Professional Documents
Culture Documents
Shaat_Amr_A_2007_11_PhD
Shaat_Amr_A_2007_11_PhD
By
Queen’s University
Kingston, Ontario, Canada
November, 2007
Abstract
Steel bridges and structures often need strengthening due to increased life loads, or repair
due to corrosion or fatigue cracking. This study explored the use of adhesively bonded
Carbon Fibre Reinforced Polymers (CFRP) flexible sheets and rigid plates in retrofitting
steel columns and girders, through experimental and analytical investigations. The first
columns comprised of square Hollow Structural Sections (HSS). Fifty columns, 175 mm
to 2380 mm long (i.e. with slenderness ratios ranging from 4 to 93), were tested under
axial compression loads to examine the effects of number and type of CFRP layers, fibre
orientation, and slenderness ratio. Transverse wrapping was shown to be suitable for
controlling outwards local buckling in HSS short columns, while longitudinal layers were
increases in axial strength observed in the experiments were 18 and 71 percent, for short
finite element model were developed for slender columns. The models account for steel
plasticity, geometric non-linearities, and residual stresses. The models were verified
using experimental results, and used in a parametric study. It was shown that CFRP
The second part of the research program investigated w-section steel-concrete composite
girders retrofitted using CFRP materials. Three girders, 6100 mm long, were tested to
study strengthening of intact girders using CFRP plates. Eleven girders, 2030 mm long,
i
Abstract
including girders artificially damaged by completely cutting their tension flanges at mid-
span, were tested to study the effectiveness of repair using CFRP sheets. The parameters
considered were the CFRP type, number of layers, number of retrofitted sides of the
tension flange, and the length of CFRP repair patch. The strength and stiffness of the
intact girders have increased by 51 and 19 percent, respectively. For the repaired girders,
the strength and stiffness recovery ranged from 6 to 116 percent and from 40 to 126
percent, respectively. Unlike flexural strength, the stiffness was not much affected by the
bond length. Analytical models were developed, verified, and used in a parametric study,
which showed that the higher the CFRP modulus, the larger the gain in stiffness and
yielding moment, but the lower the gain in strength and ductility. In general, this study
ii
Acknowledgements
Acknowledgements
First and foremost, I thank God through whom all things are possible. I would also like to
recognize and thank all the people who made my time at Queen's University during the
I would like to express my deepest gratitude to my supervisor, Dr. Amir Fam, for his
unwavering support and guidance throughout this research project. His patience,
leadership, and never ending encouragement gave me the confidence to focus and
proceed. I owe him an unbelievable amount of gratitude for his prominent role in helping
The support of the staff has been a vital part of my success. Thanks go to Fiona Froats,
Cathy Wagar, Maxine Wilson, Lloyd Rhymer, Neil Porter, Paul Thrasher, Jamie Escobar,
and Bill Boulton. Special thanks go to Dave Tryon, who provided great technical
experience and guidance to make the experimental part of this research runs efficiently.
I would also like to acknowledge my fellow graduate students, who helped me along the
way. Thanks go to Abdul Chehab, Andrew Kong, Britton Cole, Hart Honickman, Jeff
Mitchell, Siddwatha Mandal, Tarek Sharaf, Wojciech Mierzejewski, and Yazan Qasrawi.
I wish to acknowledge the financial support provided by the Natural Sciences and
Shirping Sika Inc. for providing his experience in bonding the CFRP plates of phase II. I
iii
Acknowledgements
also wish to thank Fyfe.Co.LLC, Mitsubishi Chemical, and Sika INC. for providing the
FRP materials.
I could not have survived the duration of this study without my family. I would like to
thank my parents, brother, and sisters for their on-going love, support and encouragement
throughout my entire life. Special thanks go to my uncle, Dr. Fathy Saleh, whose
example showed me the value of pursuing an academic career. Also, love and prayers of
Finally, I would like to thank my wife, Dalia, for believing in me and for all her support
throughout these years. For all your love, patience and dedication, I am grateful. I would
also like to acknowledge my son, Ibrahim, who enlightened my life with his smile.
iv
Table of Contents
Table of Contents
Abstract ....................................................................................................... i
2.1 Introduction..................................................................................................11
v
Table of Contents
2.6.2 Analysis of steel girders strengthened with FRP bonded material .........42
3.1 Introduction..................................................................................................65
vi
Table of Contents
3.2.3 Concrete.................................................................................................72
Beams ......................................................................................................................88
vii
Table of Contents
4.1 Introduction................................................................................................120
strengthening ..........................................................................................................126
5.1 Introduction................................................................................................155
viii
Table of Contents
6.1 Introduction................................................................................................202
ix
Table of Contents
7.1 Introduction................................................................................................267
Materials ...................................................................................................................268
x
Table of Contents
Materials ...................................................................................................................275
7.4.1.4 Calibration of parameter for the neglected part of the steel web .........279
xi
Table of Contents
References................................................................................................................311
xii
List of Figures
List of Figures
Figure 1.1 CFRP-strengthening of HSS columns. ............................................................9
Figure 2.2 Typical stress-strain curves for CFRP, GFRP, and steel. ..............................54
Figure 2.3 Measured and predicted strain distributions along the bonded length of a
Figure 2.5 Various techniques of introducing artificial damage to steel girders. .............56
Figure 2.11 Load-deflection response and failure mode of a tubular pole. .....................59
Figure 2.12 Effective bond length for steel tube strengthened with HM-CFRP...............60
Figure 2.14 Installation of CFRP sheets on cracked aluminum truss k-joint. ..................61
Figure 2.18 Comparisons of shear and peel stresses for plates with and without taper
xiii
List of Figures
Figure 2.19 Finite element analysis versus experimental load-deflection responses. ....63
Figure 2.20 Design guidelines for steel–concrete composite beams strengthened with
Figure 2.21 Installation of CFRP plate on the Christina Creek bridge (I-704). ................64
Figure 3.1 Different steel cross sections used in the experimental investigation. .........101
Figure 3.4 Tensile stress-strain response of a coupon cut from W250x25 ...................103
Figure 3.7 Tension coupons and test setup of FRP materials. .....................................104
Figure 3.10 Effect of FRP on local and overall buckling of short and slender HSS
columns. .................................................................................................................106
Figure 3.11 Details of FRP strengthening configurations of HSS columns in Phase I. .106
Figure 3.12 Various preparation measures of the HSS columns in Phase I. ................107
Figure 3.18 A schematic and fabrication process of girders tested in Phase II.............112
Figure 3.19 Casting concrete slabs of the girders tested in Phase II. ...........................113
xiv
List of Figures
Figure 3.21 Schematic of test setup and instrumentations of girders tested in Phase II.
................................................................................................................................115
Phase III..................................................................................................................116
Figure 3.23 Sandblasting the tension steel flanges of beams tested in Phase III. ........117
Figure 3.24 Installation process of FRP sheets on beams tested in Phase III. .............118
Figure 3.25 Test setup and instrumentations of beams tested in Phase III. .................119
Figure 4.1 Load-axial displacement responses of column sets 1 and 2 of group A. .....137
Figure 4.2 Load-axial displacement responses of column sets 3 and 4 of group A. .....137
Figure 4.3 Load-axial displacement responses of column sets 5 and 6 of group A. .....138
Figure 4.7 Load-axial strain responses based on strain gauge S1 of column sets 1 and 2
of group A. ..............................................................................................................140
Figure 4.8 Load-axial strain responses based on strain gauge S1 of column sets 3 and 4
of group A. ..............................................................................................................140
Figure 4.9 Load-axial strain responses based on strain gauge S1 of column sets 5 and 6
of group A. ..............................................................................................................141
Figure 4.10 Load-axial strain responses based on strain gauge S2 of column sets 1 and
2 of group A. ...........................................................................................................141
Figure 4.11 Load-axial strain responses based on strain gauge S2 of column sets 3 and
4 of group A. ...........................................................................................................142
Figure 4.12 Load-axial strain responses based on strain gauge S2 of column sets 5 and
6 of group A. ...........................................................................................................142
Figure 4.13 Variation of axial strength with slenderness ratio of group A columns.......143
xv
List of Figures
Figure 4.14 Effect of slenderness ratio on the CFRP effectiveness in group A columns.
................................................................................................................................143
Figure 4.16 Variation of the compressive strain of CFRP at ultimate with slenderness
ratio.........................................................................................................................145
Figure 4.24 Mid-height imperfection of specimen 7 of group B versus the applied load.
................................................................................................................................149
Figure 4.27 Load-axial strain response of control specimen 12-1 of group C...............152
columns. .................................................................................................................152
columns. .................................................................................................................153
Figure 4.30 Effect of using two types of CFRP on strengthening short specimens. .....153
Figure 5.2 Load-strain responses of the lower flange of the control specimen G1. ......178
Figure 5.3 Load-steel strain responses at the web of specimens tested in Phase II. ...179
xvi
List of Figures
Figure 5.4 Load-strain responses along the CFRP plates of specimen G2 in Phase II.179
Figure 5.5 Load-strain responses along the CFRP plates of specimen G3 in Phase II.180
Figure 5.6 Load-average shear stress responses along the CFRP plates of specimen G2
in Phase II...............................................................................................................180
Figure 5.7 Load-average shear stress responses along the CFRP plates of specimen G3
in Phase II...............................................................................................................181
Figure 5.11 Load-strain responses of the upper steel flanges of specimens B1 and B2 in
Phase III..................................................................................................................183
Figure 5.12 Load-strain responses of the lower steel flanges of specimens B1 and B2 in
Phase III..................................................................................................................184
Figure 5.13 Load-strain responses along the mid-span cross section of specimen B2 in
Phase III..................................................................................................................184
Figure 5.18 Strain distributions along the CFRP sheets of specimen B3 in Phase III...187
Figure 5.19 Strain distributions along the CFRP sheets of specimen B4 in Phase III...187
Figure 5.20 Strain distributions along the CFRP sheets of specimen B5 in Phase III...188
Figure 5.21 Strain distributions along the CFRP sheets of specimen B6 in Phase III...188
Figure 5.22 Load-strain responses of CFRP at mid-span of specimens B3, B4, and B6 in
Phase III..................................................................................................................189
xvii
List of Figures
Figure 5.23 Load-strain responses of the upper steel flange at mid-span of specimens
Figure 5.26 Strain distributions along the lower CFRP sheets of specimen B7 in Phase
III.............................................................................................................................191
Figure 5.27 Strain distributions along the upper CFRP sheets of specimen B7 in Phase
III.............................................................................................................................192
Figure 5.30 Effect of force equivalence index (ω) on the strength of the repaired beams.
................................................................................................................................194
Figure 5.34 Strain distributions along the lower CFRP sheets of specimen B8 in Phase
III.............................................................................................................................196
Figure 5.35 Strain distributions along the lower CFRP sheets of specimen B9 in Phase
III.............................................................................................................................197
Figure 5.36 Strain distributions along the lower CFRP sheets of specimen B10 in Phase
III.............................................................................................................................197
Figure 5.37 Strain distributions along the lower CFRP sheets of specimen B11 in Phase
III.............................................................................................................................198
Phase III..................................................................................................................198
Figure 5.39 Load versus slip of concrete slab of specimen B9 in Phase III. .................199
xviii
List of Figures
Figure 5.40 Maximum strains versus the bonded length of the CFRP sheets. .............199
Figure 5.41 Load-average shear stress responses along the lower CFRP sheets of
Figure 5.42 Load-average shear stress responses along the lower CFRP sheets of
Figure 5.43 Load-average shear stress responses along the lower CFRP sheets of
Figure 5.44 Load-average shear stress responses along the lower CFRP sheets of
Figure 6.2 Stress and strain distributions within the cross section of slender column at
mid-height. ..............................................................................................................237
Figure 6.4 Summary of the finite difference model and convergence study. ................238
response. ................................................................................................................239
Figure 6.6 Variation of ultimate compressive-to-tensile strain ratio of CFRP at failure with
Figure 6.12 Verification of models 1 and 2 using test results on HSS 203 x 203 x 6.3
mm..........................................................................................................................242
xix
List of Figures
Figure 6.13 Verification of models 1 and 2 using test results on HSS 152 x 152 x 4.9
mm..........................................................................................................................243
Figure 6.14 Measured and predicted load-lateral displacement responses of set 1. ....243
Figure 6.15 Measured and predicted load-lateral displacement responses of set 2. ....244
Figure 6.16 Measured and predicted load-lateral displacement responses of set 3. ....244
Figure 6.17 Measured and predicted load-lateral displacement responses of set 4. ....245
Figure 6.18 Measured and predicted load-lateral displacement responses of set 5. ....245
Figure 6.19 Measured and predicted load-lateral displacement responses of set 6. ....246
................................................................................................................................246
................................................................................................................................247
................................................................................................................................247
10............................................................................................................................248
11............................................................................................................................248
Figure 6.25 Measured and predicted load-axial displacement responses of set 1. ......249
Figure 6.26 Measured and predicted load-axial displacement responses of set 2. ......249
Figure 6.27 Measured and predicted load-axial displacement responses of set 3. ......250
Figure 6.28 Measured and predicted load-axial displacement responses of set 4. ......250
Figure 6.29 Measured and predicted load-axial displacement responses of set 5. ......251
Figure 6.30 Measured and predicted load-axial displacement responses of set 6. ......251
................................................................................................................................252
xx
List of Figures
................................................................................................................................252
................................................................................................................................253
Figure 6.34 Measured and predicted load-axial displacement responses of specimen 10.
................................................................................................................................253
Figure 6.35 Measured and predicted load-axial displacement responses of specimen 11.
................................................................................................................................254
Figure 6.36 Measured and predicted load-axial strain responses of set 1....................254
Figure 6.37 Measured and predicted load-axial strain responses of set 2....................255
Figure 6.38 Measured and predicted load-axial strain responses of set 3....................255
Figure 6.39 Measured and predicted load-axial strain responses of set 4....................256
Figure 6.40 Measured and predicted load-axial strain responses of set 5....................256
Figure 6.41 Measured and predicted load-axial strain responses of set 6....................257
Figure 6.42 Measured and predicted load-axial strain responses of specimen 7. ........257
Figure 6.43 Measured and predicted load-axial strain responses of specimen 8. ........258
Figure 6.44 Measured and predicted load-axial strain responses of specimen 9. ........258
Figure 6.45 Measured and predicted load-axial strain responses of specimen 10. ......259
Figure 6.46 Measured and predicted load-axial strain responses of specimen 11. ......259
Figure 6.47 Comparison between the deformed shapes in experiments and FEM (Model
2). ...........................................................................................................................260
xxi
List of Figures
Figure 6.53 Load-lateral displacement responses for specimens with Frs = 0.50 Fy. ....263
Figure 6.54 Load-lateral displacement responses for specimens with kL/r =160. ........264
Figure 7.1 Steel-concrete composite girder strengthened with FRP and subjected to pure
bending. ..................................................................................................................288
in Phase II...............................................................................................................290
in Phase II...............................................................................................................292
xxii
List of Figures
CFRP. .....................................................................................................................293
CFRP. .....................................................................................................................294
UHM-CFRP.............................................................................................................294
Figure 7.15 Effect of the modulus of CFRP on percentage increase in stiffness for
Figure 7.16 Effect of the modulus of CFRP on percentage increase in yielding moment
Figure 7.17 Effect of the modulus of CFRP on percentage reduction in ductility. .........296
Figure 7.18 Effect of CFRP rupture strain on percentage increase in strength for different
Figure 7.19 Effective cross sections and corresponding stress and strain distributions in
................................................................................................................................298
Figure 7.21 Schematic to illustrate the effect of damage and CFRP-repair on variation of
xxiii
List of Figures
Figure B.1 Mid-height imperfections of column sets 3 and 4 versus the applied load. .337
Figure B.2 Mid-height imperfections of column sets 5 and 6 versus the applied load. .337
Figure B.3 Mid-height imperfections of specimens 7 to 11 versus the applied load. ....338
xxiv
List of Tables
List of Tables
Table 3.1 Material properties of the W150x22 steel section used in Phase III...............95
Table 3.4 Test matrix of HSS column specimens tested in Phase I................................98
Table 3.5 Test matrix of composite girders tested in Phase II. .......................................99
Table 3.6 Test matrix of repair of artificially damaged composite beams tested in Phase
III.............................................................................................................................100
Table 4.1 Summary of test results of slender columns in group A of Phase I...............134
Table 4.2 Summary of test results of slender columns in group B of Phase I...............135
Table 4.3 Summary of test results of short columns in group C of Phase I. .................136
Table 6.2 Comparison between experimental and predicted results using Models 1 and
2..............................................................................................................................235
columns ..................................................................................................................236
girders.....................................................................................................................286
Table 7.2 Calibration of the neglected part of the steel web in repair applications. ......286
xxv
Notation
Notation
A fi Area of FRP element
e’ Initial out-of-straightness
Et Tangent modulus
xxvi
Notation
I moment of inertia
i element number
M Bending moment
moment
MD Dead moment
ML Live moment
xxvii
Notation
Pu Ultimate load
r Radius of gyration.
V Shear force
yc Distance between the extreme fibre and the effective centroid of the cross section
xxviii
Notation
yi Distance from the center of the element to the effective centroid of the cross
section
z Distance along the length of the column or slope of stress flow in damaged girders
composite girders
xxix
Notation
xxx
Chapter 1
Chapter 1
Introduction
1.1 General
safety, economy and quality of life. Steel structures comprise a large portion of the
problems in steel structures in general and steel bridges in particular. These include
damage, and lack of proper maintenance. In other cases, steel bridges may not be
deteriorated or damaged but are rather in need of upgrading to carry larger loads and
increasing traffic volumes. In most cases, the cost of retrofitting is far less than the cost of
replacement. In addition, retrofitting usually takes less construction time and therefore
heavy steel plates to the existing structures. This may require heavy lifting equipment and
1
Chapter 1
shoring systems in the site, which may cause delays and traffic interruptions. Also,
adding heavy steel plates increases the dead load of the structure, which limits the target
increase of live load carrying capacity. The continuous process of corrosion and the
reduced fatigue life associated with the welded steel plates may reduce the durability, and
The need for adopting durable materials and cost-effective retrofit techniques is evident.
One of the possible solutions is the use of high performance non-metallic materials such
as Fibre Reinforced Polymers (FRPs). Despite the higher cost of FRP materials,
compared to steel, the material cost alone generally comprises a very small portion of the
overall project cost. The superior mechanical and physical properties of FRP materials
make them quite promising for both repair and strengthening applications. They also
The use of FRP systems for retrofit of concrete structures has been quite successful (ACI
440, 2002). Today, the use of glass- and carbon-FRP materials (GFRP and CFRP) in
retrofitting concrete bridges and structures is becoming widely accepted in practice. FRP
is used in the form of sheets or plates attached to the concrete surface for flexural and
shear retrofitting or as sheets for wrapping columns to increase their ductility and axial
strength.
2
Chapter 1
The inherent high strength and stiffness of steel makes it a more challenging material to
strengthen, compared to other materials such as concrete and wood. If steel is retrofitted
using a material with a lower Young’s modulus, load transfer and hence load sharing of
the strengthening material will only be significant after the steel yields. Therefore, the
relatively low tensile modulus of glass fibres (72 GPa) makes them less desirable for
retrofitting steel structures. On the other hand, carbon fibres have outstanding mechanical
properties that could be superior to those of steel. The tensile strength and modulus of
carbon fibres could reach up to 4020 MPa and 640 GPa, respectively (Cadei et al., 2004).
In addition, a CFRP plate weighs less than one fifth the weight of a similar size steel plate
the girders (Gillespie et al., 1996a). In Delaware, USA, bridge girders with a total length
of 180 meters, were replaced due to severe and extensive damage. The replacement cost
was compared with the cost of rehabilitation using CFRP, at an assumed 25 percent
section loss. It was concluded that the total replacement cost was 3.65 times higher than
The research program carried out in this thesis is focused on retrofit of two types of steel
structures, namely, Hollow Structural Section (HSS) short and slender columns, and
3
Chapter 1
In the case of short columns, it is hypothesized that CFRP sheets attached to the surface
in the transverse and longitudinal directions could help control the outwards local
buckling of two opposite faces of the column, and thereby increase its axial strength and
In the case of HSS slender columns, it is hypothesized that CFRP sheets or plates
attached in the longitudinal direction could delay the overall buckling of the column,
particularly when using high modulus CFRP. This would enhance the axial strength of
strengthened using CFRP materials with emphasis on the gains in strength and
2. Examining the effects of slenderness ratio, CFRP type, fibre orientation, and
materials.
4. Develop analytical and numerical models to predict the behaviour and axial
4
Chapter 1
In the case of strengthening, the goal is to increase the flexural strength and stiffness of
intact girders, beyond their original capacities, using CFRP materials adhesively bonded
to the tension flange, as shown in Figure 1.2(a and c). This simulates upgrading bridge
In the case of repair, the tension flange is saw-cut at mid-span to simulate section loss
due to a fatigue crack or a localized severe corrosion. CFRP material is then adhesively
bonded to the tension flange, as shown in Figure 1.2(b and c), in order to recover and
3. Examining the effects of elastic modulus, number of layers, and bonded length of
5
Chapter 1
analytical and numerical modeling. The experimental program is planned to address the
elements (i.e. columns and beams), as mentioned in the previous section. The
experimental results of this study and other studies are used to verify the proposed
analytical and numerical models. The models are then used in parametric studies to
The experimental program includes three phases. Phase I is focused on evaluating the
behaviour of axially loaded HSS columns strengthened using CFRP sheets and plates.
This was achieved by testing 50 HSS columns of lengths varying from 175 mm to 2380
mm (i.e. slenderness ratios ranging from 4 to 93). Phase II is intended to evaluate the
Three large-scale (6100 mm long) girders have been tested in four-point bending in this
phase. Phase III is designed to evaluate the effectiveness of using CFRP sheets to repair
artificially damaged steel-concrete composite beams with a simulated section loss in the
tension flange. In this phase, a total of 11 beams, 2040 mm long, have been tested in
Four analytical and numerical models have been developed for both the axially loaded
and flexural members. The first model is an analytical fibre-element model for HSS
slender columns. The model is based on the concepts of strain compatibility and force
6
Chapter 1
equilibrium and is capable of predicting the full response of axially loaded HSS slender
columns strengthened using FRP materials. The second model is a numerical non-linear
Finite Element Model (FEM), developed as an alternative model for the HSS slender
columns. Both models were verified using experimental results and were used in
composite girders. The model is based on developing and integrating the moment-
curvature relationship of the composite cross section to predict the full load-deflection
to predict the ultimate moment capacity of steel-concrete composite girders, with a cut in
their tension flanges, and repaired using FRP materials. The model also predicts
presented.
including Phase I for column tests and Phases II and III for beam tests. The mechanical
Chapter 4 presents the results of the columns tests (Phase I) of the experimental
program, including the effects of various parameters as well as different failure modes.
7
Chapter 1
Chapter 5 presents the results of the flexural members tests (Phases II and III) of the
failure modes.
Chapter 6 presents both the analytical and numerical models developed to predict the
responses of axially loaded slender HSS columns. The models are verified using
Chapter 7 presents the analytical model developed to predict the responses of intact
steel-concrete composite girders strengthened using FRP materials. The chapter also
presents the simplified analytical model developed to predict the ultimate moment and
service load deflection of damaged steel-concrete composite girders repaired using FRP
materials.
Chapter 8 provides a summary of the thesis as well as conclusions based on both the
experimental and theoretical studies. The chapter also presents recommendations for
future work.
8
Chapter 1
T
C
Longitudinal
FRP layers
T T
Local
buckling
Transverse
FRP layer
T T C
T
Longitudinal [Key et al., 1988]
FRP layer
CFRP-strengthened
Load
Original
displacement
9
Chapter 1
Concrete slab
W-section
CFRP CFRP
Fatigue crack
[Boyd, 1970]
lost Crack
CFRP CFRP
Upgraded
Load
CFRP failure
Strengthening
(gain) ?
Original (intact)
Repair
(recovery)
?
Damaged (cracked)
deflection
10
Chapter 2
Chapter 2
2.1 Introduction
The development of metallic structures has evolved significantly over the years, from the
19th century era of construction using cast or wrought iron, to modern steel construction.
In general, the steel construction industry involves two types of steel structural members.
The first includes the hot-rolled shapes and members built-up of plates, and the second
Many of the old cast iron and steel structures and bridges that remained in service are
reasons; including deterioration due to corrosion, fatigue cracking, increased loading, and
change in design loads over the years. In most cases, maintaining this infrastructure
through retrofit to extend their service life is far more economical than replacing them.
11
Chapter 2
more detailed discussion on Fibre Reinforced Polymer (FRP) materials and their use in
retrofitting metallic structures, along with recent research advances in this field will
follow.
A wide range of metallic materials has been used in construction. This section briefly
describes two of the most commonly used structural metals, namely, cast iron and steel.
Structural cast iron was developed at the end of the 18th century and the first cast iron
bridge was built in 1779 (Cadei et al., 2004). Cast iron typically has non-linear stress-
strain relationship with low values of secant modulus ranging between 100 GPa and 145
GPa. It also has a higher compressive strength than its tensile strength. The maximum
compressive strength can reach 772 MPa, while the maximum tensile strength is limited
to 280 MPa.
2.2.2 Steel
Commercial steel has been produced since 1860; however, it is rare to find steel
structures built before 1890. Steel has become the backbone of the structural engineering
infrastructure and has remained the material of choice for various projects by engineers
due to its light weight, and favourable strength and ductility. Steel is also an ideal
12
Chapter 2
material for short, medium, and long span bridges because of its durability, ease of
maintenance, and ease of use in construction. Typically, steel has similar mechanical
properties in tension and compression, including an elastic modulus of 200 to 210 GPa
and a yield strength ranging between 230 MPa and 700 MPa (Kulak and Grondin, 2002).
Two types of buckling may occur to compression steel elements, namely, local buckling,
which occurs within the thin elements comprising the cross section of a member, and
Although there are no definite boundaries between short and long columns, it is believed
that columns with slenderness ratio values less than 20 (Fy = 300 MPa) may be
considered as short columns and will not undergo the overall buckling type of failure
(Kulak and Grondin, 2002). Local buckling of the walls of a cross section depends on the
width-to-thickness ratio and the type of support provided to the wall. For example, in
short columns consisting of rectangular Hollow Structural Sections (HSS), two opposite
sides would typically buckle outwards and the other two sides would buckle inwards
(Dawe et al., 1985 and Key et al., 1988). In thin-walled sections, this type of buckling
13
Chapter 2
Steel columns of medium to high slenderness ratios are rather susceptible to overall
buckling failure, before developing their full plastic capacity. Due to some unavoidable
disturbances during the rolling and cooling processes, the steel sections produced can
never be perfectly straight. As such, when these sections with their imperfect shapes (also
essentially the stresses remaining in an unloaded member after it has been formed into a
finished product. Examples of such stresses include but are not limited to: those induced
by cooling after rolling (as in the case of hot-rolled section) and cold bending (as in the
design as they result in reduction of flexural stiffness of the columns and consequently in
a lower buckling strength (Weng, 1984 and Key and Hancock, 1993). Although residual
stresses are self-equilibrating, the effective moment of inertia of the cross section will be
changed when parts of the section, which have residual compressive stresses, are yielded.
The studies conducted by Beedle and Tall (1960), Tebedge et al. (1973), among other
researchers, found that the magnitude of maximum residual stresses in hot-rolled sections
of a moderate steel strength is approximately equal to 30 percent of the yield strength and
are uniformly distributed across the thickness of the plate. A typical idealized residual
stress distribution in a hot-rolled I-section is shown in Figure 2.1(a), in which the shown
stresses are normal to the cross sectional plane. An extensive experimental investigation
14
Chapter 2
Birkemoe (1983) and Key and Hancock (1985). The investigations revealed that two
longitudinal residual stress gradients can in fact be found in cold-formed HSS tubes. One
gradient is known as the perimeter (membrane) residual stress and is developed parallel
to the tube wall, as shown in Figure 2.1(b). The other gradient is known as the through-
thickness residual stress and is developed across the tube wall face and along the cross
sectional perimeter, as shown in Figure 2.1(c). It was found, however, that the through-
thickness residual stress gradient is the most dominant parameter that affects the tangent
modulus and ultimate strength of HSS columns (Davison and Birkemoe, 1983). The
magnitude of this type of residual stresses varies from 25 to 70 percent of the material
Retrofit of existing structures is typically needed when live loads increase beyond those
the structures were originally designed for. It may also be required because of an
following steps are recommended for upgrading steel bridges (Bakht et al., 1979):
1. Welding cover plates to the critical flange areas of the bridge floor beams.
2. When flange material is added, the existing bolting system may become
bolts.
plates.
15
Chapter 2
5. If the web was not originally spliced to resist moment, it may be spliced by
cover plates. Cover plates must, however, be welded to the gusset plates.
reduce the width-to-thickness ratio of the plates that comprise the cross section, in
order to avoid local bucking and fully utilize their yield strength.
The previously mentioned methods of retrofitting steel bridges (and structures) typically
involve bolting or welding additional steel plates to the structure. These methods,
welding is not a desirable solution due to fatigue problems associated with weld defects
(Kulak and Grondin, 2002). On the other hand, mechanical (bolted) connections, which
have better fatigue life, are time consuming and costly. Drilling holes for bolted
connections also results in a cross sectional loss as well as the introduction of stress
raisers. Additionally, steel plates require heavy lifting equipment and may add
considerable dead loads to the structure, which reduces their strengthening effectiveness.
The added steel plates are also susceptible to corrosion, which could lead to an increase
16
Chapter 2
There is a need for adopting durable materials and cost-effective retrofit techniques to
overcome some of the drawbacks of conventional techniques stated earlier. One of the
general, FRP materials provide superior strength-to-weight ratios for retrofit of structures.
FRP rigid plates and flexible sheets are available and can easily be applied to the metallic
surface. FRP flexible sheets in particular offer a unique advantage of being able to
areas in the metal as typically occurs in welding (Grabovac et al., 1991). Bonding FRP
materials to metallic structures was first used in aerospace and mechanical engineering
applications. CFRP laminates have been successfully used to repair damaged aluminum
and steel aircraft structures (Armstrong, 1983 and Karbhari and Shully, 1995). Bonding
of composite laminates was also shown to have many advantages for marine structures
process by which they are manufactured. In the context of this thesis, CFRP material will
be referred to according to its elastic modulus. CFRP material with an elastic modulus
value less than that of steel (i.e. ECFRP < 200 GPa) will be referred to as Standard
17
Chapter 2
200 GPa and 400 GPa will be referred to as High Modulus-CFRP (HM-CFRP). CFRP
material with high value of elastic modulus larger than 400 GPa will be referred to as
fibres and resins used to develop FRP materials is reported elsewhere (Cadei et al., 2004).
