mid-span
reaches 4mm, whichever occurs first (see Fig. 4.16). The experimental data
is recorded per second.
Fig. 4.15. Short-beam shear test device

Load
Load
FMax
FMax
70%FMax
4mm
Displacement
Displacement
a) Drop to 30% of the maximum load

b) Maximum displacement of 4mm
Fig. 4.16. Termination rule of the short-beam shear tests
68
4.3.2 Experimental results and discussion

Fig. 4.17 shows the moisture absorption process of FRP short-beam
specimens immersed in water of 40C. Moisture content (calculated from
Eq. 3.15) is drawn as the function of square root of time. It can be found that
the moisture saturation level is about 0.72%. Good agreement is obtained
between experimental results and FE simulation.
Fig. 4.17. Comparison of moisture uptake curve between test results and FE analysis on
FRP specimens for short-beam shear tests
The typical failure mode of short-beam shear specimens is shown in Fig.

4.18, which is the interlaminar failure through the thickness of FRP
laminates.
69
Chapter4
interlaminar failure
Fig. 4.18 The failure mode of the short-beam shear test specimen
35
30
25
R =0.8917
20
15
2
R =0.9180
10
absorption
5
0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
Mt/M
short-beam shear strength (MPa)
40
Fig. 4.19. a) 20C
70
0.7
0.8
0.9
1.0
short-beam shear strength (MPa)
40
35
30
25
2
R =0.8805
20
15
R =0.9371
10
absorption
5
0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Mt/M
b) 40C
Fig. 4.19. Degradation on the short-beam shear strength of FRP laminates
Fig. 4.19 shows the mechanical degradation of short-beam shear strength of

FRP laminates as a function of moisture uptake content at 20C and 40C
respectively. The R-square value for each curve is also present in Fig. 4.19,
which indicates how well the curve fits test data points. The short-beam
shear strength can be calculated as follows [12]:
F sbs 0.75
Pm
bh
(4.13)
where:
Fsbs = short-beam strength, MPa,
Pm = maximum load observed during the test, N,
b = measured specimen width, mm,
h = measured specimen thickness, mm.
The predictive equations for the short-beam shear strength degradation as

the function of moisture content is curve fitted by the exponential function,
using the least square method. The obtained predictive equation is as
follows:
71
Chapter4
Shear strength, 20C, absorption process:
S 5.5
Mt
1.76)
M
11
(4.14)
Shear strength, 20C, absorption-desorption process:
S 34
Mt
0.517)
M
14.4
(4.15)
12
(4.16)
Shear strength, 40C, absorption process:
S 9.6
Mt
1.28)
M
Shear strength, 40C, absorption-desorption process:
S 46.8
Mt
0.45)
M
13.6
(4.17)
All the predictive curves are illustrated in Fig. 4.19 for comparison with the
experimental results.
The interlaminar shear modulus of the FRP laminates cannot be determined
experimentally via the short-beam three point bending. It is determined by
the FE analysis using the inverse parameter identification approach, which
is introduced in further detail hereafter.
As shown in Fig. 4.19a), in the moisture absorption process, the short-beam
shear strength is quasi-linearly decreasing from the fully dry specimens to
the specimens with about 75% moisture content of the saturated level.
Then, the test data point is distributed stably until reaching the moisture fully
saturated condition (100%M). As listed in Table 4.6, the short-beam shear
strength of the moisture fully saturated specimens is 15 MPa, which is
53.1% lower than that of the unconditioned dry specimens (32 MPa).
Furthermore, in the moisture absorption process, the test data points are
distributed more dispersively, since for the small scale short-beam
specimens, the moisture uptake process deviates significantly. Thus, under
the same water aging time, the moisture uptake content of the individual
specimen varies in a certain range. With regard to the moisture desorption
72
process, from fully saturated to dry, the short-beam shear strength is slightly
increasing, ending at 21 MPa. It is 34.4% lower than that of the
unconditioned dry specimens. This means that one cycle of moisture
absorption-desorption process degrades the shear strength of FRP
laminates by 34.4%.
Fig. 4.19b) presents the same tendency of degradation of the short-beam
shear strength of FRP laminates at 40C. As listed in Table 4.6, the higher
temperature (40C) only slightly deteriorates the shear strength of FRP
specimens, which implies that the influence of temperature is not that
significant as the influence of moisture.
Table 4.6 Short-beam shear strength degradation of FRP laminates
Specimen
Identification
S0%20C
S0%40C
S30%20C
S30%40C
S50%20C
S50%40C
S100%20C
S100%40C
S50%20CD
S50%40CD
S30%20CD
S30%40CD
S0%20CD
S0%40CD
ShearStrength*
(MPa)
32
31
26
25
19
16
15
14
15
13
16
15
21
19
Standard
Deviation
2.74
1.34
1.70
1.89
0.59
2.10
0.24
0.41
1.08
0.41
0.755
0.21
0.35
0.36
* mean value of five specimens
4.3.3 Determining the interlaminar shear modulus of FRP

laminates
Currently, standard test methods exist mostly for studies of the in-plane
normal and shear modulus, and strength parameters of FRP composite
materials [13]. However, the test method for obtaining the interlaminar shear
modulus is limited. Failure always occurred through a combination of shear
and transverse tension, indicating that a pure shear failure mode was not
achievable in the test. It is, therefore, imperative that robust methodologies
for determining the interlaminar material properties of FRP composite
materials need to be developed. In this section, an inverse parameter
identification approach for determining the interlaminar shear modulus G13
73
Chapter4
(G23) is employed, which is realized by minimizing the difference between an

experimentally measured and numerically predicted material response by
varying the interlaminar shear modulus of FE model. It has been confirmed
by the sensitivity FE analysis of Chans research [14], that the short-beam
three-point bend test (rather than the standard Iosipesu test and off-axis
tensile test) is sensitive to changes in the interlaminar shear modulus while
remaining relatively insensitive to changes in the other unknown material
properties. Hence, using the short-beam three-point bending test to study
the interlaminar shear modulus is the most suitable.
To determine the environment-dependent interlaminar shear modulus of
FRP composites, the coupled hygro-mechanical FE modeling is employed
herein, which is already well developed and has been validated by the
flexural tests in section 4.2.3. For instance, at the test temperature of 20C
and in the moisture absorption process, the flexural modulus (E11 and E22,
as indicated in Fig. 4.8a)) of FRP laminates with a nominal moisture content
of 0% to 100% can be predicted using Eq. (4.6). It is used as the fielddependent input values for material properties of the FE model. Other
material properties are used according to Table 4.7, which is supplied by the
manufacturer. Depending on the sensitivity analysis, these data cannot
significantly influence the determination of the interlaminar shear modulus of
FRP laminates.
Table 4.7. Mechanical properties of materials for FE model
Property
FRP laminates
Elastic Modulus 33 (Nmm)
11000
Poisson Ratio 12
0.33
Poisson Ratio 23
0.3
Poisson Ratio 13
0.18
Shear Modulus 12 (Nmm)
6986
1
2
Fig. 4.20. FE model of the short-beam specimen

74
Load
According to the test code ASTM D790-10 [1], as illustrated in Fig. 4.21, the
initial non-linear stage of test results is an artifact caused by a takeup of
slack and alignment or seating of the specimens, which does not represent
the properties of the material. In order to obtain correct values of material
properties, this curve must be offset to the corrected zero point (point B in
Fig. 4.21). For each test, the initial non-linear regions are different from each
other. To make easy comparisons, all the experimental curves are offset
from B to A, to make the extension line of the linear CD region exactly
through the zero point of coordinates. Original test results can be found in
the Stevin II lab report.
C
A B
displacement
Fig. 4.21. Offset of the experimental load-displacement curve
It is assumed that the degradation of interlaminar shear modulus follows a

linear relationship with the nominal moisture content, the same as obtained
for the flexural modulus of FRP laminates. Thus, the interlaminar shear
modulus of FRP specimens with the 0% moisture content (S-0%-20C) is
firstly determined by fitting the FE load-deflection curve to the test results,
as shown in Fig. 4.22a). Accordingly, the shear modulus G13 (G23) is
numerically determined as 1200MPa. In the same way, the shear modulus
G13 (G23) of FRP specimens with the 100% moisture content (S-100%-20C)
is determined as 800MPa (Fig. 4.22b)). Subsequently, the predictive
75
Chapter4
equation for the interlaminar shear modulus of FRP laminates at the test
temperature of 20C and in the moisture absorption process is as follows:
G23= G13= -400Mt/M+1200
(4.18)
3500
3000
Load [N]
2500
2000
1500
S-0%-20C-1
S-0%-20C-2
S-0%-20C-3
S-0%-20C-4
S-0%-20C-5
FEM
1000
500
0
0.0
0.2
0.4
0.6
0.8
1.0
Deflection [mm]
a) S-0%-20C
1400
Load [N]
1200
1000
800
S-100%-20C-1
S-100%-20C-2
S-100%-20C-3
S-100%-20C-4
S-100%-20C-5
FEM
600
400
200
0
0.0
0.2
0.4
0.6
0.8
1.0
Deflection [mm]
b) S-100%-20C
Fig. 4.22. Comparison of load-deflection curves between FE analysis and test results of S0%-20C specimens and S-100%-20C specimens
76
To validate Eq. 4.18, the other two exposure time intervals (30% Mt/M and
50% Mt/M ) are employed. From the moisture diffusion FE analysis, the
moisture distributions across the mid-plane of the FRP specimens are
shown in Fig. 4.23 and Fig. 4.24, which are used as the input field for the
coupled stress analysis. The field-dependent shear modulus is input as
presented by Eq. 4.18.
Fig. 4.23 Moisture concentration distribution across the mid-plane of the FRP specimen
with 30% moisture uptake content (time= 26 hours)
Fig. 4.24 Moisture concentration distribution across the mid-plane of the FRP specimen
with 50% moisture uptake content (time= 107 hours)
Comparison between FE results and test data of S-30%-20C specimens

and S-50%-20C specimens is shown in Fig. 4.25a) and Fig. 4.25b)
respectively. Good agreements on slopes of load-displacement curves are
achieved for these two groups of specimens, which prove that the predictive
equation 4.18 is accurate enough to simulate the stiffness of FRP
specimens with other moisture contents.
77
Chapter4
2500
Load [N]
2000
1500
S-30%-20C-1
S-30%-20C-2
S-30%-20C-3
S-30%-20C-4
S-30%-20C-5
FEM
1000
500
0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Deflection [mm]
a) S-30%-20C
2500
Load [N]
2000
1500
S-50%-20C-1
S-50%-20C-2
S-50%-20C-3
S-50%-20C-4
S-50%-20C-5
FEM
1000
500
0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Deflection [mm]
b) S-50%-20C
Fig. 4.25. Comparison of load-deflection curves between FE analysis and test results of S30%-20C specimens and S-50%-20C specimens
For the FRP specimens tested at 20C and in the moisture desorption
process, the same inverse parameter identification method is employed to
detemine the environment-dependent interlaminar shear modulus. The
predictive equation is developed and validated by the middle two exposure
time intervals, as follows:
78
G23= G13= -50Mt/M+850
(4.19)
2000
Load [N]
1500
1000
S-0%-20C-D-1
S-0%-20C-D-2
S-0%-20C-D-3
S-0%-20C-D-4
S-0%-20C-D-5
B2
500
0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Deflection [mm]
Fig. 4.26. a) S-0%-20C-D
Load [N]
1500
1000
S-30%-20C-D-1
S-30%-20C-D-2
S-30%-20C-D-3
S-30%-20C-D-4
S-30%-20C-D-5
FEM
500
0
0.0
0.2
0.4
0.6
Deflection [mm]
Fig. 4.26. b) S-30%-20C-D
79
0.8
1.0
Chapter4
1400
Load [N]
1200
1000
800
600
S-50%-20C-D-1
S-50%-20C-D-2
S-50%-20C-D-3
S-50%-20C-D-4
S-50%-20C-D-5
FEM
400
200
0
0.0
0.2
0.4
0.6
0.8
Deflection [mm]
c) S-50%-20C-D
Fig. 4.26. Comparison of load-deflection curves between FE analysis and test results of
specimens tested at 20C and in the moisture desorption process
For the FRP specimens tested at 40C and in the moisture absorption
process, the predictive equation is as follows:
G23= G13= -450Mt/M+1050
(4.20)
3000
Load [N]
2500
2000
1500
S-0%-40C-1
S-0%-40C-2
S-0%-40C-3
S-0%-40C-4
S-0%-40C-5
FEM
1000
500
0
0.0
0.2
0.4
0.6
0.8
Deflection [mm]
Fig. 4.27. a) S-0%-40C
80
1.0
1.2
3000
Load [N]
2500
2000
1500
S-30%-40C-1
S-30%-40C-2
S-30%-40C-3
S-30%-40C-4
S-30%-40C-5
FEM
1000
500
0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Deflection [mm]
Fig. 4.27. b) S-30%-40C

2000
Load [N]
1500
1000
S-50%-40C-1
S-50%-40C-2
S-50%-40C-3
S-50%-40C-4
S-50%-40C-5
FEM
500
0
0.0
0.2
0.4
0.6
0.8
Deflection [mm]
Fig. 4.27. c) S-50%-40C
81
1.0
1.2
Chapter4
1400
Load [N]
1200
1000
800
600
S-100%-40C-1
S-100%-40C-2
S-100%-40C-3
S-100%-40C-4
S-100%-40C-5
FEM
400
200
0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Deflection [mm]
d) S-100%-40C
specimens tested at 40C and in the moisture absorption process
For the FRP specimens tested at 40C and in the moisture desorption
process, the predictive equation is as follows:
G23= G13= -250Mt/M+850
(4.21)
1400
Load [N]
1200
1000
800
600
S-0%-40C-D-1
S-0%-40C-D-2
S-0%-40C-D-3
S-0%-40C-D-4
S-0%-40C-D-5
FEM
400
200
0
0.0
0.2
0.4
0.6
0.8
Deflection [mm]
Fig. 4.28. a) S-0%-40C-D
82
1.0
1.2
1400
1200
Load [N]
1000
800
600
S-30%-40C-D-1
S-30%-40C-D-2
S-30%-40C-D-3
S-30%-40C-D-4
S-30%-40C-D-5
FEM
400
200
0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Deflection [mm]
b) S-30%-40C-D
1400
1200
Load [N]
1000
800
S-50%-40C-D-1
S-50%-40C-D-2
S-50%-40C-D-3
S-50%-40C-D-4
S-50%-40C-D-5
FEM
600
400
200
0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Deflection [mm]
c) S-50%-40C-D
specimens tested at 40C and in the moisture desorption process
As shown in Figs. 4.26, 4.27 and 4.28, good agreement on the stiffness of
specimens is evident between the FE predicted curves and test results.
However, there are some exception (S-50%-20C-2, S-50%-20C-4, S-50%40C-2 and S-50%-40C-4), which significantly deviate from other
83
Chapter4
specimens in the same test group. It can be attributed to non-homogenity of

FRP laminate specimens, which influence the stiffness of the material. It
also influences the moisture absorption property and correspondingly
degrades the material stiffness. Excluding these exceptions, the predictive
equations of moisture-dependent interlaminar shear modulus are acceptably
reliable. Hence, they can be employed as the input material properties of an
FE model to analyse the enviroment-dependent mechanical behaviours of
FRP joints and FRP structures.
Fig. 4.29 illustrates these four predictive equations. For the specimens
tested at 20C, a dramatic drop of interlaminar shear modulus is found from
the unconditioned dry specimen (1200MPa) to the moisture saturated
specimen (800MPa). Then, after the moisture desorption process, a slight
recovery is found for the shear modulus of S-0%-20C-D specimens
(850MPa). In total 29.2% decrease of interlaminar shear modulus is
obtained for the specimens enduring one cycle of moisture absorptiondesorption process. For the specimens tested in 40C, a similar tendency of
interlaminar shear modulus loss is obtained, with a 42.9% decrease from
the unconditioned dry specimens (1050MPa) to the fully saturated
specimens (600MPa) and a 19% decrease for the specimens enduring one
cycle of moisture absorption-desorption process.
1000
800
600
400
absorption
200
0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
Mt/M
Interlaminar shear modulus (MPa)
1200
Fig. 4.29. a) 20C
84
0.7
0.8
0.9
1.0
Interlaminar shear modulus (MPa)
1200
1000
800
600
400
absorption
200
0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Mt/M
b) 40C
Fig. 4.29. Degradation on the short-beam shear strength of FRP laminates
4.4 Conclusions
This chapter describes the investigation of the environment-dependent
mechanical properties (flexural and interlaminar shear properites) of the
FRP laminate material, which is achieved by the flexural tests and shortbeam shear tests according to the ASTM test code D790-10 and
D2344/2344M-00. The hydrothermal aging condition is a typical hot/wet
aging environment (40C-water) for the application of FRP bridge decks.
The test conditions vary in terms of test temperature, moisture uptake
content and absorption/desorption process. Experimental results confirm
that the combination of moisture and temperature effects sincerely
deteriorates the mechanical properties of FRP laminates, on both strength
and stiffness.
Furthermore, a coupled hygro-mechanical FE modeling method was
developed to analyse the enviroment-dependent mechanical behaviours of
FRP lanimates. This FE model is firstly validated by the test results of
flexural tests. Subsequently, the coupled hygro-mechanical FE model is
employed in an inverse parameter identification method to determine the
elastic interlaminar shear modulus of FRP laminates. The basis of this
method is to minimize the difference between the experimentally measured
85
Chapter4
and numerically determined material response by varying the interlaminar

shear modulus of the FE model.
Finally, the predictive equations for environment-dependent mechanical
properites of FRP laminates are sustained by using the least square method
for the curve fitting. These predictive equations can be used as the input
parameters for a coupled hygro-mechanical FE model, as well as a
contribution to the design code as far as the long-term performance of FRP
structures is concerned.
86
References:
[1] D790-10 A. Standard Test Methods for Flexural Properties of Unreinforced and
Reinforced Plastics and Electrical Insulating Materials. PA, United States: ASTM
International; 2010.
[2] Gellert EP, Turley DM. Seawater immersion ageing of glass-fibre reinforced polymer
laminates for marine applications. Compos Part a-Appl S. 1999;30(11):1259-65.
[3] Loh WK, Crocombe AD, Wahab MMA, Ashcroft IA. Modelling anomalous moisture
uptake, swelling and thermal characteristics of a rubber toughened epoxy adhesive. Int J
Adhes Adhes. 2005;25(1):1-12.
[4] Ashcroft IA, Hughes DJ, Shaw SJ, Wahab MA, Crocombe A. Effect of temperature on
the quasi-static strength and fatigue resistance of bonded composite double lap joints. J
Adhesion. 2001;75(1):61-88.
[5] Ashcroft IA, Wahab MMA, Crocombe AD, Hughes DT, Shaw SJ. The effect of
environment on the fatigue of bonded composite joints. Part 1: testing and fractography.
Compos Part a-Appl S. 2001;32(1):45-58.
[6] Barjastech E, Nutt SR. Moisture absorption of unidirectional hybrid composites
Compos Part a-Appl S. 2012;43:158-64.
[7] Bank LC, Gentry TR, Barkatt A. Accelerated Test Methods to Determine the LongTerm Behavior of Frp Composite Structures - Environmental-Effects. J Reinf Plast Comp.
1995;14(6):559-87.
[8] Muliana A, Nair A, Khan KA, Wagner S. Characterization of thermo-mechanical and
long-term behaviors of multi-layered composite materials. Compos Sci Technol.
2006;66(15):2907-24.
[9] Davies P, Evrard G. Accelerated ageing of polyurethanes for marine applications.
Polym Degrad Stabil. 2007;92(8):1455-64.
[10] Moulzakis DE, Zoga H, Galiotis C. Accelerated environmental ageing study of
polyester/glass fiber reinforced composites (GFRPCs). Compos Part B-Eng.
2008;39(3):467-75.
[11] Crocombe AD, Ashcroft IA, Abdel Wahab MM. Chapter 8 Environmental
Degradation. Modeling of Adhesively Bonded joints: Springer-Verlag Berlin Heidelberg;
2008.
[12] D2344/2344M-00 A. Standard Test Methods for Short-Beam Strength of Polymer
Matrix Composite Materials and Their Laminates. PA, United States: ASTM International;
2006.
[13] Staab GH. Laminar Composites. Oxford: Butterworth-Heinemann; 1999.
[14] Chan A, Chiu WK, Liu XL. Determining the elastic interlaminar shear modulus of
composite laminates. Compos Struct. 2007;80(3):396-408.
87
Chapter4
88
Chapter 5
Mechanical behaviour of FRP-tosteel adhesively-bonded joints
before and after hydrothermal
aging*
5.1 Introduction
The research presented in this chapter is focusing on mechanical
behaviours (load-displacement behaviours and failure modes) of the
adhesively-bonded joint between FRP sandwich decks and steel girders
before and after hydrothermal aging. As stated in Chapter 1, there are two
typical stress states in the adhesively-bonded joint (Fig. 5.1): shear and
tensile stress, and combinations of both.
a) Shear stress in the longitudinal direction
b) Tensile stress in the transverse direction

Fig. 5.1. Typical stress states in adhesively-bonded joint of FRP-steel composite bridge
*
The content of this chapter is partially published in [1] Jiang X, Kolstein MH, Bijlaard FSK. Study on
mechanical behaviors of FRP-to-steel adhesively-bonded joint under tensile loading. Compos Struct.
2013;98:192-201. and [2] Jiang X, Kolstein MH, Bijlaard FSK. Experimental and numerical study on
mechanical behavior of an adhesively-bonded joint of FRP-steel composite bridge under shear loading.
Compos Struct. 2014;108:387-399.
89
Chapter5
To realize these stress states in the adhesively-bonded joint in the test

specimen, a specific loading device is developed to provide six different
loading angles, which are 0(pure tension), 18, 36, 54, 72 and 90(pure
shear). This loading device is described in detail in section 5.2.1. Firstly, the
adhesively-bonded joint is investigated experimentally under pure shear and
pure tensile loading, considering the influence of different surface
pretreatments on the FRP sandwich deck and the steel girder. The
interfacial bonding quality of different surface pretreatment methods is
studied. Subsequently, a three-dimensional FE model of tested specimens
is developed by employing ABAQUS 6.8, and validated by comparing the
relative deformation between FRP sandwich deck and steel support of the
adhesively-bonded joint with experimental results. Linear elastic simulations
are performed to characterize the stress distribution throughout the
adhesively-bonded joint. The mesh-dependency of the FE model is
addressed in the longitudinal, transverse and through-thickness direction of
the adhesive joint. The preferable mesh configuration is confirmed for the
further FE analysis of adhesive joints under other loading conditions.
Secondly, a continuous study is conducted on the adhesively-bonded joint
under the combination loading of tension and shear. Four combining ratios
are employed. Failure modes of adhesive-bonded joints are experimentally
investigated. Depending on the test results of six angle loading conditions, a
tensile/shear failure criterion of the adhesive-bonded joint before
hydrothermal aging is obtained. Based on the FE model, developed for the
pure shear and pure tensile loading condition, the three-dimensional stress
state is characterized through the interface between FRP sandwich deck
and adhesive layer for four combining loading conditions. To study the
environmental effects on the durability of adhesively-bonded joints, a further
research is focusing on mechanical behaviours of adhesively-bonded joints
after hydrothermal aging. The adhesive joint specimens have been
hydrothermally aged in water of 40C for a period of four months. Test
results address a tensile/shear failure criterion of the adhesively-bonded
joint after hydrothermal aging. To better understand the different mechanical
performances of the adhesively-bonded joint before and after hydrothermal
aging, a systematic comparison on experimental results of un-aged
specimens and hydrothermal aged specimens is conducted. The discussion
includes the ultimate failure load, failure mode and stiffness. The
90
Chapter5MechanicalbehaviourofFRPtosteeladhesivelybondedjointsbeforeand
afterhydrothermalaging
hydrothermal aging effects on the mechanical degradation of adhesive joints

are discussed. Furthermore, the post curing mechanism induced by the
elevated temperature of hydrothermal aging environment is discussed,
which significantly influences the interfacial bonding quality of adhesive
joints.
5.2 Experiment
5.2.1 Tensile-shear loading device
It is noted that, for the application of adhesively-bonded joints in aerospace
engineering, single-lap joints or double-lap joints were usually employed to
investigate the mechanical properties of joints under shear loading [3, 4].
However, it is not convincing to directly extend the experimental data and
results from the aerospace industry to the civil engineering field. FRP
composites used for civil infrastructures have essential differences as
compared to the FRP composites used in aerospace and aircraft
applications, which include geometries, types of fibres and matrix,
fabrication methods, curing process and service environmental conditions.
For instance, the FRP composite profiles and adhesive layers for
aerospace/aircraft structures are usually thin (0.11 mm); while for bridge
and building structures adherents and adhesive layers are usually
comparatively much thicker (220 mm). Thus, with the thicker adhesive
layers, the single-lap or double-lap adhesive joints will lead to a
considerable additional bending moment around the adhesive tips, which
can result in a significant increase of the peeling stress at that location (as
shown in Fig. 5.2) [5-9]. Sheppard et al. [9] confirmed that the peeling stress
at the end of a double-lap adhesive joint was relatively lower than that of the
single-lap adhesive joint, by using a damage zone FE model. Finally, the
failure of adhesive joints is initiated by the peeling stress rather than by the
shear stress.
91
Chapter5
Location of peeling stress
a) Single-lap adhesive joint
b) Double-lap adhesive joint
Fig. 5.2. Locations of peeling stress in single-lap and double-lap adhesive joints
For the application of adhesive joints in the FRP-steel composite bridge, the
adherents (FRP decks and steel beams) are usually thick profiles. Thus, the
deformation in single-lap or double-lap adhesive joints (as shown in Fig. 5.2)
is not realistic. In this research, to simulate the real stress state of the
adhesively-bonded joint for the application of FRP-steel composite bridges
as well as reduce the scale of the peeling stress, a specific loading device is
designed as shown in Fig. 5.3. The circular steel plates are separated into
two pieces. Two parts of the central area of circular steel plates are cut off to
save some space for putting up displacement sensors. The two steel blocks
are fastened to the circular steel plates by 8 bolts. The adhesively-bonded
joints are located in the middle of the loading device. Several bolt holes on
the circular steel plates are proposed to force the whole loading device by
different angles. Three bolts are employed to transfer the loading uniformly
to the circular steel plates. Dimensions of the tested adhesively-bonded joint
(more details are presented in the following section) are intentionally
designed to locate the centroid of adhesive layer exactly in line with the
shear loading axis. In this way, the additional bending moment can be
avoided.
92
18
0(tension)
36
54
72
90(shear)
Fig. 5.3. Tensile/shear loading device
Besides the pure shear loading condition, this specific tensile/shear loading
device can provide other loading conditions, such as a pure tensile loading
condition and four combinations of shear and tensile loading. Accordingly, in
total six loading conditions are feasible through this tensile/shear loading
device. The angle step between each loading direction is 18.
5.2.2 Test specimen

5.2.2.1 Un-aged specimen
To be adaptable to the loading device, the adhesively-bonded joint between
FRP deck and steel girder is extracted for experimental investigation as
shown in Fig. 5.4.
93
Chapter5
Fig. 5.4. FRP-to-steel adhesively-bonded joint
Considering the dimensional limitation of loading device and installation

convenience, the specimen geometries are determined as follows: a
190mm90mm piece of FRP sandwich bridge deck (supplied by Infra
Composites B.V.[10]) is adhesively bonded to the steel support. The area of
adhesively bonding surface on the steel support is a 90mm90mm square.
The bottom steel part is 190mm90mm, with four holes for bolting it to the
loading device. In the middle of the FRP sandwich deck, there is a Balsa
SB150 wood layer with a thickness of 38.1mm, which is a core material
produced from certified kiln-dried Balsa wood in the end-grain configuration.
The surface layer of the FRP sandwich deck is composed of three plies of
0.94mm thick EQX1200, which are the glass-fibre reinforced laminated
polymer composites (54% glass content by weight). The layup configuration
of EQX1200 is given in Table 5.1.
Table 5.1. FRP laminate properties (supplied by manufacturer)
Product
name
EQX1200
Total
weight
(g/m2)
1193
0o
283
Yarnroving
+45o
90o
45o
300
300
300
Knityarn
10
The sandwich profiles are manufactured by resin vacuum infusion. The

gluing of FRP sandwich deck to steel support is executed and subsequently
cured in a specific climate room, with the constant condition of 20 oC and
50%RH. The nominal thickness of the adhesive layer is controlled to be
94
6mm using spacers, as shown in Fig. 5.5. 6mm adhesive thickness is

supposed to be a practical value, as far as installation tolerance and easy
operation in the on-situ situation of bridge construction is concerned. The
spacers are made of Teflon, which are those for easy demoulding from the
adhesive layer after fully curing of adhesive joints. The commercial
structural adhesive used herein is BUFA-BONDING PASTE 740-0110,
which is a two-component adhesive material, based on toughened vinyl
ester resin.
Spacer
Fig. 5.5. Preparation process of adhesive joint specimens

95
Chapter5
Table 5.2. Adhesive properties (supplied by manufacturer)

Property
Viscosity(MPa.s)
Solidcontent(%)
Elongationatbreak,steel(%)
Emodulus(MPa)
Tensilestrength,steel(MPa)
Heatdeflectiontemperature(C)
Geltime(min)
Curingtime(min)
Value
1,800,000
9
3.7
3400
14.5
70
60
85
Adhesive material properties are listed in Table 5.2. The preparation

process of the specimens is shown in Fig. 5.5. To investigate the influence
of surface pretreatment methods on the interfacial bonding quality, three
different surface pretreatment methods are conducted on the surfaces of
FRP sandwich deck and steel support for the pure tensile and shear loading
conditions. The first method is the simplest one, which is only degreasing
and cleaning of the surfaces with acetone. The second one is conducted by
using sandpaper. It starts with the wiping of specimen surfaces with acetone
to eliminate any presence of oil used in the machining process. Then, the
surfaces are abraded with sandpaper to remove any impurities and oxide
layer which can potentially exist. After that, a re-degreasing and re-cleaning
of the surface is done with acetone to remove any particles that can remain
after sanding. The last surface pretreatment method is by using a sand
blasting machine. The whole process is similar to the second one, except for
using the sand blasting to remove the impurities and oxide layer. For each
surface pretreatment method, three replicated specimens are prepared. In
the following sections of this chapter, for the identification of specimens (as
listed in Table 5.3), specimens pretreated using the first method (acetone)
are indicated as AC for short, the second method (sand paper) as SP and
the last method (sand blasting) as SB. For specimens tested under the
combination of shear and tensile loading, the surface pretreatment method
is limited to the sand blasting method, based on the test results under the
pure shear and pure tensile loading. As listed in Table 5.3, in total 30 pieces
of adhesive joint specimens are prepared for the tests before hydrothermal
aging.
96
Table 5.3. Adhesively-bonded joint specimens
Specimen
identification
Loading
condition
SAC01~03
SSP01~03
SSB01~03
TAC01~03
TSP01~03
TSB01~03
18SB01~03
36SB01~03
54SB01~03
72SB01~03
SSBA01~03
TSBA01~03
18SBA01~03
36SBA01~03
54SBA01~03
72SBA01~03
90angle(shear)
90angle(shear)
90angle(shear)
0angle(tension)
0angle(tension)
0angle(tension)
18angle
36angle
54angle
72angle
90angle(shear)
0angle(tension)
18angle
36angle
54angle
72angle
Surface
pretreatment
method
acetone
sandpaper
sandblasting
acetone
sandpaper
sandblasting
sandblasting
sandblasting
sandblasting
sandblasting
sandblasting
sandblasting
sandblasting
sandblasting
sandblasting
sandblasting
Hydrothermal
aging
Numderof
specimens
no
no
no
no
no
no
no
no
no
no
yes
yes
yes
yes
yes
yes
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
In order to fix the adhesive joint to the loading system, some accessorial
components are designed as shown in Fig. 5.6 a) and b). As mentioned
above, the steel support is fastened directly by four bolts to the bottom steel
block. However, for the FRP sandwich decks no holes are proposed, since
the discontinued part in decks can cause more stress distribution distortions,
which is not the case in applications of FRP composite bridges. To fix the
FRP sandwich deck part, it is designed to be fastened with two L-shape
steel profiles by four bolts to the top steel block, as shown by the purple
coloured parts in Fig. 5.6 a). In this way, the two L-shape steel plates are
holding the FRP sandwich deck specimen. The pure shear loading condition
is realized by applying force to the L-shape steel profile, and the load is
transferred through interacted contact-surfaces from the lateral plate to the
adhesive joint. The pure tensile loading condition is realized by forcing the
bottom plate of the L-shape steel profile. The combining loading conditions
are realized by the combination of the above two loading conditions. As
shown in Fig. 5.3, the forces of six loading conditions are applied exactly via
the center line of the adhesive layer. In this way, no additional bending
moment is expected. All the accessorial components are made of steel.
Compared to the FRP composites and adhesive materials, the deformation
97
Chapter5
of steel components can be neglected during tests, due to the high stiffness
of steel material.
a) FRP Deck fixed configuration
b) Steel support fixed configuration
Fig 5.6 Fastening system for fixing the adhesively-bonded joint to the loading device
5.2.2.2 Hydrothermal aged specimen

The preparation process of FRP-to-steel adhesively-bonded joints is the
same as stated in section 5.2.2.1. The surface pretreatment method is the
sand blasting (SB) method. Based on test results under the pure tensile and
pure shear loading, this pretreatment method is considered to be the
preferable and easy method to obtain the best bonding quality and
controllable quality of adhesive layer of an FRP-to-steel adhesively-bonded
joint. The identification of specimens is given in Table 5.3. The A character
is added to the numbering of specimens, indicating that the specimens are
tested after aging. In total 18 pieces of specimens are prepared for the tests
after hydrothermal aging.
After the curing of the adhesively-bonded joints at the environment of 20C
and 50%RH, they are immersed in the water with a temperature of 40C,
which is considered as a typical hot/wet environment regrading the actual
service environment of FRP-steel composite bridges. This environmental
aging condition is simulated by putting specimens in a water filled container,
which is kept in an oven in which the temperature is set at 42C. In this way,
the water temperature is 40C. The total aging time is four months. To
simulate the real in-situ condition of FRP sandwich decks, prior to putting
98
specimens into the environmental aging condition, the four surrounding

surfaces of the FRP sandwich deck specimen are sealed by a commercial
coating Sikafloor 156. After fully curing of the first layer of coating, another
layer of water-proof adhesive tape is applied on the coating to doubly
ensure that no water can penetrate through the surrounding surfaces of
FRP sandwich decks. Because, the tested FRP sandwich deck is cut from
the whole piece of the bridge deck. In the service-life of these FRP
sandwich bridge decks, the edge surfaces are sealed to be water-resistant.
Thus, most of the moisture content should diffuse through the top and
bottom surfaces into the FRP sandwich deck specimens. The moisture
uptake of core material (Basal wood) should not directly absorb from the
environment, but absorb from the FRP laminates. Hence, during the
hydrothermal aging, the core material (Basal wood) should not be exposed
to the water aging condition.
The adhesively-bonded joint specimen after the hydrothermal aging is
shown in Fig. 5.7. To simplify the manufacturing process of the steel support,
it is made of two parts and then assembled together by 9 bolts for testing,
see Fig. 5.7.
Fig.5.7 Hydrothermal aged adhesively-bonded joint and its assembling process