Figure 2.2 shows typical stress-strain curves of commercially available SM-, HM-, and
Force transfer between FRP and steel is controlled by bond at the interface between the
two materials. Bond performance is influenced by several factors such as the bonded
length and width, type of fibres and adhesive (resin), surface preparation, thickness of
adhesive, and thickness of FRP laminate. One of the simplest test configurations for
investigating bond strength and behaviour between either similar or dissimilar materials
is the single- or double-lap shear joint test (ASTM, D5868-01). Lam et al. (2004)
investigated the tensile strength of double lap joints using SM-CFRP plates. Four
different lap lengths of 50, 75, 100, and 150 mm were employed. Although debonding
was the typical mode of failure for all lap lengths, an increase in load was obtained by
increasing the lap length. In addition, large displacement at failure was observed for
joints with longer lap lengths. Photiou et al. (2006b) investigated the behaviour of
double-lap joints using both SM-CFRP and HM-CFRP plates. The study showed that the
low value of ultimate strain of the HM-CFRP resulted in joints with lower strength
compared to using the SM-CFRP plates, which have a higher ultimate strain. Moreover,
inserting a layer of GFRP with a low value of elastic modulus (softer) between steel and
18
Chapter 2
HM-CFRP resulted in a more gradual load transfer (i.e. better shear stress distribution)
and increased the joint capacity by 26 percent. The previous lap tests attempted to
Miller (2000) examined the behaviour of a bonded CFRP system using double-lap joints.
Two 457 mm long and 37 mm wide CFRP plates were bonded on both sides of a 914 mm
long steel plate. It was found that approximately 98 percent of the total force is
transferred within the first 100 mm of the bonded plate as shown in Figure 2.3.
Al-Emrani et al. (2005) used a similar test configuration, as that shown in Figure 2.3,
however, the steel plate width was tapered from 90 mm at both ends to 36 mm at the
middle. This specific geometry was introduced to allow for possible steel yielding prior
to failure of CFRP material. The study reported that although using HM-CFRP plates (i.e.
E = 362 - 383 GPa) results in increasing the yield strength of the steel specimen, the
fibres rupture in tension in a brittle manner at low strains. On the other hand, the highest
gain in both strength and ductility was obtained by using SM-CFRP (i.e. E = 155 - 175
GPa). However, in this case, debonding occurred after the steel plate has progressively
yielded near the mid-section and the adhesive layer was no longer able to accommodate
the large difference in deformation between the steel plate and the SM-CFRP.
Although lap joint specimens are simple to investigate the bond behaviour, they examine
the adhesive under shear stresses only, which could be useful for limited number of
applications. Nozaka et al. (2005a and 2005b) developed a special test setup to simulate a
19
Chapter 2
more realistic case where the adhesive is tested while bonded to a flexural member. In
this case, the adhesive is subjected to both shear and peel stresses. The test setup
consisted of a W360 x 101 steel section with a large hole and a slit introduced at mid-
span of the girder, as shown in Figure 2.4, to represent a severe crack in the tension
flange of a fatigued girder. Two steel plates were then bolted to the bottom side of the
notched flange and then FRP strips were applied to connect these two plates together. A
total of 27 specimens were tested to study different factors affecting bond of CFRP strips,
including the CFRP and adhesive types, crack width, bond configuration, and the bonded
length. The experimental results indicated that an adhesive with relatively large ductility
is required to redistribute the stresses successfully within the thickness of the adhesive
layer.
Research efforts to examine the feasibility and effectiveness of retrofitting steel structures
20
Chapter 2
A brief summary of research activities and findings in each of these areas is given in the
Corrosion is the most common cause of deterioration in steel structures. FRP repair
Gillespie et al. (1996a). In this research program, two full-scale I-girders were removed
from an old and deteriorated bridge, repaired using FRP and then tested. The two
corroded girders were 9754 mm long, 610 mm deep, and had a flange width of 229 mm.
Both girders had uniform corrosion along their length, mostly concentrated within the
tension flange, which is typical in many bridges. Evaluation of the girders indicated,
approximately, a 40 percent loss of the tension flange. This flange loss resulted in a 29
percent reduction in stiffness. Since the webs of the girders were not severely corroded, it
was decided that only the bottom flanges would be retrofitted, along the entire length of
the girders, using a single layer of CFRP strip, 6.4 mm thick and 38 mm wide. The
repaired girders were then tested and the load was increased until local buckling failure
occurred in the compression flange, since no concrete slab was provided. As the
corrosion of the first girder was more severe than that of the second girder, it was found
that the CFRP strips have increased the elastic stiffness of the first and second girders by
10 and 37 percent, respectively. The ultimate capacities of the first and second girders
were also increased by 17 and 25 percent, respectively. These increases are with respect
to the predicted capacities of the unrepaired specimens. It is believed, however, that the
gain in ultimate capacity could have been higher, provided that local buckling of the
21
Chapter 2
compression flange was prevented. Furthermore, it was shown that the inelastic strains in
the tension flange were reduced by 75 percent in comparison to the unrepaired girder, at
Artificial damage of steel sections has been attempted to simulate corrosion or fatigue
cracks. Section loss due to corrosion is typically simulated by cutting part of the flange or
the web, or machining the tension flange to a reduced thickness throughout the entire
span, as shown in Figure 2.5. Fatigue cracks are simulated by introducing a partial or
complete saw cut in the steel flange thickness. Another method of introducing an
artificial damage is by loading the steel girder beyond yielding and then unloading. These
damaged sections are then repaired with FRP to study the effectiveness of the system in
recovering the strength and stiffness of the member, as discussed in the following
sections.
Three-point bending tests were conducted on four simply supported W310 x 21 beams
with a span of 2438 mm (Liu et al., 2001). Specimen 1, with an intact cross section, was
tested without retrofit as a control specimen. The tension flanges of specimens 2, 3, and 4
were completely cut, as shown in Figure 2.5(a) within a length of 102 mm, at mid-span.
Specimen 2 was tested without CFRP repair to serve as a damaged control specimen.
Specimens 3 and 4 were repaired by bonding a HM-CFRP plate of 100 mm width and 1.4
mm thickness to each specimen. In order to examine the effect of the bond length, the
22
Chapter 2
CFRP plate covered the entire length in specimen 3 and one quarter of the length in
specimen 4. Since no concrete slabs were provided on the compression side, the beams
were laterally braced at the supports as well as two quarter points. Both unrepaired
specimens 1 and 2 failed by lateral torsional buckling of the compression flange. The
failure mode of specimen 3 (full length repair) was due to a gradual debonding of the
CFRP laminate, which initiated at mid-span, and extended to the end as the load
increased. This behaviour was triggered by the high stress concentration and high shear
stresses near the cut part of the flange. Failure of specimen 4 was due to sudden
debonding of the CFRP laminate. Figure 2.6 shows the load versus mid-span deflection
of the four specimens. The figure shows that the strength of the intact specimen has
dropped from 200 kN to 106 kN as a result of cutting the lower tension flange. The figure
also shows that none of the two repaired specimens recovered the strength of the control
intact specimen. Test results showed, however, 56 and 41 percent recovery of the lost
The effect of partial cutting of the flange, shown in Figure 2.5(c), using different depths
of the cut was also examined, using 1220 mm long S130 x 15 beams (Tavakkolizadeh
and Saadatmanesh, 2001b). Cuts of 3.2 mm (shallow) and 6.4 mm (deep) depths were
introduced in the tension flanges, which represented 40 and 80 percent area losses of the
flange, respectively. The main difference between shallow and deep cuts was the
significant loss of ductility in the case of deep cut. SM-CFRP sheets, 7.6 mm wide and
0.13 mm thick, of different lengths were used (100 mm to 600 mm) to restore both
strength and stiffness of the damaged beams. Results indicated increases of 144 and 63
23
Chapter 2
percent in the ultimate capacities with respect to the counterpart damaged beams of 40
percent and 80 percent loss of tension flange, respectively, regardless of the length of
CFRP.
composite girders, commonly used in buildings and bridges. The presence of a concrete
slab in a composite action with a steel girder provides a continuous reinforcement and
support to the compression steel flange, which prevents the premature lateral torsional
buckling and thereby shifts failure to the tension side, at higher loads. Furthermore, the
location of the neutral axis of a composite section is normally shifted upwards, towards
the concrete slab. Therefore, it is expected that the FRP system applied to the tension
flange of composite sections will be utilized more effectively than in steel girders.
span. The steel sections were W355 x 13.6 and the concrete flanges were 910 mm x 75
mm. The area of the tension flanges of the girders was reduced in a similar fashion to that
shown in Figure 2.5(b), to simulate 25, 50, and 100 percent loss of its tensile capacity.
The specimens were repaired with SM-CFRP laminates, of 3950 mm long and various
cross sectional areas ranging between 97 mm2 for the girders with 25 percent loss and
483 mm2 for the girders with 100 percent loss, and were then tested to failure. It was
found that the strength was not only restored but also increased by 20, 80, and 10 percent,
using CFRP laminate areas of 97, 290, and 483 mm2, respectively, compared to a
24
Chapter 2
calculated value for the intact (control) specimen. On the other hand, no extra gain in
stiffness was noticed, where the measured stiffness values were 91, 102, and 86 percent,
of that of the intact girder, for the aforementioned girders. It was also found that rupture
of CFRP laminate occurred in the girder having 25 percent loss and repaired with 97 mm2
of CFRP laminate, as shown in Figure 2.7(a). The girder having 50 percent loss in tension
flange and repaired with 290 mm2 of CFRP laminate failed by crushing of the concrete
slab, followed by a limited debonding of the CFRP laminate at mid-span. In the case of
the girder having 100 percent loss in tension flange and repaired with 483 mm2 of CFRP
should be noted that the change in failure mode of the previous specimens could be
related to both the difference in the degree of damage and the area of CFRP laminate. In
both cases (50 percent and 100 percent loss), the failure of the girders was always
associated with crack propagation, as shown in Figure 2.7(b). In a different study, Al-
Saidy et al. (2004) followed a similar approach by introducing 50 percent and 75 percent
loss of the tension flange of W200 x 22 using the technique shown in Figure 2.5(b). It
was shown that repairing the girders using HM-CFRP plates was able to fully restore the
strength of the original undamaged girders, whereas the stiffness was only partially
restored. The CFRP debonding mode of failure was not observed in this study. Only
crushing of the concrete slab or rupture of the CFRP plates was reported. In fact, this
could be attributed to the HM-CFRP plates used in this particular study, unlike the SM-
CFRP plates used in the previous study by Tavakkolizadeh and Saadatmanesh (2003b).
The higher the elastic modulus of the CFRP material, the lower the rupture strain is.
25
Chapter 2
Sen et al. (2001) used a different approach to simulate severe service distress in 6100 mm
long girders. Six W200 x 36 steel girders acting compositely with concrete slabs (710
mm x 114 mm) were loaded beyond the yield stress, and then unloaded. The girders were
then repaired using SM-CFRP laminates of thicknesses 2 and 5 mm. The increases in
strength were 21 and 52 percent in the specimens repaired with 2 mm and 5 mm thick
other hand, stiffness of the repaired specimens was marginally increased. For all repaired
girders in this study, mechanical clamps were used at the CFRP laminate ends to prevent
FRP materials may not only be used to restore the capacity of damaged or deteriorated
members but can also be used to increase the strength and stiffness of intact members.
the effectiveness of using CFRP plates in strengthening applications. Colombi and Poggi
(2006) tested four HEA140 steel beams of 2500 mm span in three-point bending.
Although the tests were terminated before failure, after excessive deflections were
observed at mid-span, the beam strengthened with two 120 mm x 1.4 mm layers of HM-
CFRP plates achieved 14 and 40 percent increase in stiffness and strength, respectively. It
26
Chapter 2
should be noted that the ends of the CFRP plates were wrapped with CFRP sheets, which
were extended up along the web, to provide anchorage and prevent the debonding failure
mode. Based on the strain distribution along the CFRP plates, the study estimated that a
stresses from the tension flange to the bonded CFRP plates. This value agrees with the
observations of the experimental and analytical studies performed by Miller (2000), and
Linghoff et al. (2006) tested five HEA180 steel beams of 1800 mm span in four-point
bending. The beam strengthened with two HM-CFRP strips of 80 mm x 1.2 mm attached
to each of the bottom and top sides of the lower flange achieved 18 percent increase in
strength. The load carrying capacity of the beam was dropped after the bottom layers of
Deng and Lee (2007) tested ten 127 x 76UB13 steel beams of 1100 mm span, either in
three- or four- point bending. HM-CFRP plates, 76 mm wide, with different thicknesses
and lengths bonded to the lower flange. It was concluded that increasing the CFRP plate
thickness as well as decreasing its length initiate the debonding mode of failure at lower
load levels. The maximum gain in strength (30 percent) was achieved in the beam with
the longest bonded length (500 mm) and thinnest CFRP plate thickness (3 mm). It should
be noted that for all the previous experimental investigations, the beams were laterally
27
Chapter 2
It can be concluded for these studies that the use of CFRP materials with elastic modulus
equivalent to that of steel (i.e. HM-CFRP) and a reasonably high tensile strength would
be most suitable for increasing flexural strength of steel beams. On the other hand, CFRP
with even higher modulus (i.e. UHM-CFRP) or larger cross sectional area would be
Edberg et al. (1996) and Gillespie et al. (1996b) studied four different strengthening
schemes, as shown in Figure 2.8, applied to the tension flange of W200 x 15 steel beams
of 1372 mm span, over the middle 1219 mm. The first scheme [Figure 2.8(a)] consisted
of a 4.6 mm thick CFRP plate, bonded directly to the tension flange of the steel beam.
The second scheme [Figure 2.8(b)] consisted of a similar CFRP plate, but was bonded to
an aluminium honeycomb block, which was bonded to the steel flange. The idea was to
position the CFRP plate further away from the centroid of the steel section, to increase its
moment of inertia. In the third scheme [Figure 2.8(c)], a foam core was attached to the
tension flange, followed by wrapping the whole assembly by a GFRP sheet, which
contains fibres in the ± 45 degree directions. The fourth scheme [Figure 2.8(d)] consisted
of a GFRP pultruded channel, which was both adhesively bonded and mechanically
connected to the tension flange with self-tapping screws. Based on the test results, the
increases in stiffness were 20, 30, 11 and 23 percent, for the schemes shown in Figure
2.8(a, b, c, and d), respectively, whereas, the increases in strength were 42, 71, 41 and 37
percent, respectively. It was concluded that the sandwich CFRP-plated technique [Figure
2.8(b)], was the most efficient, while the GFRP wrapped system [Figure 2.8(c)] was the
least efficient.
28
Chapter 2
El Damatty et al. (2003) studied the effect of bonding 154 mm wide by 19mm thick
GFRP plates to both the top and bottom flanges of W150 x 37 sections. Four-point
bending tests were conducted on beams of 2800 mm span. The reported failure mode was
delamination within the GFRP plate in the tension side. No failure of the adhesive
between steel and GFRP was observed. The reported increases in stiffness, yield and
In the last two studies presented, a 4.6 mm thick CFRP plate, and two 19 mm thick GFRP
plates were bonded to steel beams of comparable sizes to enhance their structural
performance. It is worth noting that the GFRP plate, which is four times thicker than the
CFRP plate, has provided an increase in stiffness, 25 percent lower than that provided by
CFRP. On the other hand, the GFRP plate has provided an increase in strength, 36
percent higher than that provided by CFRP. This suggests that GFRP could be suitable
for strength-controlled applications, provided that a fairly thick plate is used. On the other
GFRP and CFRP systems in enhancing both the stiffness and strength, especially with
GFRP has the advantage of lower cost and also does not develop galvanic corrosion, as
As mentioned earlier, the presence of a concrete slab above the steel girder adds stability
to the compression steel flange against the lateral torsional buckling. Tavakkolizadeh and
29
Chapter 2
CFRP plates. Three 4780 mm long W355 x 13.6 steel girders with 910 mm x 75 mm
concrete slabs were tested in four-point bending. The reported increase in ultimate load
for the girders strengthened with 1.3, 3.9 and 6.4 mm thick CFRP plates were 44, 51 and
76 percent, respectively. All girders failed by crushing of the concrete slab. A sample
load versus mid-span deflection as well as the failure mode of the girder strengthened
with 6.4 mm thick CFRP plate is shown in Figure 2.9. Al-Saidy et al. (2007) followed a
similar approach but bonded an additional HM-CFRP plate on each side of the lower 50
mm of the web. Crushing of concrete slabs was also the observed mode of failure for all
specimens. The achieved increase in strength was 45 percent; however, the increase in
In all the previous studies introduced in this section, there was no remarkable increase in
the girders’ stiffness observed when CFRP reinforcement of a reasonable amount was
used. This is attributed to the value of elastic modulus of the CFRP materials used, where
only minor improvement in the transformed section properties occurred as a result of the
SM- and HM-CFRP used. A major increase in stiffness can be achieved by using UHM-
CFRP with a significantly higher modulus of elasticity than that of steel. Schnerch (2005)
used externally bonded HM- and UHM-CFRP laminates to strengthen two large-scale
steel-concrete composite beams. The beams consisted of W310 x 45 steel sections and
840 mm x 100 mm concrete slabs. The modulus of elasticity of the HM- and UHM-CFRP
materials was 229 GPa and 457 GPa, respectively. A four-point bending load
configuration was used with a 6400 mm span and a 1000 mm constant moment region.
30
Chapter 2
The CFRP plates were wrapped at their ends with 330 mm wide CFRP sheets, which
were extended up on the web from both sides. The HM-CFRP plates increased both the
elastic stiffness and flexural strength of the beams by 10 and 16 percent, respectively. On
the other hand, the UHM-CFRP strengthening (area of CFRP is 70 percent larger than the
previous case) increased both the elastic stiffness and flexural strength of the beams by
36 and 45 percent, respectively. Both beams failed by rupture of the CFRP plates.
Dawood (2005) also used UHM-CFRP strips with elastic modulus of 460 GPa to
strengthen 3050 mm long concrete-steel composite beams. CFRP end wraps were also
used in this investigation to prevent debonding of the CFRP plates. Substantial increases
in both stiffness and strength of 46 and 66 percent, respectively, were achieved. Rupture
Web damage of I-girders has been simulated and studied by Shully et al. (1994).
Specimens of 711 mm span were tested in three-point bending, where a 100 mm diameter
hole was drilled within the shear span at mid-height of the web, as shown in Figure
2.5(d). Different types of bonded FRP sheets were used to repair the web. All repaired
specimens failed in a similar fashion. As the load was increased, the FRP systems began
to buckle over the area of the hole, and then separated from the web. As such, all the
repaired specimens could not recover the strength of the undamaged specimen. The
specimens were able, however, to achieve up to 7 percent increase in strength. The fibre
31
Chapter 2
Patnaik and Bauer (2004) designed a built-up I-section with wide and thick flanges to
promote elastic buckling of the 3.2 mm thick web under shear stresses. Two CFRP-
strengthened beams of 3350 mm span were tested in four-point bending and compared to
a third control beam. The two 1270 mm long shear spans of the strengthened beams were
identically strengthened from both sides with adhesively bonded 1.4 mm thick vertical
CFRP strips. An increase of 26 percent in shear strength of the beam was reported. The
beams failed by web buckling due to high shear stresses, associated with debonding of
For shear strength applications, it is believed that the most effective fibre orientation of
the CFRP material bonded to the web would be in the directions of principal stresses (i.e.
Tubular structures have become increasingly popular in both the steel and aluminium
members can be used either as flexure members in floor beams and telecommunication
monopoles or as axially loaded members in truss structures and bridges. In tubular cross
sections, FRP sheets could be bonded in both the transverse and longitudinal directions.
The transverse wrapping could be used to minimize the debonding of the longitudinal
FRP sheets and also to provide bracing for the steel section itself against the outward
local buckling.
32
Chapter 2
length, used in telecommunication applications. The specimens were tapered with a base
diameter of 457 mm, a tip diameter of 330 mm, and a wall thickness of 4.7 mm. The
poles were strengthened using either HM-CFRP sheets (E = 229 GPa) or strips (E =338
anchorage system was used by bolting steel angles to the base plate on top of the CFRP
sheets. Additional transverse sheets were used to wrap the longitudinal sheets or strips to
prevent them from premature buckling. Figure 2.11 shows the load versus the cantilever
tip deflection of one specimen. Test results showed that both the elastic stiffness and
ultimate strength were increased by 25 and 17 percent, respectively. In all tests, the
longitudinal CFRP strips on the tension side, near the base, were ruptured and a drop in
the applied load was observed. Eventually, failure occurred by local buckling of the
monopole on the compression side, 150 mm away from the base. This was associated
with rupture of the transverse sheets, as shown in Figure 2.11. The study recommended
prestressing the HM-CFRP strips to further enhance the stiffness of the monopoles.
Seica et al. (2006) tested seven 168 x 4.8 mm circular HSS beams of 2200 mm long
spans in four-point bending. The study simulated repair of offshore pipelines, and
included one control specimen, two specimens wrapped with CFRP under standard
conditions “in air”, and four specimens wrapped with CFRP in artificial sea water. Two
layers of CFRP sheets with fibres oriented in the longitudinal direction were first bonded
to each tube and then wrapped with a third layer with fibres oriented in the transverse
direction. For the tubes wrapped and cured in air, both the stiffness and strength were
33
Chapter 2
increased by 18 and 27 percent, respectively. On the other hand, the tubes wrapped and
cured underwater were not able to attain the flexural capacity of those cured in air.
Failure typically occurred by buckling and debonding of the CFRP on the compression
side. Haeider et al. (2006) carried out a similar research on 85 mm diameter circular tubes
of 1.1 mm wall thickness and varied the number of layers in each direction. The study
showed that sheets with fibres oriented in the transverse direction have a considerable
Vatovec et al. (2002) filled the middle half of 152 x 152 x 4.8 mm HSS beams with
the tension and compression flanges. No transverse wraps were used in this study. Test
results showed that the ultimate moment capacity was increased from 6 percent for the
tube strengthened with one strip attached to the compression flange only, to 26 percent
for the specimen with two strips attached to the tension flange and one strip attached to
the compression flange. The governing failure mode of all specimens was delamination
of the CFRP strips on the compression flange, followed by delamination of the strips on
the tension flange. The CFRP strips on the compression flange buckled upwards, split
HSS beams of 1700 mm long span. The tension flange was machined to be reduced to
half of its original thickness in order to simulate material loss due to corrosion. The study
34
Chapter 2
result in a higher increase in both strength and ductility than the 60 mm x 2.40 mm HM-
CFRP, which exhibits sudden rupture of fibres due to its low ultimate strain. Failure,
generally, occurred by debonding of the CFRP sheets. The study also investigated the
effect of adding ± 45 degree GFRP U-wraps, extended to the mid-height of the webs, to
the previous repair system. The test results showed that the GFRP U-wraps prevented the
Tension tests have been done by Jiao and Zhao (2004) and Fawzia et al. (2007) on pairs
of very high strength (VHS) 38 mm diameter circular steel tubes of 1350 MPa yield
strength. The tubes were butt-welded together and strengthened with either HM-CFRP
(i.e. E = 240 GPa) or UHM-CFRP (i.e. E = 640 GPa) wraps with fibres oriented in the
longitudinal direction. The bonded length was varied among the specimens from 23 to
126 mm. Figure 2.12 shows the load versus the bond length for specimens strengthened
with HM-CFRP. The study showed that the higher the modulus of elasticity of CFRP
material, the shorter the development length required. All specimens strengthened with
HM-CFRP experienced bond failure, rather than tension failure. Nevertheless, they
achieved higher strengths than specimens strengthened with UHM-CFRP, which failed in
tension by rupture of CFRP. Both modes of failure are also shown in Figure 2.12.
Zhao et al. (2006) addressed the problem of web crippling at regions of bearing stresses,
in thin walled 100x50 mm rectangular HSS tubes with wall thicknesses of 2, 3, and 5
mm. The study investigated five different schemes of bonding SM-CFRP plates or HM-
35
Chapter 2
CFRP sheets to the rectangular HSS, as shown in Figure 2.13. The study was focused on
Type 3 and Type 5. Test results showed the Type 3 strengthening achieved 50 percent
increase in the web crippling capacity, due to the change of failure mode from web
crippling to web yielding. The study also indicated that thinner sections benefit more
Fam et al. (2006) introduced a new technique for repair of cracked welded joints of truss-
type aluminum highway overhead sign structures comprised of circular tubular members.
The repair system utilized narrow longitudinal FRP strips wrapped around the truss k-
joint and attached to the diagonal members in alternating v-shape patterns, as shown in
Figure 2.14. The longitudinal layers were then wrapped with additional layers in the
circumferential direction for anchorage. The study showed a complete restoration of the
joint capacity that had a 90 percent loss in the weld perimeter, when CFRP sheets were
used. When a similar number of GFRP layers was used, only 79 percent of the joint
Steel plates have conventionally been welded to steel girders for retrofit applications.
However, the welded detail of steel plates is sensitive to fatigue failure. Several
experimental studies involving fatigue testing of (a) tension coupons (Buyukozturk et al.,
2003 and Jones and Civjan, 2003), (b) notched beams (Tavakkolizadeh and Saadat-
manesh, 2003c), and (c) beams removed from old bridges (Gillespie et al., 1996a and
Bassetti et al., 1998) have been performed to investigate the effectiveness of bonded
36
Chapter 2
CFRP plates in improving fatigue life of steel structures. For the different stress ranges
considered in these studies, the CFRP retrofitting techniques have improved fatigue life
of specimens by a factor ranging from 1.2 to 5.7 times that of the unretrofitted specimens,
Dawood (2005) applied three million fatigue cycles to three steel-concrete composite
beams of a 3050 mm span. The beams consist of W200 x 19 steel section acting
compositely with 525 mm wide and 65 mm thick concrete slabs. At the end of the applied
fatigue cycles, the control unstrengthened beam exhibited a 30 percent increase in the
mean deflection, whereas the two strengthened beams (same amount of CFRP but
different thickness of the epoxy adhesive) exhibited only 10 percent increase of their
mean deflection. The degradation of the mean deflection of the beams, normalized to the
mean deflection of their first fatigue cycle, versus the number of cycles is plotted in
Figure 2.15.
Bassetti et al., (2000) reported that applying prestress forces to the CFRP plates has
significantly decreased the crack growth rate and increased the fatigue life by a factor as
In general, the studies have demonstrated the effect of increasing both the width and
length of FRP laminates (i.e. bonded area) on increasing the fatigue lives of the steel
specimens.
37
Chapter 2
The load resistance of CFRP-strengthened steel frame connections was also studied under
cyclic loading (Mosallam et al., 1998). Two strengthening details were investigated,
stiffener. The effectiveness of the two techniques was compared to that of a fully welded
control specimen. Test results indicated that using CFRP stiffeners has resulted in
increasing ductility. The CFRP bonded stiffener provided the highest ductility with an
increase of more than 1.25 times that of the fully welded control specimen.
Surface preparation is the key for a strong and durable adhesive bond. Since
Surface grinding or sandblasting is recommended to remove all rust, paint, and primer
from the steel surface. Additionally, the bare steel surface may be pre-treated using either
metal oxide surface (AASHTO, 2000). This type of treatment significantly improves the
long-term durability of the bond as it prevents water from penetrating through to the
surface. The bonded side of the FRP plates may be sanded to increase the surface
roughness, using medium grit sandpaper or a sandblaster, and then wiped clean with
acetone. However, excessive surface preparation of FRP plates may cause damage and
expose the carbon fibres, leading to a possible galvanic corrosion if it becomes in direct
contact with the steel surface. The adhesive is then applied to the pre-treated steel
38
Chapter 2
surface. The adhesive typically used is a two-component viscous epoxy. A less viscous
epoxy resin is typically used in the case of bonding flexible sheets. It is generally
recommended to leave the bonded FRP plates or sheets to cure for a sufficient time, not
less than 48 hours. Application of an accelerated curing method such as heating blankets
Generally, the adhesive must perform three functions. It must have adequate bond
strength so that the FRP plates or sheets can be optimally utilized. This requires the
failure mode of the system to be governed by the ultimate strength of the FRP and not by
a premature bond failure. The adhesive must also be sufficiently durable in the
environment of the structure to match the extended life expectancy of the structure.
Finally, the adhesive must also be easy to handle and apply under field conditions.
Analysis has also shown that bond failure of the FRP sheets or plates could occur due to
high peeling stresses normal to the surface. In order to prevent peel-off failure, different
techniques have been proposed. Vinson and Sierakowski (1987) stated that tapering the
thickness of the CFRP plates to a 45o angle at all terminations, as shown in Figure
2.16(a), could effectively limit the peeling stresses. Furthermore, Schnerch et al. (2007)
reported that reverse tapering [Figure 2.16(b)] could even enhance the performance of the
bonded joint more. Mechanical clamping [Figure 2.16(c)] can also be applied over the
ends of the laminates to withstand the peeling stresses (Sen et al., 2001). Bolts could be
used to augment the load transfer capacity of the epoxy adhesive, especially with thicker
laminates. Liu et al. (2001) suggested wrapping GFRP sheets around the tension flange
39
Chapter 2
and part of the web, perpendicular to the longitudinal CFRP laminates. These transverse
sheets would be applied along the length of the girder to avoid delamination of the CFRP
laminates. Schnerch (2005) has also indicated that spew fillets [Figure 2.16(b)], which
result from excess epoxy being squeezed out of the joint when pressed or clamped can
significantly reduce the shear stresses in the adhesive. An extensive review of surface
preparation and FRP bonding problems can also be found in Hollaway and Cadei (2002)
One of the problems associated with adhesive bonding is the complexity of the stress
model to investigate the adhesive shear and normal stresses along the bonded length of
two joint types. These types included the double-sided reinforcement joint and double-
sided splice joint, as shown in Figure 2.17. The model included a correction to account
for shear lag in the adhesive and was capable of analysing joints with tapered
reinforcement (or splice) ends. Figure 2.17 also shows the typical shear stresses of the
adhesive and normal stresses of the adherent plates for both joint types. The model was
experimentally verified by Miller et al. (2001), as shown in Figure 2.3, and also by
Colombi and Poggi (2006), and showed good correlation. Schnerch et al. (2006) also
verified the model when splicing two FRP plates using another FRP plates in long beams.
The model showed accurate predictions of the tensile stress distribution along the bonded
40
Chapter 2
Finite element modeling (FEM) of bonded joints is very sensitive to the number of
elements used in the regions of the expected high shear and peel stresses. The need for
large number of elements is necessary to account for the very small thickness of adhesive
relative to the member’s size, while maintaining reasonable aspect ratios of the elements.
This makes the computational effort very tedious. Attempts to develop finite element
models were, however, made by few researchers. Photiou et al. (2006b) developed a two
dimensional linear elastic FEM to simulate a double lap joint with two different adhesive
thicknesses, namely, 0.1 mm and 0.5 mm. The study showed that the peak shear stress
value in the adhesive layer [Figure 2.17(b)] increases as the adhesive thickness decreases.
The study recommended performing extra experimental verification for the results.
Linear elastic analytical models for the interfacial shear and peel stresses arising when a
thin FRP plate is bonded to the soffit of a steel beam have also been developed by several
researchers, in lieu of the finite element methods. Figure 2.18 shows the typical shear and
peel stress distributions at the cut-off point of CFRP bonded plate. Taljsten (1997)
strengthened beam, adhesive, and FRP plate. The derivation was based on a single point
load acting on the beam. The model, however, does not account for bending of the FRP
plate and neglects the shear variation through the thickness of the adhesive. Due to the
previous assumptions, the solution does not satisfy the zero shear boundary condition at
the ends of the adhesive layer. This approximation is acceptable and the model gives
reasonable results, except at a very small zone near the ends of the adhesive layer, which
is equal to the adhesive thickness (Buyukozturk et al., 2004). Smith and Teng (2001)
41
Chapter 2
developed a model which covers all three common load cases, namely, single point load,
double point load, and uniformly distributed load. Additionally, the model accounts for
the bending deformation of the FRP plate when calculating the normal stresses. Deng et
al. (2004) presented a model that includes the thermal effects as well as tapered ends of
the FRP plates. A parametric study was carried out and showed that the maximum shear
and peel stresses decrease as: (a) the thickness of the adhesive increases, (b) the shear
modulus of the adhesive decreases, or (c) the thickness of the FRP plate decreases. The
tapers were also found to reduce both the maximum shear stresses and peel stresses by
about 30 and 50 percent, respectively, as shown in Figure 2.18. Al-Emrani and Kliger
(2006) developed a model to determine the shear and normal stresses when prestressed
FRP plates are used. The results suggest that using a mechanical anchorage device is
recommended to avoid premature failure of the adhesive, due to the high shear stresses at
used basic principles of mechanics (i.e. equilibrium and strain compatibility) to develop
zadeh and Saadatmanesh, 2003a, Al-Saidy et al., 2004, and Schnerch, 2005) or the finite
element method (El Damatty et al., 2003, and Deng et al., 2004).
In order to establish the moment-curvature response, the cross section is divided into
layers. The principles of strain compatibility and internal force equilibrium are then
42
Chapter 2
applied by varying the depth of the neutral axis and summing the forces acting on the
cross section until equilibrium is satisfied for a predefined strain value. In this process,
the FRP material is assumed to be fully bonded to the steel flange. The internal moment
is calculated by summing the moments of the internal forces. The curvature is determined
as the slope of the strain profile. The process is repeated for different strain values, until
the full response of the section is determined. The deflection can then be calculated by
integrating the curvature along the span. Failure is considered when FRP is ruptured in
Three-dimensional finite element analysis has also been conducted to simulate the FRP-
strengthening technique, where the adhesive bond between steel and FRP plates was
modeled using a spring system with two constants (El Damatty et al., 2003). The first
constant is in-plane of the steel surface to simulate the shear resistance, while the second
constant is normal to the plane to simulate the peel resistance. It should be noted that the
two constants were experimentally obtained from shear lap tests (El Damatty and
Abushagur, 2003). Results obtained from the finite element analysis showed an excellent
agreement with experimental results in both the elastic and inelastic ranges, as shown in
Figure 2.19. The model was also able to predict the distribution of the peeling stresses of
FRP plate bonded to the tension flange of a beam subjected to four-point bending. The
model showed a symmetric behaviour about the mid-span with the critical sections of the
43
Chapter 2
Modeling cracked steel sections and capturing their full behaviour, including the effect of
crack propagation is a complex fracture mechanics problem. The problem has not been
The strength of a bonded joint is generally dictated by the strength of the adhesive. In the
presence of combined shear and peel stresses, failure of the adhesive layer can be
characterised by the maximum principal stress σ1 (Cadei et al., 2004), as given below:
2
σ ⎛σ ⎞
σ 1 = + ⎜ ⎟ +τ 2 ≤ σ (2.1)
2 ⎝2⎠
where, σ and τ are the maximum normal and shear stresses of the joint. σ is the
The design of reliable bonded joints, in general, requires limiting the stresses in the
adhesive material to its proportional limit of the elastic range (Hart-Smith, 1980). This
solutions may be reasonable for design of bonded joints. More detailed design guidelines
for adhesive joints can also be found in “A guide to the structural use of adhesives”
44
Chapter 2
Cadei et al., (2004) have introduced design guidelines for strengthening of steel structures
using FRP materials. Section analysis was recommended to calculate the amount of FRP
material required to achieve the desired strength. It was noted, however, that sectional
analysis should take into account the initial stress in the structure at the time of
such as historic structures constructed using cast iron, were proposed by Cadei et al.,
(2004). They also gave design guidelines for strengthening metallic structures acting
Schnerch et al. (2007) proposed another set of design guidelines for strengthening bridge
on the use of HM-CFRP, as summarized in Figure 2.20. The guidelines stated that the
allowable increase of live load for a steel–concrete composite beam strengthened with
1. The flexural yield load of the strengthened beam should be greater than the
2. The strengthened member should remain elastic under the effect of the increased
live load. This is achieved by insuring that the total service load of the
strengthened beam, including the dead load and the increased live load should not
exceed 60% of the calculated new yield capacity of the strengthened beam.
3. To satisfy the ultimate strength requirements, the total factored load based on the
appropriate dead load and live load factors should not exceed the ultimate
45
Chapter 2
factor.
4. To ensure that the structure remains safe in the case of a possible loss of the
strengthening system, the total load, including the dead load and the increased live
Durability of FRP materials bonded to metallic structures, combined with fatigue loading,
has been carefully studied in the aircraft industry. Armstrong (1983) reported on the
condition of an FRP repair that was used for 20 months to patch cracks on the leading
edge of the aluminium wing of a Concord that was flown for 2134 hours and subjected to
576 supersonic flights. At the end, the repair was in such an excellent shape, that it had to
be chiselled off, in order to be removed. It was clear that this repair appeared to still be
One of the most important factors affecting durability is the environmental surroundings.
The FRP retrofitting system itself is non-corrosive, however, when carbon fibres become
in contact with steel, a galvanic corrosion process may be generated. Three requirements
are necessary for galvanic corrosion to occur between carbon and steel: (a) an electrolyte
(such as salt water) must bridge the two materials, (b) there must be an electrical
connection between the materials, and (c) there must also be a sustained cathodic reaction
on the carbon (Mays and Hutchinson, 1992). By eliminating any of these requirements,
the galvanic cell is disrupted. A good selection of adhesives with inherent durability and
46
Chapter 2
high degree of resistance to chlorides, moisture, and freeze-thaw cycles is also very
important. For example, for repairs of steel ships with CFRP patches, Allan et al. (1988)
reported that a moisture barrier comprised of an additional GFRP sheet could be used to
cover the CFRP patch, which is attached directly to the metal using an adhesive. In this
case, in addition to the electrical isolation of the carbon fibres from the metal surface by
the resin matrix, two of the three conditions required for galvanic corrosion to occur were
controlled.
In order to test the durability of the bond between composite materials and steel, the
wedge test method (ASTM, D3762-03) is used. This test has great sensitivity to
environmental attack on the bond and is considered more reliable than conventional lap
shear or peel tests (Scardino and Marceaue, 1976). Shulley et al. (1994) performed wedge
tests on five different types of carbon and glass fibres bonded to steel surfaces.
Specimens were placed in five different environmental conditions (hot water, freezing,
freeze/thaw, salt water, and room temperature water) for two weeks before initiation of
the wedge test. After the wedge was inserted into the bond line, the specimens were
returned to their respective environments. The recorded crack growth rate after seven
days showed no dominance of one environmental effect over the others. There was
evidence that the GFRP reinforced systems have a more durable bond with steel than
CFRP. Also, the most durable bond systems were those subjected to a sub-zero
environment.
47
Chapter 2
Brown (1974) studied the corrosion of CFRP bonded to metals in silane environments.
The metals investigated included aluminium, steel, stainless steel and titanium.
Specimens were fabricated by either bolting the CFRP laminate to the metal or by
bonding with epoxy resin. Accelerated testing was performed by placing the specimens in
days. It was found that for all the metals studied, there was no accelerated deterioration
due to galvanic coupling for the adhesively bonded specimens. However, considerable
deterioration occurred for the bolted specimens. Since most structural adhesives are
insulators, and provided that a continuous film of adhesive can be maintained over the
bonded region, galvanic corrosion should not occur. As indicated earlier, numerous
studies have been conducted on aluminium and steel structures retrofitted with CFRP for
aerospace and marine applications. The studies showed that coupling CFRP with
aluminium is rather a more critical test for durability against galvanic corrosion,
compared to steel, since the electrode potential between carbon and aluminium is even
greater than the potential between carbon and mild steel (Francis, 2000).
The effects of the thickness of epoxy coating and salt water on galvanic corrosion have
also been investigated (Tavakkolizadeh and Saadatmanesh, 2001a). Test results showed
that applying a thin film of epoxy coating (0.1 mm) decreased the corrosion rate in
seawater sevenfold, relative to the specimens with direct contact (i.e. no epoxy) between
steel and CFRP. Furthermore, by applying a thicker epoxy coating (0.25 mm) the
48
Chapter 2
corrosion, including the use of a nonconductive layer of fabric between the carbon fibres
and steel, an isolating epoxy film on the steel surface, or a moisture barrier applied to the
bonded area. The use of a glass fabric layer between CFRP and steel during the bonding
process has been shown to be effective in preventing galvanic corrosion (Karbhari and
Other durability issues such as temperature, creep, ultraviolet, and fire also have
significant effects on the overall response and life-cycle durability of the FRP system. Up
cure for the FRP composite and adhesive. However, beyond a certain level of elevated
temperature, resins and adhesives can soften, which causes an increase in the visco-
The mechanical properties of polymers have characteristics of both elastic solids and
viscous fluids, and hence they are classified as visco-elastic materials. The ambient
operating temperatures of these materials are very close to their visco-elastic phase. Thus,
A basic requirement to minimize creep is to ensure that the service temperatures do not
approach the glass transition temperature of the polymer. The creep characteristic of a
polymer composite is also dependent upon the direction of alignment, the type of fibres,
and the fibre volume fraction. Furthermore, it is also dependent upon the time-dependent
49
Chapter 2
nature of the micro-damage in the composite material under stress. Glass, carbon and
aramid fibres are considered to have small creep component. For composite plate bonded
structures, the resin at the interface between the composite plate and the structure
composites will have glass transition temperature values about 20–30oC above the cure
temperature (i.e. 45–55oC for pre-impregnated materials cured at site temperatures and
by a discoloration of the polymer and a breakdown of the surface of the composite. The
inclusion of ultraviolet stabilizers into epoxy resin formulations seems to have little effect
on discoloration, but there is no evidence that continuous exposure to sunlight affects the
All structural materials undergo some degree of mechanical degradation when exposed to
a severe fire. For FRP-strengthened structures under fire, the resin can neither protect the
fibres nor transfer the load between them. In addition to degradation within the FRP
composite itself, the bond between the FRP and the substrate will eventually fail and
expose the member to the full fire effect. Supplementary insulation systems can,
however, significantly improve the FRP performance at high temperatures (Bisby, 2003).
50
Chapter 2
More details regarding the durability of FRP composite materials used in retrofitting steel
structures can also be found in Karbhari and Shully (1995) and Hollaway and Cadei
(2002).
Field installations to date demonstrate that retrofit of steel structures using FRP materials
can indeed be applied under actual field conditions. This section provides examples of
several field applications utilizing the use of FRP systems in upgrading steel structures.
The Christina Creek bridge (I-704), just outside of Newark (New Jersey State), was
process conducted by the University of Delaware (Miller et al., 2001). A 5.25 mm thick
SM-CFRP plate was bonded to the outer face of the tension flange of the W610x150 steel
girder, which has a span of 7500 mm. Six CFRP plates were placed side-by-side to cover
the entire flange width. The CFRP plates were installed over the full length by using four
overlapped 1500 mm long pieces, as shown in Figure 2.21. Consecutive CFRP plates
were tapered at a 45o angle to form a scarf joint instead of a typical butt joint. Load tests,
using a three-axle dump truck, were performed on the retrofitted girder, prior to and after
the rehabilitation. A comparison between the load test data indicated that adding a single
layer of CFRP plates resulted in 12 percent increase in the girder’s stiffness, and 10
51
Chapter 2
Several metallic bridges in the UK were also strengthened with CFRP plates. The Hythe
Bridge had eight inverted Tee sections (cast iron beams) of 7800 mm span (Luke, 2001).
Four prestressed HM-CFRP plates were bonded to each beam using epoxy adhesive, in
addition to the mechanical end anchorages. The prestressing level was designed to
remove all tensile stresses under service loads. The prestressing technique of FRP was
recommended for the cast iron beams because of their brittleness and limited tensile
In order to overcome fatigue problems in The Acton Bridge of the London Underground,
it was decided to reduce the live load stresses by 25 percent. Prefabricated UHM-CFRP
plates were epoxy bonded to the underside of the girders supporting the track. The post-
installation monitoring verified that the desired reduction in stresses was achieved (Moy
The Tickford Bridge, Newport Pagnell, Buckinghamshire, UK, was built in 1810 and is
the oldest cast iron bridge in service. The bridge was strengthened by bonding wet lay-up
CFRP sheets instead of plates, in order to conform to the curved surfaces. The restoration
was successful to increase the bridge capacity and also in terms of visible effect on the
In the Slattocks Canal Bridge, Rochdale, UK, the steel girders were 510 mm deep and
191 mm wide, and supported a reinforced concrete deck. HM-CFRP plates, 8 mm thick,
were bonded to the bottom flanges of the 12 innermost girders. Repair of the bridge
52
Chapter 2
allowed for upgrading its load capacity from 17 to 40 tonnes. A feasibility study
indicated that it would have cost much more to install a set of special traffic lights for
traffic control, if traditional bridge repairs were used as compared to the cost of
strengthening using CFRP plates. Repair using CFRP allowed for traffic on the bridge
On September 2003, an aluminium truss overhead sign structure with cracked welded
joints due to fatigue has been successfully repaired using FRP sheets (Fam et al., 2006).
The structure is located over Route 88 (westbound direction) in New York State, east of
Other field applications, including King Street Railway Bridge, Bid Bridge, and Bow
Road Bridge involving retrofitting using FRP materials have also been reported by
53
Chapter 2
+
+Frs
+Frs
-Frs
+ _
- -
-Frs -Frs
+Frs t/3 t/3 t/3
-Frp
-
t = wall thickness
t/3
+Frp
+
t/3
t/3
-
-Frp
+Frs
2000
14
1800
14
r H5
712
M9
Carb RP
arb FRP
oDu
1600
ur
CF
ad K63
a C M-C
oD
HM-
1400
hi Diale
Sika
S ik
Stress (MPa)
1200
itsubis
1000
FRP M
800
UHM-C
600 Steel
400
GFRP
HE -51A
200 Tyfo S
0
0 5 10 15 20 25 30 35
Strain x 10-3 (mm/mm)
Figure 2.2 Typical stress-strain curves for CFRP, GFRP, and steel.
54
Chapter 2
3.5
Adhesive 3 Adhesive 2
3.0
Adhesive 1
2.0 100 mm
1.5
0.5
0.0
-250 -200 -150 -100 -50 0 50 100 150 200 250
Distance (mm)
Figure 2.3 Measured and predicted strain distributions along the bonded length of a
double lap joint. [Miller et al., 2001]
2032
Stiffeners
hole
W360x101
x x
4064
Elevation
55
Chapter 2
Hole
diameter
(a) Complete (b) Partial cutting (c) Partial cutting of (d) Web cutting
flange cutting of the flange width the flange thickness
250
Specimen 1
200 100 mm
Debonding
long cut
Specimen 2
Specimen 3
150
Load (kN)
Specimen 4
100
L/4 Load L/4
Lateral
support
50
Deflection
L
0
0.0 1.0 2.0 3.0 4.0 5.0
Deflection (mm)
56
Chapter 2
57
Chapter 2
GFRP
Aluminum (±45 degree)
Honeycomb Foam Core GFRP section
CFRP
700
600 Strengthened
Concrete
crushing (experiment)
500
Strengthened
Load (kN)
400 (calculated)
Control
(calculated)
300
500 mm
200
100
4780 mm CFRP
(th.=6.4 mm)
0
0 10 20 30 40 50 60 70
Deflection (mm)
58
Chapter 2
Debonding/rupture
of strips on tension
side near the base
Load (kN)
59
Chapter 2
160
Effective bond length = 75 mm Test Data
140
100
Load (kN)
80
Rupture of CFRP
60
Debonding
40 of CFRP
20
0
0 50 100 150
Figure 2.12 Effective bond length for steel tube strengthened with HM-CFRP.
[Jiao and Zhao, 2004 and Fawzia et al., 2007]
60
Chapter 2
1.2
Normalized deflection
Control beam
1.1
1.0 CFRP-Strengthened
beams
0.9
0.0 1.0 2.0 3.0 4.0
61
Chapter 2
Spew fillet
CFRP plate
P P P P
Tapered edge
shear stress
shear stress
Adhesive
Adhesive
Uniform thickness
Tapered edge
Uniform thickness
Distance Distance
Normal stress in
reinforcing plate
Normal stress in
splice plate
Distance Distance
40
Normal stress
Normal stress
in substrate
Tapered edge
in substrate
30
Uniform thickness
20
Tapered edge 10
Uniform thickness
Distance Distance
62
Chapter 2
(a) Uinform thickness FRP plate (b) Tapered thickness FRP plate
Figure 2.18 Comparisons of shear and peel stresses for plates with and without
taper under UDL. [Deng et al., 2004]
Steel
beam
Load (kN)
63
Chapter 2
Figure 2.20 Design guidelines for steel–concrete composite beams strengthened with
HM-CFRP materials. [Schnerch et al., 2007]
Figure 2.21 Installation of CFRP plate on the Christina Creek bridge (I-704).
[Miller et al., 2001]
64
Chapter 3
Chapter 3
Experimental Program
3.1 Introduction
structures retrofitted using carbon-FRP (CFRP) sheets and plates, in flexure and under
axial compressive loads. The experimental program consisted of three phases. The first
phase, Phase I, was focused on the axial compression behaviour of 50 short and slender
Hollow Structural Section (HSS) steel columns strengthened using different types of
CFRP sheets and plates. The parameters considered were the effect of CFRP
reinforcement ratio, the effect of fibre orientation, namely, in the longitudinal and
transverse directions, and the slenderness ratio of the columns. The columns were
The other two phases of the experimental program were essentially focused on the
flexural behaviour of steel W-sections acting compositely with concrete slabs and
retrofitted with different CFRP materials. In Phase II, three large-scale girders, scaled
65
Chapter 3
down from an actual bridge, were tested in four-point bending to investigate the
effectiveness of CFRP plates in strengthening intact girders. Phase III was focused on the
repair of artificially damaged beams using CFRP sheets. The tension flanges of 10
fatigue crack or a severe loss of the cross section due to corrosion. The beams were then
repaired with CFRP sheets of different configurations and tested in four-point bending.
The parameters considered were the type of CFRP sheets, force equivalence ratio,
number of CFRP-bonded sides of the tension flange (i.e. upper and lower sides), and the
length of CFRP repair patch. Test specimens in phases II and III were instrumented to
measure their flexural behaviour in terms of both the load-mid span deflection, and the
load-strains responses.
This chapter presents properties of the materials used to fabricate the specimens, details
3.2 Materials
This section describes the properties of various materials used in the experimental
program, namely, steel, concrete, and FRP of different types. Cold-formed HSS sections
were used in Phase I, whereas hot-rolled W-sections were used in Phases II and III. Five
different types of CFRP sheets and plates as well as one type of glass-FRP (GFRP) sheets
were used. Three different concrete batches were prepared for composite girders of
66
Chapter 3
In this section, a detailed description of the cold-formed HSS sections used in Phase I and
Phase I of the research program included two different cross sections of cold-formed
HSS, namely, HSS1 and HSS2. Both sections were manufactured according to CSA
and HSS2 are 44 x 44 x 3.2 mm and 89 x 89 x 3.2 mm, respectively. A schematic of their
Stub-column tests are typically used in lieu of coupon tests to provide the average
compressive stress-strain curves (Bjorhovde and Birkemoe, 1979). This type of test
demonstrates the overall column performance at very low slenderness ratio. The yield
strength criterion is normally used when there is a gradual yielding without a distinct
yield point, and is defined as the stress at a 0.2 percent strain offset. Short columns with
built-in residual stresses typically show a gradual transition from the linear elastic
behaviour to the fully plastic plateau, as a result of gradual yielding. The magnitude of
residual stresses (Frs) can be estimated as suggested by Salmon and Johnson (1980), as
the difference between the yield strength and the proportional limit stress (i.e. the stress at
67
Chapter 3
Two HSS1 and one HSS2 stub-column specimens, 150 mm and 175 mm long,
respectively, were tested. The lengths of the stubs were measured using a measuring tape
with an accuracy of ±1.00 mm. The ends of the stubs were machined to ensure flat and
square faces. Electrical resistance strain gauges were used to measure the longitudinal
strains at mid-height, at the middle of the flat sides of the columns. The strain gauges
were installed on the four sides of HSS1 and on two adjacent sides of HSS2. The
specimens were tested under stroke control at a rate of 0.20 mm/min, using a Riehle
The stress-strain curve of the HSS1 and HSS2 sections based on the average of strain
gauges measurements is shown in Figure 3.3(a and b). The behaviour of HSS1 shows a
proportional limit stress (Fp) of 257 MPa and yield strength (Fy) of 504 MPa, which
indicates that the magnitude of the residual stress is approximately 49 percent of the yield
strength. On the other hand, the behaviour of HSS2 shows a proportional limit stress (Fp)
of 255 MPa and yield strength (Fy) of 382 MPa, which indicate that the magnitude of
specify the limit for the flat width-to-thickness ratio (b t ) of HSS subjected to
(
compressive stresses as 670 )
F y , in order to permit yielding of steel prior to local
buckling under axial compression, where Fy is the yield strength of the steel in MPa. For
the HSS1 and HSS2 types of steel, these limits are 29.8 and 34.3, respectively, whereas,
68
Chapter 3
the actual (b/t) ratios are 9.75 and 23.8, respectively. Therefore, both sections satisfy the
W250x25 hot-rolled sections of weldable steel (Type W) were used for the girders of
Phase II of the research program. A schematic of the cross section is shown in Figure
3.1(b). Tension tests were performed on coupons cut from the web and flange of the same
sections by Savides (1989). The reported average stress-strain diagram is shown in Figure
3.4.
W150x22 hot-rolled steel sections were used for the beams of Phase III. A schematic of
the cross sections is shown in Figure 3.1(b). Uniaxial tension tests were performed
cut from the flanges (with a thickness of 6.5 mm), whereas the other three coupons, W1
to W3, were cut from the web (with a thickness of 5.8 mm). The dimensions were
measured using a digital calibre with an accuracy of ±0.01 mm. A typical coupon is
shown in Figure 3.5. The stress-strain plots for all the six steel coupons are shown in
Figure 3.6. The average yield strengths of the flange and the web were 386 and 406 MPa,
respectively. The average modulus of elasticity of both flange and web was 197 GPa.
Details of the tensile properties of both the flange and web coupons are also presented in
Table 3.1.
69
Chapter 3
Both FRP flexible sheets and FRP rigid plates were used in the experimental program.
Three different types of unidirectional CFRP sheets, referred to as C1, C2, and C3, were
used. The commercial names of C1, C2, and C3 types are Tyfo SCH-35, Tyfo SCH-41
(Tyfo Co. LLC, San Diego, California) and Dialead F637400 (Mitsubishi Chemical,
commercially known as Tyfo SHE-51A, was used. A layer of this GFRP was typically
installed on the steel surface prior to the application of the CFRP sheets to prevent direct
contact between steel and CFRP. This practice has been recommended by many
researchers (Allan et al., 1988, Karbhari and Shully, 1995, and West, 2001) to prevent
galvanic corrosion. Although this was a short term study, and galvanic corrosion was
unlikely to occur, the study was intended to simulate the actual practice that is likely to
take place. The typical thicknesses of a lamina (a single layer of dry fabric, wetted with
resin and cured) is 0.89 mm, 1.11 mm, 0.54 mm, and 1.46 mm for the C1, C2, C3, and G
Two types of 1.4 mm thick pultruded CFRP plates were used (Sika Canada Inc., Pointe-
Claire, QC). The first type, C4, is commercially known as Sika CarboDur M914, and the
plate is 90 mm wide. The second type, C5, is commercially known as Sika CarboDur
70
Chapter 3
H514, and the plate is 50 mm wide. Since the rigid CFRP plates are typically installed
using a relatively thick epoxy layer (3.2 mm), no GFRP layer was used between the
CFRP plate and the steel surface as in the case of CFRP sheets.
Two types of epoxy resins were used. Resin 1 is a two-component Tyfo S (Tyfo Co.
LLC, San Diego, California) epoxy matrix, and was used in the case of flexible sheets.
The resin was used to bond the dry fabric layers to each other and to the steel surface
through a wet-lay up process. The mixing ratio of the epoxy is (2.9:1) of component A
(resin) and component B (hardener), by weight. The Tyfo S epoxy is relatively ductile
Resin 2, commercially known as Sikadur-30 (Sika Canada Inc., Pointe-Claire, QC), was
used in the case of CFRP rigid plates and is essentially a thixotropic adhesive mortar,
based on a two-component solvent free epoxy resin. The mixing ratio is (3:1) of
Several coupons of each FRP type were prepared and tested according to ASTM
D3039/D 3039M, as shown in Figure 3.7(a). The coupons were 250 mm long, with end
tabs of 60 mm long each, in order to minimize the effect of gripping stresses. The tabs
71
Chapter 3
each side of the coupons. Tension tests were performed using an Instron Model 1350
testing machine with wedge-type mechanical grips, as shown in Figure 3.7(b). The
mechanical properties of all coupons of the different types of FRP sheets and plates are
listed in Table 3.2. Figure 3.8 shows the average stress-strain response for each type of
FRP material. A typical tension failure mode of the tested coupons is shown in Figure
3.7(c).
3.2.3 Concrete
One concrete batch was prepared for Phase II (batch 1), while two batches were prepared
for Phase III (batches 2 and 3), as shown in Table 3.3. The first batch of Phase III (batch
2) was mixed at the laboratory of Queen’s University with a target compressive strength
of 46 MPa after 28 days. The other two concrete batches were ordered from a ready-mix
plant, with target strengths of 40 MPa and 45 MPa for batches 1 and 3, respectively. It
should be noted that batch 1 was ordered with high slump, not less than 150 mm, to
increase its workability while pouring in the special formwork constructed for Phase II,
as will be described later. Three 100 x 150 mm cylinders were prepared for each test
specimen at the time of casting, and were kept at room temperature to the date of testing
of the respective specimen. The cylinders were tested using a 1300 kN Reihle testing
machine, as shown in Figure 3.9. The concrete age at the time of testing of each specimen
and its cylinders was at least 2 months. At this age, the concrete strength has already
stabilized. Table 3.3 shows the concrete strength measured for each batch, based on
cylinders tests. Generally, the average measured compressive strengths for batches 1, 2,
and 3 are 38.9 MPa, 49.9 MPa, and 50.2 MPa, respectively.
72
Chapter 3
compressive stresses, two opposite sides would typically buckle outward and the other
two sides would buckle inwards. Therefore, it is hypothesized that externally bonded FRP
sheets, particularly wraps oriented in the transverse direction, could help brace the flat
sides of the column and control the outward buckling, as shown in Figure 3.10(a). On the
other hand, FRP sheets may not contribute much on the sides that buckle inwards and
may in fact debond from the steel surface. In long columns, where global buckling takes
place, it is hypothesized that FRP sheets or plates oriented in the longitudinal direction
could provide tension reinforcement on the outer surface, as shown in Figure 3.10(b).
evaluate the effect of strengthening axially loaded square HSS members with CFRP
sheets or plates. The fabrication and FRP installation processes, instrumentation, and test
In total, 50 square HSS columns were tested in compression. The specimens were divided
into 20 sets, falling into three groups, A, B, and C, as shown in Table 3.4. Groups A and
B included slender columns, whereas group C included short columns. Table 3.4 provides
all details of the columns, namely, the set number, number of similar specimens per set,
steel cross section type, length, and slenderness ratios (kL/r). The table also provides
information on strengthening schemes, including FRP type, number of layers, width, and
73
Chapter 3
fibre orientation. Groups A and C include three identical specimens in each set, whereas
group B includes one relatively large size specimen in each set. The identification of each
specimen in further sections of this thesis will indicate the set number, followed by the
specimen number (for example, 6-1 refers to the first specimen of set number 6). As
shown in Table 3.4, group A columns were fabricated using the HSS1 section (44 x 44 x
3.2 mm), while groups B and C columns were fabricated using the HSS2 section (89 x 89
x 3.2 mm). In group A, sets 1, 3, and 5 served as control (unstrengthened) sets with
slenderness ratios of 46, 70, and 93 respectively, while sets 2, 4, and 6 were the
corresponding strengthened sets. In groups B and C, sets 7 and 12, respectively, served as
control (unstrengthened) sets with slenderness ratios of 68 and 4, while sets 8 to 11 and
13 to 20 were the corresponding strengthened sets, using different CFRP schemes and
number of layers. The type of CFRP used, number of layers, and the fibre orientation are
type C5, is applied on two opposite sides of columns with kL/r ranging from 46 to
93.