99
Chapter5
5.2.3 Experimental procedure

The series of experiments are carried out in the Stevin II laboratory of the
Faculty of Civil Engineering and Geosciences at Delft University of
Technology. All the aged and un-aged specimens are tested by the specific
tensile/shear loading device under the ambient environment. The whole test
set-up is shown in Fig. 5.8. The tensile/shear loading device can provide six
loading conditions by forcing the different angles of semi-circle steel plates.
A SCHENCK Hydropuls testing machine with a loading capacity of 600 kN
in tension is employed and controlled by the INSTRON 8400 controller,
which can provide both load- or displacement-controlled test procedures.
The whole tensile/shear loading device is loaded by jacks by means of two
hinged joints, which avoid the additional bending moment due to the
eccentric loading from the specimen misalignment. Two LVDTs are fixed on
each side of the loading system, as shown in Fig. 5.9, to measure the
displacement between the top and bottom semi-circle loading device, for
checking equal distribution of the load. The measuring range of the LVDT is
0~10mm. The quasi-static experiments of adhesively-bonded joints are
displacement controlled by LVDTs (linear variable differential transformer) at
a rate of 0.001mm/sec. When the adhesive joint specimen is installed in the
loading device, the four bolts connecting the L-shape steel profiles (see
Fig.5.6 a)) are not fully fastened firstly. A preload of 1kN is applied to make
every loading component touch each other. In this way, the load can
transfer smoothly from the loading device to the adhesive joint. After that,
these four bolts are fully fastened and then the preload is unloaded. The
tests start at a load of 0kN. All the measured test data can be found in the
Stevin II lab report.
100
18
0 angle (tension)
54
36
18 angle
72
54 angle
36 angle
90
72 angle
90 angle (shear)
Fig. 5.8. Six loading angles of the tensile/shear loading device
101
Chapter5
LVDT-02
LVDT-01
LVDT
Fig. 5.9. Test set-up for the shear loading condition
For the shear loading condition, two displacement sensors are secured on
both sides of adhesive joints, as shown in Fig. 5.10, to track the relative
deformation between FRP sandwich deck and steel support (indicating the
shear deformation in the adhesive layer) during the whole test process. The
measuring range of the displacement sensor is 0~2mm. For the tensile
loading condition, the locations of displacement sensors are illustrated in Fig.
5.11. Two displacement sensors are secured on each side of adhesive joint,
to track the vertical deformation between FRP sandwich deck and steel
support during the whole test process.
102
D S -0 2 (0 4 )
F R P s a n d w ic h d e c k
D S -0 1 (0 3 )
a d h e s iv e la y e r
s te e l s u p p o rt
a) Schematic representation of the displacement sensor location
Displacement sensor
b) Lateral view of the displacement sensor location
DS-01
DS-03
DS-02
DS-04
c) Front view of the displacement sensor location

Fig. 5.10. Location of four displacement sensors for the adhesive joint under the pure shear
loading (DS=displacement sensor)
103
Chapter5
DS-01(03)
DS-02(04)
10
10
FRP sandwich deck

adhesive layer
steel support
DS-03
DS-01
DS-04
DS-02
b) Top view of the displacement sensor location
c) Lateral view of the displacement sensor location

Fig. 5.11. Location of four displacement sensors for the adhesive joint under the pure
tensile loading (DS=displacement sensor)
Fig. 5.12 shows the location of displacement sensors for the combination
loading condition. The ones on the left (DS-01 and 03) measure the vertical
104
deformation between FRP sandwich deck and steel support, which indicates
the tensile deformation, while the ones on the right (DS-02 and 04) measure
horizontal deformation, which indicates the shear deformation.
tensile deformation
shear deformation
DS-02(04)
FRP sandwich deck

8
DS-01(03)
10
10
adhesive layer
steel support
DS-1(3)
DS-2(4)
b) Lateral view of the displacement sensor location

Fig. 5.12. Location of four displacement sensors for the adhesive joint under the
combination loading of shear and tension (DS=displacement sensor)
Before testing, all instrumentations (LVDTs and displacement sensors) are

calibrated. Experimental data is recorded per second by a PC. In the tests,
the local strain through the adhesive layer and FRP laminates is difficult to
be measured by the strain gauges. Hence, the stress-strain distribution in
the adhesive joint is analysed by a three-dimensional FE analysis as
discussed hereafter.
105
Chapter5
5.3 Mechanical behaviour of adhesively-bonded joints

before hydrothermal aging
5.3.1 Shear loading
5.3.1.1 Experimental results and discussion
For the S-AC-specimens, S-SP-specimens and S-SB-specimens, the load
increase is almost linear up to failure. The ultimate failure of the adhesive
joints always occurs in a brittle and sudden manner, without any visible
signs or sound warning from fibre breaking. Fig. 5.13 shows the loaddisplacement curves of all nine specimens under the pure shear loading
condition, which are the average values measured by the LVDTs. It can be
found that, at the initial stage of the test process, a large deviation is
observed. This is because at the beginning of the loading process, the shear
load is transferred by the friction between the L-shape steel profile and the
FRP deck. As the load increases, it is beyond the friction and then the
adhesive joint starts to move under the constant loading until the FRP deck
touches the L-shape steel profile. Subsequently, the shear load is
transferred by the interacted contact-surfaces. The friction between the Lshape steel profile and the FRP deck cannot be controlled during the tests,
which causes the deviations at the initial stage of loading. It is also found
that, when the load is below the friction, the slope of the curve is higher than
that beyond the friction. This is because the friction between the L-shape
steel profile and the FRP deck makes the deformation of joint distributing
uniformly. After the load beyond the friction, the shear load is transferred
mainly by the vertical contact surface of L-shape steel profiles. Then, the
deformation on the load directly applied side dominates the global
deformation of the adhesive joint.
106
100
S-SB-02
S-SP-03
S-SP-01
Load [kN]
80
S-SB-03
60
S-SB-01
S-SP-02
40
S-AC-02
S-AC-03
S-AC-01
20
0
0
10
Displacement [mm]
Fig. 5.13. Load-displacement curves based on average values of LVDT-01 and LVDT-02
measurements
The average shear stress-displacement curves are shown in Fig.5.14. The

average shear stress is calculated by eq. (5.1) as follows:
average
Fshear
A
(5.1)
where average is the average shear stress, Fshear is the shear load applied on
the adhesive joint, and A is the adhesively-bonding area.
107
Chapter5
Average shear stress [MPa]
14
12
S-SB-02
S-SP-03
S-SP-01
10
8
S-SB-03
S-SP-02
S-SB-01
6
4
S-AC-02
S-AC-03
S-AC-01
2
0
0
10
Displacement [mm]
Fig. 5.14. Average shear stress-displacement curves
Fig. 5.14 shows that specimens with different surface pretreatment methods
agree well with each other for the slopes of curves, which indicates that the
stiffness of adhesive joints under shear loading hardly relates to surface
pretreatment methods. For the ultimate failure loads, as listed in Table 5.4,
the average ultimate failure loads of S-SP-specimens and S-SB-specimens
obtain almost the same value of about 70 kN, which is more than three
times of that of AC-specimens. The deviation of test results is established by
Eq. (5.2):
|
|
|
|
(5.2)
Here, the deviation indicates the spread of the test data. It can be found
from Table 5.4, a relatively large deviation of test results is obvious,
especially for the S-SP-specimens and S-SB-specimens, which is due to the
artificial surface pretreatment and porosity in the adhesive layer. The
bonding quality is not easily controlled. More test results are needed to gain
a better statistical determination of characteristic values. According to EN
1990 Eurocode 0: Basis of structural design-Annex D: Design assisted by
testing [11], for the test data of only three specimens, the value of Kn
(characteristic fractile factor) is 3.37. This will decrease to 1.73, if the
108
number of specimens increases to 30. The value of Kn will continue

decreasing to 1.64, if the number of specimens is infinite.
Moreover, Table 5.4 also implies that there is no significant difference on
load-bearing capacities of adhesive joints between the surface pretreatment
methods using the sandpaper and the sand blasting technique. Both these
methods achieve a similar bonding quality between adhesive layer and steel
support. As compared to S-AC-specimens, the failure plane of S-SPspecimens and S-SB-specimens move from the interface between adhesive
layer and steel support to the adhesive layer, which is discussed in detail
hereafter.
Table 5.4. Ultimate failure loads of nine adhesive joints
SACspecimen
Failureload(kN)
SSPspecimen
Failureload(kN)
SSBspecimen
Failureload(kN)
01
18.1
01
76.4
01
51.0
02
25.0
02
51.2
02
92.9
03
22.7
03
82.4
03
64.1
Average
21.9
Average
70.0
Average
69.3
Deviation
17.5%
Deviation
26.9%
Deviation
34.0%
Figs. 5.15, 5.16 and 5.17 show the failure modes of all adhesive joints. As to
the S-AC-specimens, there is no damage occurring in the adhesive layer.
The failure of adhesive joints take place through the interface between the
adhesive layer and the steel support, as presented in Fig. 5.18 a). This
failure mechanism is due to the lack of sufficient surface pretreating on the
surface of steel support, in this way losing the reliable bonding quality
between adhesive layer and steel support. From the failure surface of the
adhesive layer, as shown in Fig. 5.15, it can be clearly seen that the residual
rust (dark-colour stuff) is torn away from the steel surface and left on the
adhesive surfaces, which is not observed on the failure surfaces of S-SPspecimens and S-SB-specimens. For S-SP-specimens and S-SBspecimens, the failure modes are the same, which are cohesive fractures
(near the interface between adhesive layer and steel support) in the
adhesive layer, as clearly shown in Figs. 5.18 b) and c). The failure modes
of S-SP-specimens and S-SB-specimens also indicate that the fracture
initiates at the edge zone (about 10mm away from the end of adhesivelybonding area) of the adhesive layer, and then propagates to trigger the final
failure of the whole adhesive joint. The 10mm distance is indicated in Fig.
109
Chapter5
5.16 and Fig. 5.17, where the fracture initiated lines (white-colour lines) on
the adhesive layer are visible. However, the S-SP-02 specimen is an
exception. There is no obvious slowly fracturing area on the surface of the
steel support, where cracks gradually develop. Instead, some relatively
large cracks can be found at the area close to the adhesive layer edge (Fig.
5.16, middle picture). It can be attributed to the fact that cracks are initiated
by the stress concentration at the non-homogeneous area of adhesive
material, which induces the premature failure of the adhesive joint.
Accordingly, the S-SP-02 specimen obtains the lower load-bearing capacity
(51.2kN) as compared to the other two S-SP-specimens.
S-AC-01
S-AC-02
Fig. 5.15. Failure mode of S-AC-specimens
110
S-AC-03
Chaapter5Me
echanicalbe
ehaviourof
fFRPtoste
eeladhesive
elybonded jointsbefo
oreand
afterhyydrotherma
alaging
fracture
e
initiated line
cracks
10mm
10mm
fracture
initiated line
S-SP-01
10mm
S-SP-02
S--SP-03
Fig. 5.16. Failuure mode off S-SP-speciimens
fracture
e
initiated line
fracture
initiated line
10mm
10mm
S-SB
B-01
fractu
ure
initiated
d line
10
0mm
S-SB-02
2
S-SSB-03
Fig. 5.17.
5
Failur
ure mode of S-SB-speciimens
Fig. 5.18 clea

arly show
ws the diifferent fa
ailure mo
odes amo
ong the S-ACS
speciimens, S--SP-specim
mens and
d S-SB-sp
pecimens, from view
ws of the
e steel
suppo
ort surfacces. The adhesive
a
jjoints with
hout surfa
ace pretre
eatment (S
S-ACspeciimens) fa
ail at extrremely lo
ower values of ulttimate faiilure load
ds, as
mentioned be
efore, less
s than o
one-third of S-SP-specimen
ns and S-SBS
speciimens. Co
onsequenttly, it is off great im
mportance to executte the suffficient
111
Chapter5
surface pretreatment on both FRP sandwich deck and steel girder in

practice to ensure reliable mechanical performance of the adhesive joints
between them.
no residual adhesive material
a) Failure mode interfacial failure of S-AC-specimen
residual adhesive material
b) Failure mode adhesive failure of S-SP-specimen

c) Failure mode adhesive failure of S-SB-specimen

Fig. 5.18. Different failure modes from the view of steel support surfaces
Fig. 5.19 shows the measurement of relative deformation between FRP

sandwich deck and steel support on the S-SB-specimens from four
displacement sensors. In this respect, reference is made to the location of
displacement sensors (Fig. 5.9). The measured test data from the DS-02
112
and DS-04 displacement sensors are negative values, since these sensors
are under compression during the tests. The test data from the DS-01 and
DS-03 displacement sensors, on the other hand, are positive, since they are
extending during the tests. However, the shear deformation in the adhesive
layer is in the same direction. The positive/negative test data from
displacement sensors are only due to their opposite locations. It can be
seen that the measured deformation from DS-02 and DS-04 significantly
deviate from each other, which indicates that the shear force is not loaded
perfectly centrally. Besides this, the non-homogeneous characteristic of the
adhesive layer can also make the load distribution non-uniformly. However,
for the displacement sensors DS-01 and DS-03, the measured deformations
match each other well, indicating that the load is balanced on both sides of
the specimen at the far end of the adhesive joint from the loading edge. The
absolute values of displacement are less than half of those from DS-02 and
DS-04. Moreover, all the load-deformation curves are not increasing linearly,
since the stresses in the adhesive layer keep redistributing during the whole
test process, which is supposed to be due to the non-homogeneous
property of the adhesive layer. Porosity of the adhesive layer is not
avoidable during the preparation process of the adhesive joint. Test results
on S-AC-specimens and S-SP-specimens present similar mechanical
behaviours regarding the relative deformation between the FRP sandwich
deck and the steel support. The eccentric loading cannot be avoided in such
a small scale test. Even the load is compelled to be exactly centric;
afterwards, it will be eccentric as the load is increasing. This is because the
stress in the adhesive joint keeps redistributing during the whole loading
process.
113
Chapter5
60
DS-04
DS-02
DS-03
DS-01
Load [kN]
50
40
30
DS-01
DS-02
DS-03
DS-04
mean value of DS-01,03
20
10
0
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
0.20
Deformation [mm]
a) S-SB-01
100
DS-02
DS-04
DS-03
DS-01
Load [kN]
80
60
DS-01
DS-02
DS-03
DS-04
40
20
0
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
Deformation [mm]
b) S-SB-02
70
DS-04
DS-02
DS-03 DS-01
60
Load [kN]
50
40
30
DS-01
DS-02
DS-03
DS-04
20
10
0
-0.4
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
Deformation [mm]
c) S-SB-03
Fig. 5.19. Load-deformation curves measured from four displacement sensors on the S-SBspecimens (DS=displacement sensor)
114
5.3.1.2 FE analysis
1) FE model
As mentioned before, the experimental investigations usually present
limitations to obtaining the stress-strain distribution throughout the adhesive
joint. While the FE analysis of adhesive joints can provide more information
about the stress-strain state. Compared with the two-dimensional (2D) plane
strain analysis, three-dimensional analysis results can reveal the existence
of a complex multi-axial stress/strain state at the ends of the overlap in the
bondline. Thus, the 3D FE model of the adhesive joint loaded under shear
loading is developed by employing the FE package ABAQUS 6.8, as shown
in Fig. 5.20 and Fig. 5.21.
loading
area
X
Y
reaction
area
Fig. 5.20. Loading and boundary condition of FE model
Adhesive
layer
Fig. 5.21. FE model of adhesive joint
As shown in Fig. 5.20, to simplify the FE model and save computational time,
only the L-shape steel parts and bolts for fastening the adhesive joint are
115
Chapter5
included in the FE model. Other accessorial load transferring devices are

equivalently replaced by two big-size stiff blocks. Eight-node linear brick
elements with reduced integration and hourglass control (C3D8R) are
employed for all the components. The total number of elements is 284593.
With regard to the adhesive layer, a six-layer discretization is used in the
through-thickness direction to guarantee accurate stresses and flexible
deforming capacity, as well as a three-layer discretization for the FRP
laminates close to the adhesive layer, as shown in Fig. 5.21. There is no
specific element defined at the interphase between the FRP sandwich deck
and the adhesive layer, or between the adhesive layer and the steel support.
All these surfaces are fixed to each other, assuming that no cracks or
relative slip occur during the loading procedure. As shown in Fig. 5.20,
dimensions of the loading area and the reaction area (three-direction fully
restricted as a boundary condition) are intentionally designed, with the
center of these two areas exactly in the middle plane of the adhesive layer
to avoid an additional bending moment. The total loading of 50kN is applied
by the surface tractions on the loading area. The contact pairs are created
between the L-shape steel part and the sandwich FRP deck. Eight bolts are
fixed to the top and bottom stiff blocks by TIE command. The material
properties of FRP laminates and core material are anisotropic, while
adhesive and steel are assumed to be isotropic. The input material
properties supplied by manufacturers are listed in Table 5.5. The value of
E11, E22, G13 and G23 is from the test results of Chapter 4.
116
Table 5.5. Mechanical properties of materials for FE model
Property
FRP laminates
Elastic Modulus 11 (MPa)
16609
16609
11000
Poisson Ratio 12
0.33
Poisson Ratio 23
0.3
Poisson Ratio 13
0.18
Shear Modulus 12 (MPa)
6986
1200
1200
adhesive
Elastic Modulus (MPa)
3400
Poisson Ratio
0.37
core material
5759
Poisson Ratio
0.35
Shear Modulus (MPa)
309
steel
206000
Poisson Ratio
0.3
The whole analysis process is linear elastic. No geometric non-linearity or

elastic-plastic material properties are included in FE analysis. Thus, the FE
results can be amplified by specific factors to be comparable with the
experimental results. The automatic incrementation scheme is active to
make the analysis easily convergent as well as optimize the increment sizes
based on computational efficiency [12].
2) Validation of FE model
To validate the FE model, the FE results of relative deformation between
FRP sandwich and steel support at the locations of experimental
displacement sensors are presented in Fig. 5.22 for comparison with test
results. The FE results are amplified to make the comparison much clearer.
The FE results from two symmetric locations of DS-01 and DS-03 are
117
Chapter5
almost the same (the same case goes for the locations of DS-02, DS-04),
since the FE model is perfectly symmetric, centrally loaded and there is no
experimental installation inaccuracy. When the FE result is compared with
the mean value of experimental results from DS-01, 03 and DS-02, 04, good
agreements are achieved. The experimental curves are not ideally linear,
which is attributed to the stress redistribution occurring during the tests. But
the experimental curve still revolves around the linear FE curve. Based on
the above comparison between FE modeling and experimental results, it
can be confirmed that the FE model is reasonably accurate for predicting
the mechanical behaviours of the adhesive joint under shear loading. Hence,
further analysis and discussion depending on this FE model are reliable.
100
Load [kN]
80
60
40
S-SB-01
S-SB-02
S-SB-03
FEM
20
0
-0.5
-0.4
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
Deformation [mm]
Fig. 5.22 Comparison with experimental results and FE analysis at the locations of
displacement sensors
3) FEA results and discussion

Based on the failure mode of specimens, it is of interest to further
investigate the stress distribution across the interface between the steel
support and the adhesive layer, which is supposed to be the weakest plane
of the adhesive joint. Contour maps of shear stress and tensile stress are
drawn through this plane, as shown in Fig. 5.23 and Fig. 5.24. The X-axis is
along the longitudinal direction of the FRP sandwich deck, as indicated in
Fig. 5.20, while the Y-axis is in the transverse direction. These two specific
118
contour maps are drawn by employing the MATLAB software (R2011b

version), using the FE results from ABAQUS 6.8.
8.4mm
y/90mm
x/90mm
Fig. 5.23. Contour map of shear stress on the interface between steel support and adhesive
layer of the adhesive joint under the shear load of 50kN
y/90mm
x/90mm
Fig. 5.24. Contour map of tensile stress on the interface between steel support and adhesive
layer of the adhesive joint under the shear load of 50kN
119
Chapter5
For the shear stress distribution along the longitudinal direction, there are
two stress peaks throughout the interface, with the one on the right side
(loaded side) considerably higher (absolute value) than the left side. The
locations of stress peaks are not exactly at the ends of the adhesive overlap,
but 8.4mm away from the ends, which is approximately the same place of
adhesive fracture initiated lines for S-SP-specimens and S-SB-specimens,
as shown in Fig. 5.16 and Fig. 5.17. In the transverse direction, the shear
stress is distributed in a saddle shape, with a less steeper gradient. The
stress peaks occur at the transverse edges. For the tensile stress
distribution in the X direction, it is more stable with the values around zero,
except that at the edge zone of the interface the stress increases steeply,
which results in very high tensile stress at the end of the adhesive overlap.
With regard to the transverse direction, stress distributes at a rather steady
gradient. The maximum absolute values of stress are not located exactly at
the ends of interface, but a small distance away from the corners.
Further investigation is conducted on mesh-dependency of this FE model.
The part concerned is the adhesive layer in both longitudinal direction and
transverse direction, as shown in Fig. 5.25. Herein, four mesh configurations
are selected with mesh scales (length and width of elements) of 2.00mm,
1.50mm and 1.25mm and 1.00mm.
a) Mesh scale = 2.00mm
b) Mesh scale =1.50mm
c) Mesh scale = 1.25mm
d) Mesh scale = 1.00mm

Fig. 5.25. Different mesh configurations at the adhesive layer
The shear and tensile stress distributions are extracted for a comparison
from a longitudinal path, as shown in Fig. 5.26. The location of the
120
longitudinal path is through the interface between the adhesive layer and the
steel support.
Fig. 5.26. Longitudinal path for analysing the stress distribution
For the shear stress distribution (Fig. 5.27), a large deviation is evident
between the FE results of 2.00mm mesh scale and the other three meshrefined FE models. For the FE model with the 2.00mm scale mesh, the
elements defined along the longitudinal path are not enough, which tend to
constrain the deformation in the adhesive layer, in this way making the
adhesive layer much stiffer. Thus, the elements in the adhesive layer cannot
deform appropriately to release the stress concentration on the left side of
the adhesive layer. Due to the additional load-carrying capacity of the left
side of the adhesive layer, the absolute value of stress peak on the right
side (directly forced side) is relatively lower than expected. On the contrary,
the other three FE models with 1.50mm, 1.25mm and 1.00mm mesh scales
have sufficient deformation ability in the adhesive layer. The stress peak on
the right side (directly forced side) is significantly higher than that of the left
side, and these three curves overlap each other along the whole longitudinal
path, except the end nodes, where the mesh refined FE models tend to
slightly decrease the absolute stress values. It does not influence the shear
stress peak value. The peak value of shear stress is 12.52 MPa (under the
total loading of 50kN).
121
Chapter5
mesh scale=2.00mm
mesh scale=1.50mm
mesh scale=1.25mm
mesh scale=1.00mm
Shear stress [MPa]
-2
-4
end nodes
-6
-8
-10
stress peak
-12
-14
0
20
40
60
80
100
Distance [mm]
Fig. 5.27. Shear stress distribution along the longitudinal path depending on mesh scale
For the tensile stress distribution along the longitudinal path (Fig. 5.28),
deviation between FE results of 2.00mm mesh scale and the other three
mesh-refined FE models occurs again. Due to limitation of deformation
ability in the adhesive layer, the FE curve of 2.00mm mesh scale develops a
lower absolute value of compressive stress on the right side (directly forced
side) of adhesive layer. Nodes at the right end of the longitudinal path (as
shown in Fig. 5.28), lead to different tensile values. The tensile stress of
1.50mm, 1.25mm and 1.00mm scale mesh FE models is 20.07MPa,
21.37MPa and 22.88MPa respectively. This mesh dependent stress
singularity at the end of adhesively bonding overlap was also found by the
researches [13, 14]. The smaller mesh scale used, the higher tensile stress
was obtained, until FE analysis cannot converge at the location of end
nodes.
122
25
20
tensile stress [MPa]
end nodes
mesh scale=2.00mm
mesh scale=1.50mm
mesh scale=1.25mm
mesh scale=1.00mm
15
10
5
0
-5
-10
0
20
40
60
80
100
Distance [mm]
Fig. 5.28. Tensile stress distribution along the longitudinal path
Further investigation on mesh dependency of FE model through the

thickness of the adhesive layer is carried out. Eight-layer discretization is
defined in the adhesive layer to be compared with the six-layer discretization,
as illustrated in Fig. 5.29.
a) six-layer discretization
b) eight-layer discretization
Fig. 5.29. Mesh configuration through the thickness of adhesive layer
Comparison of the stress distribution between six-layer discretization mesh

and eight-layer discretization mesh is presented in Fig. 5.30 and Fig. 5.31.
Two curves are in conformity with each other, which mean the refined mesh
configuration (eight-layer discretization) through the thickness of the
adhesive layer does not make any difference in the stress distribution along
the concerned longitudinal path. Thus, the six-layer discretization mesh is
enough to obtain reasonably accurate FE results.
123
Chapter5
-2
edge zone
edge zone
Shear stress [MPa]
-4
-6
average shear stress
-8
-10
-12
eight-layer
six-layer
-14
0
20
40
60
80
100
Distance [mm]
Fig. 5.30. Shear stress distribution along the longitudinal path depending on the different
through-thickness meshes
25
edge zone
edge zone
Tensile stress [MPa]
20
eight-layer
six-layer
15
10
-5
-10
0
20
40
60
80
100
Distance [mm]
Fig. 5.31. Tensile stress distribution along the longitudinal path depending on the different
through-thickness meshes
As shown in Fig. 5.30 and Fig. 5.31, the shear stress and tensile stress
distribute relatively stable in the middle part of the bondline, but dramatically
bend over at the edge zone (approximately 10mm from the ends of
adhesive layer). The shear stress peak is approximately twice the average
124
shear stress (calculated by Eq. 5.1). It implies that the failure of the
adhesive joint is triggered by the stress concentration at the edge zone of
the directly forced part (right side in Fig. 5.30 and Fig. 5.31). It can be found
that, even under the proposed pure shear loading condition, the failure of
an adhesive joint is attributed to a combination of shear stress and tensile
stress. Furthermore, the singularity of tensile stress is relatively larger than
the peak of shear stress and they are located at different places of the edge
zone. The FE results also indicate that the bonding quality at the edge zone
of adhesive layer calls for more attention to avoid premature failure of the
adhesive joint under shear condition.
5.3.2 Tensile loading

The ultimate failure loads of nine adhesive joints are given in Table 5.6. It
can be found that the average failure load of T-SP-specimens is close to
that of T-SB-specimens, with a value of about 17.5kN. For T-AC-specimens,
the failure load is slightly lower at 16.05kN, which means the additional
surface pretreatments (using sandpaper or sand blasting) cannot
significantly increase the load-bearing capacity of an adhesive joint under
tensile loading (only 9% higher). However, as stated in Section 5.3.1 under
shear loading condition, both sandpaper and sand blasting surface
pretreatments increase the failure load of adhesive joints more than twice
under shear loading condition. With regard to deviations of test results
(calculated by Eq. 5.2), T-SP-specimens show the largest deviation of all. It
is due to that, the surface pretreatment quality of sandpaper method is
difficult to be controlled. The roughness of surfaces after pretreating is not
uniform throughout the whole section. It definitely influences the bonding
quality between adhesive and adherent, and accordingly deviates the
ultimate failure load and even failure modes (this is discussed in more detail
hereafter). Failure modes of nine adhesive specimens are shown in Fig.
5.32. For all the T-AC-specimens, the loads increase almost linearly up to
failure. The ultimate failure of three specimens always occurs in a brittle and
sudden manner, through the bondline between FRP laminates and adhesive
layer, without any visible signs or sound warning from fibre breaking, as
125
Chapter5
shown in Fig. 5.32 a). For T-SP-specimens and T-SB-specimens, there is

always some sound warning from fibre breaking when the applied load
approaches the ultimate failure load. Failure modes of these are a
combination of fibre breaking and interfacial failure between adhesive layer
and FRP laminations. The typical failure mode is shown in Fig. 5.33. Also,
FRP delamination failure is observed on the T-SP-01 (Fig. 5.34), which
achieves the highest failure load (19.37kN) of adhesive joints. This indicates
that the failure threshold of FRP delamination is a little higher than that of
interfacial failure between adhesive layer and FRP laminates. Furthermore,
from Fig. 5.32 it can be found that the areas of fibre breaking or FRP
delaminating parts are different among T-SP-specimens and T-SBspecimens. The T-SP-01 specimen attains the largest FRP delaminated
area which almost covered the whole bonding area. Therefore, it achieves
the highest load-bearing capacity of adhesive joints. Thus, it can be found
that the area of fibre breaking and FRP delamination is the direct factor
influencing the final strength of an adhesive joint under tensile loading.
Table 5.6. Ultimate failure loads of nine adhesive joints under tensile loading
TACspecimen
Failureload(kN)
TSPspecimen
Failureload(kN)
TSBspecimen
Failureload(kN)
01
15.69
01
19.37
01
17.53
T-AC-01
02
16.43
02
17.93
02
16.05
T-AC-02
03
16.04
03
15.57
03
18.72
Average
16.05
Average
17.62
Average
17.43
T-AC-03
Fig. 5.32. a) Failure mode of T-AC-specimens

126
Deviation
2.37%
Deviation
11.63%
Deviation
7.94%
T-SP-01
T-SP-02
T-SP-03
b) Failure mode of T-SP-specimens
fracture
initiated area
T-SB-01
T-SB-02
T-SB-03
c) Failure mode of T-SB-specimens

Fig. 5.32. Failure modes of adhesive joints
127
Chapter5
Fibre breaking
Fig. 5.33. Combination of fibre breaking and interfacial failure between adhesive layer and
FRP laminates (T-SB-03)
FRP Delamination
Fig. 5.34. Delamination failure in FRP laminates (T-SP-01)
Fig. 5.35 shows the load-displacement curves of nine adhesive joint

specimens under tensile loading, which are measured by the LVDTs. Fig.
5.36 shows the average tensile stress-displacement curves of these
specimens. The average tensile stress is calculated by Eq. 5.3. It is obvious
that curves are almost parallel to each other in the stable load increasing
stage. However, for the initial part slopes of nine curves are different. It can
be explained that at the beginning of loading, the friction between each
component of loading device makes the initial stiffness of specimens
different from each other. However, when the loading is large enough
beyond the friction, the deformation measured by LVDTs is dominated by
the adhesive joint specimen and the stiffness of adhesive joints is
approximately the same. For the T-SP-specimens and T-SB-specimens,
there is a certain extent of residual load-bearing capacity achieved after
reaching the maximum loading, especially for the T-SP-01 and T-SB-03. It is
contributed by the mechanism of fibre breaking and FRP delamination, of
which the failure process is not brittle but gradual until fibres from the whole
128
bonding area disconnect from the fracture interface. As already proven for
the adhesive joints under the shear loading condition (section 5.3.1), the
surface pretreatments cannot significantly influence the stiffness of the
adhesive joint, whereas they change the failure modes and subsequently
increase the load-bearing capacity.
average
Ftension
A
(5.3)
where average is the average tensile stress. Ftension is the tensile load applied
to the adhesive joint.
Load [kN]
20
T-SB-02
T-SB-01
T-SP-01
T-SP-02
T-AC-02
T-AC-03
T-AC-01
15
T-SP-03
10
T-SB-03
0
0.0
0.1
0.2
0.3
0.4
0.5
Displacement [mm]
Fig. 5.35. Load-displacement curves measured from LVDTs
129
0.6
average tensile stress [MPa]
Chapter5
2.5
T-SB-02
T-SB-01
T-SP-01
T-SP-02
T-AC-02
T-AC-03
T-AC-01
2.0
T-SP-03
1.5
1.0
T-SB-03
0.5
0.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
Displacement [mm]
Fig. 5.36. Average tensile stress-displacement curves
Fig. 5.37 shows the measurement of vertical deformation between FRP

sandwich deck and steel support on the T-SB-specimens from four
displacement sensors. It can be clearly seen that the measured data from
displacement sensors DS-01, DS-02, DS-03 and DS-04 are different from
each other, which implies that the loads are not centrally applied. The
relative deformation between FRP sandwich deck and steel support
measured by displacement sensors is rather small, on the order of
magnitude of 0.01mm. With this loading device, the unexpected eccentric
loading cannot be avoided. Even the load is compelled to be exactly centric
on the adhesive joint (as illustrated at the initial stage of the curve of Fig.
5.37 (c)). Afterwards it will be eccentric as the load increased. Fig. 5.37 also
indicates, besides the pure tensile loading, a certain amount of additional
bending moment is also applied during the whole testing process in both
longitudinal and transverse directions of the adhesive joints. Test results on
other specimens present similar mechanical behaviours regarding the
vertical deformation of adhesive joints. Furthermore, the vertical deformation
between FRP sandwich deck and steel support is not increasing linearly
during the whole test. It indicates that the stress redistribution occurs
through the adhesive joint when the applied load is approaching the failure
load. As shown in Fig. 5.32c), it can be found that on the fracture surface of
adhesive layer there are some fracture initiated areas at the edge zone,
130
which reconfirms that the stress redistribution takes place due to the
partially fractured area of the adhesive layer. Until the remaining part of the
adhesively-bonded area cannot carry the total loading anymore, the
adhesive joint fails in a brittle mode.
20
DS-03
DS-01
DS-04
DS-02
Load [kN]
15
10
0
0.000
0.005
0.010
0.015
0.020
Deformation [mm]
5.37. a) T-SB-01
20
DS-02
DS-03
DS-01
DS-04
Load [kN]
15
10
0
0.000
0.005
0.010
0.015
0.020
Deformation [mm]
5.37. b) T-SB-02
131
0.025
0.030
Chapter5
20
DS-02 DS-03
DS-01
DS-04
Load [kN]
15
10
0
0.000
0.005
0.010
0.015
0.020
Deformation [mm]
c) T-SB-03
Fig. 5.37. Load-deformation curves measured from four displacement sensors on the
T-SB-specimens (DS=displacement sensor) under tensile loading
5.3.2.2 FE analysis
1) FE model
Similar to what is stated in section 5.3.1, the 3D FE model is more
preferable for the stress-strain analysis on the adhesive joint than the twodimensional FE model. By employing the three-dimensional FE models, the
behaviour outside the plane can be addressed in a more reliable way. As
shown in Fig. 5.38, the 3D FE model is built up by using ABAQUS 6.8
software. To simplify the FE model and save computational time, only the Lshape steel parts and four bolts for fastening the sandwich deck are
involved in the FE model. Other accessorial steel parts are equivalently
replaced by the corresponding boundary conditions. Eight node linear brick
elements (C3D8R) with reduced integration and hourglass control are
employed for all the modeling works. The total number of elements is
189828. Regarding the adhesive layer, a six-layer discretization is used in
the through-thickness direction to provide accurate stresses and flexible
deforming capacity as well as a three-layer discretization for the FRP
laminates close to the adhesive layer. There is no specific element defined
at the interphase between FRP sandwich deck and adhesive layer, or
between adhesive layer and steel support. All these surfaces are fixed to
each other, assuming that no cracks or relative slip occur during the loading
132
procedure. For the boundary condition of the FE model, as shown in Fig.