74
Chapter 3
slenderness ratio, through sets 7 to 11 of group B. In this case, one, three, and five
layers of CFRP type C3, 75 mm wide, were installed on two opposite sides of the
columns. Also, another column included three layers attached to the four sides.
All columns had kL/r of 68 and had one layer, 75 mm wide, of GFRP between the
comparing sets 13 and 14 relative to 12 for C1 type and sets 17 and 18 relative to
12 for C3 type.
and 16 relative to 12 for CFRP type C1, and sets 17 and 19 relative to 12 as well
3. Effect of CFRP type (i.e. C1 and C3) in group C, by comparing sets (13 and 17)
and sets (14 and 18), relative to the control set 12.
The columns were cut to the desired lengths as listed in Table 3.4. The (L/r) values of the
specimens were 46, 70, 93, 68, and 5 for sets (1 and 2), (3 and 4), (5 and 6), (7 to 11), and
(12 to 20), respectively. The effective length factors (k), required to determine the
slenderness ratios (kL/r), depend on the end conditions and will be discussed later in the
test setup description (3.3.3). The ends of the columns in groups A and C were machined
75
Chapter 3
flat and perpendicular to the longitudinal axis, using a milling machine, as shown Figure
3.12(a). Although this type of machining could not be performed on the columns in group
B, due to their larger size and the limitations of the milling machine, every effort was
made during the cutting process of those particular specimens to ensure a flat and
perpendicular cut.
The overall buckling direction and strength are generally influenced by the out-of-
straightness geometric imperfection of slender columns (Allen and Bulson, 1980). The
ILD1400 laser optical displacement sensor, as shown in Figure 3.12(b). The sensor
operates with a semiconductor laser having a wavelength of 670 nm, which classifies the
sensor in Laser Class II. The out-of-straightness profiles of longitudinal lines at the mid-
width of two perpendicular sides of a sample specimen (specimen 6-3) are shown in
Figure 3.13. The figure shows a single curvature along side “a” of the cross section with a
maximum value of 0.54 mm, and a triple curvature along side “b” with a maximum value
of 0.29 mm. The complete out-of-straightness profiles for column sets 1 to 6 are
presented in Appendix A. It should be noted, however, that the measured values of out-
of-straightness of the bare steel columns are very small, as shown in Figure 3.13. In fact
the installation process of CFRP on two opposite sides, using the hand lay up technique,
perfect symmetry of the CFRP installation. As such, another method will be described in
based on strain measurements. Group B columns were relatively large in size and could
76
Chapter 3
not be fit on the moving bed of the laser sensor. Their out-of-straightness imperfections
were interpreted from the strain measurements. For group C columns, the out-of-
straightness measurements have not been conducted since the columns were very short
and were expected to have local buckling, rather than overall buckling.
Prior to bonding the CFRP sheets or plates, the outer surface of all HSS columns was
sandblasted as shown in Figure 3.12(c), to remove mill-scale, rust, and debris and also to
roughen the steel surface in order to improve the mechanical interlock between the steel
surface and the adhesive. The outer surface was then cleaned using pressurized air to
remove any impurities remaining on the surface from the sandblasting process. The steel
surface was also wiped with acetone to remove any chemical impurities on the surface.
Two CFRP plates (C5 type) of widths 25 and 16 mm were bonded on each of the two
Figure 3.11. For sets 8 to 10 of group B, 70 mm wide CFRP sheets (C3 type) were also
applied to two opposite sides. The plates and sheets in all these specimens were applied
to the two opposite surfaces perpendicular to the plane in which global buckling was
allowed in the test setup. Set 11 was strengthened with three layers of 70 mm wide sheets
(C3 type), applied to all four sides of the column. In all slender column specimens of
groups A and B, the CFRP plates or sheets were installed with the fibres oriented in the
77
Chapter 3
For the short column sets 13 to 20 of group C, CFRP sheets of different orientations and
patterns were applied using either C1 or C3 types, as shown schematically in Figure 3.11
and Table 3.4. In sets 16 and 20, the longitudinal layer of CFRP was installed prior to
A single layer of GFRP (G) was first installed on the steel surface with fibres oriented in
the longitudinal direction for sets 8 to 11 of group B, and oriented in the transverse
direction (complete wrap) for sets 13 to 20 of group C. The FRP plates and sheets were
cut 25 mm shorter than the slender steel columns, from both ends in groups A and B,
mainly to simulate an actual case where access to the column ends may not be feasible.
The second layer of CFRP plates (16 mm wide) applied in column sets 2, 4, and 6 was
cut 50 mm shorter than the steel column from both ends. Loading was thus applied to the
steel cross-section only, without any contact with the CFRP plates or sheets. The ends of
CFRP plates in sets 2, 4, and 6 were wrapped with 50 mm wide GFRP sheets (G), as
shown in Figure 3.14 to hold the CFRP plates at the ends. For short column sets 13 to 20
of group C, the longitudinal and transverse FRP layers were completely wrapped and
at the round corners, as shown in Figure 3.11. Also, the FRP jacket was 5 mm shorter
than the steel specimens from both ends, for the same reasons indicated earlier for slender
columns.
78
Chapter 3
All column specimens were tested under concentric loading using three different test
setups. Test setup A was prepared for group A using a 1000 kN Riehle testing machine,
as shown in Figure 3.15(a and b). The load was applied using stroke control at a rate of
0.50 mm/min, except for specimen (1-1), where a 0.20 mm/min rate was used. Lubricated
cylindrical bearings were used at both ends of the specimen to allow for free end rotation,
in one plane only, as shown in Figure 3.15(b and c). The specimens were braced against
out-of-plane displacement, using two L-shaped frames, as shown in Figure 3.15(a, b, and
d), in order to promote in-plane buckling only. Each L-shaped frame consists of a heavy
square HSS column, attached to the base of the Riehle machine, and a horizontal
cantilevered arm. One end of the cantilevered arm is attached to the column using two
plates and threaded rods, while the other end is welded to a 25 x 25 x 3.2 mm HSS
section to guide the column’s buckling in one plane only, as shown in Figure 3.15(d).
The tests on group B columns were carried out in a specially constructed horizontal setup
B, as shown in Figure 3.16(a and b), since they were too large to be accommodated in the
testing machine. The specimens were placed between two rigid steel reaction columns
anchored to the floor. A 1500 kN hydraulic loading ram was used to apply the load.
Lubricated cylindrical bearings were used at both ends of the specimen to allow for free
end rotation, in the horizontal plane only, as shown in Figure 3.16(c). The specimens
were braced against out-of-plane displacement, using another set of free sliding rollers, as
shown in Figure 3.16(d), in order to promote buckling in the horizontal plane only. The
79
Chapter 3
in-plane effective length factor (k) for the columns tested in setups A and B is assumed
The short columns in group C were tested under concentric loading in test setup C, using
the same Riehle machine used in test setup A, as shown Figure 3.17. The load in this case
was applied using a semi-spherical head on the top end of the specimen whereas the
bottom end was supported by a fixed flat plate, as shown in Figure 3.17. This setup is
analogous to a fixed-hinged condition and therefore the effective length factor (k) can
(kL/r) of 4. The load was applied using a stroke control at a rate of 0.20 mm/min.
3.3.4 Instrumentation
For group A, one horizontal Linear Potentiometer (LP), with a range of 100 ± 0.01 mm,
Figure 3.15(a and d). Vertical displacement was measured directly through the moving
cross head of the testing machine. The longitudinal strains at mid-height of the columns
were also measured using two electric resistance strain gauges, attached directly to the
two opposite sides of the specimens. The strain gauges were 5 mm long with a gauge
resistance of 119.8 ± 0.2 Ω. The gauges were attached to the outer surface of the CFRP
plate (or to the steel surface of the control columns). The load was measured using a load
80
Chapter 3
For group B columns, which were tested in a horizontal plane, six LPs were used to
record both axial and lateral displacements, as shown in Figure 3.16(a). Two LPs were
mounted at each end of the specimen, parallel to the longitudinal direction, to measure its
net axial displacement. The two LPs at each end were placed on the opposite sides of the
specimen in the plane of the buckling. Two transverse LPs were mounted at mid- and
quarter-points of the height of the specimen to measure lateral displacements and capture
the buckling shape. The longitudinal strains at mid-height were measured using two 5
mm electric resistance strain gauges, attached to the two opposite sides of the specimens.
were also attached to the specimen at mid-height (PI1 and PI2), over a gauge length of
200 mm, as shown in Figure 3.16(e). The PI gauges have a displacement range of ± 5.0 ±
0.005 mm. The load was measured using a 2000 kN load cell positioned between the end
of the reaction frame and the hydraulic jack, as shown in Figure 3.16(a).
For group C columns, three LPs were mounted around the specimen, in a vertical
position, to provide a reliable average for the axial displacement of the short columns and
also to check if any unintended eccentricity existed early during the test and correct the
resistance strain gauges, installed in the longitudinal direction on two adjacent sides of
the specimen, 30 mm below the top surface, where local buckling was anticipated [see
Figure 3.10(a)]. The same load cell used for group B columns was also used for group C
81
Chapter 3
The data measured throughout all the tests by the LPs, PI-gauges, load cells, and strain
gauges as well as the load and stroke of the Riehle machine were recorded using a Vishay
System 5000 Data Acquisition System (DAS). The DAS receives the load and stroke of
the testing machine as well as the LPs readings through a high-level input card with an
accuracy of ±10 mV. Both strain gauges and PI gauges readings are transmitted through a
strain gauge card with an accuracy of ± 5 mV. Test data were collected and stored using
Girders
This section describes the second phase of the experimental program undertaken to
girders. Test specimens used in this part of the study were scaled down (4:1) from an
actual bridge, in order to provide realistic proportions of section size, concrete slab size,
and the span. Description of test specimens, and fabrication processes, including CFRP
installation, instrumentation, as well as test setup and procedures are presented in the
following sections.
A total of three steel-concrete large-scale composite girders were fabricated and tested to
failure in four-point bending. Each girder consists of 6100 mm long W250 x 25 hot-
rolled steel section acting compositely with a 65 mm thick, 500 mm wide, concrete slab,
as shown in Figure 3.18(a). The girders include one intact control (unstrengthened)
82
Chapter 3
specimen (G1) and two CFRP-strengthened specimens (G2 and G3). The strengthening
scheme of G2 and G3 consisted of one 90 mm x 1.4 mm layer of CFRP type C4, 4000
mm long, which covers 67 percent of the span. A second layer of 50 mm x 1.4 mm and
1500 mm long was installed. This layer was of CFRP type C4 in the case of G2 and type
C5 in the case of G3. Table 3.5 provides all details of the girders, including their
identification number as well as type, width, and bonded length of CFRP plates.
plates.
2. Examine the effectiveness of the CFRP bonded plates when they cover a length
The steel girders were previously fabricated as part of experimental research programs
by Savides (1989) and He (1992). The purpose of those studies was to investigate
transversely prestressed concrete bridge decks. The girders were used to support the deck
slabs, and failure occurred in the slabs under relatively low load levels such that the steel
girders were still elastic. In this study, the same girders were used with new concrete
83
Chapter 3
slabs. In the following section, a brief summary of the history of the girders, in terms of
The prototype bridge selected to be modeled by Savides (1989) and He (1992) is a two-
lane bridge under loads specified by Ontario Highway Bridge Design Code (OHBDC) for
class A highway. The bridge is 6.6 m wide, with a simply supported span of 24.0 m long
and is supported with three steel girders at 2.3 m spacing and a 1.0 m overhang on each
side. The prototype deck slab was 175 mm thick, which is thinner than the minimum
requirements (225 mm) specified by the OHBDC. A scale factor of 1:4 was chosen for
the model bridge, based on the available laboratory space, testing apparatus, and the need
to model all the bridge components and details. Therefore, the scale bridge model
simply supported steel girders spaced at 569 mm with a 248 mm slab overhang on each
side. The model was tested under statically applied concentrated loads. Each girder has
four pairs of stiffener plates welded to the web. The stiffener plates are 240x40x10 mm
and located at the end supports and at third points of each girder. Shear studs of 8.8 mm
diameter and 35 mm long were welded in pairs at a longitudinal and transverse spacing of
3.18(b). All tested slabs failed in punching shear at a maximum load of 95 kN. This level
of load produced tensile strains of 0.11 percent in the tension flange of the steel section,
comparing this strain to the stress-strain curve shown in Figure 3.4, it was concluded that
these girders remained fully elastic and hence could be reused in the current study. A
84
Chapter 3
detailed design of both the prototype and the model bridges can be found in Savides
(1989). Prior to reusing the girders in the current study, the old concrete slab was
In the current study, due to the large size of the girders and space limitations in the
laboratory, it was decided to complete the CFRP installation process before casting the
new concrete slabs. The underside of the tension flanges were sandblasted, as shown in
Figure 3.18(c). Also, parts of the upper side of the tension flange and the web were
termination points of the CFRP strips. The CFRP plates were cut to the desired
dimensions listed in Table 3.5, using a guillotine cutter. Prior to bonding the plates, dust
was removed from the steel surface by thoroughly blowing compressed air. The side of
the CFRP plate, which was to receive the adhesive, was rubbed with a fine sand paper to
remove all residual carbon dust. The same side was then thoroughly wiped with acetone
using a clean white cloth. Adhesive 2 (Sikadur 30) was applied to the steel surface as a
prime coat using a saw-tooth spatula, as shown in Figure 3.18(d). In order to apply the
adhesive to the CFRP plate, a wooden hopper was specially fabricated and used, as
shown in Figure 3.18(e). A roof shaped spatula with 3 mm height at the edges and 5 mm
height at the middle was placed at one end of the hopper. The CFRP plate was then
pulled through the spatula, under the adhesive, to produce a regular cross section of the
adhesive layer. The plate was then placed on the steel surface and pressed with a rubber
roller, as shown in Figure 3.18(f), using enough pressure to squeeze the adhesive out
from both sides. This particular procedure was to provide a maximum bond line of 3.2
85
Chapter 3
mm thick. All the required tools and recommended procedures were provided by Sika
Canada Inc. A 50 mm wide GFRP sheet was used at all termination points of the CFRP
plates as a transverse wrap around the tension flange and also extended 50 mm within the
The concrete slabs were cast in an inverted position on a smooth flat floor for
convenience, as shown in Figure 3.19. Simple wooden forms were fabricated, and a
double layer of 150x150x5 mm welded wire mesh reinforcement was provided at mid-
thickness of the concrete slab. The steel girders were supported on the edges of the
formwork in an inverted position with the shear studs projecting downwards into the
forms. High slump concrete (batch 1) was then poured into the formwork, vibrated, and
then the surface was troweled. Immediately after finishing the concrete surface, the
specimens were covered using a plastic sheet. The concrete surface was kept wet for
seven days following casting. After seven days, the specimens were released from the
All three girders were tested in a simply supported configuration with a span of 5940 mm
between the centerlines of the supports. Tests were performed using four-point bending
with a distance of 1000 mm between the two applied loads. The loads were applied using
a stiff HSS steel spreader beam. The two point loads were applied over two transverse
rectangular HSS section, which covered the entire width of the concrete slab. Steel rollers
were placed between the spreader beam and each of the rectangular HSS sections, as
86
Chapter 3
shown in Figure 3.20 and 3.21. A quick setting plaster was placed under the transverse
HSS section to avoid any stress concentrations associated with the irregularity of the
concrete surface and to uniformly distribute the load. Due to the length of the specimens,
which is longer than the base of the available Riehle machine, a 6200 long and stiff
welded wide flange (WWF350 x 263) reaction beam was first placed on the testing
machine base. The test setup was then assembled on top of the WWF beam. The details
of the test setup are shown in Figure 3.20 and 3.21. The specimen was supported on a
roller support at one end and on a hinged support at the other end. Both supports were
elevated using heavy HSS square stubs to accommodate the expected large deflection at
mid-span. Two 25 x 25 x 3.2 mm HSS vertical posts were mounted under the concrete
slab, on each side of the web, at the two ends for bracing, as shown in Figure 3.20(c). The
girders were monotonically loaded under stroke control at a rate of 1.75 mm/min, using
3.4.4 Instrumentation
Two LPs were placed at both sides of the girders, at mid-span to measure vertical
deflection and monitor any torsional rotation due to any misalignment. Another two LPs
were also mounted under the tips of the WWF steel reaction beam, below the support
locations to measure any settlement of the supports. The longitudinal strains along the
steel girder and CFRP plates were measured using several 5 mm long electric resistance
strain gauges. Three strain gauges were attached directly to the steel surface at mid-span,
including two strain gauges attached to the web, 50 mm and 150 mm above the tension
flange, while the third one was attached to the underside of the compression flange, as
87
Chapter 3
shown schematically in Figure 3.21(a). Several strain gauges were attached to the CFRP
plates to measure the longitudinal strains, and are spaced as shown schematically in
Figure 3.21(b). Two PI gauges, installed over a gauge length of 100 mm, were also
attached to both the top and bottom sides of the concrete slab, as shown schematically in
Figure 3.21(a). The same data acquisition system used in Phase I was also used to
Composite Beams
This section describes the third phase of the experimental program undertaken to evaluate
composite beams with a simulated section loss in the tension flange. This may be the case
fabrication process, including cutting the tension flange, welding the studs, casting the
concrete slab, installing FRP sheets, instrumentation, and testing of the specimens are
also presented.
configuration. The cross section of the beams consists of W150x22 hot rolled steel
sections acting compositely with a 75 mm thick and 465 mm wide concrete slabs, as
shown in Figure 3.22(a). It should be noted that the size of the concrete slabs in these
specimens was over designed, relative to the size of the W150 x 22 sections. This was
88
Chapter 3
intended to avoid concrete crushing failure and ensure that failure would occur at the
tension side, either by CFRP rupture or debonding. Table 3.6 provides all details of the
beams, including their identification number (B1 to B11), CFRP type, total cross
sectional area of CFRP, the force equivalence index, which is discussed later, and the
FRP sheet dimensions including their width and bonded length. The specimens include
one control intact (undamaged) beam (B1) and ten artificially damaged beams (B2 to
B11), where the steel tension flanges were completely cut at mid-span, throughout the
entire thickness and width, to simulate a severe section loss in bridge girders. Specimens
B1 and B2 were tested without FRP material to serve as control intact and damaged
bonding CFRP sheets of different types and configurations to the artificially damaged
flanges. The force equivalence index (ω) is introduced to quantify the amount of FRP
reinforcement on the basis of a relative axial strength of the flange, as given by:
n
[
∑ A fi F fi ]
i =1
ω= (3.1)
A sf F y
where Ffi and Fy are the strength of FRP layer i, and the yield strength of steel,
respectively. A f i and A sf are the cross sectional areas of FRP layer i and the steel
flange, respectively.
The objective of Phase III is to study the effect of the following parameters:
includes beams B3 and B4, which were repaired by bonding CFRP type C3 sheets
of different cross sectional areas on the bottom side of the tension steel flange, as
89
Chapter 3
well as beams B7 and B8 with CFRP type C2 sheets bonded on both the bottom
2. The effect of number of bonded sides of the steel flange (i.e. the bonded surface
includes beam B6 with CFRP sheets bonded on one side only and beam B7
having a samilar CFRP force equivalence index but the CFRP sheets are bonded
Beams B8 to B11 were repaired using the same CFRP type C2 and the same force
equivalence index, applied to both sides of the tension flange. The total bonded
lengths of B8 to B11 were 1900, 1000, 250, and 150 mm, respectively.
comparing B5 to B6, which had the CFRP installed immediately and 22 months
equivalence index and although two types (C1 and C2) were used, they had
The W150x22 steel sections were first cut to 11 beams, each 2030 mm long. One beam
was left intact to serve as a control specimen (B1). The tension flanges of the remaining
10 beams were completely cut (i.e. through the entire width and thickness) at mid-span,
as shown in Figure 3.22(b). The cut was done using a band saw with a 1.4 mm thick
blade. Four pairs of 76 x 76 x 9.5 mm angles, 130 mm long each, were bolted to the web
90
Chapter 3
at both the loading and supporting points to prevent web buckling. Each pair of angles
was bolted to the web through two 12 mm diameter holes drilled through the web. These
stiffener angles were reused in all test beams. Conventional Nelson shear studs, 41 mm
long and 9.5 mm in diameter, as shown in Figure 3.22(c), were welded to the
compression flange using Nelson Stud Welding’s Series 4500 welder. One electrode of
the welder was grounded to the steel beam and a special welding gun was attached to the
other electrode. The head of the stud was placed into the end of the welding gun and a
small porcelain ferrule was placed on the tip of the stud to contain the weld. The stud was
pushed against the top surface of the flange and the trigger of the welding gun was pulled
to activate the welder, as shown in Figure 3.22(d). The studs were welded in pairs at
The concrete slabs of beams B1, B3, and B4 were cast using the same concrete batch
(batch 3), while the concrete slabs of the remaining nine beams were cast at a different
time using concrete batch 2. After assembling the formwork, a double layer of
the concrete slab, as shown in Figure 3.22(e). Concrete was then poured, vibrated,
troweled, and cured in a similar procedure to that described in Phase II, except that
casting was done in a normal position in this phase (i.e. the slab is cast above the steel
section).
In order to prepare the surface of the beams for FRP sheet installation, beams B7 to B11
were sandblasted along both the underside and top side of the steel tension flange,
91
Chapter 3
whereas beams B3 to B6 were sandblasted along the underside of the steel tension flange
In the cases of test beams accommodating FRP sheets on both sides of the flange, one
side was usually covered with plastic sheets and taped along the edges with masking tape
for protection, as shown in Figure 3.24(a), prior to installing the FRP on the other
surface. The flanges were cleaned with air pressure and wiped with acetone just before
applying the FRP sheets. The dry fabric sheets were cut to the desired dimensions shown
in Table 3.6 and were laid down on a plastic sheet and completely wetted with epoxy
adhesive 1 on both sides, as shown in Figure 3.24(b). The sheets were then carefully
lifted and attached to the surface of the steel tension flange and pressed with a roller to
squeeze out the excess epoxy resin, as shown in Figure 3.24(c). Long flat aluminum
plates and heavy steel blocks were placed on top of the wet sheets to apply some pressure
and ensure a finished flat surface. For beams designed to have additional FRP sheets on
the other side of the flange, the FRP installed on the first side was left for one day to
ensure that the adhesive had set sufficiently, before the beam was flipped upside down to
bond the FRP sheets on the other side of the flange. It should be noted that the entire
installation process of FRP was always completed within the first 8 to 36 hours after
sandblasting in the cases of one side or double sides installation. Only one beam (B5) was
left after sandblasting without applying the FRP sheets for a period of 22 months. The
steel surface of that beam, however, was covered with a thin layer of oil to be protected
against possible corrosion. Before applying the FRP sheets, the flange was cleaned with
92
Chapter 3
acetone. This beam was intended to investigate the effect of delayed application of FRP
All 11 beams were tested in four-point bending, using a simply supported configuration
with a span of 1960 mm between the centerlines of the supports and a 400 mm distance
between the two loads, as shown in Figure 3.25. The beams were monotonically loaded
under stroke control, at a rate of 1.75 mm/min, using the 1000 kN Riehle machine. The
loads were applied using a stiff HSS steel spreader beam. The two point loads were
applied over two transverse rectangular HSS sections, which covered the entire width of
the concrete slab. Steel rollers were placed between the spreader beam and each of the
rectangular HSS sections, as shown in Figure 3.25. A quick setting plaster was placed
under the transverse HSS section to avoid any stress concentrations associated with the
3.5.4 Instrumentation
Figure 3.25 shows the test setup of beams tested in Phase III. Two LPs were placed at
both sides of the beams, at mid-span, to measure vertical deflection and monitor any
torsional rotation due to any misalignment. The longitudinal strains along the tension
flange were measured using several 5 mm long electric resistance strain gauges, spaced
as shown in Figure 3.25(b). Another two strain gauges were attached directly to the steel
surface at mid-span. One strain gauge was attached to the web, right above the cut to
monitor the stress concentration, while the other strain gauge was attached to the
93
Chapter 3
underside of the compression flange. A different arrangement of strain gauges was used
for the damaged control beam (B2), as shown in Figure 3.25(c). This arrangement was
used to monitor the stress concentration in the web at the vicinity of the flange cut. Two
PI gauges, installed over a gauge length of 100 mm, were also attached to the top of both
the concrete slab and steel tension flange for all the beams, as shown schematically in
Figure 3.25(b). The same data acquisition system used in Phases I and II was also used to
94
Chapter 3
Table 3.1 Material properties of the W150x22 steel section used in Phase III.
95
Chapter 3
Ultimate
Elastic Ultimate Strain, εult
FRP Coupon Width, No. of Thickness, Modulus, Strength, x 10-3
type number w (mm) layers t (mm) E (GPa) Fult (MPa) (mm/mm)*
1 19.00 3.11 18.3 381 20.8
2 19.20 3.05 18.1 341 18.8
3 19.40 2 2.98 16.5 N/A N/A
GFRP
4 19.50 2.95 N/A 268 N/A
(G)
5 19.25 2.90 17.4 352 20.2
average 19.27 3.00 17.6 336 20.0
St. Dev. 0.19 0.08 0.8 48 1.0
1 19.43 1.17 111.3 1235 11.1
2 19.20 1.20 102.8 1062 10.3
3 19.34 1 1.19 117.0 1201 10.3
CFRP1
4 19.20 1.19 N/A 1029 N/A
(C1)
5 19.14 1.20 127.9 N/A N/A
average 19.26 1.19 114.8 1132 10.6
St. Dev. 0.12 0.01 10.5 101 0.5
1 25.10 2.21 90.7 1117 12.3
2 24.80 2 2.23 94.4 966 10.2
CFRP2
3 24.90 2.22 85.7 878 10.3
(C2)
average 24.93 2.22 90.3 987 10.9
St. Dev. 0.15 0.01 4.3 121 1.2
1 19.40 1.02 222.3 564 2.5
2 19.40 1.03 212.1 N/A N/A
3 19.43 2 1.15 202.3 509 2.5
CFRP3 4 19.36 1.11 298.5 492 1.7
(C3)
5 19.40 1.09 218.0 473 2.2
average 19.40 1.08 230.6 510 2.2
St. Dev. 0.02 0.05 38.6 39 0.4
1 24.92 1.40 154.2 1823 11.8
2 24.91 1 1.40 145.4 2203 15.2
CFRP4
3 24.85 1.40 157.3 1716 10.9
(C4)
average 24.89 1.40 152.3 1914 12.6
St. Dev. 0.04 0.00 6.2 256 2.2
1 16.51 2.98 325.9 1456 4.5
2 16.41 2 2.95 305.2 1397 4.6
CFRP5
3 16.37 3.09 308.2 1572 5.1
(C5)
average 16.43 3.01 313.1 1475 4.7
St. Dev. 0.07 0.07 11.2 89 0.3
* All listed strain values in this column must be multiplied by 10-3.
96
Chapter 3
Concrete Standard
Phase Batch Average
strength, fc’ (MPa) deviation
39.0
II 1 36.9 38.9 1.9
40.7
50.1
49.1
49.4
50.1
2 49.1 49.9 1.0
49.4
III
51.1
51.6
48.8
51.8
3 49.5 50.2 1.4
49.2
97
Chapter 3
No. of strengthened
Reinforcement ratio
HSS Designation
Length, L (mm.)
ρ = Af / As x 100
No. of layers
No. of layers
orientation
orientation
Columns
Set no.
Group
width
width
kL / r
Type
type
1 3 762 46 --
HSS1 (44x44x3.2)
6 3 1528 93 2 C5 2 25, 16 L 23
7 1 2380 68 --
(89x89x3.2)
8 1 2380 68 2 G 1 75 L C3 1 75 L 9
HSS2
B 9 1 2380 68 2 G 1 75 L C3 3 75 L 25
10 1 2380 68 2 G 1 75 L C3 5 75 L 43
11 1 2380 68 4 G 1 75 L C3 3 75 L 54
12 3 175 4 --
13 3 175 4 4 G 1 cw T C1 1 cw T
HSS2 (89x89x3.2)
14 3 175 4 4 G 1 cw T C1 2 cw T
15 3 175 4 4 G 1 cw T C1 1 cw L
Short
C 16 3 175 4 4 G 1 cw T C1 2 cw L, T
17 3 175 4 4 G 1 cw T C3 1 cw T
18 3 175 4 4 G 1 cw T C3 2 cw T
19 3 175 4 4 G 1 cw T C3 1 cw L
20 3 175 4 4 G 1 cw T C3 2 cw L, T
L = longitudinal,
T = transverse,
L, T = two layers, one longitudinal, followed by one transverse,
cw = complete wrap,
# See Figure 3.11
98
Chapter 3
99
Chapter 3
Table 3.6 Test matrix of repair of artificially damaged composite beams tested in Phase III.