5.38, all degrees of freedom on the cross-section of four bolts are restricted,
and the loading of 15kN is applied by the surface tractions through the four
bolt holes on the bottom steel support. The contact pairs are created
between L-shape steel parts and sandwich FRP deck. Four bolts are fixed
to the bolt hole surfaces of L-shape steel parts by TIE command, since it is
assumed that no relative slip occurred between bolts and L-shape steel
parts during the whole loading process. The properties of materials involved
in the FE model are the same as presented in section 5.3.1 (Table 5.5). The
whole analysis process is linear elastic. No geometrical non-linearity or
elastic-plastic material properties are involved in FE analysis. The FE results
can be easily amplified to a suitable value to compare with the test results.
The automatic incrementation scheme is active to make the analysis easily
convergent as well as to optimize the increment sizes based on
computational efficiency [12].
boundary condition
X
Y
Fig. 5.38. Mesh, loading and boundary conditions of FE model under tensile loading
To validate the FE model, the FE results of vertical displacement between
FRP sandwich and steel support at the locations of experimental
displacement sensors are shown in Fig. 5.39, comparing these with
experimental results. The FE results from four locations of DS-01, 02, 03
and 04 deviate much less from each other, since the FE model is perfectly
symmetric, centrally loaded and there is no experimental installation
133
Chapter5
inaccuracy. When it is compared with experimental results of T-SBspecimens, good agreement is obtained. The T-SB-01 and T-SB-02 curves
are close to each other, with a higher stiffness than that of FE model. While
the T-SB-03 curve develops lower than the FE curve. This deviation from
experimental results can be due to the non-homogeneity of adhesive layer,
bonding quality of interphases, as well as additional bending moment
induced by inevitable eccentric loading. But still, the FE results are
acceptably accurate for predicting the mechanical behaviours of the
adhesive joint under tensile loading. Further FE analysis and discussion
depending on this FE model is convictive.
20
Load [kN]
16
12
FEM
T-SB-01
T-SB-02
T-SB-03
0
0.000
0.003
0.006
0.009
0.012
0.015
0.018
0.021
Deformation [mm]
Fig. 5.39. Comparison with FE analysis and experimental results at the locations of
3) FEA results and discussion
Fig. 5.40. Von Mises stress distribution and deformation throughout the adhesive joint
134
Fig. 5.40 shows the global Von Mises stress distribution and deformation
throughout the adhesive joint. The Von Mises stress v is calculated by the
equation (5.4), where 1, 2 and 3 are principal stresses in three directions.
Deformation is amplified to 50 times of the original value, the purpose of
which is to clearly present the local deformation at the corners of the
adhesive layer. A large portion of deformation is evident in the FRP
sandwich deck as well as at the edges of the adhesive layer. There is,
however, no visible deformation occurring in the steel supports, since the
stiffness of steel material is non-comparatively higher than that of other
composed materials. With regard to the adhesive layer, stress concentrates
at the ends against the surface of FRP laminates, where stress singularity
takes place and cracks initiate.
v
( 1 2 ) 2 ( 2 3 ) 2 ( 1 3 ) 2
2
(5.4)
In order to better understand the stress distribution across the interface

between FRP laminates and adhesive layer, contour maps of tensile stress
and shear stress are drawn through this plane, as shown in Fig. 5.41 and
Fig. 5.42. The X-axis is along the longitudinal direction of the FRP sandwich
deck (the ordinate is shown in Fig. 5.38), while the Y-axis is in the
transverse direction. In the longitudinal direction, as can be predicted, the
tensile and shear stress concentrates at the ends of the interface. The
variation rate of the stress field is rough at the edge zone, whereas both
tensile and shear stress distribution in the central part of interface is more
stable, with the values round zero. This means that the load is dominantly
carried through the two highly loaded end zones, separated by a lightly
loaded trough. This result confirms the conclusion that increasing the length
of overlap of adhesively-bonded joints cannot always result in the strength
improvement of the whole adhesive joint. There is an upper limit of the joint
strength as a function of overlap length. Thus, the conclusion can be drawn
that the bonding defect in the central zone of adhesive layers does not
significantly influence the ultimate failure loads of joints. On the contrary, the
bonding quality at the edge zone of the adhesive layer calls for more
attention to avoid premature failure of the adhesive joint.
135
Chapter5
y/90mm
x/90mm
Fig. 5.41. Contour map of tensile stress on the interface between FRP laminates and
adhesive layer under the tensile loading of 15kN
y/90mm
x/90mm
Fig. 5.42. Contour map of shear stress on the interface between FRP laminates and
adhesive layer under the tensile loading of 15kN
Regarding the transverse direction, both tensile and shear stress are
distributed at a rather steady gradient. Worthy mentioning here is that the
maximum absolute values of stress are not located exactly at the ends of
136
the adhesive layer, but a small distance away from the corners. Comparing
Fig. 5.42 with Fig. 5.41, shows that even under the pure tensile loading
condition there still is a comparative scale of shear stress distributed across
the interface between FRP laminates and adhesive layer, which implies that
the failure of adhesive joints in the experiments is induced by the
combination of tensile stress and shear stress, but not only the tensile stress.
Further investigation was conducted on mesh-dependency of the FE model
under tensile loading. The same condition as the shear loading, the part
concerned is the adhesive layer in both longitudinal direction and transverse
direction, as shown in Fig. 5.43. Four mesh configurations were selected
with mesh scales of 2.00mm, 1.50mm and 1.25mm and 1.00mm.
a) Mesh scale = 2.00mm
b) Mesh scale =1.50mm
c) Mesh scale = 1.25mm
d) Mesh scale = 1.00mm

Fig. 5.43. Different mesh configurations in the adhesive layer
Based on the failure modes of tested adhesive joints under tensile loading,
the concerned location of the longitudinal path is on the interface line
between adhesive layer and steel support. As shown in Fig. 5.44, the shear
and tensile stress distributions are extracted for comparison from the
longitudinal path.
137
Chapter5
Fig. 5.44. Longitudinal path for analysing the stress distribution
Tensile stress and shear stress distributed along the longitudinal path are
shown in Fig. 5.45 and Fig. 5.46. For the tensile stress distribution, mesh
refined FE models do not show a highly effective improvement, which
means the 2.00mm scale mesh can offer sufficient deformation flexibility of
the adhesive layer. But at the ends of the adhesive layer, the small scale
mesh tends to increase the tensile stress singularity, with a value of the
1.00mm scale mesh (21.58 MPa) 19.5% higher than that of the 2.00mm
scale mesh (18.06 MPa). It can be predicted that the stress singularity will
go to higher values when much smaller elements are proposed at the ends
of the adhesive layer. But based on the former researches [13, 14], this
stress singularity is not realistic. For the shear stress distribution (Fig. 5.46),
the curves of 1.00mm, 1.25mm and 1.5mm scale mesh are almost the same
along the whole longitudinal path, even at the ends of the adhesive layer. A
slight deviation is evident for the 2.00mm scale mesh FE model. However,
differences between the 2.00mm scale mesh curve and the other three
curves are very limited.
138
25
edge zone
edge zone
mesh scale=2.00mm
mesh scale=1.50mm
mesh scale=1.25mm
mesh scale=1.00mm
tensile stress [MPa]
20
15
10
average tensile stress
20
40
60
80
100
Distance [mm]
Fig. 5.45. Tensile stress distribution along the longitudinal path

edge zone
edge zone
8
6
Shear stress [MPa]
4
2
0
-2
mesh scale=2.00mm
mesh scale=1.50mm
mesh scale=1.25mm
mesh scale=1.00mm
-4
-6
-8
0
20
40
60
80
100
Distance [mm]
Fig. 5.46. Shear stress distribution along the longitudinal path
To fully investigate the mesh dependence of the FE model, mesh

configurations through the thickness of the adhesive layer are also
139
Chapter5
considered. As shown in Fig. 5.47, six-layer discretization and eight-layer

discretization are selected for the study. The FE results are presented in Fig.
5.48 and Fig. 5.49. Similar to what is stated in section 5.3.1, the FE model
with eight-layer elements in the through-thickness direction develops almost
the same curve as that of the six-layer discretization mesh, except a slight
deviation at the ends of the adhesive layer. Depending on the investigations
of mesh dependency of FE model, the preferable mesh configuration is the
1.5mm element scale with six-layer discretization through the thickness of
the adhesive layer, which can present the satisfactory solution and
meanwhile optimize the computational time. Thus, this mesh configuration is
used for further FE analysis on adhesive joints under monotonic loading in
combination of shear and tension, which is addressed in the following
section.
a) six-layer discretization
b) eight-layer discretization
Fig. 5.47. Mesh configuration through the thickness of adhesive layer
20
eight-layer
six-layer
15
10
20
40
60
80
100
Distance [mm]
Fig. 5.48. Tensile stress distribution along the longitudinal path (through-thickness mesh)
140
8
Shear stress [MPa]
6
4
2
0
-2
-4
-6
eight-layer
six-layer
-8
0
20
40
60
80
100
Distance [mm]
Fig. 5.49. Shear stress distribution along the longitudinal path (through-thickness mesh)
5.3.3 Combination of shear and tension loading

Table 5.7 lists failure loads of adhesive joints under four load angles (18,
36, 54 and 72). For comparison, test results of adhesive joints (S-SBspecimens and T-SB-specimens) under pure tensile and pure shear loading
conditions are also listed in Table 5.7. The lowest load-bearing capacity of
adhesive joints is obtained under the 18 angle loading condition, with an
average failure load of 11.9kN. The load-bearing capacity is gradually
increasing from 11.9kN to 23.27kN as the loading angle increases from 18
to 72. The largest deviation obtained by Eq. 5.2 is 16.5% under the 36
loading angle, while the smallest deviation obtained is 5.55% under the 54
loading angle.
141
Chapter5
Table 5.7. Ultimate failure loads of adhesive joints under six loading conditions
Puretensile
Failureload(kN)
18angle
Failureload(kN)
36angle
Failureload(kN)
54angle
Failureload(kN)
72angle
Failureload(kN)
Pureshear
Failureload(kN)
01
17.53
01
12.7
01
11.2
01
16.5
01
26.6
01
51.0
02
16.05
02
11.2
02
14.6
02
17.7
02
20.0
02
92.9
03
18.72
03
11.8
03
11.8
03
16.1
03
23.2
03
64.1
Average
17.43
Average
11.9
Average
12.53
Average
16.77
Average
23.27
Average
69.3
Deviation
7.94%
Deviation
6.72%
Deviation
16.5%
Deviation
5.55%
Deviation
14.3%
Deviation
34.0%
To easily recognize the load combining effects, the total failure load is
vectorially separated into shear load and tensile load, with regard to the
loading angle of each loading condition, as shown in Fig. 5.50. The
horizontal axis represents the shear load applied to the adhesive joint, while
the vertical axis is the tensile load. It is apparent that, under combination of
tensile and shear loading, the load-bearing capacity of adhesive joints
decreases as compared to that of pure tensile and pure shear loading
conditions. From Fig. 5.50, the failure load of adhesive joints under pure
shear loading (69.3kN) is considerably higher than other loading conditions,
due to the different failure mode, which is discussed in detail hereafter. It is
manifest that the failure load of an adhesive joint under different combined
loading conditions is more sensitive to the ratio of tensile load vectorially
separated from the resultant force.
142
100
90
o
80
tensile load [kN]
70
18
60
50
36
40
54
30
20
72
10
0
0
10
20
30
40
50
60
70
80
90
100
90
shear load [kN]
Fig. 5.50. Failure loads of adhesive joints under the combining loading of tensile and shear
To investigate the failure criterion of adhesive joints, an ellipsoid function is

employed for the curve fitting by the least square method. Firstly,
considering all the test results of six loading conditions, the predictive curve
(Eq. 5.5) is obtained and indicated as a solid line in Fig. 5.51. It can be
found that the agreement between the test results and the predictive
equation is very bad. All the test data of four combining loading conditions
are below the predictive curve, which implies that the predictive equation is
not conservative for the design of adhesive joints. To solve this issue, only
the test results of four combinations of shear and tensile loads are selected
to perform as the basic curve fitting data, since involving the test results of
the pure tensile and pure shear loading conditions will make the predictive
curve deviate too much from the four combination loading conditions. The
modified predictive equation is expressed by Eq. 5.6, and indicated as a
dashed line in Fig. 5.51. In this way, the failure criterion is more
conservative and practical, since in the real application of an adhesive joint
between FRP decks and steel girders, the joint mainly serves under the
combination loading of tension and shear.
2
Predictive curve:

1
1.54 9.59
143
(5.5)
Chapter5
2
Modified predictive curve:

Average tensile stress [MPa]

1
1.41 3.18
(5.6)
test data
predictive curve (Eq.5.5)
modified predictive curve (Eq.5.6)
5
4
3
2
1
0
0
10
11
Fig. 5.51 Shear-tensile failure criterion for the adhesively-bonded joint
It is known that the stress distribution is not uniform either through the
interface between the adhesive layer and the steel support or through the
interface between the FRP sandwich deck and the adhesive layer. Regularly,
the stress concentrates at the ends of the adhesive layer. Thus, the sheartensile failure criterion needs to be modified with the stress non-uniform
distribution factor, which needs to be determined by the FE parametric study
of adhesive joints in future work.
The failure modes of adhesive joints under four combinations of tensile and
shear loads are illustrated in Fig. 5.52. All the fracture planes are through
the interface between the FRP sandwich deck and adhesive layer, which is
the same as that observed for the adhesive joint under pure tensile loading
(section 5.3.2). Some area of fibre breaking or FRP delamination is evident
from the view of the FRP laminate failure surface. But these areas do not
fully cover the adhesively-bonding area. This failure mode can be defined as
the combination of fibre breaking (or FRP delamination) and interfacial
adhesion failure between FRP sandwich deck and adhesive layer. It is
worthwhile mentioning that, for the 72 angle loading condition, there are
cracks observed in the adhesive layer of the 72-SB-01 and propagate
through the interface between adhesive layer and steel support, as
illustrated in Fig. 5.53. It indicates that the failure plane almost switches to
the interface between adhesive layer and steel support, which occurred for
the adhesive joints under the pure shear loading condition. This
144
phenomenon suggests that the upper and lower interfaces between

adhesive layer and FRP laminates or steel support almost achieve the
failure homogenously. However, for the other two specimens under the 72
loading angle, cracks in the adhesive layer are not visible. Instead, a large
portion of FRP delamination or fibre breaking area is evident in Fig. 5.54.
Once again, the test results confirm that the controllable adhesively-bonding
technique is essential to guarantee the mechanical performance of the
adhesive joints.
18-SB-01
18-SB-02
Fig. 5.52. a) 18 angle
36-SB-01
36-SB-02
Fig. 5.52. b) 36 angle
145
18-SB-03
36-SB-03
Chapter5
54-SB-01
54-SB-02
c) 54 angle
54-SB-03
72-SB-02
d) 72 angle
72-SB-03
cracks
72-SB-01
Fig. 5.52. Failure modes of adhesive joints under four loading angles
146
cracks
Fig. 5.53. Cracks in the adhesive layer of 72-SB -01 under the 72 angle loading
fibre breaking
FRP delamination
Fig. 5.54. FRP delamination and fibre breaking in the FRP sandwich deck of 72-SB-03
under 72 angle loading
5.3.3.2 FE analysis
1) FE model
The FE model of adhesive joints under the combination of tensile and shear
loads is developed by using ABAQUS 6.8. The simplification of an FE model
is the same as that of the shear loading condition (section 5.3.1.2), except
the loading and boundary condition definition. As shown in Fig. 5.55, the
shear and tensile loads are applied by the surface tractions on the loading
area respectively, while all the degrees of freedom are restricted for the
reaction area. The center of shear and tensile load is exactly through the
centroid of the adhesive layer, in such a way as to confirm that the resultant
force is also through the centroid of the adhesive layer and no additional
bending moment is proposed. Subsequently, four loading angles can be
147
Chapter5
realized by varying the ratio between shear load and tensile load with
specific tangent values. Depending on the investigations of mesh
dependence of an FE model under shear and tensile loading conditions
(sections 5.3.1 and 5.3.2), the FE model with 1.50mm mesh scale and sixlayer discretization through the thickness of the adhesive layer is preferable
to achieve reasonable accuracy as well as to save computational time. Thus,
this mesh configuration is continuously employed for FE analysis in this
section, as shown in Fig. 5.56 and Fig. 5.57. All the elements used are
C3D8R. No geometric non-linearity or elastic-plastic material properties are
involved in the FE analysis. The input of material properties is the same as
listed in Table 5.5. The resultant force of 15kN (vectorial combination of
tensile and shear load) is applied on the FE model. The FE results can be
amplified with any ratio to be comparable with test results.
loading
area
reaction
area
Fig. 5.55. Loading and boundary condition of FE model
148
X
Y
Fig. 5.56. FE model
Adhesive
layer
Fig. 5.57. Mesh configuration of adhesive joint
Fig. 5.58 illustrates the FE results of relative deformation between FRP
sandwich deck and steel support at the locations of experimental
displacement sensors under four loading angles, compared with the mean
value of experimental results. The deformation values in the minus area
represent the mean value of measured deformation from DS-02 and DS-04,
while the positive area represents the mean value of measured deformation
from DS-01 and DS-03. The location of displacement sensors is illustrated
in Fig. 5.12. For the 54-SB-02 of the 54 angle loading condition, DS-01
and DS-03 displacement sensors are not installed firmly and they drop off
during the test. Thus, the test data are absent in Fig. 5.58 c). From Fig. 5.58
it can be found that at the initial stage of loading process good agreement is
obtained between FE results and experimental data. However, as the test
proceeds, some deviations occur for both tensile and shear deformation.
149
Chapter5
The curves of test results are nonlinear, especially for the mean values of
DS-02 and DS-04. This is due to the inhomogeneity of adhesive material
and Basal wood. During the whole test, the stress in the adhesive joint is not
ideally distributed, but keeps redistributing depending on the stiffness of the
components. The three test curves of each loading angle do not exactly
match with each other, since the extent of porosity in the adhesive layer and
Basal wood deviate from each other. Based on the above discussion,
considering the manufacture tolerance, inhomogeneity of materials involved
and the scale of measured deformation, the agreement obtained between
FE results and experimental data is reasonably acceptable. The FE model is
therefore reasonably acceptable to be used in further analysis.
16
14
Load [kN]
12
10
8
6
SB01
SB02
SB03
FEM
4
2
0
-0.10
-0.08
-0.06
-0.04
-0.02
0.00
0.02
0.04
Deformation [mm]

16
14
Load [kN]
12
10
8
6
SB01
SB02
SB03
FEM
4
2
0
-0.10
-0.08
-0.06
-0.04
-0.02
0.00
Deformation [mm]

150
0.02
0.04
18
16
14
Load [kN]
12
10
8
6
SB01
SB02
SB03
FEM
4
2
0
-0.10
-0.08
-0.06
-0.04
-0.02
0.00
0.02
0.04
Deformation [mm]
c) 54 angle
30
25
Load [kN]
20
15
10
SB01
SB02
SB03
FEM
0
-0.10
-0.08
-0.06
-0.04
-0.02
0.00
0.02
0.04
Deformation [mm]
d) 72 angle
Fig. 5.58. Comparison on experimental results and FE analysis at the locations of
3) FE results and discussion

Fig. 5.59 and Fig. 5.60 illustrate the contour map of tensile and shear stress
on the interface between FRP laminates and adhesive layer under four
loading angles. For comparison, the tensile and shear stress states under
the shear and tensile loading conditions are also included. For all the six
angle loading conditions, in the transverse direction (the Y axis as indicated
in Fig. 5.56), tensile and shear stress are distributed more uniformly. The
maximum absolute stress values are not located exactly at the ends of the
151
Chapter5
interface, but at 9mm distance from the end. To further investigate the
stress distribution in the longitudinal direction (the X axis as indicated in Fig.
5.56), the tensile and shear stress is extracted through the longitudinal path
at the location of 9mm distance from the end, as shown in Fig. 5.61 and Fig.
5.62.
a) 0 angle loading (tensile)
b) 18 angle loading
c) 36 angle loading
d) 54 angle loading
e) 72 angle loading
f) 90 angle loading (shear)
Fig. 5.59. Contour map of tensile stress on the interface between FRP laminates and
adhesive layer under six loading conditions
152
a) 0 angle loading (tensile)
b) 18 angle loading
c) 36 angle loading
d) 54 angle loading
e) 72 angle loading
f) 90 angle loading (shear)
Fig. 5.60. Contour map of shear stress on the interface between FRP laminates and
adhesive layer under six loading conditions
153
Chapter5
25
tensile
o
18
o
36
o
54
o
72
shear
20
15
10
5
0
-5
-10
-15
0
20
40
60
80
100
Distance [mm]
Fig. 5.61. Tensile stress distribution in the longitudinal path at a location 9mm from the
end of the interface
tensile
o
18
o
36
o
54
o
72
shear
10
Shear stress [MPa]
-5
-10
-15
0
20
40
60
80
100
Distance [mm]
Fig. 5.62. Shear stress distribution in the longitudinal path at a location 9mm from the end
of the interface
From Fig. 5.61, it can be found that for all six loading conditions the tensile
stress is approximately zero in the central part of the interface in the
154
longitudinal direction (the X axis as indicated in Fig. 5.56). However, in the

vicinity of adhesive layer ends, the tensile stress concentration is evident.
As the loading angle rotates from 0 (tensile) to 90 (shear), the tensile
stress singularity decreases at both ends of the interface, since the
vectorially separated tensile load applied on the FE model of adhesive joint
decreases. For the right end (shear load applied side), the extent of tensile
stress decreasing is more significant than that on the left side, which
introduces an asymmetrical tensile stress state throughout the longitudinal
path. Subsequently, the tensile stress singularity at the right end drops to
the minus zone for the shear loading condition.
Fig. 5.62 illustrates the shear stress distribution along the longitudinal path.
The tensile and shear loading condition achieve the two extreme shear
stress states. The shear stress distributions of the other four angle loading
conditions regularly transit from tensile loading condition to shear loading
condition. In the central part of the interface, the shear stress values of four
loading angles are not zero anymore, the absolute values of which are
increasing with the vectorially separated shear load increased from the 18
angle loading to 72 angle loading. Without any doubt, the absolute value of
shear stress achieves the maximum for the shear loading condition (90
angle). For the right end (shear load applied side), the variation of the shear
stress state is more significant than that on the left side, since the shear
stress at the load directly forced side is more sensitively influenced. To the
contrary, the shear stress distribution at the left side does not vary too much
among six loading conditions. The three curves of 0 (tensile), 18 and 36
loading conditions almost cover each other.
5.4 Mechanical behaviour of adhesively-bonded joints

after hydrothermal aging
5.4.1 Shear loading
For the aged adhesive joints under shear loading, the failure mode is brittle,
without any visible signs or sound warning from fibre breaking. The loaddisplacement curves of three specimens are shown in Fig. 5.63. Similar to
what was discussed in section 5.3.1.1 for the un-aged specimens under
155
Chapter5
shear loading, the initial stage with a higher slope is due to the friction
between the L-shape steel profile and the FRP sandwich deck. Fig. 5.63
also indicates that the slopes of the two-stage load-displacement curves are
close to each other. Only the friction of each specimen test is a bit different,
which compels the test curves to bend at different locations.
50
Load [kN]
40
30
20
S-SB-A-1
S-SB-A-2
S-SB-A-3
10
0
0
Displacement [mm]
Fig. 5.63. Load-displacement curves of aged specimens under shear loading
As shown in Fig. 5.64, the failure of adhesive joints is a cohesive fracture

(adjacent to the interface between the adhesive layer and the steel support)
in the adhesive layer, which confirms that the sand blasting surface
pretreatment method results in a good quality bonding between the
adhesive layer and the steel surface, even in the hydrothermal aging
environment. The failure mode also indicates that the absorbed moisture in
the hydrothermal aging environment does not significantly degrade the
interfacial strength between the adhesive layer and the steel support.
156
S-SB-A-01
S-SB-A-02
S-SB-A-03
Fig. 5.64. Failure mode of the aged adhesively-bonded joint under shear loading
Table 5.8 lists the ultimate failure loads of three adhesive joint specimens.
Their average value of them is 41.9 kN with the deviation (calculated by Eq.
5.2) of 18.1%.
Table 5.8. Ultimate failure loads of three aged adhesive joints under shear loading
SSBAspecimen
Failureload(kN)
01
49.5
02
37.7
03
38.4
Average
41.9
Deviation
18.1%
Fig. 5.65 shows the failure modes from the view of steel supports. From Fig.
5.65 a) it can be easily found that there is a fracture initiated line (white
colour) locating a small distance from the edge of the steel support. As
indicated by the FE analysis of stress distribution on the interface between
steel support and adhesive layer (section 5.3.1.2), the shear stress peak is
more or less at the same location as this yielding line. This means the failure
of the whole adhesive joint is initiated at this spot and develops to the
vicinity to trigger the final failure of the joint. However, significantly different
from the S-SB-A-01 specimen, there is no visible fracture initiated line on
the surface of the S-SB-A-03 specimen (see Fig. 5.65c)). Instead, there is a
big adhesive fracture initiated area at the right corner of the whole adhesivebonding area, which suggests that cracks firstly occur at the right corner and
157
Chapter5
gradually propagate into the middle of adhesive layer as the load increases
until the rest of the adhesively-bonding area cannot carry the load anymore.
Then, the whole adhesive joint fails in a sudden way. The failure mode of
the S-SB-A-03 specimen may be due to the local non-homogeneity of the
adhesive layer, where the stress concentration is located and subsequently
changes the stress distribution across the section and induces the
premature failure of the adhesive joint. As can be seen in Fig. 5.63, the
load-displacement curve of the S-SB-A-03 specimen is a little more flexible
than that of the other two specimens, which can be attributed to its specific
failure mode. Furthermore, for the S-SB-A-02 specimen (Fig. 5.65b)), both
the fracture initiated line and fracture initiated area are visible on the surface
of the steel support. This failure mode can be considered to be a
combination of the S-SB-A-01 specimen and the S-SB-A-03 specimen.
Clearly, the failure mode of the S-SB-A-01 specimen is more preferable,
which results in the highest load-carrying capacity (49.5kN) of the three
adhesive joints under shear loading.
158
fracture initiated line
a) S-SB-A-01
fracture initiated line and
fracture initiated area
b) S-SB-A-02
fracture initiated area
c) S-SB-A-03
Fig. 5.65. Failure modes of aged specimens from the view of steel surfaces
5.4.2 Tensile loading

Load-displacement curves of aged adhesively-bonded joints under tensile
loading are shown in Fig. 5.66. It can be seen that the curves of the T-SB-A2 and T-SB-A-3 specimens are very close to each other both for slopes and
maximum loads. However, the T-SB-A-1 specimen develops a rather
flexible curve and the maximum load obtained is also quite low. This may be
159
Chapter5
attributed to the non-homogeneity of FRP laminates and adhesive materials.

The ultimate failure loads of three aged adhesive joints under tensile loading
are summarized in Table 5.9. Their average value is 11.0 kN with a
deviation of 16.4% (calculated by Eq.5.2). The average tensile strength is
1.36MPa (calculated by Eq.5.3).
12
Load [kN]
10
T-SB-A-1
T-SB-A-2
T-SB-A-3
0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
Displacement [mm]
Fig. 5.66. Load-displacement curves of aged specimens under tension loading

Table 5.9. Ultimate failure loads of three aged adhesive joints under tensile loading
TSBAspecimen
Failureload(kN)
01
9.2
02
11.7
03
12.0
Average
11.0
Deviation
16.4%
The failure mode of the aged adhesively-bonded joint under tensile loading
is shown in Fig. 5.67. It is the combination of FRP delamination and fibre
breaking on the surface of an FRP sandwich deck, as illustrated in Fig. 5.68.
The ultimate failure of these three specimens does not occur suddenly, but
there is always some sound warning from fibre breaking or FRP
delamination when the applied load is approaching the ultimate failure load.
After achieving the ultimate failure load, there is still some residual loadbearing capacity of the adhesive joints (as indicated by the loaddisplacement curves in Fig. 5.66), due to the gradual delamination of FRP
laminates. From Fig. 5.67 it can be seen that the fibre breaking area is
uniformly distributed and almost covers the whole adhesively-bonding area.
160
T-SB-A-01
T-SB-A-02
T-SB-A-03
Fig. 5.67. Failure mode of the aged adhesively-bonded joint under tensile loading
fibre breaking
FRP delamination
Fig. 5.68. FRP delamination and fibre breaking in the FRP sandwich deck of the aged
specimen under tensile loading
161
Chapter5
5.4.3 Combination loading of shear and tension

Failure loads of adhesive joints under four load angles (18, 36, 54 and
72) are listed in Table 5.10. For comparison, the test results of shear and
tensile loading conditions are also involved. During the test of the 54-SB-A03 specimen, some operation mistakes occur. Hence, the test result of the
54-SB-A-03 specimen is not valid and excluded from the Table 5.10.
Beside the tensile loading condition, the lowest load-bearing capacity of
adhesive joints after hydrothermal aging is obtained under the 18 angle
loading condition, with the average failure load of 11.5kN. The load-bearing
capacity is increasing from 11.5kN to 25.6kN as the loading angle increases
from 18 to 72. The deviations (calculated by Eq. 5.2) of all these four
groups of specimens are within 10%, which indicates that the test results
are repeatable and reliable.
Table 5.10. Ultimate failure loads of aged adhesive joints under four loading conditions
TSBAspecimen
Failureload(kN)
18SBAspecimen
Failureload(kN)
36SBAspecimen
Failureload(kN)
54SBAspecimen
Failureload(kN)
72SBAspecimen
Failureload(kN)
SSBAspecimen
Failureload(kN)
01
9.2
01
12.1
01
13.5
01
13.9
01
26.5
01
49.5
02
11.7
02
11.5
02
14.9
02
14.6
02
26.4
02
37.7
03
12.0
03
10.8
03
13.7
03
03
23.8
03
38.4
Average
11.0
Average
11.5
Average
14.0
Average
14.3
Average
25.6
Average
41.9
Deviation
16.4%
Deviation
6.1%
Deviation
6.4%
Deviation
2.8%
Deviation
7.0%
Deviation
18.1%
To easily recognize the combining effects of tensile and shear loading, the
total failure load is vectorially separated into shear load and tensile load,
with regard to the loading angle of each loading condition, as shown in Fig.
5.69. The horizontal axis represents the shear load applied to the aged
adhesive joint, while the vertical axis represents the tensile load. It can be
found that the vectorially separated tensile loads are close to each other for
the aged specimens tested under tensile, 18 angle and 36 angle loading.
This implies that under these three loading conditions, the vectorially
separated tensile load dominates the failure load. For the other two loading
conditions (54 and 72), it is the combination of both tensile and shear
162
loading that triggers the final failure of aged adhesive joints, but not only the
vectorially separated tensile load.
50
tensile load [kN]
40
18
30
36
20
54
10
72
0
0
10
20
30
shear load [kN]
90
40
50
Fig. 5.69. Failure loads of aged adhesive joints under a combination of tensile and shear
loading
To investigate the failure criterion of aged adhesive joints, an ellipsoid

function is employed for the curve fitting by the least square method. All the
test results of the aged specimens are included to perform as the basic
curve fitting data. The average shear stress is calculated by Eq. 5.1, while
the average tensile stress is calculated by Eq. 5.3. The best fit shear-tensile
failure criterion of the aged adhesive joints is as follows:
2

1
1.3 4.6
(5.7)
Fig. 5.70 shows the predictive curve of the shear-tensile failure criterion of
the aged adhesive joints together with the test results.
163
Chapter5
Average tensile stress [MPa]
test data
predictive curve
0
0
Fig. 5.70. Shear-tensile failure criterion for the aged adhesively-bonded joint
The failure modes of adhesive joints under four combinations of tensile and
shear loads are illustrated in Fig. 5.71. The failure mode of the aged
adhesive joint under 18, 36 and 54 angle loading conditions is the
combination of FRP delamination and fibre breaking, which is the same as
that of specimens under tensile loading. The fibre breaking area always fully
covers the whole adhesively bonded area. For the 72 angle loading
condition, the failure mode of the 72-SB-A-01 specimen is the cohesive
failure in the adhesive layer, which is as the same as that of specimens
under shear loading. It indicates that the failure plane switches from the
FRP delamination to the cohesive fracture in the adhesive layer. The
vectorially separated shear load dominates the failure load of the 72-SB-A01 specimen. Fig. 5.72 clearly shows the cohesive failure in the adhesive
layer and residual adhesive material left on the surface of the steel support.
The other two aged specimens tested under the 72 angle loading condition
fail in the same mode (FRP delamination and fibre breaking) as other
loading angles. This phenomenon suggests that the 72 angle loading
condition is approximately the marginal loading angle of the tensile-shear
combining ratio, which switches the failure mode from the typical tensileloading failure mode (FRP delamination and fibre breaking) to the typical
shear-loading failure mode (cohesive facture in the adhesive layer).
164
18-SB-A-01
18-SB-A-02
18-SB-A-03
36-SB-A-01
36-SB-A-02
165
36-SB-A-03
Chapter5
54-SB-A-01
54-SB-A-02
c) 54 angle
72-SB-A-01
72-SB-A-02
72-SB-A-03
d) 72 angle
Fig. 5.71. Failure modes of aged adhesive joints under four load angles
166
Fig. 5.72. Cohesive failure in the adhesive layer of 72-SB-A-01 specimen
5.5 Comparison of mechanical behaviours of FRP-to-steel

adhesively-bonded joints before and after hydrothermal
aging
5.5.1 Ultimate failure load and shear-tensile failure criterion
Ultimate failure loads of the un-aged and aged specimens under six angle
loading conditions are listed in Table 5.11. For the tensile and shear loading
conditions, the failure loads achieved by the un-aged adhesive joint
specimens are definitely higher than those of the aged specimens, 58.2%
higher for tensile loading and 65.4% higher for shear loading. This indicates
that the absorbed moisture content in FRP laminates and adhesive layer
tend to deteriorate the mechanical performance of the adhesively-bonded
joints. However, for the tensile/shear combining loading conditions, the unaged and aged specimens achieve almost the same failure loads. For some
cases (36 and 72 angle loading), the failure loads of aged specimens are
even a little higher than those of un-aged specimens. Unlike the tensile and
shear loading, the hydrothermal aging does not dramatically decrease the
failure loads of adhesive joints under the combining loading of tensile and
shear, but to the contrary, it slightly increases the load-bearing capacity of
joints under 36 and 72 angle loading. These results can be due to the post
curing mechanism. Post curing is the process of exposing polymeric
materials to elevated temperatures to accelerate the curing process and to
maximize some of the materials physical properties by expediting the cross167
Chapter5
linking process and properly aligning the polymer's molecules. This is

usually done after the material has cured at room temperature. In general,
post curing is employed to achieve the full cure of polymeric materials,
which cannot be easily realized at room temperature. As stated in
researches [15, 16], the post curing can produce significant improvements
on the mechanical properties of polymeric materials. Here, the hydrothermal
aging environment (40C-water) offers the elevated temperature to postcure the FRP laminates and adhesive layer, in this way increasing the loadbearing capacity of adhesive joints. For the shear loading condition, the
post-curing efforts are not that obvious. It may be due to the fact that, under
shear loading, the failure mode of adhesive joints is the cohesive fracture in
the adhesive layer. As compared to the improvement by the post curing, the
absorbed moisture is more dramatic in degrading the mechanical properties
of adhesive materials.
Also, deviations of test results of the aged specimens are significantly lower
than those of un-aged specimens in almost all loading conditions except the
tensile loading, which also indicates that the post curing mechanism
improves the bonding quality of adhesive joints to minimize the test result
deviations.
Table 5.11. Comparison of ultimate failure loads of un-aged and aged adhesive joints
Loading
condition
Tensile
18angle
36angle
54angle
72angle
Shear
Unagedspecimen
Agedspecimen
Failureload(kN) Deviation Failureload(kN) Deviation
17.4
7.9%
11.0
16.4%
6.1%
11.9
6.7%
11.5
12.5
16.5%
14.0
6.4%
16.8
5.6%
14.3
2.8%
23.3
14.3%
25.6
7.0%
69.3
34.0%
41.9
18.1%
Fig. 5.73 illustrates the vectorially separated tensile and shear failure loads
of un-aged and aged adhesive joints under six loading conditions. Fig. 5.74
shows the comparison of the shear-tensile failure criterions for un-aged and
aged adhesively-bonded joints. In the tensile load dominated area, the
predictive failure criterion curve of un-aged specimens is slightly higher than
that of the aged specimens. To the contrary, in the shear load dominated
area, the predictive failure criterion curve of aged specimens is higher than
168
the modified predictive curve of un-aged specimens. In general, the fourmonth hydrothermal aging does not significantly deteriorate the loadcarrying capacity of adhesive joints. However, in Chapter 4, a relatively
larger loss of strength of FRP laminates after hydrothermal aging is
observed. In conclusion, the absorbed moisture content tends to degrade
the mechanical properties of FRP laminates and adhesives, while the post
curing process tends to upgrade the mechanical properties of FRP
laminates and adhesives, as well as the interfacial bonding quality. These
two mechanisms occur homogeneously when the adhesive joints are
exposed to the hydrothermal aging environment (40C-water). The failure
load-bearing capacity as well as failure modes of adhesive joints are
controlled by whichever mechanism (moisture absorption or post curing) is
dominant. Further discussion about this is to be found hereafter in section
5.5.3 of Failure mode.
100
90
tensile load [kN]
80
un-aged
aged
70
18
60
50
36
40
54
30
20
72
10
0
0
10
20
30
40
50
60
shear load [kN]
70
80
90
90
100
Fig. 5.73. Comparison on vectorially separated failure loads of un-aged and aged adhesive
joints under six loading conditions
169
Chapter5
average tensile stress [MPa]
test data of un-aged specimens

test data of aged specimens
predictive curve of un-aged specimens
modified predictive curve of un-aged specimens
predictive curve of aged specimens
6
5
4
3
2
1
0
0
10
11
average shear stress [MPa]
Fig. 5.74. Comparison on the shear-tensile failure criterions for un-aged and aged
adhesively-bonded joints
5.5.2 Failure mode