Area Bonded Configurations of FRP layers [ Σ (number of layers x width x length) ]
Type of sides of
Specimen of CFRP ωa the tension GFRP base
I.D. CFRP 2
(mm ) %age flange layer CFRP layers
B1 Intact (control 1)
B2 Damaged (control 2)
1 x 150 x 1900 + 2 x 150 x 1850 + 2 x 150 x 1800 +
B3 C3 648 87 Lower 1 x 150 x 1900
2 x 150 x 1750 + 1 x 150 x 1700
1 x 150 x 1900 + 2 x 150 x 1850 + 2 x 150 x 1800 +
B4 C3 1134 152 Lower 1 x 150 x 1900 2 x 150 x 1750 + 2 x 150 x 1700 + 2 x 150 x 1650 +
2 x 150 x 1600 + 1 x 150 x 1550
B5b C1 668 198 Lower 1 x 150 x 1900 1 x 150 x 1900 + 2 x 150 x 1850 + 2 x 150 x 1800
B6 C2 744 193 Lower 1 x 134 x 1900 1 x 134 x 1900 + 2 x 134 x 1850 + 2 x 134 x 1800
B7 C2 716 185 Lower 1 x 128 x 1900 1 x 128 x 1900 + 1 x 104 x 1850 + 1 x 87 x 1850
Upper 2 x 64 x 1900 2 x 64 x 1900 + 2 x 46 x 1850 + 2 x 35 x 1850
B8 C2 813 210 Lower 1 x 128 x 1900 1 x 128 x 1900 + 1 x 128 x 1850 + 1 x 128 x 1850
Upper 2 x 64 x 1900 2 x 64 x 1900 + 2 x 64 x 1850 + 2 x 46 x 1850
B9 C2 813 210 Lower 1 x 128 x 1000 1 x 128 x 1000 + 1 x 128 x 950 + 1 x 128 x 950
Upper 2 x 64 x 1000 2 x 64 x 1000 + 2 x 64 x 950 + 2 x 46 x 950
B10 C2 813 210 Lower 1 x 128 x 250 1 x 128 x 250 + 1 x 128 x 200 + 1 x 128 x 200
Upper 2 x 64 x 250 2 x 64 x 250 + 2 x 64 x 200 + 2 x 46 x 200
B11 C2 813 210 Lower 1 x 128 x 150 1 x 128 x 150 + 1 x 128 x 145 + 1 x 128 x 145
Upper 2 x 64 x 150 2 x 64 x 150 + 2 x 64 x 145 + 2 x 46 x 145
Af Ff
a
force equivalence index ω = * 100 b
Late application of FRP
A flange F y
100
Chapter 3
t = 8.4
w = 6.1
d = 257
b = 76.2
x x w = 5.8 t = 6.6
b = 31.2
d = 152
t = 3.2
x x
t = 3.2 x x
x x R = 6.4
R = 6.4
B = 44 B = 89 b = 102 b = 152
2 2
A = 495 mm A = 1072 mm A = 3230 mm2 A = 2840 mm2
Ix = 0.13x106 mm4 Ix = 1.31x106 mm4 Ix = 34.2x106 mm4 Ix = 12.0x106 mm4
HSS 44x44x3.2 mm HSS 89x89x3.2 mm W 250x25 W 150x22
Figure 3.1 Different steel cross sections used in the experimental investigation.
Semi-spherical
loading head
Strain
gauges
Flat plate
101
Chapter 3
600
500
Yield strength Fy= 504 MPa
Frs= 49% Fy
400
Stress (MPa)
300
100
600
500
400
Frs= 33% Fy
Stress (MPa)
300
100
102
Chapter 3
Stress (MPa)
12.5
12.5
20
50 100 50
Dims. are in mm.
Strain gauge
103
Chapter 3
550
500
450
400
350
Stress (MPa)
Web
Flange
300
250
200
150
100
50
0
0
0 0.005
5 0.01
10 0.015
15 0.02
20 0.025
25 0.03
30
Extensometer
C3
Typical failure of CFRP sheets (Type C3) Typical failure of CFRP plates (Type C4)
(c) Typical failure of CFRP coupons
104
Chapter 3
2000
1800
1600
CFRP
C4
C5
1400
1200
Stress (MPa)
C1
1000
C2
800
600 C3
400 G GFRP
200
0
0 5 10 15 20 25
Strain x 10-3 (mm/mm)
105
Chapter 3
T
T T C
Longitudinal
Transverse FRP layers
FRP layer
T T
C
Longitudinal T
FRP layer
Figure 3.10 Effect of FRP on local and overall buckling of short and slender HSS
columns.
Slender Short
1 Layer
of C3 1 Layer of G 5 Layers of C3
L T
L
2 Layers of C5 CW
(25x1.4 & 16x1.4)
Set 8 Set 10 75
L
L L CW
50
3 Layers of C3
Set 9 Set 11 Sets 13 to 20
106
Chapter 3
Laser sensor
Coordinates of
the laser point
en
cim
Spe
Distance traveled
Specimen by the laser beam
Moving bed
Before After
sandblasting sandblasting
107
Chapter 3
0.6
a
Side (a)
0.5 b
Out-of-Straightness (mm)
0.4
0.3
Side (b)
0.2
0.1
0
0 400 800 1200 1600
-0.1
Length (mm)
Set 6
C5 s
inst trips b
alla efor
ti e Set 2
othe on on t
r sid he
e
Set 4
50 mm
Specimen 8
Specimen 9
108
Chapter 3
Hinged
end
Specimen
Two
threaded
Plate Cross
rods
attached to head
roller Hinged
the machine
LP cross head end
10 mm deep
sleeve
Specimen Specimen
CFRP
2 plates plates
CFRP
plates
Steel
Two plates
t column
threaded
rods S1 S2
LP
Strain
gauge
(S1 or S2)
Hinged roller
end Hinged
end
10 mm
deep
sleeve LP
r arm
ti leve
Can
109
Chapter 3
PI2 S2
Fixed column Hydraulic LP2 LP6 LP5 LP4 Fixed column
Steel tie jack
2380 mm
110
Chapter 3
Semi spherical
Specimen loading head
LPs
LP 1
LP 3 LP 2
111
Chapter 3
studs 500
65
8.4
6.1
257 W250x25
102
FRP plate
covered with
epoxy
Wooden hopper
(b) Shear studs
Pulling direction
(c) Sandblasting
Figure 3.18 A schematic and fabrication process of girders tested in Phase II.
112
Chapter 3
Figure 3.19 Casting concrete slabs of the girders tested in Phase II.
113
Chapter 3
2 HSS
25x25x3.2
114
Chapter 3
1000
Spreader
beam
Concrete slab
500 mm x 65 mm
PI gauges Stiffener
W250x25 Strain gauges 50 mm wide
GFRP wraps plate
hinge roller
CFRP plates 1500
4000
5940
WWF350x362
6200
LP LP
5940
500 50 mm wide
675 GFRP wraps
support support
1000
1500
1750
1925
Figure 3.21 Schematic of test setup and instrumentations of girders tested in Phase
II.
115
Chapter 3
465
studs
75
6.6
5.8 Dims. are
152 in mm. 1.40 mm
thick cut
W150x22 Porcelain
Shear ferrule
152
stud
(a) Typical cross section of
beams B1 to B11 (b) Flange cut (c) Shear stud and ferrule
116
Chapter 3
Before
sandblasting
After
sandblasting
Figure 3.23 Sandblasting the tension steel flanges of beams tested in Phase III.
117
Chapter 3
FRP sheet
(a) Covering one side of the tension flange (c) FRP sheets installation
Figure 3.24 Installation process of FRP sheets on beams tested in Phase III.
118
Chapter 3
Spreader
400
beam
Concrete slab
75 mm thick.
75 25
225 125
625 325
875
1960
Figure 3.25 Test setup and instrumentations of beams tested in Phase III.
119
Chapter 4
Chapter 4
Experimental Results and Discussion of Phase I:
Axial Compression Members1
4.1 Introduction
This chapter presents the results of Phase I of the experimental program, including
discussion of the behaviour and failure modes. Phase I was focused on strengthening
short and slender HSS steel columns using CFRP sheets and plates. A total of 50 HSS
columns of class 2 square sections were tested under a concentric compression loading.
The slenderness ratios of the columns ranged from 4 to 93. The study was intended to
evaluate the effect of the CFRP strengthening system on the axial load capacity, stiffness,
1
Most of content of this chapter has been published as follows:
Shaat, A. and Fam, A. (2006) “Axial Loading Tests on Short and Long Hollow Structural Steel Columns
Retrofitted using Carbon Fibre Reinforced Polymers.” Canadian Journal of Civil Engineering, 33(4):458-
470.
120
Chapter 4
The tests carried out in Phase I are divided into three groups A, B, and C, as shown in
Table 4.1 to 4.3. Group A includes slender column sets 1 to 6, which are intended to
evaluate the effect of slenderness ratio of the columns on the effectiveness of the CFRP
strengthening system, for a given CFRP reinforcement ratio. Group B includes slender
column sets 7 to 11, which are intended to evaluate the effect of CFRP reinforcement
ratio on the effectiveness of the CFRP strengthening system, for columns of a given
slenderness ratio. Group C includes short column sets 12 to 20, which are intended to
investigate the effect of the CFRP strengthening system on axial strength and stiffness of
short columns. This includes the effect of CFRP fibre orientation (i.e. in both the
longitudinal and transverse directions), number of layers, and the CFRP type (i.e.
standard and high modulus). A summary of the research findings for groups B and C
As shown in Table 4.1, this group of specimens consists of three pairs of sets, namely, (1,
2), (3, 4), and (5, 6) of slenderness ratios of 46, 70, and 93, respectively. The first set of
each pair of each set consists of three similar control specimens (i.e. unstrengthened),
while the second set consists of three similar specimens strengthened using CFRP. A
summary of test results, including the measured geometric imperfections (i.e. the
capacity and the elastic stiffness of the columns is presented in Table 4.1. Typically, the
axial strength of slender columns is inversely proportional to the magnitude of initial out-
of-straightness (Allen and Bulson, 1980). However, the results suggest that this is not the
121
Chapter 4
case in all the specimens. It is believed that several possible factors have resulted in final
imperfection values different from the initial ones. These factors are related to the
installation process of CFRP and include the accuracy of alignment of the CFRP plates,
slight variation in the adhesive thickness of the plates on the two opposite sides, and the
possibility of inevitable minor misalignment in the test setup. Nevertheless, the testing of
three specimens in each set can provide a reliable average of test results, as shown in
Table 4.1. The table also shows the standard deviation of each set, which suggests a good
level of repeatability of test results. Also given in Table 4.1, are the percentage increases
in both the maximum axial load and stiffness, for the CFRP−strengthened column sets 2,
Table 4.1 shows that the percentage increases in axial strength of these sets were 6, 35,
and 71, respectively. It is also shown in Table 4.1 that the axial stiffness has increased by
12, 16, and 17 percent, respectively. All the percentage increases are generally based on
The load versus net axial displacement responses of column sets 1 to 6 are shown in
Figure 4.1 to 4.3, respectively. In each figure, the responses of both the control and
strengthened sets, including the three similar specimens in each set, are plotted. The
figures generally show reasonable repeatable responses for similar specimens. The load
versus lateral displacement responses at mid-height of the columns are shown in Figure
4.4 to 4.6. The low values of lateral displacement up to the peak load are a result of very
small values of out-of-straightness and precision in the test setup alignment. The load
versus axial strain responses are shown in Figure 4.7 to 4.9 for gauge S1 and Figure 4.10
122
Chapter 4
to 4.12 for gauge S2 (see Figure 3.15). The figures show that both sides of the columns
are under compression up to a certain load level, at which excessive buckling occurs. At
this point, the strains on one side (S2) revert to tension. Figure 4.7 to 4.12 also show the
effect of CFRP plates on reducing the axial strain values on both sides of the
level.
Figure 4.13 shows the variation of the axial strength of the columns with slenderness
ratio. The variation is given for both the control and the CFRP-strengthened columns. For
each case, the average strength of the repetitive specimens within one set is reported,
along with the error bars. The figure shows that the axial strength of the control
specimens reduces as the slenderness ratio increases, which is expected according to the
Euler’s equation (Equation 6.16). Also, the axial strength of the CFRP-strengthened
specimens reduced as the slenderness ratio increases but at a much lower rate than the
columns, as will be discussed later. This behaviour clearly suggests that the effectiveness
of the CFRP system increases for higher slenderness ratios. It is also noted that the axial
equivalent to that of a control column of a much lower slenderness ratio (point “b” in
Figure 4.13). This effect becomes more pronounced as the slenderness ratio of the
123
Chapter 4
To summarize the effect of slenderness ratio, Figure 4.14 shows the variation of the
percentage increases in both the axial strength and axial stiffness with the slenderness
ratio. All columns have the same CFRP reinforcement ratio (ρ = 23 %). Figure 4.14
clearly demonstrates the effectiveness of the CFRP system, reflected by the percentage
increase in strength, as the slenderness ratio increases. The figure shows strength
increases of 5.5, 34.9, and 70.7 percent for columns with slenderness ratios of 46, 70, and
93, respectively. On the other hand, the figure shows increases in axial stiffness of 10.3,
15.5, and 17.4 percent, for the same slenderness ratios, respectively. It is clear that the
increased, whereas the increase in axial stiffness seems to be only slightly affected by
slenderness ratio.
In all specimens (i.e. sets 1 to 6), failure was mainly due to excessive overall bucking of
the columns for both the shortest columns with kL/r = 46 (i.e. sets 1 and 2) and the
longest columns with slenderness ratio kL/r = 93 (i.e. sets 5 and 6), as shown in Figure
4.15(a and b). For the CFRP-strengthened columns, two different failure modes were also
associated with overall buckling. The first mode was observed in sets 2 and 4 of
slenderness ratios of 46 and 70, respectively, where the CFRP layers on the inner side of
the buckled column debonded from the steel surface, as shown in Figure 4.15(c). Also,
the GFRP wraps at the ends partially ruptured, as shown in Figure 4.15(d). In set 2,
debonding was associated with a load drop, followed by a load increase and a second
peak, and then descending due to overall buckling, as shown in Figure 4.1 and 4.4. This
124
Chapter 4
suggests that debonding at this slenderness ratio occurred prematurely at one side, before
reaching the peak load, which is associated with overall buckling. In set 4, debonding of
CFRP on one side occurred almost simultaneously when the peak load was reached and
overall buckling occurred. This is evident by a clear load drop near the peak with
insignificant rising of the curves after the drop, as shown in Figure 4.2 and 4.5. The
second mode of failure was observed in set 6 with slenderness ratio of 93, where the
CFRP layers on the inner side have crushed at mid-height, as shown in Figure 4.15(e).
Crushing of CFRP in set 6 occurred when the load was already descending, long after
reaching the peak load and overall buckling has occurred, as shown in Figure 4.3 and 4.6.
By carefully examining the strains measured on the inner sides by gauge S1 at failure
(Figure 4.7 to 4.9), an average strain values of 0.161 and 0.226 percent can be observed
in sets 2 and 4, respectively, when debonding occurred. On the other hand, the strain at
which the CFRP crushed was 0.274 percent. It should be noted that this compressive
strain is only 58 percent of the tensile rupture strain given in Table 3.2. This compressive
failure strain is limited to this particular CFRP, and may vary for different types of CFRP
or HSS sections with different width-to-thickness ratios. Figure 4.16 shows the variation
of the compressive strains of CFRP at ultimate with slenderness ratio. The figure shows
that higher strains (and hence higher effectiveness) are developed as slenderness ratio
increases. No signs of CFRP failure by rupture or debonding have been observed on the
125
Chapter 4
strengthening system of different reinforcement ratios (i.e. different number of layers) for
columns of the same slenderness ratio. Due to the relatively large size of these specimens
(2380 mm long) tests were conducted horizontally using a specially designed setup as as
steel specimens could not be measured as the specimens did not fit in the laser sensor
apparatus. Instead, they were estimated from strain measurements as will be discussed
later. The load versus net axial displacement and lateral displacement at mid-height of
column sets 7 to 11 are shown in Figure 4.17 and 4.18, respectively. The lateral
displacements in Figure 4.18 are offset by the values of the estimated out-of-straightness.
The figures show that the gain in axial strength of the CFRP-strengthened specimens
Figure 4.19 to 4.23 show the load versus axial strain at the two opposite sides of all the
five specimens, based on both electrical resistance strain gauges and the 200 mm PI
gauges. The strain values indicate that global buckling started at strain values very close
CFRP-strengthening
Table 4.2 shows that the gains in axial strength among the different specimens do not
correlate with the number of CFRP layers used. For example, specimen 10 (with 5 layers)
126
Chapter 4
shows only 13 percent increase in axial strength, compared to specimen 8 (with 1 layer),
which showed a 20 percent increase. This discrepancy is attributed to the variation of the
Figure 4.19 to 4.23, at low load levels before buckling, it becomes clear that both sides of
the column are not equally strained, right from the beginning. Furthermore, the difference
in strain values on both sides varies from one specimen to the other. This suggests that
the columns varied in their geometric out-of-straightness values, which are essentially
eccentricities relative to straight lines connecting both ends. The strain gradient through
the cross section of the column has been used to estimate this imperfection, in terms of
the initial eccentricity (e’), which is assumed to represent a maximum amplitude at mid-
height (i.e. at the location of strain gauges), as reported in Table 4.2. The strain gradient
is established using the following load-strain relationship on both sides of the loaded
columns:
1 ⎛ − P P (e'+δ ) y ⎞
ε c or t = ⎜ m ⎟ (4.1)
Es ⎜ A It ⎟
⎝ t ⎠
where, εc or t are the strains at the high compression side or the low compression (or
tension) side of the column, P is the applied load, e’ is the imperfection at mid-height, δ
is the lateral deflection at mid-height due to the applied load, and y is the distance
between the extreme CFRP surface and the centroid of the cross section, and At and It are
the transformed cross sectional area and moment of inertia, respectively, and are given
by:
n ⎡Ef ⎤
At = A s + ∑ ⎢ i A f ⎥ (4.2)
i =1 ⎢ E ⎥⎦
i
⎣ s
127
Chapter 4
n ⎡Ef ⎤
It = Is + ∑⎢ i I fi
⎥ (4.3)
i =1 ⎢ E
⎣ s ⎥⎦
where Is and I fi are the moments of inertia of the steel section and CFRP layer i,
strains (εc or t) and the corresponding lateral deflection in Equation (4.1), within the elastic
range. It is noted that the calculated e’ differs when calculated using εc or εt and also
varies with the applied load P. This is probably attributed to the assumptions made,
including those used in the calculations of section parameters involved in Equations 4.1
to 4.3. A relation between the imperfection e’, based on the average value obtained from
the strains on both sides, and the applied load P is established. Figure 4.24 shows the
variation of the average e’ with the applied load for the column sets 1 and 2, as a sample.
The initial imperfection is then estimated as the y-intercept (i.e. the value at P = 0). The
average imperfection plots versus the applied loads for the other column set 3 to 11 are
given in Appendix B. It is believed that this method is more reliable than measuring the
out-of-straightness of the columns before applying the CFRP material as the method
accounts for all possible sources of imperfection combined in the final situation,
including the initial out-of-straightness, the variability of adhesive thickness in the FRP
system, the possibility of minor misalignment of the fibres in CFRP sheets, unintended
misalignment within the test setup, or a combination of all. The method, however,
128
Chapter 4
Figure 4.18 shows the load versus total lateral deflection responses of specimens 7 to 11,
including the estimated imperfection values (e’) at mid-height of the columns. The curves
are offset with the value of their respective imperfections (e’). The figure indicates that
for specimens of comparable imperfections (i.e. specimens 7 and 9), the behaviour of
control specimen, which indicates that the CFRP has improved the stability of the column
against lateral deflections. Specimen 8 showed a higher peak load than specimen 9,
despite the lower number of CFRP layers, due to its smaller imperfection. Typically, the
larger the imperfection, the lower the peak load for a given slenderness ratio. It is
hypothesised that the effectiveness of CFRP would increase for columns with larger
4.25(a and b). Unlike the specimens in sets 1 to 6, the overall buckling in this case was
followed by a secondary local buckling failure at the inner side, at or near the mid-height
of the specimen, as shown in the close up picture of Figure 4.25(a). The local buckling
was attributed to the relatively thin walls of the columns in this case (b/t = 23.8), and was
in the form of an inward buckling of the compression flange and outward buckling of the
two side webs. This was clearly revealed after the test by cutting the specimen, as shown
in Figure 4.25(c). For the CFRP-strengthened specimens, the secondary local buckling of
the compression flange was associated with a combined local crushing and debonding
between the FRP sheets and the steel surface, as shown in Figure 4.25(b). By carefully
129
Chapter 4
examining the longitudinal strains on the failed compression face, in Figure 4.20 to 4.23,
it can be seen that strains at failure were very close (0.143, 0.125, 0.148, and 0.118
percent for specimens 8, 9, 10, and 11, respectively). Therefore, an average strain of
0.133 percent can be defined as the strain at which the CFRP sheets failed in
compression. No signs of FRP failure have been observed on the tension side. For
specimen 11, strengthened on four faces, the CFRP on the side faces have also fractured
modulus (SM-) and high modulus (HM-) CFRP sheets in short HSS columns. The effects
of fibre orientation and number of CFRP layers are also examined. The load−axial
displacement curves for all short column specimens, sets 12 to 20, are shown in
Figure 4.26. In each graph, three curves are presented for three similar specimens. A
summary of test results, including the ultimate load capacity, axial displacement at
maximum load and the elastic stiffness, is presented in Table 4.3. Also given in Table
4.3, are the percentage increases in the axial load and stiffness as well as the percentage
In order to understand the behaviour of HSS short columns and to assess the contribution
that CFRP sheets might provide, the load-axial strain behaviour of the control specimen
130
Chapter 4
12-1 is examined. Figure 4.27 shows the load-axial strain behaviour on two adjacent
sides of this specimen. Both gauges S1 and S2 showed an increase in the compressive
strains, up to point “a”, where substantial local buckling took place. At this point, one
side buckled outwards, as indicated by the strains measured by S2, which reverse
direction sharply, as a result of the reduction of compressive strain, and the other side
compression. It is, therefore, believed that the CFRP could brace the two opposite sides
that buckle outwards. It is also noted that CFRP would unlikely have any significant
contribution to the sides that buckle inwards. In fact, it debonds from the steel surface as
The effect of CFRP wrapped sheets on the behaviour of HSS short columns is presented
in terms of the load-axial displacement responses, as shown in Figure 4.28 and 4.29 for
the SM-CFRP (types C1) and HM-CFRP (type C3), respectively. The curves in Figure
4.28 and 4.29 are each based on the average of the responses of the three similar
specimens in each set. The maximum load values of the three similar specimens, of each
set, show good repeatability, as listed in Table 4.3. However, the repeatability of the
elastic stiffness was not as good. Figure 4.28 and 4.29 show that the axial strength and
stiffness of short HSS columns is increased with different degrees, depending on the
number of layers, fibre orientation, and type of CFRP, as will be discussed in the
following sections. Figure 4.28 and 4.29 also show that, while the CFRP system increases
both the strength and stiffness, it does not have much effect on the post-peak softening
rate.
131
Chapter 4
Test results in Figure 4.28 and 4.29 as well as Table 4.3, based on the average values;
suggest that CFRP layers with fibres oriented in the transverse direction are more
efficient than those with fibres oriented in the longitudinal direction. This is evident from
comparing column sets 13 and 15 for the SM-CFRP (types C1) and column sets 17 and
19 for the HM-CFRP (type C3). All column sets had one CFRP layer. Also, results of
column sets 14 and 16 as well as 18 and 20 suggest that two transverse CFRP layers are
more efficient in increasing the strength than one longitudinal and one transverse layer,
for both types of CFRP. Given the level of variability among repeated tests, it is difficult
The effect of CFRP type is examined by comparing specimens of the same number of
layers in Figure 4.30. Test results in Table 4.3 and Figure 4.30 indicate that the SM-
CFRP (sets 13 and 14) resulted in better strengthening than the HM-CFRP (sets 17 and
18). This is attributed to the fact that each SM-CFRP lamina has 10 percent higher
stiffness (Ef Af) than the HM-CFRP. It was also noted that because of the very stiff
characteristics of the HM-CFRP (see properties of C3 in Table 3.2), fibres have fractured
at the round corners near the ultimate loads, which have reduced their efficiency. The
results also indicate that set 14 with two transverse layers of SM-CFRP achieved the
highest gain in strength (18 percent), among the short columns. Figure 4.28 and 4.29
show that adding a second layer enhances the strength, particularly if both layers are in
132
Chapter 4
the transverse direction as evident by comparing sets 13 and 14 for the SM-CFRP and
The typical failure mode of all short column specimens was essentially yielding, followed
by symmetric local buckling, where two opposite faces would buckle inwards and the
other two faces would buckle outwards, as shown in Figure 4.31(a). In all specimens
strengthened with CFRP layers oriented in the longitudinal direction, debonding occurred
between CFRP and steel at one end, as shown in Figure 4.31(b), even in the specimens
with additional transverse CFRP outer layer. In specimens with all CFRP layers oriented
in the transverse direction, debonding occurred between CFRP and steel, only on the two
opposite faces that experienced inward local buckling. This was revealed after the test
was completed by cutting the specimen as shown in Figure 4.31(c). This failure mode, in
Figure 4.31(c), supports the hypothesis described in Figure 3.10(a). For specimens with
HM-CFRP layers oriented in the transverse direction, rupture of the fibres was observed
near the corners, as shown in Figure 4.31(d). This is likely attributed to the very stiff
nature of this high modulus CFRP. In general, none of the short column specimens failed
133
Chapter 4
134
Chapter 4
135
Chapter 4
Displacement
@ Pmax (mm)
@ Pmax (mm)
Specimen
Specimen
Maximum
Maximum
Stiffness
Stiffness
(kN/mm)
(kN/mm)
Set
Set
12-1 385 1272 0.62 17-1 455.5 1271 0.62
12 12-2 411 1161 0.59 17 17-2 408 1335 0.51
12-3 393 954 0.66 17-3 420 1027 0.62
Average 396 1129 0.62 Average 428 1211 0.59
St. dev. 13 161 0.04 St. dev. 25 163 0.06
% gain 8 7 -5
13-1 453 1405 0.47 18-1 474 971 0.67
13 13-2 454 1161 0.57 18 18-2 412 1271 0.57
13-3 458 954 0.55 18-3 434 1376 0.59
Average 455 1173 0.53 Average 440 1206 0.61
St. dev. 3 226 0.05 St. dev. 31 210 0.05
% gain 15 4 -15 % gain 11 7 -2
14-1 511 1571 0.65 19-1 441 1335 0.54
14 14-2 444 1214 0.6 19 19-2 438 1335 0.49
14-3 447 1068 0.63 19-3 421 971 0.58
Average 467 1284 0.63 Average 433 1214 0.53
St. dev. 38 259 0.03 St. dev. 11 210 0.05
% gain 18 14 2 % gain 9 8 -15
15-1 444 1571 0.49 20-1 444 1214 0.57
15 15-2 420 1068 0.59 20 20-2 440 1405 0.54
15-3 429 1214 0.5 20-3 440 1469 0.51
Average 431 1284 0.52 Average 441 1363 0.54
St. dev. 12 259 0.06 St. dev. 2 133 0.03
% gain 9 14 -16 % gain 11 21 -13
16-1 453 1907 0.47
16 16-2 454 1161 0.57
16-3 458 1271 0.56
Average 455 1446 0.54
St. dev. 3 403 0.06
% gain 15 28 -13
136
Chapter 4
220
Set 2
P
2-2
(Pavg. )strengthened= 192 kN
2-13
200 (strengthened)
2-
(Pavg. )control= 182 kN
180 Debonding of ∆
CFRP on
inner side
160
140 1-2
1-3
Load (kN)
Set 1
120 (Control)
100
80
60
40
HSS 44 x 44 x 3.2 mm
20 kL/r = 46
0
0 3
1.5 6
3.0 9
4.5 12
6.0 15
7.5
Axial displacement ∆ (mm)
220
(Pavg. )strengthened= 200 kN Debonding
200 of CFRP on P
inner side
180 ∆
160 (Pavg. )control= 148 kN
140 4-3
Set 4
4-2
Load (kN)
4-1 (strengthened)
120
2
3- -3
100 3 -1
3
Set 3
80 (Control)
60
40
HSS 44 x 44 x 3.2 mm
20 kL/r = 70
0
0 1.5
3 3.0
6 4.5
9 6.0
12 7.5
15
Axial displacement ∆ (mm)
137
Chapter 4
220
200 P
Overall Buckling
180 (Pavg. )strengthened= 175 kN ∆
6-3
120 5-3 6-1 (strengthened)
(Pavg. )control= 5-2
100 103 kN 5-1
80
60 Set 5
(Control)
40
HSS 44 x 44 x 3.2 mm
20
kL/r = 93
0
0 1.5
3 3.0
6 4.5
9 6.0
12 7.5
15
Axial displacement ∆ (mm)
(Control)
120
100
80
60
40
HSS 44 x 44 x 3.2 mm
20 kL/r = 46
0
-5 0 5 10 15 20 25 30 35 40 45 50
Lateral displacement δ (mm)
138
Chapter 4
220
CFRP debonding on inner side
P
200 4-3 Set 4
4-1
4-2 (strengthened)
180
3-3
160 3-1 Set 3 e’ δ
3-2 (Control)
140
Load (kN)
120
100
80
60
40
HSS 44 x 44 x 3.2 mm
20 kL/r = 70
0
-5 0 5 10 15 20 25 30 35 40 45 50
Lateral displacement δ (mm)
220
6-3
P
200 6-2 Set 6
6-1 (strengthened)
180
120
100
5-3
5-2 Set 5
80 5-1 (Control)
60
40
HSS 44 x 44 x 3.2 mm
20 kL/r = 93
0
-5 0 5 10 15 20 25 30 35 40 45 50
Lateral displacement δ (mm)
139
Chapter 4
220
Set 2 2-1
2-3
200 Set 1 1-3 (strengthened)
2-2
(Control) 1-2
180
160
140
Load (kN)
120
(εavg.)strengthened = -1.61x10-3
100
80
60 S1
40
HSS 44 x 44 x 3.2 mm
20 kL/r = 46
0
-16 -14 -12 -10 -8 -6 -4 -2 0
-3
Axial strain x 10 (mm/mm)
Figure 4.7 Load-axial strain responses based on strain gauge S1 of column sets 1
and 2 of group A.
220
200 4-3
Set 4 4-1
(strengthened) 4-2
180
S1
160 Set 3
(Control)
3-1
140 3-3
Load (kN)
3-2
120
(εavg.)strengthened = -2.26x10-3
100
80
60
40
HSS 44 x 44 x 3.2 mm
20 kL/r = 70
0
-16 -14 -12 -10 -8 -6 -4 -2 0
-3
Axial strain x 10 (mm/mm)
Figure 4.8 Load-axial strain responses based on strain gauge S1 of column sets 3
and 4 of group A.
140
Chapter 4
220
200 6-3
Set 6 6-2
180 (strengthened) 6-1
160 S1
140
Load (kN)
5-3
120 Set 5 5-2
(Control) 5-1
100
(εavg.)strengthened = -2.74x10-3
80
60
40
HSS 44 x 44 x 3.2 mm
20 kL/r = 93
0
-16 -14 -12 -10 -8 -6 -4 -2 0
-3
Axial strain x 10 (mm/mm)
Figure 4.9 Load-axial strain responses based on strain gauge S1 of column sets 5
and 6 of group A.
220
2-1 2-3 Set 2
200 2-2 (strengthened)
180
S2
160
140
Load (kN)
120
1-3
100 1-2
80 Set 1
(Control)
60
40
HSS 44 x 44 x 3.2 mm
20 kL/r = 46
0
-3 -1.5 0 1.5 3 4.5 6 7.5 9
-3
Axial strain x 10 (mm/mm)
Figure 4.10 Load-axial strain responses based on strain gauge S2 of column sets 1
and 2 of group A.
141
Chapter 4
220
200
4-1
180 4-3 Set 4
4-2 (strengthened)
160 S2
140
Load (kN)
120
3-1
100 3-3
Set 3 3-2
80 (Control)
60
40
HSS 44 x 44 x 3.2 mm
20 kL/r = 70
0
-3 -1.5 0 1.5 3 4.5 6 7.5 9
-3
Axial strain x 10 (mm/mm)
Figure 4.11 Load-axial strain responses based on strain gauge S2 of column sets 3
and 4 of group A.
220
200
180 6-1
Set 6
6-3
160 6-2
(strengthened)
S2
140
Load (kN)
120
100 5-3
Set 5 5-2
(Control) 5-1
80
60
40
HSS 44 x 44 x 3.2 mm
20 kL/r = 93
0
-3 -1.5 0 1.5 3 4.5 6 7.5 9
-3
Axial strain x 10 (mm/mm)
Figure 4.12 Load-axial strain responses based on strain gauge S2 of column sets 5
and 6 of group A.
142
Chapter 4
250
CFRP-strengthened
200
a
b
Axial strength (kN.)
Effect
150 of
Bare steel CFRP
50 HSS 44x44x3.2 mm
ρ = 23 %
0
40 50 60 70 80 90 100
Slenderness ratio (kL/r)
Figure 4.13 Variation of axial strength with slenderness ratio of group A columns.
80
70.7 %
%age increase in axial strength (or stiffness)
60 th
e ng
s tr
HSS 44x44x3.2 ial
50 ρ = 23 % Ax
40
34.9 %
30
20 17.4 %
15.5 % Axial stiffness
10.3 %
10
5.5 %
0
40 50 60 70 80 90 100
Slenderness ratio (kL/r)
143
Chapter 4
Debonded
CFRP plates
crack
144
Chapter 4
4.5
3.5
2.5
2
Crushing
1.5 Debonding
Debonding
1
0.5
0
0 20 40 60 80 100 120
Slenderness ratio (kL/r)
400
Set 9
7 Control 11
350 8 1 layer/ 2 sides
9 3 layers / 2 sides
10 5 layers / 2 sides
300 11 3 layers / 4 sides
7
250 10
Load (kN)
200 8
P
150 ∆
100
50
0
0 1 2 3 4 5 6 7
Axial displacement ∆ (mm)
145
Chapter 4
400
400
HSS 89 x 89 x 3.2 mm
8 11
350 kL/r = 68
350
300 10
300
9
7
250
250
Load (kN)
P
200
200
150
150
e’ δ
100 Set
100 7 Control
8 1 layer/ 2 sides
50 9 3 layers / 2 sides
10 5 layers / 2 sides
11 3 layers / 4 sides
00
-5
-5 00 55 10
10 15
15 20
20 25
25 30
30 35
35 40
40 45
45
Total lateral deflection (e’+δ) (mm)
400
350
300 PI2
S1
250 PI1 S2
Load (kN)
200
εy = 1.94 x 10-3
150
PI1 PI2
100
S1 S2
50
0
-0.01
-10 -0.008
-8 -0.006
-6 -0.004
-4 -0.002
-2 00 0.002
2 0.004
4
146
Chapter 4
400
400
350
350
300
S1
300
PI1 PI2
S2 250
250
200
200
150
150
50 S1 S2 50
0 0
-0.01 0
-10 -8 -6 -4 -2 0 2 4
Axial strain x 10-3 (mm/mm)
400 400
350
350 S1
PI1
300
PI2
300
S2
250
250
Load (kN)
-3
(εcrushing )CFRP= 1.25 x 10
200
200
150
150
PI1 PI2
100 100
S1 S2 50
50
0 0
-10
-0.01
-8 -6 -4 -2 00 2 4
147
Chapter 4
400 400
350
350
PI1 PI2
300
300
S1 S2
250 250
Load (kN)
150 150
50 S1 S2 50
0 0
-0.01 0
-10 -8 -6 -4 -2 0 2
Axial strain x 10-3 (mm/mm)
Figure 4.22 Load-axial strain responses of specimen 10 of group B.
400 400
350
350 PI1
S1 PI2
300
300 S2
250 250
Load (kN)
200 200
150 150
50 S1 S2 50
0 0
-0.01 0
-10 -8 -6 -4 -2 0 2 4
148
Chapter 4
1.8
Note: imperfection measurements of specimen 1-1 are not included because the bracing guides were
overclamped, which caused partial restraints at mid-height. This has resulted in development of additional
Mid-height Imperfection, e’ (mm) 1.6 bending moments on the specimen
1.4
1.2
1
Specimen 1-2, eP=0= 0.79
0.8
Specimen 2-3, eP=0= 0.78
0.6
Specimen 2-1, eP=0= 0.47
0.4 Specimen 1-3, eP=0= 0.47
0
0 10 20 30 40 50
Applied load, P (kN)
149
Chapter 4
Local
buckling
CFRP
sheets on
2 sides
Debonding and
fracture of
sheets
CFRP
sheets on
4 sides
150
Chapter 4
600
600 600
600 600600
P
∆ 500
450
450 450
450 450
400
300
300 300
300 300300
200
150
150
150
150 150
100
0 0 0.5
0.5 1.0
1 1.5
1.5 0
0 0.5
0.5
1
1.0
1.5
1.5 0
0 0.5
0.5 1.0
1 1.5
1.5
600 600
600600 600600
450 450
450450 450450
Load (kN)
300 300
300300 300300
150 150
150150 150150
0
0 0.5 1 1. 5
600
600 600
600 600
6 00
450
450 450
450 450
450
300
300 300
300 300
3 00
150
150 150
150 150
150
0 0
0.5
0.5
1.0 1
1.5
1.5
0 0.5 1.0 1.5 0
0 0.5
0.5 1.0
1 1 .5
1.5
0 0.5 1 1.5
Displacement (mm)
151
Chapter 4
450
a Specimen 12-1 (S1)
400 Specimen 12-1 (S2)
Local
buckling
350
300
Load (kN)
250
200
150 S2 S1
100
50
0
0 1 2 3 4 5 6 7
Axial strain x 10-3 (mm/mm)
500
14
400 15
16 13
300
Load (kN)
12 (control)
200
100
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
152
Chapter 4
500
18
17
400 19
12 (control) 20
300
Load (kN)
200
100
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
500
14 (
SM-
C F RP
400 )
18 (H
M-CF
RP)
1 2 (c o
300 ntrol)
Load (kN)
200
100
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Axial displacement (mm)
Figure 4.30 Effect of using two types of CFRP on strengthening short specimens.
153
Chapter 4
Debonding of CFRP
Local
buckling
Rupture
of CFRP
(d) Set 18
Local Debonding of CFRP
buckling
(c) Set 14
154
Chapter 5
Chapter 5
Experimental Results and Discussion of Phases II
5.1 Introduction
This chapter presents the results of Phases II and III of the experimental program,
including discussion of the behaviour and failure modes. Phase II was focused on
strengthening intact steel-concrete composite girders. Three girders, scaled down (4:1)
from an actual bridge, were tested in four-point bending. Phase III was focused on repair
tension flange at the mid-span section. Eleven beams were tested in four-point bending in
Phase III. The flexural behaviour in both phases is evaluated in terms of the flexural
stiffness and strength. The force transfer between steel and CFRP material is evaluated in
Three large scale steel-concrete composite girders (G1 to G3) were tested in this phase.
Each girder consisted of 5940 mm long W250x25 hot rolled steel section acting
155
Chapter 5
compositely with a 65 mm thick and 500 mm wide concrete slab. The girders included
specimens (G2 and G3). The strengthening scheme of specimens G2 and G3 consisted of
one 90 mm x 1.4 mm layer of CFRP type C4 and a second layer of 50 mm x 1.4 mm,
which was CFRP type C4 for G2 and CFRP type C5 for G3. The lengths of layers 1 and 2
were 4000 mm and 1500 mm, respectively, in both G2 and G3. The objectives of this
phase were to investigate the effectiveness of CFRP bonded plates in strengthening intact
composite girders, examine the effectiveness of the bond between the steel surface and
CFRP plates with lengths shorter than the full span of the girder, and also compare CFRP
A summary of test results, including the flexural stiffness, yield load, and maximum load
of these girders, and their percentage increases relative to the control specimen, is
presented in Table 5.1. The flexural stiffness is calculated based on the slope of the load-
deflection curve within the linear elastic part. The results show that CFRP has indeed
increased both the flexural strength and stiffness. The flexural strength has increased by
50 and 51 percent for girders G2 and G3, respectively, relative to G1, whereas, the
Figure 5.1 shows the load versus mid-span deflection of the three girders of Phase II. The
figure shows a yield load of 85 kN for the control specimen, which is essentially the load
at the end of the linear part of the load-deflection curve. Figure 5.2 shows the load versus
156
Chapter 5
strain measurements of two strain gauges attached to the underside of the lower steel
flange of the control specimen. The figure shows a slight difference in the two strain
gauge measurements. The yield strain of the steel girder, based on coupon tests (Figure
3.6), is about 0.17 percent. Figure 5.2 shows that at this strain the behaviour indeed
changes, suggesting that yielding has started. This occurs at the 85 kN load, which agrees
with the estimated yielding load at the end of the linear part in Figure 5.1. The small
the location within the flange, where the level of residual stress varies. It is also possible
that a slight relative slip between the concrete slab and the steel beam has contributed to
the rather unusual behaviour just before the yielding flat plateau in Figure 5.2. The yield
loads of specimens G2 and G3 have also been obtained from Figure 5.1 and are 97 kN
and 103 kN, respectively. Table 5.1 shows that CFRP plates have increased the yield load
by 14 and 21 percent in the cases of G2 and G3, respectively, which satisfies the design
particularly important from the design point of view as it clearly reflects some increase in
The outer CFRP layers (1500 mm long) in both G2 and G3 girders were debonded early
from the inner CFRP layer (4000 mm long), at loads of 129 kN and 111 kN, respectively,
because of their short length. It is also noted that debonding in specimen G3 with two
different types of CFRP occurred earlier than in G2 of the same type of CFRP for the two
layers. The load drop after debonding of the second layer was very small due to the small
cross sectional area of the second CFRP layer, relative to the total area of CFRP. After
157
Chapter 5
debonding has occurred, both girders showed a similar trend, where they both behaved as
a girder strengthened with one layer of CFRP type C4, 4000 mm long. Both girders
achieved almost the same strength (216 kN and 217 kN for G2 and G3, respectively), as
shown in Figure 5.1. This level of strength represents 51 percent increase in ultimate
strength, as given in Table 5.1. The similarity in behaviour of both G2 and G3 after
debonding of the outer layer also reflects good repeatability. It is clear that the effect of
the CFRP type on the ultimate flexural strength could not be assessed in this phase of
Figure 5.3 shows the load versus the longitudinal strains in the web, measured at a height
of 50 mm above the lower flange, for the three girders. The figure clearly shows that
strains in the steel cross section have stabilized and remained almost constant at about
0.24 percent up to failure, due to bonding CFRP plates to the lower flanges of G2 and G3.
The second (outer) layer of CFRP was type C4 in the case of G2 and type C5 in the case
of G3. Although the value of elastic modulus of CFRP type C5 is almost double that of
type C4 (Ec4 = 152 GPa and Ec5 = 313 GPa) the difference in the calculated transformed
moment of inertia of girders G2 and G3 is in fact quite small, as shown in Table 5.1. The
calculations accounted for the concrete slab, steel section, and the CFRP. The table shows
respectively, relative to G1. The observed increases in flexural stiffness, within the elastic
158
Chapter 5
region, for girders G2 and G3 (17 and 19 percent, respectively) are almost proportional to
Figure 5.4 and 5.5 show the load versus strain at different locations along the CFRP
plates of both strengthened girders G2 and G3, respectively. The strain gauges and curves
are marked by their distance from mid-span of the girder. It should be noted that in the
case of G2, the strain gauge at zero distance was attached to the inner CFRP layer, which
remained bonded to the steel substrate until failure, whereas in G3 it was attached to the
outer CFRP layer, which debonded early. Figure 5.4 shows a linear relationship to the
end, at the location where the steel cross sections have not yielded (a cross section at
1925 mm from the mid-span). On the other hand, bi-linear relationships are observed at
all other locations, where the steel cross sections were at various stages of yielding. The
change of the strain behaviour to the reverse direction of the gauge at 675 mm (Figure 5.4
and 5.5), which is bonded at the end of the outer CFRP layer, indicates the initiation of
debonding.
The difference in tensile forces between two locations on the CFRP plate must be
balanced by the shear force acting between the CFRP plate and steel substrate, as noted
by Garden et al. (1998). The average shear stress could then be determined between two
⎡ε 2 − ε1 ⎤
τ avg = E f t p ⎢ ⎥ (5.1)
⎣ x 2 − x1 ⎦
159
Chapter 5
(x 2 − x1 ) is the distance between the gauges, Ef and tp are Young’s modulus and
Figure 5.6 and 5.7 show the load versus the average shear stresses, based on Equation
5.1, at discrete locations representing mid-distances between the strain gauges along the
CFRP plates of girders G2 and G3, respectively. Insignificant shear stresses are observed
at a distance of 250 mm, since it is within the constant moment region. On the other hand,
high shear stresses have developed near the end of the outer CFRP plate, at a distance
588 mm from mid-span, which indicate the initiation of debonding. The shear stresses at
this location were linear up to load levels of 96 kN and 83 kN for G2 and G3,
respectively. Beyond these load levels, the adhesive yielded and a nonlinear behaviour
was observed. It can be noticed that debonding of the outer CFRP layer of both G2 and
G3, which is only 1500 mm long, occurred at load levels of 129 kN and 111 kN,
respectively, after the interfacial shear stresses reached a maximum average value of 2.25
MPa. It is clear that the CFRP type C5, used in G3, debonded at a lower load level than
type C4, used in G2. As indicated earlier, this is probably attributed to the large
difference in the elastic modulus values between the CFRP types C4 and C5.
The inner layer of CFRP of specimens G2 and G3 was 4000 mm long, followed by a
second (outer) layer, 1500 mm long. The strain gauges attached to the short layers (at
zero, 500, and 675 mm) indicate that the bonded length of the second layer was
160
Chapter 5
insufficient to fully utilize the CFRP plates. The maximum recorded strains in the outer
layer at mid-span for CFRP type C4 (in G2) and CFRP type C5 (in G3) just prior to
debonding are 2.8 and 1.75 percent, respectively, as shown in Figure 5.4 and 5.5. These
values represent only 22 and 37 percent, respectively, of the ultimate strains of CFRP
The inner layer for both G2 and G3 specimens, on the other hand, had a sufficient bonded
length, 4000 mm long, to keep the layer attached to the steel substrate until failure of the
girders. Figure 5.4 shows that the mid-span strain of the inner CFRP layer of girder G2
reached 13 percent, which is slightly higher than the average ultimate rupture strain (12.6
percent), based on the coupon tests (Table 3.2 and Figure 3.8). However, failure of the
girder occurred due to crushing of the concrete slab, which was associated with lateral
The unstrengthened control girder G1 exhibited large deflection at mid-span (142 mm),
as shown in Figure 5.9(a). This large deflection was associated with excessive yielding of
the steel cross section. Based on the strain measurements of gauges PI2 and S1, and
assuming a linear strain distribution, it is estimated that 90 percent of the depth of the
steel cross section (d) was yielded before the girder failed. Failure occurred due to
crushing of the concrete slab, as shown in Figure 5.9(b), when its compressive strains,
161
Chapter 5
In both girders G2 and G3, the outer CFRP layer was prematurely debonded first. The
debonding was associated with fracture of the GFRP transverse end wraps, as shown in
Figure 5.9(c). This premature debonding occurred at 60 and 51 percent of the maximum
loads of G2 and G3, respectively. The debonding was also associated with insignificant
drops of 3 kN and 4 kN in the load in girders G2 and G3, respectively. Both girders
continued to sustain an increasing load up to failure. The two girders became vulnerable
to lateral torsional buckling at high load levels, particularly because of the unsupported
and relatively long span, as shown in Figure 5.9(d). These girders were braced against
torsional rotation at the supports only. The lateral buckling of the compression flange
produced additional lateral bending stresses on the concrete slab. For this reason, the
concrete slabs in G2 and G3 girders crushed on one side only in an unsymmetric manner,
as shown in Figure 5.9(d and e). Also, this transverse gradient of compressive stresses in
the concrete flange explains the relatively low strains measured by gauge PI1 at failure,
which were lower than the typical 0.35 percent (Figure 5.8). This is because PI1 was
positioned at the mid-width of the concrete slab, while maximum strains were at one
edge. It is very important to note, however, that this failure should not be considered
premature from an ultimate load point of view; because the tensile strains measured in the
CFRP at ultimate (Figure 5.4) suggests that tension failure of the CFRP was indeed quite
imminent.
A total of 11 steel-concrete composite beams (B1 to B11) were tested in this phase. Each
beam consisted of 1960 mm long W150x22 hot rolled steel section acting compositely
162
Chapter 5
with a 75 mm thick and 465 mm wide concrete slab. Table 5.2 shows a summary of the
test results of Phase III, in terms of the flexural stiffness and strength of the control intact,
control damaged, and CFRP-repaired specimens. Also given in Table 5.2, are the
percentage differences of both the stiffness and strength for the control damaged
specimen B2 and repaired specimens B3 to B11 with respect to the control intact
specimen B1. The following sections present test results, comparisons, and discussions,
Figure 5.10 shows the load versus mid-span deflection of the control intact beam B1 and
the control damaged (unrepaired) beam B2. The figure shows that both the strength and
stiffness of beam B2 have been severely degraded as a result of the complete cutting of
the lower steel flange at mid-span. Table 5.2 shows 60 percent reduction in flexural
Figure 5.11 shows the load versus longitudinal strains of the underside of the upper steel
flange for both beams B1 and B2. The figure shows small tensile strain values initially at
the upper flange of both beams, which indicate that the neutral axis is inside the concrete
slab. The figure also shows that the strain of beam B2 is greater than that of beam B1, at
the same load levels, within the linear elastic part. This indicates a substantial upward
shift of the neutral axis as a result of cutting the lower flange of beam B2. In both cases, a
163
Chapter 5
sudden increase in the tensile strain is observed upon yielding of the lower part of the
beam, indicating spread of yielding throughout the steel cross section. However, the top
flange of B2 only is yielded at ultimate. Figure 5.12 shows the load versus strains of the
underside of the lower steel flange of both beams B1 and B2. Two strain gauges (S50-5
and S100-5) were attached to beam B2 at distances of 50 and 100 mm from mid-span,
while the strain of B1 was measured at mid-span. The figure shows that the strain values
of B2 at a distance of 100 mm from mid-span are higher than those at 50 mm and are
close to those of B1 up to a load level of 69 kN. At higher loads, the pre-cut in the flange
develops into a crack propagating within the steel web. This results in further spreading
of the stress flow away from mid-span. As a result, the strains along the steel flange of
As discussed earlier, cutting the lower steel flange at mid-span creates regions of stress
concentration, which leads to crack propagation within the web as the load increases.
This is discussed further in detail, in this section. Figure 5.13 shows the load versus
strains at four different locations along the depth of a mid-span section of specimen B2.
The figure shows that the entire steel section is under tensile stresses. The behaviour also
suggests a highly nonlinear strain distribution along the web at mid-span, as also shown
in the small diagram within Figure 5.13. It is clear from this figure that the closest region
to the cut flange yields much earlier than farther regions along the web, due to stress
concentration. The strains of gauge (S0-4), just above the cut, increase rapidly and reach
the yield strain at a very low load level of about 10 kN. The figure also indicates that the
entire steel section at mid-span, including the upper flange, yields excessively before the
164
Chapter 5
beam attains its maximum load. After this excessive yielding, the parts above the cut
completely separate leading to propagation of the crack from the flange into the web.
Figure 5.14 shows the load versus strains in the web, at heights of 20 mm and 80 mm
above the underside of the lower steel flange, at three different locations in the transverse
direction. The figure shows that just above the cut (i.e. at 20 mm from the bottom) the
evident by the readings of S50-4 and S100-4 strain gauges, which show significantly
lower strain (below yielding) than that measured by S0-4, just above the cut. The figure
also shows that at a height of 80 mm from the bottom, the stress concentration effect
becomes more pronounced at a load level of about 50 kN, which is indicated by the
departure of strain readings of S0-2 gauge from the trend of the other gauges S50-2 and
S100-2. This is quite different from the strains at the 20 mm height, where stress
The failure mode of the intact control beam B1 was yielding of the steel cross section,
followed by concrete crushing, as shown in Figure 5.15(a). For the damaged beam B2,
significant yielding associated with crack propagation from the cut flange into the web, as
shown in Figure 5.15(b), was observed. At the end of the test, the measured crack width
and height were 14 and 67 mm, respectively. Because of the crack propagation within
almost 44 percent of the depth of the steel section, the neutral axis was significantly
shifted upwards, inside the concrete slab. Therefore, the bottom of the concrete slab
165
Chapter 5
experienced high tensile stresses, which led to tension cracks, as shown in Figure 5.15(b).
Beams B3 and B4 were repaired by bonding high modulus- (HM-) CFRP type C3 (Ec3 =
231 GPa) of different number of layers, whereas beams B5 and B6 were repaired by
bonding standard modulus- (SM-) CFRP type C1 (Ec1 = 115 GPa) and type C2 (Ec2 = 90
GPa), respectively. It should be noted that for all four beams the CFRP sheets were
bonded on the lower side of the tension steel flange, along the full span. Figure 5.16
shows the load-deflection responses of beams B3 and B4, compared to the reference
(control) beams B1 (intact) and B2 (damaged). The figure shows that both the HM-CFRP
repaired beams (B3 and B4) reached flexural stiffness values higher than the intact beam
B1, which is attributed to the high value of elastic modulus of CFRP type C3. Table 5.2
shows gains in stiffness of 13 and 26 percent for beams B3 and B4, respectively. Figure
5.17 also shows a comparison between the load-deflection responses of beams B5 and B6
and control beams B1 and B2. The figure shows that beam B6, which has a larger area of
CFRP (i.e. higher force equivalence index, ω), but lower CFRP elastic modulus, could
not achieve the full stiffness of the intact beam B1 and was 14 percent lower, as indicated
in Table 5.2. Figure 5.16 and 5.17 show that only beam B4 (repaired with HM-CFRP and
ω = 152 percent) was able to recover the original strength of B1 and even exceed it by 10
percent. Beam B6 had an ω = 193 percent, higher than that of B4, but only achieved a
strength 13 percent lower than that of the intact beam B1. This is attributed to the
166
Chapter 5
different failure modes as will be discussed in section 5.3.2.2. It is also noted that the
Beam B5, which is almost similar to B6, but the CFRP sheets were applied 22 months
after sandblasting, clearly shows a much lower strength than B6. Table 5.2 shows that the
strength of B5 is 50 percent lower than that of the intact beam B1. The lower bond
integrity between the steel surface and the CFRP sheets has promoted the premature and
progressive debonding along the interface, which affected the load-deflection response,
Figure 5.18 and 5.19 show the strain distributions along part of the length of the CFRP
sheets of beams B3 and B4, respectively, at different load levels. The figures show that
the CFRP sheets type C3 of both beams reached their ultimate (rupture) strain (εult = 2.2 x
10-3, as reported in Table 3.2). It is also clear that at any load level the strains in B3 are
Figure 5.20 and 5.21 show similar strain distributions along the CFRP sheets of beams
B5 and B6, respectively, at different load levels. The figures show that both beams
reached their maximum loads before utilizing the full tensile strength of the CFRP sheets.
It is also clear that the maximum strain value in the CFRP sheets of B5, at failure, is
much smaller than that of B6, due to the bond deficiency and the early initiation of
debonding.
167
Chapter 5
Figure 5.22 compares the load versus strains of the CFRP at mid-span of beams B3, B4,
and B6, relative to the control beam B1. The figure clearly shows that the HM-CFRP
sheets bonded on both B3 and B4 have reached their rupture strain. On the other hand,
the SM-CFRP sheets bonded on B6 reached only 48 percent of their rupture strain. Figure
5.23 shows the load versus strains of the upper steel flange at mid-span of beams B3, B4,
and B6, relative to control beams B1 and B2. The figure shows initial compressive strains
at the upper flanges of B3, B4, and B6, which suggest that adding the CFRP sheets to the
damaged beams has shifted the neutral axis down (i.e. within the steel cross section),
particularly in B4 with HM-CFRP and ω = 152 percent. The figure also shows that after
rupture of the CFRP sheets in beams B3 and B4 the strains of the upper steel flanges
convert suddenly to tension, whereas in the case of B6 the strains of the upper steel
flange convert gradually to tension as the debonding of the CFRP sheets progresses.
The failure mode of beams B3 and B4 occurred by rupture of the CFRP sheets at mid-
span, as shown in Figure 5.24(a and b), after reaching the ultimate strain of CFRP type
C3. There were no signs of local debonding in the vicinity of mid-span. After rupture of
the CFRP sheets, the cut in the lower flange developed into a crack that propagated up
into the web, as shown in Figure 5.24(a). The load capacity of both beams dropped after
the rupture of CFRP sheets, as shown in Figure 5.16, and then increased again, slightly
until the concrete slabs crushed. After crushing of the slabs, the behaviour followed the
168
Chapter 5
The failure of beam B5 occurred by premature debonding along the interface between the
steel and the FRP layers at a very early stage, as shown in Figure 5.24(c). Failure of beam
B6 occurred in multiple steps. The first step was debonding of the CFRP sheets within
the shear span at one end of the beam, as shown in Figure 5.24(d and e). The debonding
of the CFRP sheets was associated with complete and sudden separation of the debonded
portions of the sheets, followed by a drop in the load from 298 kN to 265 kN, as shown in
Figure 5.17. After this drop, the specimen gained additional load and reached its
maximum capacity of 311 kN when the concrete crushed. Local debonding was observed
in the vicinity of the steel flange cut (i.e. mid-span) for both the GFRP sheets and the
remaining portion of the CFRP sheets, as indicated by the discoloration in Figure 5.24(e).
In order to study the effect of the bonded surface area, a comparison was performed
between beams B6 and B7 with almost similar force equivalence indices of 193 and 185
percent, respectively. In beam B6, the CFRP sheets were bonded on the bottom side of
the steel flange, whereas in B7 the CFRP sheets were bonded on both the top and bottom
sides of the flange. Figure 5.25 shows the load-deflection responses of both beams. The
figure shows that increasing the bonded surface area in B7 enhanced the overall stiffness
of the beam, compared to B6. Table 5.2 shows that B7 achieved the same stiffness as the
control intact beam B1 (33.8 kN/mm), which is 14 percent higher than B6. Both beams
achieved very similar strength (311 and 307 kN, respectively). These strengths resemble
169
Chapter 5
However, debonding in beam B7 started earlier than in B6 and affected the behaviour as
Figure 5.26 and 5.27 show the strain distributions along the lower and upper CFRP sheets
bonded on both sides of the flange of beam B7, at different load levels. The figures
generally show that beam B7 reached its maximum capacity before utilizing the full
strength of the CFRP sheets. The smaller strain gradient of the upper CFRP sheets, with
respect to that of the lower CFRP sheets; indicate that debonding along the upper side
started earlier than in the lower side. The earlier debonding on the upper side is attributed
Peeling stresses on the upper side were developed as a result of the beam’s curvature.
The maximum load capacities of beams B6 and B7 were controlled by crushing of the
concrete slabs, as indicated in Figure 5.25. The detailed failure description of B6 was
discussed earlier in section 5.3.2.2. For B7, a stationary camera was mounted near the
mid-span to record the progressive failure of the beam as the load increases. Figure
5.28(a) shows four images throughout the history of loading of B7. Image 1 shows the
beam before loading (P = 0), while image 2 captures the initiation of debonding of the
upper CFRP layers and the opening of the steel flange cut (P = 180 kN). In image 3,
crack propagation in the web and crushing of the concrete slab as well as complete
debonding of the upper CFRP layers are observed (P = 300 kN). The load capacity of B7
was then dropped from 307 kN to 238 kN, as shown in Figure 5.25. After excessive
170
Chapter 5
deflection of the beam (19.8 mm), the lower CFRP layers was completely debonded, as
shown in image 4 (P = 130 kN). An overall picture of beam B7 after failure is shown in
Figure 5.28(b).
Figure 5.29 shows the load-deflection behaviour of beams B7 and B8, relative to control
specimens B1 and B2. Beams B7 and B8 were repaired by bonding SM-CFRP sheets on
both the lower and upper sides of the steel flange, and have force equivalence indices of
185 and 210 percent, respectively. Also, beams B3 and B4, repaired by bonding HM-
CFRP sheets on the lower side of the steel flange, had force equivalence indices of 87
and 152 percent, respectively, and their load-deflection responses are given in Figure
5.16. Table 5.2 shows that for both types of CFRP, beams with the smaller force
equivalence indices (i.e. B3 and B7) did not reach the ultimate strength of the intact beam
B1. The recorded strengths of B3 and B7 were only 166 and 306 kN, respectively. These
strengths, however, represent increases of 15 and 112 percent relative to the strength of
the damaged beam B2, for B3 and B7 respectively. On the other hand, beams with higher
force equivalence indices (i.e. B4 and B8) were able to reach the strength of the intact
beam B1, and even exceeded it by 10 and 16 percent, respectively. Figure 5.30 shows the
flexural strengths of the beams versus the force equivalence indices for the two types of
CFRP. The figure shows similar trends of strength increase by increasing the amount of
CFRP used (i.e. increasing the force equivalence index). It is also noted that as the elastic
171
Chapter 5
restore the strength. This is a result of the change in failure mode from debonding to
rupture. The figure also indicates that the optimum force equivalence indices that would
just recover the strength of the undamaged beam are 142 and 197 percent for the HM-
In both beams B3 and B4 with HM-CFRP, failure was due to rupture of CFRP as
discussed before. The failure mode of the beams repaired using SM-CFRP (i.e. B7 and
B8) was initiated by debonding of CFRP on the upper side of the flange. However, the
load continued to increase until crushing of concrete slabs occurred and the peak load
was reached. The load then dropped gradually until the CFRP on the lower side of the
flange debonded. The behaviour of beams B7 and B8 was significantly nonlinear, relative
to B3 and B4, which were quite linear elastic. Figure 5.31 shows a picture of beam B8
after failure.