For the aged and un-aged adhesive joints tested under shear loading, the
failure mode is the cohesive failure in the adhesive layer, located close to
the steel support surface (Fig. 5.75). The locations of the fracture initiated
lines are close to each other. As already discussed in section 5.5.1, the
failure load of aged specimens is significantly lower than that of un-aged
specimens, which means that for the adhesive material, the mechanical
degradation due to the absorbed moisture content is much larger than the
property improvement by the post curing process.
a) Un-aged specimen
b) Aged specimen
Fig. 5.75. Failure mode of un-aged and aged adhesive joints under shear loading from the
view of the steel support surface
For the other five angle loading conditions, the failure mode is different
between the aged and un-aged specimens. The failure mode of un-aged
adhesive joints is the combination of partial fibre breaking or local FRP
170
delamination and interface failure between FRP laminates and adhesive

layer, as shown in Fig. 5.76 a). The interfacial fracture is due to the
insufficient bonding quality between FRP laminates and adhesive layer.
However, for the adhesive joints after four-month hydrothermal aging, the
weakest plane of adhesive joints moves to the FRP laminates. The failure
mode of aged specimens is definitely the FRP delamination through the
thickness of FRP laminates combined with fibre breaking on the surface ply.
The area of the FRP delamination and fibre breaking distributes more
uniformly and almost covers the whole adhesively bonding area, as shown
in Fig. 5.76 b). As stated in Section 5.5.1, the load-bearing capacity of aged
adhesive joints does not significantly decrease, and even slightly increases
for the 36 and 72 angle loading conditions. This implies that the elevated
temperature of the hydrothermal aging environment makes the adhesive
joint post-cured and in turn significantly improves the bonding quality
between FRP laminates and adhesive layer, while the absorbed moisture is
less inclined to deteriorate this bonding line. In this way, the failure plane of
aged specimens switches to the FRP laminates. Furthermore, the
secondary reason for the aged specimens undergoing this failure mode is
because the absorbed moisture content degrades the through-thickness
mechanical property of FRP laminates, which propels the FRP plies to
delaminate.
a) Un-aged specimen
b) Aged specimen
Fig. 5.76. Failure mode of un-aged and aged adhesive joints under tensile loading
171
Chapter5
It is worthwhile mentioning here that, for the 72 angle loading condition,

there is one exceptional failure mode both for the un-aged specimens and
the aged specimens. They are the 72-SB-01 specimen (Fig. 5.77 a)) and
72-SB-A-01 specimen (Fig. 5.77 b)).
cracks
a) 72-SB-01 specimen
b) 72-SB-A-01 specimen
Fig. 5.77. Different failure modes of adhesive joints under the 72 loading condition
From Fig. 5.77 a) it can be found that, beside the common failure mode of
the adhesive joints tested under other tensile/shear combining loading,
several cracks are observed in the adhesive layer of the 72-SB-01
specimen and propagate through the interface between adhesive layer and
steel support. It indicates that the failure plane almost switches to the
cohesive failure in the adhesive layer located close to the steel support,
which occurred for the adhesive joints under the shear loading condition.
For the 72-SB-A-01 specimen (Fig. 5.77 b)), the failure mode of the aged
172
adhesive joint is the same as that observed for the aged specimens under
shear loading. This means that the absorbed moisture degrades the
mechanical property of adhesive materials, which in turn influences the
failure mode of adhesive joints. Clearly, the 72 loading angle offers the
critical combining ratio of tensile and shear load. Under a larger (than 72)
angle loading, the failure of the adhesive joint may occur in the adhesive
layer but not in the FRP laminates. However, only one exceptional failure
mode is observed both for the un-aged specimens and the aged specimens.
Hence, the current test results are too limited to draw further conclusions.
5.5.3 Stiffness
Fig. 5.78 and Fig. 5.79 show the load-displacement curves (measured by
LVDTs) of un-aged and aged specimens. It seems that the stiffness of
adhesive joints does not decrease significantly under the influence of
hydrothermal aging. However, it should be noted that the displacement
measured by the LVDTs includes the deformation of the whole loading
device and not only the deformation of the adhesive joint. Thus, to further
understand the hydrothermal aging effects on the stiffness of the adhesive
joint, more comparison should be made on the local deformation of
adhesive joints measured by the displacement sensors.
100
Load [kN]
80
60
40
un-aged
aged
20
0
0
10
Displacement [mm]
Fig. 5.78. Comparison of load-displacement curves of un-aged specimens and aged

specimens under shear loading
173
Chapter5
Load [kN]
20
15
10
un-aged
aged
0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
Displacement [mm]
Fig. 5.79. Comparison of load-displacement curves of un-aged specimens and aged

specimens under shear loading
Fig. 5.80 shows the comparison of load-deformation curves measured by

four displacement sensors on the un-aged and aged specimens under six
angle loading conditions. For the tensile loading condition, each loaddeformation curve of Fig. 5.80 a) represents the average test result of four
displacement sensors. The location of the displacement sensors is
illustrated in Fig. 5.11. It can be found that the stiffness of aged adhesive
joints is significantly lower than that of the un-aged specimens. After the
failure load has been achieved, the curves of aged specimens do not drop
immediately but continue with a plateau and then gradually go down. This
phenomenon is due to the failure mode of FRP delamination. For the
tensile/shear combining loading conditions (Fig. 5.80 b), c), d) and e)), the
minus values of displacement sensors represent the average horizontal
relative deformation between the FRP sandwich deck and the steel support
(DS-2(4), as indicated in Fig. 5.12). The positive value, however, is the
average vertical deformation (DS-1(3), as indicated in Fig. 5.12). It is clear
that the curves of aged specimens are always more flexible than the unaged specimen curves. For 18 angle loading condition, the stiffness
difference in the vertical direction (positive zone) between un-aged and
aged specimens is more significant than the horizontal direction, since the
load is separated more in the tensile direction. For the 72 angle loading
condition, the situation is quite opposite. The stiffness difference in the
174
horizontal direction (minus zone) between un-aged and aged specimens is

more significant than that in the vertical direction, since the load is
separated more in the shear direction. For all these four loading conditions,
the vertical deformation curves of aged adhesive joints experience a plateau
after reaching the failure load. This confirms that these aged joints
experienced the same failure mode (FRP delamination) as mentioned in
section 5.5.2. For the shear loading condition, the displacement sensors are
located as shown in Fig. 5.10. In Fig. 5.80 f), the minus deformation curves
represent the average horizontal deformation between the FRP sandwich
deck and the steel support at the loading directly applied side (DS-02(04)).
Once again, the stiffness of aged specimens is lower than that of un-aged
specimens. What is more, for the un-aged adhesive joints, the absolute
value of minus deformation is dramatically larger than the positive
deformation, which indicates that the deformation distribution along the
adhesive layer is very different and the loading directly applied side carries
more load than the other side. However, for the aged specimens, the slopes
of minus and positive deformation curves are relatively close to each other,
which suggests that the deformation distribution in the aged adhesive joints
is much more balanced. Further, it indicates that the stiffness of the
adhesive layer is decreased by the absorbed moisture content, which
makes the adhesive layer rather flexible and subsequently the load is easily
transferred from the loading directly applied side to the other side.
Load [kN]
20
un-aged
aged
15
10
0
0.00
0.02
0.04
0.06
Deformation [mm]
Fig. 5.80. a) Tensile
175
0.08
0.10
Chapter5
14
12
Load [kN]
10
un-aged
aged
0
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
Deformation [mm]

16
14
Load [kN]
12
10
8
6
4
un-aged
aged
2
0
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
Deformation [mm]
Fig. 5.80. c) 36 angle

18
16
Load [kN]
14
12
10
8
6
un-aged
aged
4
2
0
-0.20
-0.15
-0.10
-0.05
0.00
Fig. 5.80. d) 54 angle

176
0.05
0.10
0.15
30
Load [kN]
25
20
15
10
un-aged
aged
5
0
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
Deformation [mm]
e) 72 angle
100
Load [kN]
80
un-aged
aged
60
40
20
0
-0.4
-0.2
0.0
0.2
0.4
Deformation [mm]
f) Shear
Fig. 5.80. Comparison of load-deformation curves measured by four displacement sensors
on the un-aged and aged specimens
5.6 Conclusions
In this chapter, mechanical behaviours of FRP-to-steel adhesively-bonded
joints are studied before and after the hydrothermal aging. A specific tensileshear loading device was designed and then employed to offer six different
angle loading conditions. The FE model of the adhesive joint was developed
by using ABAQUS 6.8 and subsequently employed to better understand the
stress distribution throughout the adhesive joint. Finally, comparison
177
Chapter5
between the experimental results of un-aged joints and hydrothermal aged

joints was conducted, with regard to the ultimate failure load, failure criterion,
failure mode and stiffness. The following conclusions are drawn:
For un-aged adhesively-bonded joints under shear and tensile loading,
three types of surface pretreatment methods were investigated: acetone
(AC), sand paper (SP) and sand blasting (SB). The surface pretreatment
cannot improve the stiffness of the adhesive joint. However, for the shear
loading condition, the surface pretreatment methods using sand paper (SP)
and sand blasting (SB) increase the ultimate failure load of specimens by
more than twice of specimens pretreated only using acetone (AC), while for
the tensile loading condition the corresponding increase of the ultimate
failure load is 9.5%;
Under shear loading, the un-aged adhesively-bonded joints fail in a brittle
way. The failure of S-SP-specimens and S-SB-specimens occurs in the
adhesive layer, at a location close to the steel support, while the S-ACspecimens fail through the interface between the adhesive layer and steel
support;
Under tensile loading, the un-aged adhesively-bonded joints pretreated by
using sand paper (SP) and sand blasting (SB) fail in a gradual failure
process, due to the FRP delamination and fibre breaking. The failure load of
the adhesive joints is closely related to the FRP delaminated area or the
fibre breaking area. The larger the area is, the higher load-bearing capacity
that is achieved, while the T-AC-specimens fail in a brittle and sudden
manner, and the failure mode is the interfacial fracture between the FRP
laminates and the adhesive layer;
In practice, the sufficient and quality-controllable surface pretreatment on
FRP sandwich decks and steel girders is required to improve the
mechanical performance of the adhesively-bonded joints. The sand blasting
method is considered to be a preferable surface pretreatment method,
based on the researches in this thesis;
The three-dimensional FE model of un-aged adhesive joints is developed
by using ABAQUS 6.8, and validated by the experimental results of relative
deformation between the FRP sandwich deck and the steel support. The
stress distribution in the adhesive joint proves that the failure is induced by
the combination of both tensile and shear stress peaks, but not only the
178
shear stress or normal tensile stress, even under the expected pure shear
or pure tension loading condition. The edge zone (approximately 10mm
from the ends of the adhesive layer) is the most sensitive area to initiate the
failure, where both the shear stress peak and the tensile stress singularity
are located;
Investigations on mesh dependency of the FE model confirm that the
1.50mm mesh scale model with six-layer discretization through the
thickness of adhesive layer is reasonably accurate and optimizes the
computational time;
For the un-aged adhesive joints under the combing loading of shear and
tension, the failure mode is the combination of fibre breaking (or FRP
delamination) and interfacial adhesion failure between FRP sandwich deck
and adhesive layer, except the 72-SB-specimen-01, the failure of which
combines the cracks in the adhesive layer and propagation through the
interface between adhesive layer and steel support. The shear-tensile
failure criterion of the un-aged adhesive joints is addressed;
For the hydrothermal aged adhesive joints, under shear loading, the
adhesive joint specimens fail in a brittle mode, with the failure mode of
cohesive fracture in the adhesive layer. For the tensile loading and
shear/tensile combining condition, the failure mode of aged adhesive joints
is a combination of FRP delamination and fibre breaking and it occurs
gradually, except for the 72-SB-A-01 specimen, which fails in the cohesive
fracture of the adhesive layer. The shear-tensile failure criterion of the aged
adhesive joints is addressed;
Comparison of the mechanical behaviours of adhesively-bonded joints
before and after a four-month hydrothermal aging shows that the
hydrothermal aging significantly decreases the ultimate failure loads of aged
adhesive joints under shear and tensile loading. However, for the
tensile/shear combining loading conditions, the environmental degradation
is not that obvious, which is due to the post curing mechanism, improving
the bonding quality between the FRP sandwich deck and the adhesive layer.
The failure criterion curves of un-aged and aged adhesive joints are close to
each other. However, the stiffness of adhesive joints is significantly
influenced by the hydrothermal aging;
179
Chapter5
With regard to the failure modes, for the tensile and tensile/shear combing
loading conditions, the hydrothermal aging switches the failure mode of
adhesive joints, from the partial interfacial failure between the FRP
sandwich deck and the adhesive layer to the full FRP delamination in FRP
laminates and fully covered fibre breaking area. For the shear loading
condition, the same failure mode (cohesive fracture in the adhesive layer) is
obtained before and after the four-month hydrothermal aging.
180
References:
[1] Jiang X, Kolstein MH, Bijlaard FSK. Study on mechanical behaviors of FRP-to-steel
adhesively-bonded joint under tensile loading. Compos Struct. 2013;98:192-201.
[2] Jiang X, Kolstein MH, Bijlaard FSK. Experimental and numerical study on mechanical
behavior of an adhesively-bonded joint of FRP-steel composite bridge under shear loading.
Compos Struct. 2014;108:387-399.
[3] da Silva LFM, das Neves PC, Adams RD, Wang A, Spelt JK. Analytical models of
adhesively bonded joints-Part II: Comparative study. Int J Adhes Adhes. 2009;29(3):33141.
[4] da Silva LFM, das Neves PJC, Adams RD, Spelt JK. Analytical models of adhesively
bonded joints-Part I: Literature survey. Int J Adhes Adhes. 2009;29(3):319-30.
[5] Diaz J, Romera L, Hernandez S, Baldomir A. Benchmarking of three-dimensional
finite element models of CFRP single-lap bonded joints. Int J Adhes Adhes.
2010;30(3):178-89.
[6] He XC. A review of finite element analysis of adhesively bonded joints. Int J Adhes
Adhes. 2011;31(4):248-64.
composed of pultruded GFRP profiles part I: Experimental and numerical investigations.
Compos Sci Technol. 2006;66(13):1903-14.
[8] Zhang Y, Vassilopoulos AP, Keller T. Effects of low and high temperatures on tensile
behavior of adhesively-bonded GFRP joints. Compos Struct. 2010;92(7):1631-9.
[9] Sheppard A, Kelly D, Tong LY. A damage zone model for the failure analysis of
adhesively bonded joints. Int J Adhes Adhes. 1998;18(6):385-400.
[10] http://www.infracomposites.com.
[11] EN 1990:2002(E) Eurocode - Basis of structural design. Brussels2002.
[12] Abaqus Analysis User's Manual, 6.8 version.
[13] Sancaktar E, Narayan K. Substrate volume and stress gradient concepts in mechanical
adhesion: analysis of single straight sections. J Adhes Sci Technol. 1999;13(2):237-71.
[14] Gleich DM, van Tooren MJL, Beukers A. Analysis of bondline thickness effects on
failure load in adhesively bonded structures. Int Sampe Tech Conf. 2000;32:567-79.
[15] Ferracane JL, Condon JR. Postcure Heat-Treatments for Composites - Properties and
Fractography. Dent Mater. 1992;8(5-6):290-5.
[16] Soares CJ, Pizi ECG, Fonseca RB, Martins LRM. Mechanical properties of lightcured composites polymerized with several additional post-curing methods. Oper Dent.
2005;30(3):389-94.
181
Chapter5
182
Chapter 6
Conclusions and
recommendations
This research work is divided into two parts: material level research (Part I)
and joint level research (Part II). The following sections present the main
outcome of each part, while the final section provides some
recommendations for future research work.
6.1 Conclusions
6.1.1 Part I: Material level research
The main aim of this part is to study the moisture diffusion characteristics of
two types of FRP composite materials and the environmental degradation
on mechanical properties of FRP laminates. The following brief conclusions
can be drawn:
Moisture diffusion coefficients are determined using the test data fitting
method based on the one-dimensional Fickian diffusion theory for pultruded
FRP composites and three-dimensional Fickian diffusion theory for resininfusion FRP laminates. Using these moisture diffusion coefficients as input
values for material properties, the FE model was developed for simulating
the moisture diffusion process in FRP composite materials and structures
and subsequently validated by the gravimetric experimental data and
analytical solution. It can be considered as a first step towards the coupled
hygro-thermal mechanical FE analysis;
For the two FRP composite materials studied in this thesis, the resininfusion FRP laminates obtain much lower values for both moisture diffusion
rates and saturated contents than those of the pultruded FRP composites,
which indicates good corrosion resistance from environmental effects;
It is confirmed that the elevated temperature could speed up the moisture
diffusion process, and the maximum moisture equilibrium contents are
dominated by the humidity of aging environments;
183
Chapter6
For the pultruded FRP composites, no significant variations on moisture

diffusion characteristic of specimens from different parts of the FRP deck
profile exist;
For the resin-infusion FRP laminates, the mass loss mechanism is
observed for Square and Rectangular specimens under 40C-water aging
condition;
Three-point bending test results of FRP lamiantes confirm that the
combination of moisture and temperature effects seriously deteriorates the
mechanical properties of FRP laminates on both strength and stiffness;
A coupled hygro-mechanical FE model was developed to analyse the
enviroment-dependent mechanical behaviours of FRP lanimates. This FE
model was first validated by the test results of flexural tests and
subsequently employed in an inverse parameter identification method to
determine the elastic interlaminar shear modulus of FRP laminates;
The
predictive
equations
for
environment-dependent
mechanical
properties (flexural and interlaminar) of FRP laminates were sustained by

using the least square method for the curve fitting.
6.1.2 Part II: Joint level research

In Part II, mechanical behaviours of FRP-to-steel adhesively-bonded joints
are studied before and after the hydrothermal aging. A specific tensile-shear
loading device was designed and then employed to offer six different angle
loading conditions. A comparison between the experimental results of unaged joints and hydrothermal aged joints is made with regard to the ultimate
failure load, failure criterion, failure mode and stiffness. The following brief
conclusions can be drawn:
For un-aged adhesively-bonded joints under shear and tensile loading,
three types of surface pretreatment methods were investigated: acetone
(AC), sand paper (SP) and sand blasting (SB). The surface pretreatment
cannot improve the stiffness of the adhesive joint. However, for the shear
loading condition, the surface pretreatment methods using sand paper (SP)
and sand blasting (SB) increase the ultimate failure load of specimens by
more than twice of specimens pretreated only using acetone (AC), while for
the tensile loading condition the corresponding increase of the ultimate
failure load is 9.5%;
184
Chapter6Conclusionsandrecommendations
Under shear loading, the un-aged adhesively-bonded joints fail in a brittle

way. The failure of S-SP-specimens and S-SB-specimens occurs in the
adhesive layer, at a location close to the steel support, while the S-ACspecimens fail through the interface between the adhesive layer and steel
support;
Under tensile loading, the un-aged adhesively-bonded joints pretreated by
using sand paper (SP) and sand blasting (SB) fail in a gradual failure
process, due to the FRP delamination and fibre breaking. The failure load of
the adhesive joints is closely related to the FRP delaminated area or the
fibre breaking area. The larger the area is, the higher load-bearing capacity
that is achieved, while the T-AC-specimens fail in a brittle and sudden
manner, and the failure mode is the interfacial fracture between the FRP
laminates and the adhesive layer;
In practice, the sufficient and quality-controllable surface pretreatment on
FRP sandwich decks and steel girders is required to improve the
mechanical performance of the adhesively-bonded joints. The sand blasting
method is considered to be a preferable surface pretreatment method,
based on the researches in this thesis;
The three-dimensional FE model of un-aged adhesive joints is developed
by using ABAQUS 6.8, and validated by the experimental results of relative
deformation between the FRP sandwich deck and the steel support. The
stress distribution in the adhesive joint proves that the failure is induced by
the combination of both tensile and shear stress peaks, but not only the
shear stress or normal tensile stress, even under the expected pure shear
or pure tension loading condition. The edge zone (approximately 10mm
from the ends of the adhesive layer) is the most sensitive area to initiate the
failure, where both the shear stress peak and the tensile stress singularity
are located;
Investigations on mesh dependency of the FE model confirm that the
1.50mm mesh scale model with six-layer discretization through the
thickness of adhesive layer is reasonably accurate and optimizes the
computational time;
For the un-aged adhesive joints under the combing loading of shear and
tension, the failure mode is the combination of fibre breaking (or FRP
delamination) and interfacial adhesion failure between FRP sandwich deck
185
Chapter6
and adhesive layer, except the 72-SB-specimen-01, the failure of which

combines the cracks in the adhesive layer and propagation through the
interface between adhesive layer and steel support. The shear-tensile
failure criterion of the un-aged adhesive joints is addressed;
For the hydrothermal aged adhesive joints, under shear loading, the
adhesive joint specimens fail in a brittle mode, with the failure mode of
cohesive fracture in the adhesive layer. For the tensile loading and
shear/tensile combining condition, the failure mode of aged adhesive joints
is a combination of FRP delamination and fibre breaking and it occurs
gradually, except for the 72-SB-A-01 specimen, which fails in the cohesive
fracture of the adhesive layer. The shear-tensile failure criterion of the aged
adhesive joints is addressed;
Comparison on the mechanical behaviours of adhesively-bonded joints
before and after a four-month hydrothermal aging shows that the
hydrothermal aging significantly decreases the ultimate failure loads of aged
adhesive joints under shear and tensile loading. However, for the
tensile/shear combining loading conditions, the environmental degradation
is not that obvious, which is due to the post curing mechanism, improving
the bonding quality between the FRP sandwich deck and the adhesive
layer. The failure criterion curves of un-aged and aged adhesive joints are
close to each other. However, the stiffness of adhesive joints is significantly
influenced by the hydrothermal aging;
With regard to the failure modes, for the tensile and tensile/shear combing
loading conditions, the hydrothermal aging switches the failure mode of
adhesive joints, from the partial interfacial failure between the FRP
sandwich deck and the adhesive layer to the full FRP delamination in FRP
laminates and fully covered fibre breaking area. For the shear loading
condition, the same failure mode (cohesive fracture in the adhesive layer) is
obtained before and after the four-month hydrothermal aging.
6.2 Recommendations for future research work

The following recommendations for future work can be given as follows:
For material level researches, more test data are needed contribute to the
statistical determination of characteristic values and appropriate safety
186
Chapter6Conclusionsandrecommendations
factors of material properties considering the moisture and temperature

effects;
Further research work is needed to study more moisture absorption/
desorption cycles, to understand how the residual damage develops and
what is the maximum scale of it;
The brittle character of FRP composite materials and adhesive joints need
to be better understood, especially under the colder temperature and
freezing condition;
Researches presented in this thesis are limited in types of FRP composite
and adhesive materials used for composing the adhesively-bonded joints.
Other types of FRP composite and adhesive materials can lead to different
failure modes, stiffness, load-bearing capacity of adhesive joint and
workability, due to different material properties and interfacial bonding
quality;
The post curing mechanism is evident for adhesive joints exposed to the
hydrothermal aging environment. It can definitely be used in practice to
improve the interfacial bonding quality between FRP decks and adhesive
layers. Future work can be conducted on how to realize the post-curing
process in the in-situ condition of FRP-steel composite bridges;
Environmental effects on the fatigue and creep performance of adhesivelybonded joints need to be investigated in the future. On the one hand, fatigue
cracks can enable moisture penetration. On the other hand, the absorbed
moisture content can result in the stress redistribution and in turn influence
the fatigue performance. Meanwhile, the absorbed moisture content and
elevated temperatures can influence the viscoelastic properties of FRP
composite and adhesive materials, which accordingly influence the creep
behaviours of adhesive joints;
Based on the research results of material level and joint level researches
obtained in this thesis, full-scale tests on the FRP-steel composite bridge
should be proposed in future research work, to investigate environmental
effects on the degree of composite action between FRP decks and steel
girders as well as the effective width of FRP decks;
Parametric studies are of interest to be conducted on the dimensions of
the adhesively-bonded joint and full-scale FRP-steel composite bridge with
regard to the thickness of the adhesive layer, FRP laminates and core
187
Chapter6
materials, to study their influence on the stress distribution, stiffness, degree

of composite action and effective width.
188
Acknowledgements
After five years of studying, researching and living in Delft, I am going to
defend my PhD thesis. At this moment, I would like to give my gratitude
to my supervisors, colleagues and friends, for their technical
suggestions, encouragement and accompanying.
First and foremost, I would like to thank my supervisors Prof. Frans
Bijlaard and Dr. Henk Kolstein, for their patient guidance and cordial
encouragement. The most research freedom they offered is highly
appreciated, which fully explored my research interests and independent
research ability. Before I come to the Netherlands, I never heard of FRP
bridge decks. In the first one or two years, they always encourage me to
be patient on literature study and explore the interesting research points
in this field. It is really my great fortune to conduct my PhD research
under their supervisions.
Furthermore, I would like to acknowledge the valuable discussions from
the committee members. Special appreciation goes to Prof. Airong
Chen, Prof. Philippe Van Bogaert and Prof. Dick Hordijk, for their careful
review and helpful comments.
The experimental research in this thesis was carried out in Delft
University of Technology, at Stevin II Macrolab and Delft Aerospace
Structures and Materials Laboratory. I would like to give my gratitude to
Arjen van Rhijn, John Hermsen, Kees van Beek, Fred Schilperoort and
Bob de Vogel. They realized my imagination in fantastic ways. The test
works would not have been so successful without their help. Moreover, I
would like to thank Infra Composite B.V. for test specimen manufacture
and financial support.
189
I would also like to thank my colleagues Sofia Teixeira de Freitas,

Maarten Mulder, Richard Pijpers, Carmen Sandhaas, Ayse Nesen
Surmeli-Anac, Wolfgang Gard, Roland Abspoel, Geert Ravenshorst,
Peter de Vries for their help, patience in academic discussion, culture
talk, group excursion, city tour organization, lunch time and coffee break.
Special gratitude goes to Caspar Groot, who brings me into the world of
art, share life attitude and treat me in a Chinese way. Caspar, I feel very
lucky to meet you in the Netherlands, and hopefully we will extend our
museum journey in the future.
What is more, I would like to extend my faithful thanks to my dear friends
Mingliang Li, Ning Li, Zhuqing Yu, Tingting Jiang, Yue Xiao, Jitang Fan,
Jinlong Li, Zhan Zhang, Hailing Zhang, Yuan Zhang, Gang Liu, Zhiyi
Chen, Ming Liu, Hui Yu, Yunan Gao etc. I enjoy the great time with you
by cooking, drinking, talking, kidding, sporting and travelling.
Finally, I am deeply grateful to my parents for their love, patience,
encouragement and understanding. Their amiable smiles always easily
refresh me from massive test data and complicated equations. Special
thanks goes to my beloved Xuhong Qiang, who shares my sorrow and
happiness forever.
Xu Jiang
Delft, the Netherlands
190
Curriculum vitae
Personal Information
Name
Date of birth
Gender
Nationality
Email
Xu Jiang
August 8th, 1982
Male
China
[email protected]
Education Background
Sep. 2001 July 2005
Bachelor in Civil Engineering with honors

Dalian University of Technology
Dalian, Liaoning Province, China
Sep.2005 Mar. 2008
MSc in Structural Engineering with honors

Major in Steel Structures
Tongji University
Shanghai, China

Work Experience
Apr. 2008 Oct. 2008

Nov. 2008 Oct. 2013

Structural engineer
East China Architectural Design & Research Institute
Shanghai, China

PhD researcher in Structural Engineering
Research on mechanical performance and durability
of the adhesivelybonded joint between FRP bridge
deck and steel girder
Delft University of Technology
Delft, the Netherlands

191
192
ListofPublications
I.
Journalpublications:
1) Xu Jiang, Henk Kolstein, Frans S.K. Bijlaard. Moisture Diffusion and Hygrothermal
Aging in Pultruded Fibre Reinforced Polymer Composites of Bridge decks. Materials
& Design, 2012, 37: 304-312.
2) Xu Jiang, Henk Kolstein, Frans S.K. Bijlaard. Moisture diffusion in glass-fiberreinforced polymer composite bridge under hot/wet environment. Composites Part BEngineering, 2013, 45: 407-416.
3) Xu Jiang, Henk Kolstein, Frans S.K. Bijlaard. Study on mechanical behaviors of FRPto-steel adhesively-bonded joint under tensile loading. Composite Structures, 2013,
98: 192- 201.
4) Xu Jiang, Henk Kolstein, Frans S.K. Bijlaard. Experimental and Numerical Study on
Mechanical Behavior of an Adhesively-bonded joint of FRP-steel Composite Bridge
under Shear Loading. Composite Structures, 2014, 108: 387-399.
5) Xu Jiang, Henk Kolstein, Frans S.K. Bijlaard. Effects of hygrothermal aging on glassfibre reinforced polymer laminates and adhesive of FRP composite bridge: Moisture
diffusion characteristics. Composite Part A - Applied Science and Manufacturing,
2014, 57: 49-58.
6) Xuhong Qiang, Frans S.K. Bijlaard, Henk Kolstein, Xu Jiang. Behaviour of beam-tocolumn high strength steel endplate connections under fire conditions - Part 1:
Experimental study. Engineering Structures. (minor revision)
7) Xuhong Qiang, Frans S.K. Bijlaard, Henk Kolstein, Xu Jiang. Behaviour of beam-tocolumn high strength steel endplate connections under fire conditions - Part 1:
Numerical study. Engineering Structures. (minor revision)
II.
ConferencePublications
1) Xu Jiang, Henk Kolstein and Frans S.K. Bijlaard. Coupled Hygro-mechanical Stress
Analysis on Adhesive Material under Hot/wet Environments. In Proceeding of the 4th
International Conference on Structural Engineering, Mechanics and Computation,
2010, Cape Town, South Africa.
2) Xu Jiang, Henk Kolstein, Frans Bijlaard. Moisture Diffusion in FRP AdhesivelyBonded Joints under Hot/wet Environments. Proc. of the 5th International Conference
on FRP Composites in Civil Engineering. 2010, Beijing, China.
3) Xu Jiang, Henk Kolstein and Frans S.K. Bijlaard. Study on Moisture Fickian
Diffusion Process of a Pultruded FRP Composite Material under Hot/wet Environment.
In Proceeding of the Fourth International Conference on Durability & Sustainability of
Fibre Reinforced Polymer (FRP) Composites for Construction and Rehabilitation.
2011, Quebec, Canada.
4) Xu Jiang, Henk Kolstein, Frans S.K. Bijlaard. Experimental Investigation on FRP to
Steel Adhesively-bonded Joint under Tensile Loading. Proceedings of the 6th
International
Composites
Conference
(ACUN-6):
COMPOSITE
&
NANOCOMPOSITES IN CIVIL, OFFSHORE AND MINING INFRASTRUCTURE.
2012, Melbroune, Australia.BEST PAPER AWARD
5) Xu Jiang, Henk Kolstein and Frans S.K. Bijlaard. Numerical Analysis and Parametric
Study on Composite Action Between Fiber-reinforced Polymer Bridge Decks and Steel
Girder. ASCCS: 10th International Conference on Advances in Steel Concrete
Composite and Hybrid Structures, 2012, Singapore.
6) Xu Jiang, Henk Kolstein and Frans S.K. Bijlaard. Mechanical Behaviour of Adhesive
Joint under Tensile and Shear Loading. IABSE, 2013, Rotterdam, the Netherlands.
193

Mechanical Behaviour and Durability of FRP To Steel Adhesively Bonded Joints

Uploaded by

Copyright:

Available Formats

Mechanical Behaviour and Durability of FRP To Steel Adhesively Bonded Joints

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mechanical Behaviour and Durability of FRP To Steel Adhesively Bonded Joints

Uploaded by

Copyright:

Available Formats

MechanicalBehaviourandDurabilityof

Dit proefschrift is goedgekeurd door de promotoren:

environments. The FRP composite profiles and adhesive layers for

1.2 Aim of the research

a) Shear stress in the longitudinal direction

b) Tensile stress in the transverse direction

Different surface pretreatment methods are compared with regard to the

gives a brief state-of-the-art in FRP bridge decks, the FRP-steel composite

adhesively-bonded joints is addressed. Furthermore, a three-dimensional

PartII: Joi ntlevelressearch

Chapterr 6 Concluusions and reecommendaations

(c) ASSET (Fiberline, Denmark)

(d) DuraSpan (Martin Marietta Composite, USA)

Fig. 2.1. FRP bridge decks

The growing acceptance of FRP bridge decks can be attributed to their

High strength to weight ratio.

2.2 FRP-steel composite bridge deck system

Fig. 2.2. FRP-to-Steel composite girder system [12]

The plate-bending behaviour of a pultruded GFRP bridge deck system was

Fig. 2.4. FE model of test set-up

Fig. 2.4. FE model of test set-up

Fig. 2.5. Normalized through-thickness tensile stress distribution in adhesive layer/bridge

Fig. 2.6. Normalized through-thickness tensile stress distribution in adhesive layer/bridge

2.3 Adhesively-bonded joint

Fig. 2.7. Geometry of lap joint specimens (not to scale)

Parametric studies were conducted experimentally and numerically on the

a) Shear stress distribution

b) Through-thickness stress distribution

The stress analysis on the adhesively-bonded joints is usually approached

linear elastic analysis is supposed to be more appropriate for both

2.4 Environmental effects

(b) Relative humidity

The hot/wet exposure is supposed to be the severest environmental

decrease the service life of FRP composite structures. The absorbed

Generally, the moisture diffusion experiments for FRP composites are

3.2 Moisture diffusion theory

where c represents the moisture concentration, x the space coordinate

the x direction. D is supposed to be independent of the spatial and temporal

where c is the maximum equilibrium moisture concentration. The

where M is the equilibrium amount of absorption. The theoretical Fickian

Fig. 3.1. Fickian diffusion process.

In practice, this simple procedure is mostly used for the thin-section of an

With the boundary conditions as:

A closed-form solution to the above is given as:

By integrating on the space variables, the moisture uptake content can be

It is noted that the three-dimensional moisture diffusion analytical

FRP plates was developed by Pierron [5], which is based on an optimization

These triangular shape profiles are manufactured by the pultrusion process,

Fig. 3.3. Typical cross-section view of pultruded FRP composites [7]

For resin-infusion FRP laminates, specimens are manufactured by resin

were deliberately prepared with three different aspect ratios, as listed in

Fig. 3.4. FRP laminate specimens

3.3.2 Gravimetric test process

3.4 Experimental results and discussion

Fig. 3.6. Illustration of coordinates in the pultruded FRP specimen

Comparing with the gravimetric experiments in the vapour aging conditions,

condition, which is reasonable according to the aforementioned conclusion.

Herein, only the moisture diffusion in the 40C-water aging condition is

Fig. 3.8 shows the comparison of analytical solutions and FE analysis