The repair scheme used in beam B8 achieved the best results in terms of restoring both
the strength and stiffness, among the beams repaired using the SM-CFRP type C2.
Therefore, the same repair scheme of B8 was chosen to investigate the effect of varying
the bonded length of CFRP sheets. Beams B9, B10, and B11 were repaired with a similar
system of the same force equivalence index but of various lengths, 1000, 250, and 150
mm, respectively. It is clear that the bonded lengths of beams B10 and B11 fall inside the
172
Chapter 5
constant bending moment zone, which is 400 mm long. Figure 5.32 shows the load-
deflection responses of the beams. It clearly shows a consistent decrease in the flexural
capacity of the beams with shortening the bonded length of the CFRP sheets. As
discussed earlier, beam B8 with CFRP sheets bonded along almost the full span (1900
mm) achieved a 16 percent higher strength than the intact beam B1. Table 5.2 indicates
that beam B9, which is bonded with 1000 mm long CFRP sheets achieved a strength of
353 kN, which is almost equal to that of the intact beam B1. Beams B10 and B11 with
bonded lengths of 250 and 150 mm, respectively, failed at lower load levels before they
reach the strength of the intact beam B1. Table 5.2 shows that the flexural strengths of
B10 and B11 are 319 and 256 kN, respectively. The table also shows that the stiffness of
the intact beam B1 has been fully achieved in the case of B9, and very closely achieved
in B10 and B11, regardless of the short bond length. Figure 5.33 shows a summary of the
effect of bond length. The figure gives the variation of ultimate load achieved relative to
that of control beam B1 versus the bonded length of CFRP, normalized to the span
length. Figure 5.32 also shows the effect of the bonded length on the ductility of the
repaired beams. The deflection at ultimate and the length of the linear part of the curve
increase by increasing the bonded length of CFRP. The maximum loads attained by
beams B8 to B11 were achieved at mid-span deflections of 19, 16, 13, and 9 mm,
respectively.
Figure 5.34 to Figure 5.37 show the axial strain distributions at different locations along
the lower CFRP sheets of beams B8 to B11, respectively, at different load levels. The
figures show sudden changes in the trends at certain locations in beams B8 to B10, which
173
Chapter 5
indicate the initiation of CFRP debonding and the associated redistribution of stresses in
the CFRP sheets. The consistent pattern of strain distribution of beam B11 up to failure,
without signs of redistribution (Figure 5.37), indicates a complete and sudden debonding
of the CFRP sheets. The load versus the strain readings of gauge S0-5 for beams B8 to
B11 are plotted in Figure 5.38. The figure shows higher strain values in the CFRP sheets
of the repaired specimens, compared to the strain values of the steel flange of the control
intact beam B1. This is attributed to the lower elastic modulus of CFRP type C2 with
respect to the steel. The figure also shows that the maximum strain attained by the CFRP
sheets at mid-span, at debonding failure, varies among the beams of different bonded
lengths. A small jump in the strain values can be observed in the behaviour of the
repaired beams at a load level of about 50 kN. This jump is attributed to a minor slip
between the concrete slab and the steel beam. This slip was detected using a 25 mm
horizontal LP to measure the relative displacement between the concrete slab and the
steel beam in specimen B9, and was only about 0.15 mm at ultimate, as shown in Figure
5.39.
Figure 5.40 shows the variation of the maximum strain reached in the CFRP sheets with
the bonded length of CFRP. The figure clearly shows a bilinear behaviour. The first
linear part represents the effect of the CFRP bonded length within the constant moment
zone in absence of shear, whereas the second part represents the effect of the bonded
length within the shear span, in presence of both bending and shear. The first line can be
extended down to the origin (point “a”), which represents a hypothesized case without
CFRP sheets. If the line is also extended up, to the rupture strain of the CFRP sheets
174
Chapter 5
(point “b”), the optimum bonded length required for repair in a constant moment zone
can be estimated, which is 181 mm in this case. The second linear part of the curve
clearly has a shallower slope due to the presence of additional high shear stresses within
the shear span of the beam, which necessitate longer bonded length. By extending the
second line to point “c”, the sufficient bonded length required to achieve rupture of CFRP
can be obtained. Obviously it is longer than the span of the beams tested in this study.
Figure 5.41 to Figure 5.44 show the load versus the average shear stresses of beams B8 to
B11, respectively. The shear stresses were calculated based on Equation 5.1, at discrete
locations along the CFRP sheets on both sides of the mid-span. Unlike the strengthening
situation (Figure 5.6 and Figure 5.7), significantly higher shear stresses are developed
along the interface between the CFRP sheets and the steel substrate near the edge of the
terminated steel flange (i.e. mid-span), in the case of repair. Figure 5.41 to 5.44 show that
a typical shear stress curve reaches a peak value then reverses direction after the
maximum stress, which is about 25 MPa, is reached. It is believed that once this peak is
reached, local debonding occurs at this specific location. It is noted then that these peaks
towards the ends. This behaviour agrees well with the bi-linear bond-slip model
developed by Xia and Teng (2005), and discussed earlier in Chapter 2. It is clear that
locations. For beam B11 with the shortest CFRP sheets (Lsheet = 150 mm), the shear
stresses near the ends of the sheets were higher than those near the mid-span, as shown in
175
Chapter 5
Figure 5.44. This behaviour indicates the inadequacy of the bonded length of the CFRP
sheets in B11.
The failure mode of beams B9 and B10 was debonding of both the upper and lower
CFRP layers, which started at mid-span (i.e. from the edge of the cut steel flange). After
debonding of the CFRP sheets, a large drop in the load was clearly observed. Loading
was then continued until the concrete slab crushed, as indicated in Figure 5.32. In the
case of B11, debonding occurred in the lower CFRP layers as well as on one side of the
two upper CFRP layers. The lower CFRP layers were debonded from one end, while the
upper CFRP layers were completely separated. Also, the tests of beams B10 and B11
176
Chapter 5
moment of inertia
Specimen I.D.
Ultimate load
Transformed
x 106 (mm4)
%age gain
%age gain
%age gain
%age gain
Yield load
(kN./mm.)
Stiffness
(kN.)
(kN.)
G1a 8.55 -- 3.67 -- 85 -- 144 --
G2 9.59 12 4.29 17 97 14 216 50
G3 9.88 16 4.36 19 103 21 217 51
a
Control (unstrengthened) girder
difference
Specimen
Maximum
load (kN)
Stiffness
(kN/mm)
%age
%age
I.D.
a
%age difference is calculated with respect to beam
B1 (intact control beam).
177
Chapter 5
225
P = 217 kN. P = 216 kN.
200
G3 failure
Debonding of G2
175
outer layer
(1500 mm long)
150 G1 (Control) P = 144 kN.
125
P = 111 kN. crushing
100
Py = 85 kN
Load
75
50 δ
5940 mm
25
0
0 20 40 60 80 100 120 140 160 180
Deflection (mm)
160
140
S1
120
S2
100
Load (kN)
Py = 85 kN.
80
60
40
20 S2 S1
50 mm
0
0 2 4 6 8 10 12 14 16
-3
Strain x 10 (mm/mm)
Figure 5.2 Load-strain responses of the lower flange of the control specimen G1.
178
Chapter 5
225
200 G3
175
G2
150
G1 (Control)
Load (kN)
125
100
75
Strain
50 gauge
50 mm
25
0
0 2 4 6 8 10 12 14 16
-3
Strain x 10 (mm/mm)
Figure 5.3 Load-steel strain responses at the web of specimens tested in Phase II.
225
(εmax)actual = 13 x 10-3
1000
200 1500
1925
0
175
(εc4 )ult = 12.6 x 10-3
Debonding of
150
675
outer layer
(coupons)
500
Load (kN)
125
Max. strain reached in the
75
1500
1000
1925
0
500
675
50
On inner
On inner
On outer
25
layer
layer
layer
0
0 2 4 6 8 10 12 14
-3
Strain x 10 (mm/mm)
Figure 5.4 Load-strain responses along the CFRP plates of specimen G2 in Phase II.
179
Chapter 5
225
1925 1500
200
175
150
Load (kN)
125 675
Debonding of
outer layer mid-span P/2
100
0 and 500
75
1500
1925
1000
0
500
675
50
On outer
On inner
layer
layer
25
0
00 2
2000 4
4000 6000
6 8
8000 10
10000 12
12000 14
14000
-3
Strain x 10 (mm/mm)
Figure 5.5 Load-strain responses along the CFRP plates of specimen G3 in Phase II.
225
P/2
mid-span
200
175
588
250
150 Debonding
(P = 129 kN)
Load (kN)
125 588
250
strength of adhesive
Ultimate shear
100
Adhesive yielding @
75 ( τ , P ) = ( 0.16 MPa , 96 kN)
50
25
0
0 0.5 1 1.5 2 2.5
Average shear stress (MPa)
Figure 5.6 Load-average shear stress responses along the CFRP plates of specimen
G2 in Phase II.
180
Chapter 5
225
200 P/2
mid-span
175
150
588
250
Debonding
Load (kN)
250 588
strength of adhesive
100
Ultimate shear
75
Adhesive yielding @
( τ , P ) = ( 0.37 MPa, 83 kN)
50
25
0
0 0.5 1 1.5 2 2.5
Average shear stress (MPa)
Figure 5.7 Load-average shear stress responses along the CFRP plates of specimen
G3 in Phase II.
225
G3 (PI 1) G2 (PI 1) G3 G2
(PI 2) (PI 2)
200
Concrete
175 crushing
150 G1 (PI 1)
G1 (PI 2)
Load (kN)
125
PI1
(yielded)
100
0.90 d
75
S1
50 PI2
25
0
-4 -3 -2 -1 0 1
-3
Strain x 10 (mm/mm)
181
Chapter 5
cross
loading Spreader beam
Rupture of
beam
GFRP end
wraps
Mid-span
Debonding
pan
Mid-s
Lateral
deflection of the
concrete slab Concrete
crushing in
Concrete crack compression
in tension
182
Chapter 5
400
Concrete crushing
B1 (intact) P = 357 kN
350
300 Py = 270 kN
250
Load (kN)
150 B2 (damaged)
Load
100
50 δ
1960 mm
0
0 5 10 15 20 25 30 35
Deflection (mm)
400
350 B1
εy = 1.9 x 10-3
S0-1
300
Load (kN)
250
200
150 B2
100
50
0
0 0.5 1 1.5 2 2.5 3 3.5
Strain x 10-3 (mm/mm)
Figure 5.11 Load-strain responses of the upper steel flanges of specimens B1 and B2
in Phase III.
183
Chapter 5
400
350 B1 (mid-span)
300
250
Load (kN)
cut
mid-span
200 B2 P/2
S 50-5 S 100-5
150
152
100
50 50
50
S100-5
0 S 50-5
-2 0 2 4 6 8 10 12 14 16
Strain x 10-3 (mm/mm)
Figure 5.12 Load-strain responses of the lower steel flanges of specimens B1 and B2
in Phase III.
160
Load, kN
S0-1 160 25
140 S0-1
140
50
75
120
100
120 S0-4 30 40
S0-2
20
20 0
0 4 8 12
εy = 2x10-3
100
Load (kN)
S0-3
80
60
S0-4
40
20
0
0 2 4 6 8 10 12 14 16 18 20
Figure 5.13 Load-strain responses along the mid-span cross section of specimen B2
in Phase III.
184
Chapter 5
160
mid-span
S100-2
140 S50-2
S0-2 S50-2 S100-2
152
S0-2
120 60
S0-4 S50-4 S100-4
20
S100-4
100 50 50
Load (kN)
S50-4
80
60
εy = 2x10-3
S0-4
40
20
0
0 2 4 6 8 10 12 14 16 18 20
Strain x 10-3 (mm/mm)
Mid-span
concrete
crack
n
pa
d -s
i
M
width
widtof
h
opening
of cut
185
Chapter 5
450
Rupture of
400 CFRP sheets
B4
B1
350 (ω = 152%)
Load
300
Rupture of
CFRP sheets δ
250
Load (kN)
1960 mm
B3
200 (ω = 87%) Concrete
Concrete
crushing
crushing
150
100 B2
50
0
0 5 10 15 20 25 30 35
Deflection (mm)
450
Load
400
First debonding
δ of CFRP sheets B1
350 1960 mm
Concrete
crushing
300 B6 complete debonding
(ω = 193 %) of CFRP sheets
Load (kN)
250
progressive complete debonding
200 debonding of CFRP sheets
B5 (ω = 198 %)
150 (late application of CFRP)
100
B2
50
0
0 5 10 15 20 25 30 35
Deflection (mm)
186
Chapter 5
2.5
165
1.5
0.5
Mid-span
0
0 50 100 150 200 250
Distance (mm)
Figure 5.18 Strain distributions along the CFRP sheets of specimen B3 in Phase III.
2.5
(εc3)ult = 2.27 x 10-3
Load (kN)
50 100 150
2
200 300 394
Strain x 10-3 (mm/mm)
1.5
Mid-span
0.5
0
0 50 100 150 200 250
Distance (mm)
Figure 5.19 Strain distributions along the CFRP sheets of specimen B4 in Phase III.
187
Chapter 5
11
8 100
Strain x 10-3 (mm/mm)
Mid-span 150
7 179
mid-span
6
5 P/2
G0-5
2
0
0 50 100 150 200 250
Distance (mm)
Figure 5.20 Strain distributions along the CFRP sheets of specimen B5 in Phase III.
11
(εc2)ult = 10.9 x 10-3
10
Mid-span
9
8
Strain x 10-3 (mm/mm)
6
Load (kN)
5 50 100 150 200
4
250 300 311
3
0
0 50 100 150 200 250
Distance (mm)
Figure 5.21 Strain distributions along the CFRP sheets of specimen B6 in Phase III.
188
Chapter 5
450
400
350 Yield of B1
steel flange
εmax = 5.2 x 10-3
300 B4
Load (kN)
mid-span
250
B6
200 P/2
150 B3
100
S0-5
50
0
0 2 4 6 8 10 12 14 16
Strain x 10-3 (mm/mm)
Figure 5.22 Load-strain responses of CFRP at mid-span of specimens B3, B4, and
B6 in Phase III.
450
mid-span
400 B4
P/2
B1
350
B6 S0-1
300
Load (kN)
250
200
150 B3
100
B2
50
0
-0.5 0 0.5 1 1.5 2 2.5
Strain x 10-3 (mm/mm)
Figure 5.23 Load-strain responses of the upper steel flange at mid-span of specimens
B3, B4, and B6 in Phase III.
189
Chapter 5
Concrete slab
Lower flange
Rupture of
CFRP sheets
(a) Typical failure mode of B3 and B4 (b) Downside view of CFRP rupture in B3
Loading beams
pan
h e ar s
s
in the
with
ng
o ndi
Deb
Deb
on
she ded
ets
Debonded area
(discoloration)
CFRP sheets
GFRP sheet
190
Chapter 5
450
400
CFRP Concrete
Steel crushing
350 CFRP
B1
Debonding of
(two sides) lower CFRP
200
B6 (ω = 193 %)
150 (one side) B2
100 Load
50 δ
1960 mm
0
0 5 10 15 20 25 30 35
Deflection (mm)
11
(εc2)ult = 10.9 x 10-3
10 Load (kN)
Mid-span
9 50
100
8
Strain x 10-3 (mm/mm)
150
7 200
6 250
307
5
0
-150 -100 -50 0 50 100 150 200 250
Distance (mm)
Figure 5.26 Strain distributions along the lower CFRP sheets of specimen B7 in
Phase III.
191
Chapter 5
11
Load (kN)
(εc2)ult = 10.9 x 10-3
10 50
Mid-span
100
9
150
8
Strain x 10-3 (mm/mm)
200
7 250
307
6
0
-150 -100 -50 0 50 100 150 200 250
Distance (mm)
Figure 5.27 Strain distributions along the upper CFRP sheets of specimen B7 in
Phase III.
192
Chapter 5
Stiffener
angle
Mid-span cut
Crack
opening
Initiation of
debonding
propagation of
crack in the web
complete
debonding Crack
height
Debonding of
lower layers
Upper CFRP
layers
193
Chapter 5
450
B8
400 (ω = 210 %)
350 B1
Concrete
Debonding of crushing
300
upper CFRP Debonding of
B7 lower CFRP
250 (ω = 185 %)
Load (kN)
200
150
100 B2 Load
50 δ
1960 mm
0
0 5 10 15 20 25 30 35
Deflection (mm)
1.2
Af x F f B8
Ultimate load ratio relative to control beam B1
ω= x 100
1.1 Aflange x Fy B4
Pa
Intact beam (B1)
0G
1
=9
c2
E
0.9
B7
a
GP
0.8
0
23
c3
=
E
0.7
ω = 142 %
0.6
ω = 197 %
0.5
B3 Damaged beam (B2)
0.4
50 70 90 110 130 150 170 190 210 230
Force equivalence index, ω (%)
Figure 5.30 Effect of force equivalence index (ω) on the strength of the repaired
beams.
194
Chapter 5
Concrete slab
Debonding of
the upper
CFRP layers support
450
B8 (Lsheets = 1900mm) P = 415 kN
ω = 210 %
400 Concrete crushing
B9 (Lsheets = 1000mm)
P = 353 kN
350 B1
Debonding
300
B11 P = 256 kN
Load (kN)
200
150
B2
100 Load
50
δ
1960 mm
0
0 5 10 15 20 25 30 35
Deflection (mm)
195
Chapter 5
1.4
0.8
B11
0.6
B2
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
Lsheet / Span
10
10000
9
9000 Load (kN)
Load 50
8
8000 Mid-span
100
Strain x 10-3 (mm/mm)
7
7000 150
1900 mm
6
6000 1960 mm 200
250
5
5000
300
4
4000 350
3 415
3000
2
2000
1
1000
00
-350 -250 -150 -50 50 150 250 350
Distance (mm)
Figure 5.34 Strain distributions along the lower CFRP sheets of specimen B8 in
Phase III.
196
Chapter 5
10000
10 Load
Mid-span
9000
9 Load (kN)
50
8000
8 1000 mm 100
1960 mm 150
7000
Strain x 10-3 (mm/mm)
7 200
250
6000
6 300
353
5
5000
4
4000
3
3000
2
2000
1
1000
00
-350 -250 -150 -50 50 150 250 350
Distance (mm)
Figure 5.35 Strain distributions along the lower CFRP sheets of specimen B9 in
Phase III.
10
10000 Load
9
9000 Load (kN)
8 50
8000 250 mm Mid-span
1960 mm 100
Strain x 10-3 (mm/mm)
7
7000
150
6
6000
200
5
5000 250
4
4000 320
3
3000
2
2000
1
1000
00
-350 -250 -150 -50 50 150 250 350
Distance (mm)
Figure 5.36 Strain distributions along the lower CFRP sheets of specimen B10 in
Phase III.
197
Chapter 5
10000
10 Load
Mid-span
9000 Load (kN)
9
50
8000
8 150 mm
100
1960 mm
7000
Strain x 10-3 (mm/mm)
7 200
6000
6 256
5000
5
4000
4
3000
3
2000
2
1000
1
00
-350 -250 -150 -50 50 150 250 350
Distance (mm)
Figure 5.37 Strain distributions along the lower CFRP sheets of specimen B11 in
Phase III.
450
400 B8
350 B10 B1
B9
300
(εc2)ult = 10.9 x 10-3
B11
Load (kN)
250
mid-span
200
P/2
150
100
50
S0-5
0
0 2 4 6 8 10 12 14 16
Strain x 10-3 (mm/mm)
198
Chapter 5
400
350
300
250
Load (kN)
200 LP
150 LP
100
50 kN
50
0
0 0.05 0.1 0.15 0.2 0.25
Slip (mm)
Figure 5.39 Load versus slip of concrete slab of specimen B9 in Phase III.
12
b (εc2)ult = 10.9 x 10-3
Mid-span
10
Maximum strain x 10-3 (mm/mm)
B8
8 B9
B10
6 ω = 210 %
B11
4
181 mm 1175 mm
2
Constant Combined moment
moment zone and shear zone
a
0
0 200 400 600 800 1000 1200 1400
Distance from mid-span (mm)
Figure 5.40 Maximum strains versus the bonded length of the CFRP sheets.
199
Chapter 5
450
475 -275 -175
275
400
175
350
300
100 -100
50
250
Load (kN)
200
-50
150
-12.5
12.5
100
50
0
-30 -20 -10 0 10 20 30
Figure 5.41 Load-average shear stress responses along the lower CFRP sheets of
specimen B8 in Phase III.
450
400
350 175
300
250
Load (kN)
50
12.5
200
100
150
100
50
0
-30 -20 -10 0 10 20 30
Average shear stress (MPa)
Figure 5.42 Load-average shear stress responses along the lower CFRP sheets of
specimen B9 in Phase III.
200
Chapter 5
450
400
350
95
300 -95
250
Load (kN)
50 -50
200
150
100
12.5 -12.5
50
0
-30 -20 -10 0 10 20 30
Figure 5.43 Load-average shear stress responses along the lower CFRP sheets of
specimen B10 in Phase III.
450
400
350
300
Load (kN)
250
200
50
150
12.5
-50
100 -12.5
50
0
-30 -20 -10 0 10 20 30
Figure 5.44 Load-average shear stress responses along the lower CFRP sheets of
specimen B11 in Phase III.
201
Chapter 6
Chapter 6
6.1 Introduction
This chapter presents analytical and numerical models developed to predict the behaviour
and strength of concentrically loaded HSS slender steel columns strengthened using
The first model is an analytical fibre model based on the concepts of strain compatibility
and force equilibrium (Shaat and Fam, 2007a). The second model is an alternative
numerical non-linear finite element model (FEM) (Shaat and Fam, 2007b). Both models
1
Most of contents of this chapter have been published as follows:
i. Shaat, A. and Fam, A. (2007a) “Fiber-Element Model for Slender HSS Columns Retrofitted with
Bonded High Modulus Composites.” Journal of Structural Engineering, ASCE, 133(1):85-95.
ii. Shaat, A. and Fam, A. (2007b) “Finite Element Analysis of Slender HSS Columns Strengthened with
High Modulus Composites.” Steel & Composite Structures, 7(1):19-34.
202
Chapter 6
specimens 7 to 11, of different CFRP reinforcement ratios have resulted in gains in the
axial strength that do not correlate to the reinforcement ratios. This led to difficulties in
assessing the effect of CFRP reinforcement ratio experimentally. As such, the models
presented in this chapter are quite useful, and in fact essential, to uncouple the two
effects, namely, the out-of-straightness imperfection and the CFRP reinforcement ratio. A
In order to predict the load versus axial and lateral displacement responses of CFRP-
strengthened slender HSS steel columns, a non-linear fibre model has been developed.
The model accounts for both material and geometric (second order effects) non-linearities
as well as residual stresses. An incremental approach is used, where the concepts of force
equilibrium and strain compatibility are satisfied at each loading step. The stress-strain
curve of steel is assumed to follow an elastic-perfectly plastic model. FRP materials are
different components of the model, whereas section 6.2.7 provides the procedures of
203
Chapter 6
In the proposed fibre model, the through-thickness residual stress distribution is idealized
Davison and Birkemoe (1983) and by Chan et al. (1991). The residual stress, Frs is
chapter 3 and shown in Figure 3.3. The values of Frs are equal to 0.49 Fy and 0.33 Fy for
HSS1 and HSS2 sections, respectively. The effect of varying the level of residual stresses
In order to generate the full load-displacement responses, the residual stress pattern
across the wall thickness was first defined by dividing the steel wall into three equal
layers, as shown in Figure 6.1. A uniform compressive stress value of (-Frs) was assigned
to the inner layer, while a tensile value of (+Frs) was assigned to the outer layer. The
middle layer was divided into two equal halves. The inner half was assigned a uniform
value of (-0.5 Frs), while the outer half was assigned a uniform value of (+0.5 Frs). This
distribution was used to simulate the prescribed residual stress pattern shown in Figure
2.1. It should be noted that the second type of residual stresses, namely, the perimeter
(membrane) residual stresses was not considered in the analysis since it was deemed
areas of steel and FRP. The cross section is divided into four areas (A1 to A4), as shown
204
Chapter 6
in Figure 6.1. The flat part of the flanges oriented normal to the plane of buckling (A1) is
divided into 12 strips through the thickness, where the strain is assumed constant across
the width and thickness of each strip. The flat part of the flanges parallel to the plane of
buckling (A2) is divided into 12 x 80 elements to capture the strain gradient across the
depth of the section and also to capture the residual stress distribution within the
thickness. The corner section (A3) is idealized as a square and is divided into 12 x 12
elements. Area A4 represents the FRP layers attached to area A1, and is divided into 1 x
n elements, where n is the total number of FRP layers. To model specimens with FRP
material bonded on four sides, an additional area of FRP (A5) attached to area A2, is also
considered. The centroid of each element is located at its mid thickness and the stress is
assumed constant within the element area. Linear strain distribution and strain
compatibility (i.e. full bond between steel and FRP) are also assumed in the analysis.
buckling, the axial force P is offset relative to the effective centroid of the mid-height
section by eccentricity e. For a given strain gradient induced by the eccentric load P, and
based on a strain level ε at the extreme compression side, and a neutral axis depth c, the
strain εi in each steel or FRP element i, located at a distance yi from the effective centroid
⎡ y y ⎤
ε i = ⎢1 − i − c ⎥ ε (6.1)
⎣ c c ⎦
where yc is the distance between the extreme fibre in compression and the effective
centroid of the cross section, as shown in Figure 6.2. It should be noted that before
205
Chapter 6
yielding of steel or crushing of FRP, yc = h/2, where h is the depth of the section. The
effective centroid only shifts when parts of the section yield in an unsymmetrical manner,
due to the strain gradient. The stress in steel elements Fsi is then calculated as Fsi = Es εi,
where Es is Young’s modulus of steel, and is added to the residual stress Frs to obtain the
total stress Fsi + rs for every steel element, as given by Equation 6.2. The total stress Fsi + rs
is used to check whether the element has yielded or not. The effect of yielded elements is
F si + rs = F si + Frs (6.2)
Possible stress distributions at various stages of loading are shown in Figure 6.2. The
total axial load P at a given stage of loading (i.e. for a given ε and c) can be obtained by
numerical integration of stresses over the cross section, for both the yielded and elastic
steel elements as well as for the FRP elements, based on a linear stress-strain response, as
follows:
P= ∑
elastic steel
(F si )
Asi + ∑
plastic steel
(F y ) FRP
(
Asi + ∑ ε i E f A f i ) (6.3)
M = ∑
elastic steel
(F si )
Asi y i + ∑
plastic steel
(F y ) FRP
(
Asi y i + ∑ ε i E f A f i y i ) (6.4)
where A s i and A f i are the areas of steel and FRP elements, respectively, Ef is Young’s
modulus of the FRP element, and yi is the distance between the element i and effective
centroid.
206
Chapter 6
Figure 6.3(a) shows a prismatic elastic pin-ended slender column, which is slightly
z, measured from the bottom, the lateral displacement due to out-of-straightness, before
loading, is wo. As the column is loaded, it deflects further and the additional lateral
expression can then be derived for the net lateral deflection d at mid-height under axial
load P, following the procedure suggested by Allen and Bulson (1980). The bending
moment in the loaded column at any height z is M = Pw. The bending moment is also
⎛1 1 ⎞
M = EI ⎜⎜ − ⎟⎟ (6.5)
⎝ R R o ⎠
where R and Ro are the radii of curvature for the loaded and unloaded columns,
1 d 2w 1 d 2 wo
=− 2 , =− (6.6)
R dz Ro dz 2
d 2w d 2 wo P
+ µ 2w = where µ2 = (6.7)
dz 2 dz 2 EI
For any imperfect shape of the unloaded column, the column’s profile can be represented
∞ ⎡ iπ z ⎤
w o = ∑ ⎢ a i sin (6.8)
i =1 ⎣ L ⎥⎦
207
Chapter 6
where L is the length of column. The amplitudes a i are known or can be measured.
∞ ⎡ iπ z ⎤
w = ∑ ⎢ a i sin (6.9)
i =1 ⎣ L ⎥⎦
in which the amplitudes a i are to be found. Substitution for wo and w from Equations 6.8
∞ ⎡ ⎛ i 2π 2 ⎞ iπ z ⎤ iπ z ⎤ ∞ ⎡ ⎛ i 2π 2 ⎞ iπ z ⎤
∞
⎡
− ∑ ⎢ai ⎜ ⎟ sin ⎥ + µ 2
∑ ⎢ a sin = − ∑ ⎢ ai ⎜ ⎟ sin ⎥ (6.10)
⎜ L2 ⎟ L ⎥⎦ ⎜ L2 ⎟
i
⎢
i =1 ⎣ ⎝ ⎠ L ⎦⎥ i =1 ⎣ ⎢
i =1 ⎣ ⎝ ⎠ L ⎦⎥
⎛ i 2π 2 ⎞ iπ z iπ z ⎛ i 2π 2 ⎞ iπ z
− a i ⎜⎜ 2 ⎟ sin
⎟ + µ 2
a i sin = − ai ⎜
⎜ L2
⎟ sin
⎟ (6.11)
⎝ L ⎠ L L ⎝ ⎠ L
If the previous equation is satisfied for all values of i, then Equation 6.10 is automatically
ai ai i 2 π 2 EI
ai = , or ai = where Pi = (6.12)
(
1 − µ 2 L2 i 2 π 2 ) 1 − ( P Pi ) L2
The effect of the load P is to increase the amplitude of the ith term of the original Fourier
approaches Pi. Provided that the shape of the unloaded column is known and can be
broken down into its Fourier components ( a i ), then the Fourier components for the
loaded column ( a i ) can be found from Equation 6.12 and the total deformation of the
loaded column can be found from Equation 6.9. Assuming that the load is increased
208
Chapter 6
steadily from zero, as it approaches the first critical load (P1 or Pcr), the amplitude of the
first mode becomes very large, larger than all the other amplitudes, which can be
close to Euler buckling load, the lateral deflection at any point along the column’s axis
a1 ⎛π z ⎞
w= sin ⎜⎜ ⎟⎟ (6.13)
1 − ( P Pcr ) ⎝ L ⎠
Due to the approximation stated above, Equation 6.13 provides acceptable deflections up
At mid-height (i.e. z = L/2) the total lateral deflection w takes the following form:
e'
w= (6.14)
1 − ( P Pcr )
where e' is the imperfection at mid-height. The net lateral displacement at mid-height δ is
An expression that relates the net lateral displacement δ at mid-height of the column to
⎡ 1 ⎤
δ = e'⎢ − 1⎥ (6.15)
⎣ 1 − ( P Pcr ) ⎦
π 2 EI
Pcr = (6.16)
( kL ) 2
209
Chapter 6
and moment of inertia I of the column’s cross section. It should be noted that the
effective length factor k in Equation 6.16, which accounts for the column’s boundary
It is important to note that Equations 6.15 and 6.16 assume linear elastic behaviour of the
material and that the residual stresses and the bonded FRP are not accounted for in these
expressions. In the following sections, methods are proposed to account for residual
In order to account for gradual yielding of different parts of the cross section under the
applied loads, which are essentially the axial load and the associated bending moment
induced from the P-δ effect, the concept of “effective moment of inertia” is incorporated
in this analysis (Salmon and Johnson, 1980). The location of the effective centroid of the
section is first determined, using the first moment of areas after discounting the yielded
effective centroid of the cross section (Figure 6.2), to the flexural rigidity ( EI ) s i is the
product of the tangent modulus and the element’s moment of inertia, as follows:
( EI ) s i = E t As i y i2 (6.17)
where Et is the tangent modulus of steel. If the idealized elastic-plastic stress-strain curve
with Young’s modulus Es is used, then: for |Fsi + rs| < Fy, Et = Es and for |Fsi + rs| ≥ Fy , Et =
210
Chapter 6
0. This indicates that the flexural rigidity of the yielded parts becomes zero. Therefore,
the stress level in each steel element must be checked for yielding at each load level to
determine whether the area of the element will be included in the effective bending
stiffness or not. Consequently, the effective bending stiffness ( EI ) s eff of the entire steel
( EI ) s eff = Es . ∑
elastic steel
(A si y i2 ) (6.18)
The effective moment of inertia I s eff for the section can then be introduced in terms of
I s eff = ∑
elastic steel
(A si yi2 ) (6.19)
First yielding will typically occur at the inner side of the buckled column, at mid-height.
As the axial load and corresponding lateral deflection increase, yielding spreads within
the cross-section and also in the longitudinal direction of the column, as shown in Figure
6.4(a). The spreading of yielding indicates that the effective moment of inertia I s eff
varies from one section to the other within the yielded length and also varies with the
applied load. The length of the partially yielded part of the column in the middle zone
depends on many factors, including the slenderness ratio, out-of-straightness profile, and
level of residual stresses. In the current model, an average value of 0.40 L for the length
of the partially yielded portion of the column is assumed. This assumption is made based
on the results of an independent nonlinear finite element analysis (Model 2) for columns
211
Chapter 6
It has been established that the effective moment of inertia of the section varies along the
length as a result of yielding, in spite of the prismatic geometry of the section. Therefore,
a more general expression for the Euler buckling load is needed in lieu of the
order to account for the variable cross sectional inertia along the length of the column, the
finite-difference method is used (Ghali and Neville, 1989), where the column is divided
into a number of segments of equal length ∆L, as shown in Figure 6.4(a), and the
equivalent concentrated elastic loads at each of the m internal nodes can be obtained. The
variation of moment of inertia of the steel section within the middle partially yielded zone
(0.4L) is assumed to follow a parabolic curve with minimum and maximum values
section. A series of simultaneous equations representing the elastic load at each node of
the internal m nodes are then written in the following matrix form:
Pcr ∆L
[ A ]m x m {δ }m x 1 = [ B ]m x m [C ]m x m {δ }m x 1 (6.20)
12
⎡ 2 −1 ⎤ ⎡10 1 ⎤
⎢ −1 2 −1 ⎥ ⎢ 1 10 1 ⎥
⎢ ⎥ ⎢ ⎥
[ A] = 1 ⎢ ... ... ... ⎥, [ B ] = ⎢ ... ... ... ⎥ ,
∆L ⎢ ⎥ ⎢ ⎥
⎢ − 1 2 − 1⎥ ⎢ 1 10 1 ⎥
⎢⎣ − 1 2 ⎥⎦ ⎢⎣ 1 10 ⎥⎦
⎡ (1 EI 1 ) ⎤
⎢ (1 EI 2 ) ⎥
[C ] = ⎢ ⎥ , and ∆L =
L
(6.21)
⎢ ... ⎥ m +1
⎢
⎣ (1 EI m ) ⎥⎦
212
Chapter 6
[ H ] {δ } = γ {δ } (6.22)
12
and γ = (6.24)
Pcr ∆L
used to satisfy Equation 6.22, by assuming a reasonable eigenvector {δ} in the left-hand
side and comparing both sides until the equation is satisfied. As a starting point, the
eigenvector {δ} was assumed to follow a second degree parabola with its apex at the
column’s mid-height. The buckling load Pcr can then be calculated from the largest
For HSS sections with FRP layers, the transformed effective moment of inertia
It should be used in lieu of I s within the middle 0.4L zone and the transformed
eff eff
gross moment of inertia I t should be used in lieu of I s outside the 0.4L zone. I t is
g g eff
⎡Ef ⎤
= I s eff + ∑ ⎢ ⎥
i
It I (6.25)
FRP⎢ ⎥
fi
eff Es
⎣ ⎦
213
Chapter 6
where I fi
( i
)
= A f y i2 . Ifi, Efi and Afi are the moment of inertia, Young’s modulus, and
the area of intact FRP element i, respectively. I t is calculated using the following
g
Equation:
⎡Ef ⎤
I t g = I s g + ∑ ⎢ i I fi ⎥ (6.26)
FRP ⎣ E s ⎦
The lateral displacement of the column can now be calculated at any point along the
In order to establish the appropriate number of segments for the finite difference method,
a convergence study was carried out using 1, 5, 10, 15, and 20 segments along the entire
length L of one of the experimentally tested columns, specimen 9, and the full load-lateral
displacement response was predicted for each case. Details of the full procedure of
prediction are given later. It should be noted that treating the column as one segment can
of inertia ( I t eff ) along the entire length of the column. Figure 6.4(b) shows the predicted
the experimental response. Figure 6.4(c) shows the variation of the peak load with
number of segments. The figures show that convergence occurred when using 15 or more
segments as a very similar behaviour was observed when using 15 and 20 segments,
which also showed good agreement with the experimental response. As such, it was
214
Chapter 6
The axial displacement ∆ is the sum of two components, referred to as ∆a and ∆b, as
∆ = ∆ a + ∆b (6.27)
where ∆a and ∆b are the displacements due to axial shortening and curvature from the P-δ
PL
∆a = (6.28)
Es At
⎡Ef ⎤
At = As + ∑ ⎢ i A fi ⎥ (6.29)
FRP ⎣ E s ⎦
∆b = L − ∆ a − S (6.30)
where S is the chord length of the deformed column [Figure 6.3(b)]. ∆b is calculated
based on a sine curve of an arc length (L-∆a) and amplitude (δ +e’), using the following
equation:
S
L − ∆ a = ∫ 1 + w ' 2 dz (6.31)
0
where w’ is the first derivative of the lateral displacement function w(z) given by
Equation 6.13. Since the integration limit S is unknown, Equation 6.31 was solved
215
Chapter 6
In order to illustrate and verify this simplified approach, the load axial displacement
components ∆a and ∆b, as shown in Figure 6.5. The figure shows that the contribution of
the ‘curvature’ component ∆b is only significant near and after the peak load, when
overall buckling occurs, whereas the axial shortening component ∆a is dominant before
excessive buckling.
In the proposed fibre model, the strain values in the steel cross section are incrementally
increased, until the section reaches its full plastic capacity in the case of bare steel
column, as shown in Figure 6.2. Elements with compressive residual stresses would
typically yield before elements with tensile residual stresses. Eventually, all elements
yield and the effective moment of inertia of the steel cross section becomes zero, based
on Equation 6.19. Consequently, a value for the lateral displacement δ can no longer be
For steel columns with bonded FRP material, a complex failure criterion involving
localized debonding associated with local buckling and crushing after the occurrence of
CFRP is provided (i.e. in columns that are sufficiently long), crushing of CFRP becomes
the dominant failure mode. In the experimental phase of the current study, crushing of
CFRP without debonding was observed in the case of column sets 6 (CFRP plates of type
216
Chapter 6
relatively shorter column sets 2 and 4 (CFRP plates of type C5) at average strains of
0.161 and 0.226 percent, respectively. Debonding associated with crushing occurred in
column sets 8 to 11(CFRP sheets of type C3) at an average strain of 0.133 percent, which
was actually independent of the number of layers. It is noted that for CFRP plates the
scatter in failure strain values increases as slenderness ratio gets higher. This may be
attributed to the larger bending associated with the axial loads in larger slenderness ratios,
which introduces a larger strain gradient through the thickness of CFRP plates. The
ultimate tensile strains (εult), based on tension coupon tests, of the two CFRP types C3
and C5 are 2.22 and 4.72 percent, respectively. The compressive strains of CFRP at
failure (εf cu), in the cases of either crushing or debonding, have been normalized with
respect to their respective ultimate tensile strains, and plotted versus slenderness ratio in
debonding and crushing of CFRP occurring at a slenderness ratio of 76. The following
expressions may be used to calculate the compressive strain of CFRP at failure (εf cu), as a
function of kL/r:
⎛ ε f cu ⎞
⎜ ⎟ = 7.8 x 10 − 3 kL for
kL
≤ 76
⎜εf ⎟ r r
⎝ tu ⎠
(6.32)
⎛ ε f cu ⎞ kL
⎜ ⎟ = 0.59 for > 76
⎜εf ⎟ r
⎝ tu ⎠
It is noted that these limiting strain expressions based on Figure 6.6 have been established
based on relatively few data points and two different types of CFRP. Further research
may be needed to enhance the database in Figure 6.6 and to examine the applicability of
these expressions to other types of CFRP or when CFRP is bonded to HSS sections of
217
Chapter 6
different (b/t) ratios, where secondary local buckling may or may not occur after overall
buckling.
Once CFRP fails in compression, it is eliminated from the cross section at mid-height
cross section. A gradual change in the values of the effective moment of inertia,
following a second degree polynomial, is then defined with a value of the reduced I t at
eff
mid-height and a value of I t g at both ends of the 0.4L middle zone. This is particularly
accurate in case of local crushing, whereas in case of debonding, the CFRP should have
been completely removed along the entire height of the column. This is ignored in this
model, since the presence of CFRP outside the middle zone is likely to provide
insignificant contribution.
After excessive overall buckling, FRP on the outer surface could be subjected to some
tensile strains, which are well below the ultimate tensile strain values.
In order to obtain the full load-lateral displacement (P-δ) response, the procedure of using
1. Assume a value of the extreme compressive strain ε and a neutral axis depth c (Figure
6.2). The full strain gradient across the section at mid-height is then defined.
218
Chapter 6
2. For each steel element of the cross section at mid-height, calculate its strain εsi,
(Equation 6.1), its stress Fsi = Es εsi and add the residual stress Frs (Equation 6.2).
Compare the total stress to the yield stress Fy to check for yielding. If F si + rs ≥ F y ,
3. Calculate the strain εfi (Equation 6.1 also), and the corresponding stress Ffi = Ef εfi for
each FRP element. Compare the strain to the ultimate compressive value εfcu (Equation
4. Calculate the axial load P and bending moment M for the entire section, for the
5. Calculate the eccentricity e = M/P induced by the non uniform stress distribution,
6. Calculate the transformed effective moment of inertia I teff , excluding both the yielded
steel elements and failed FRP elements (Equation 6.25). This is used to calculate the
critical buckling load Pcr (Equations 6.20 to 6.24), which is then used to calculate the
6.15).
7. Compare the eccentricity e calculated in step 5 with (δ + e'). If the two values are
different, assume a new value of neutral axis depth c and repeat steps 2 to 6 until the
values are equal. The load P calculated in step 4 and displacement δ calculated in step
8. Use a larger value of strain ε in step 1 and repeat the process until the complete P-δ
response is established.
219
Chapter 6
In order to generate the full load-axial displacement P-∆ response, for a given axial load
P and corresponding lateral deflection δ (obtained earlier), the following procedure can
be followed:
2. For a given δ, establish a deformed sine curve of the column with mid-height
amplitude of (δ + e’) and an arc length of (L-∆a), and calculate the chord length S of
5. Repeat the previous steps for each P and δ, until the complete P-∆ response is
established.
The fibre model developed has several significant features, namely, accounting for the P-
features, the load-lateral displacement response of one of the test specimens, specimen 9
of Phase I of the current study has been predicted based on the estimated initial out-of-
straightness at mid-height (e’ = 7.04 mm) and residual stresses (Frs = 0.33 Fy). The
predictions are executed for four different cases. In case 1, Equations 6.15 and 6.16 have
been used in their original elastic form (i.e. ignoring steel yielding and residual stresses).
In case 2, plasticity of steel is considered, however, residual stresses are ignored. In case
220
Chapter 6
3, both steel plasticity and residual stresses are accounted for but the failure criterion of
FRP in compression is not applied. In case 4, all the features of the model presented
earlier are applied. Figure 6.7 shows the experimental response and the analytical
responses for the four cases. The figure clearly shows that ignoring the plasticity of steel
(case 1) would grossly overestimate the axial strength as the ultimate load approaches the
Euler elastic buckling load. Ignoring residual stresses (case 2) would overestimate the
load at which the behaviour departs from the elastic range. Also, by assuming that FRP
remains intact in compression throughout the full response (case 3), the ultimate load is
somewhat overestimated. It is then clear that case 4 represents the most accurate
prediction out of the four cases, using the full capabilities of the model. Therefore, it is
used later for the predictions of the behaviour of all cases in section 6.4 and also in the
The finite element analysis program ANSYS (revision 10) was used to model the non-
linear behaviour of the pin-ended HSS slender steel columns strengthened by CFRP
material. The primary objectives of the FEM are to introduce an alternative tool that may
be used in analysis or design and also for verifying the fibre model (Model 1). Both
models will also be verified using experimental results in section 6.4 and will be
The finite element simulation consisted of two stages. In the first stage, an eigenvalue
221
Chapter 6
modeling of the entire cross section (i.e. not utilizing the symmetry) of the HSS2 column
(89 x 89 x 3.2 mm), to establish the probable buckling modes of the column for different
lengths (i.e. slenderness ratios of 5, 30, and 60). The analysis showed that the columns
with slenderness ratios of 5 and 30 experienced local buckling, whereas the column with
behaviour was also experimentally demonstrated for slenderness ratios of 4 (set 12) and
68 (specimen 7). In the second stage, a non-linear analysis was performed on slender
predicted buckling shape (geometric nonlinearity) established through the first stage of
analysis. In this stage, the columns were loaded to failure to predict their full responses
and ultimate loads. Also in this stage, the analysis incorporated the material non-linearity
(plasticity of steel) and residual stresses. The centerline dimensions of the cross-sections
and the base metal thickness were used in the geometric modeling, based on the
various aspects of the finite element model such as element type, mesh density, boundary
As mentioned earlier, the first stage of the numerical simulation was essentially a linear
elastic analysis of the control column, in which the stiffness of the structure remained
unchanged. As such, only the values of Young’s modulus (200 GPa) and Poisson’s ratio
of steel (0.30) were defined. On the other hand, the second stage of the numerical
simulation comprised a non-linear analysis, in which the stiffness of the structure changes
222
Chapter 6
as it deforms. The steel non-linearity (plasticity) was accounted for in the FEM by
specifying a bi-linear isotropic hardening model, as shown in Figure 6.9. The tangent
modulus for the steel was assumed equal to 0.5 percent of its elastic modulus as
suggested by Bruneau et al. (1998). For the FRP materials, unidirectional elastic
properties were assigned, namely, Young’s moduli of 20 GPa and 230 GPa for GFRP
type (G) and CFRP type (C3), respectively, as the model was used to predict the
behaviour of specimens 7 to 11. The ultimate compressive strain value of the CFRP type
(C3) was limited to 60 percent of its ultimate tensile strain to account for crushing of
reached, the stress level in CFRP is locked, however, the model is incapable of
eliminating the FRP material from the global stiffness matrix beyond this point, which
An eight-node quadrilateral layered shell element (SHELL91) was used for the steel
section in this model. The element configurations as well as its coordinate system are
shown in Figure 6.10(a). Each node has six degrees of freedom, namely, three
translations (Ux, Uy, and Uz) and three rotations (Rx, Ry, and Rz). The multiple layers of
the element were utilized to account for the residual stress distribution through the steel
wall thickness, as will be discussed later. When FRP sheets were used, the FRP was
Figure 6.10(b). This is considered reasonable because of the small flexural rigidity of the
thin FRP layers. Each node has three degrees of freedom, namely, translations in the
223
Chapter 6
nodal x, y, and z directions (Ux, Uy, and Uz). Perfect bond between steel and FRP sheets
was assumed by defining one node for both the SHELL91 and the LINK8 elements
having the same coordinates. This assumption is quite reasonable when thin FRP sheets
were used as no signs of debonding were observed experimentally, except at the very
end, well beyond the peak load when local debonding and crushing occurred as a result of
One quarter of the specimen was modeled, as shown in Figure 6.11(a), by taking
advantage of the double symmetry of the column. This symmetry was simulated by
introducing two planes of symmetry, one vertical plane in the longitudinal direction along
the full length and another horizontal plane in the transverse direction at mid-height of
the column. The final mesh configuration of the model was established after a mesh
different mesh densities, namely, mesh 1, mesh 2, and mesh 3, were first carried out on
the control steel specimen, as shown in Figure 6.11(b). Different sizes of the elements in
the HSS flanges and their curved corners are also shown in Figure 6.11(b). The number
of elements varied from 1050 elements in the first mesh to 4730 elements in the third
mesh. The predicted maximum axial load for each mesh configuration as well as their
mathematical average is plotted in Figure 6.11(b). The figure shows almost identical
results with minor changes in the axial load capacities when refining the model beyond
mesh 1. However, the computer run-time dramatically increases with refining the mesh
size. As such, mesh 1 was deemed sufficient and was used in all the analyses that
followed.
224
Chapter 6
In order to model the hinged end condition of the columns, shown in Figure 3.16, a rigid
end plate was simulated with controlled degrees of freedom. The translational degrees of
freedom in the transverse direction (Ux, and Uz) along the middle line of the rigid plate
were restrained, whereas the translational degrees of freedom in the longitudinal direction
(Uy) along the same line were released. The rotational degrees of freedom of the entire
plate in all directions (Rx, Ry, and Rz) were released. Loading was modeled by two lines
of point loads, spaced by a distance equal to the width of the hinged end of the test setup
introduced in the FEM to initiate the overall buckling mode of failure, indicated by the
buckling analysis (first stage analysis). The imperfect profile of the unloaded column was
assumed to follow a sine curve with its apex at the mid-height of the column.
In the finite element model, the through-thickness residual stress distribution is idealized
as previously shown in Figure 6.1. In order to model the residual stress pattern, four
layers were defined in the multi layer steel shell element (SHELL91). A simplified and
approximate approach was used, in which the residual stress was defined by shifting the
origin of the axes of the stress-strain curve of the steel material of each layer along the
linear part of the curve upwards or downwards, depending on whether the residual stress
225
Chapter 6
is compression or tension. The magnitude of the shift is equal to the residual stress Frs ,
as shown in Figure 6.9. This means that, the origin of the ‘pre-tensioned’ steel at the two
outer layers of the wall thickness are defined as (-εrs , -Frs) and (-0.5εrs , -0.5Frs),
respectively, whereas the origin of the ‘pre-compressed’ steel at the inner two layers of
the wall thickness are defined as (0.5εrs , 0.5Frs) and (εrs , Frs), respectively.
Both models were verified using two independent experimental studies. First, the models
were verified using test results reported by Key and Hancock (1985) on 152 x 152 x 4.9
mm and 203 x 203 x 6.3 mm conventional HSS steel columns. The columns were not
strengthened with FRP materials. The pin-ended columns had kL/r ranging from 66 to 98
and imperfection values e’ ranging from 0.3 mm to 1.75 mm, as shown in Table 6.2. The
reported Fy and Frs by Key and and Hancock (1985) were 350 MPa and 200 MPa,
respectively. The average axial stresses, based on the applied load divided by the cross
sectional area of the column, versus the normalized lateral displacement (δ/L) responses
for both the experiments and the two models are plotted in Figure 6.12 and 6.13. The
figures show that both models provide reasonable agreement with the experimental
The models were then verified using the experimental results of the current study. Model
1 was verified using all columns’ sets (1 to 11); whereas Model 2 was verified using
columns’ sets 7 to 11. The initial imperfections used in the predictions for column sets (1
226
Chapter 6
to 6) were obtained using two methods, namely, the laser sensor (section 3.3.2), before
CFRP installation, and the measured strain gradient through the cross section, under
loading (section 4.3.1). For columns sets’ 7 to 11, laser profiling could not be used, as
indicated earlier. A comparative study was performed to investigate the effect of each
method on the predicted axial strength. Table 6.1 shows the predicted axial strengths of
column sets 1, 2, 5, and 6, using Model 1, for both the minimum and maximum
imperfection values obtained from each method (i.e. the values at mid height and
maximum amplitude, respectively, in laser profiling and the values based on the strains
from both sides, respectively, in the strain gradient method). Sets 1 and 2 represent
unstrengthened and strengthened columns with the minimum slenderness ratio, while sets
5 and 6 represent the same for the maximum slenderness ratio. The table shows that, for
each method, the minimum and maximum imperfections yielded very close results. Also,
obtained in set 6, which represents a maximum difference of 2.8 percent between the four
predicted strengths. This suggests that estimating the imperfection using either method is
reasonable. Therefore, it was decided to use the maximum imperfection values obtained
The full responses of load versus: lateral displacements, axial displacements, and axial
strains on two opposite sides of the columns, have been predicted and compared with the
experimental results. Figure 6.14 to 6.24 show the predicted versus experimental load-
lateral displacements of column sets 1 to 11. Figure 6.25 to 6.35 show the predicted
versus experimental load-axial displacements and Figure 6.36 to 6.46 show the predicted
227
Chapter 6
versus experimental load-axial strains on both side, for the same specimens. A summary
of axial strength and stiffness is also given in Table 6.2. While the models appear to
somewhat overestimate the axial strength and stiffness at very low slenderness ratios
(kL/r = 46) for the small scale columns (sets 1 and 2), they generally showed very good
agreement with the experimental results for all other column sets within a maximum
was in fact not symmetric, and the maximum lateral displacement and failure occurred
near the quarter length point and not at the middle. For this reason, the predicted
maximum axial load, which is based on symmetric buckling, is higher than the
The failure mode predicted by the FEM (Model 2), which is an overall buckling [Figure
6.47(a)] is quite similar to the buckling failure mode observed in the tests [Figure
is shown in Figure 6.47(c). As discussed in Chapter 4, after overall buckling took place,
inward local buckling occurred in the compression flange, whereas outward local
buckling occurred in the two side webs for specimens 7 to 11 with a relatively large b/t
ratio. This deformed shape was revealed after the test by cutting the specimen at mid-
height. Figure 6.47(d, e and f) show that the same pattern of deformation has been
predicted by the FEM, in terms of the displacement contours in both x- and z-directions,
Ux and Uz (i.e. displacements within the cross sectional plane) as well as the nodal
rotations about the longitudinal axis, Ry. The resemblance of deformations in Figure
228
Chapter 6
Columns
One of the objectives of developing the models is to assess the effect of individual
parameters that control the behaviour of CFRP-strengthened slender HSS columns, each
independently. It was shown in Section 4.3.1 that specimens 7 to 11 had different levels
CFRP reinforcement ratio (i.e. number of layers) exclusively, which was the main
a) CFRP reinforcement ratio ρ, which is defined as the ratio of total CFRP to steel
areas Af /As (based on one to five layers of CFRP type C3 bonded on two opposite
sides).
Also, Model 2 was used to predict the strength and stiffness for some of the cases being
A total of 28 HSS columns with the same cross sectional dimensions and material
properties as those used in the tested specimens 7 to 11 of the experimental program are
analyzed. The following identification system was adopted to distinguish the various
229
Chapter 6
cases. As shown in Table 6.3, the first number represents the slenderness ratio of the
column (kL/r), while the second number specifies the geometric out-of-straightness
imperfection as a ratio of (L/e’). These two numbers are followed by the CFRP
reinforcement ratio to quantify the amount of CFRP layers bonded on two opposite sides
of the HSS. Another number describing the residual stress level as a percentage of the
yield stress is also added at the end. For the unstrengthened control columns, the CFRP
imperfection of (Length/500), CFRP reinforcement ratio of 0.43 (i.e. five layers of CFRP
type C3), and has a residual stress level of 25 percent of its yield stress. Figure 6.48 to
6.54 show the predicted load-lateral responses for all cases. The effect of each parameter
Figure 6.48 to 6.54 clearly show that bonding longitudinal CFRP sheets to slender steel
columns can indeed increase both their strength and stiffness. For example, the
percentage increases in axial strength of specimens with one, three and five CFRP layers
are 11, 26, and 39 percent for specimens with e’=L/500 and are 5, 9, and 11 percent,
respectively, for specimens with e’=L/2000. These increases in strength are equivalent to
other hand, the percentage increases in axial stiffness are in the order of 13, 31, and 45
percent for specimens strengthened with one, three and five CFRP layers, respectively.
Figure 6.55(b) shows that increasing the reinforcement ratio (ρ = Af /As) decreases the
230
Chapter 6
It is clear that the reinforcement ratio and the value of out-of-straightness have a
combined effect. The results listed in Table 6.3 suggest that the CFRP system is more
effective for columns with higher levels of out-of-straightness, as also shown in Figure
6.55(a and c), particularly for higher reinforcement ratios. Figure 6.55(c) shows the
variation of percentage increase of axial strength with the out-of-straightness values, for
various numbers of CFRP layers. Depending on the number of layers, there is a certain
level of out-of-straightness (for example L/750 for one layer of CFRP), before which, the
this range, both sides of the column are under compression and the CFRP fails at both
and the CFRP crushes only at the inner curved side. Table 6.3 summarizes the axial load
capacities and stiffness of columns with different number of CFRP layers, for various
out-of-straightness values. The table shows that the out-of-straightness has a negligible
effect on the percentage increase of the columns’ axial stiffness. As such, the percentage
As shown in Table 6.3 and Figure 6.55(d), the through-thickness residual stress has a
little effect on the gain in axial strength of CFRP-strengthened HSS columns, and that the
lower the residual stress, the higher the gain in columns’ strength. The values listed in
Table 6.3 indicate that if the value of residual stress increased from 25 to 50 percent of
231
Chapter 6
the yield stress, the maximum load of the control specimen reduces from 297 kN to 290
kN, which represents only a 2.4 percent reduction. This reduction reaches 5.6 percent for
Figure 6.55(d) shows that the slenderness ratio perhaps had the most pronounced effect
ratio, the strength gain increases substantially as the slenderness ratio is increased,
particularly for higher CFRP reinforcement ratios. Table 6.3 indicates that the percentage
increase in axial stiffness due to increasing the number of CFRP layers is slightly affected
by slenderness ratio.
As the two developed models were completely independent, it was not expected that they
will both yield identical results. The difference in the solution approach between both
models as well as the inherent differences in the assumptions in each model explain the
small differences in the results. In order to further assess the difference in results, the first
500-43-25” have also been analysed using Model 2. Figure 6.56 and 6.57 show a
comparison between the predicted axial strength and stiffness using both models,
respectively. Also, the predictions for the experimental specimens 7 to 11 are compared
in Figure 6.56 and 6.57. The figures show comparable results for both models with
Model 1 giving slightly higher axial strengths than Model 2 for most of the cases. The
232
Chapter 6
maximum difference in strength was seven percent for specimens with CFRP-
reinforcement ratios of 9, 25, and 43 percent. On the other hand, Model 2 gave slightly
higher axial stiffness than Model 1 for most of the cases. The maximum difference in
233
Chapter 6
(Ppred)avg
Standard
(kN)
Set deviation
emin Ppred emax Ppred emin Ppred emax Ppre1
(kN)
(mm) (kN) (mm) (kN) (mm) (kN) (mm) (kN)
Set 1 0.14 247 0.25 246 0.46 247 0.78 245 246 0.96
Set 2 0.06 261 0.26 254 0.22 253 0.85 253 255 3.86
Set 5 0.53 107 0.96 106 0.37 108 0.57 107 107 0.82
Set 6 0.28 185 0.84 181 0.17 186 1.22 175 182 4.99
234
Chapter 6
Table 6.2 Comparison between experimental and predicted results using Models 1 and 2.
Axial strength, P Axial stiffness, k
Model 1 Model 2 Model 1 Model 2
identification
Specimen
Experiment
Experiment
Ppred1 (kN)
Ppred2 (kN)
(kN/mm)
(kN/mm)
(Ppred1 /
(Ppred2 /
(kpred1 /
(kpred2 /
kpred1
kpred2
Pexp)
Pexp)
kexp)
kexp)
Source
Specimen e’
cross section kL/r (mm)
i 68 0.30 898 831 0.93 933 1.04 N/A N/A N/A
Hancock
(1985)
Set 2 46 0.26 192 254 1.32 N/A N/A 118 184 1.56 N/A
Set 3 70 0.36 148 175 1.18 N/A N/A 86 88 1.02 N/A
44 x 44 x 3.2
Set 4 70 0.32 200 236 1.18 N/A N/A 100 121 1.21 N/A
Set 5 93 0.96 103 106 1.03 N/A N/A 68 65 0.96 N/A
Set 6 93 0.84 175 181 1.03 N/A N/A 80 88 1.10 N/A
Set 7 68 6.60 295 268 0.91 267 0.91 90 88 0.98 85 0.94
Set 8 68 0.92 355 360 1.01 359 1.01 89 102 1.15 98 1.10
89 x 89 x 3.2 Set 9 68 7.04 335 343 1.02 326 0.97 88 109 1.24 109 1.24
Set 10 68 2.04 332 407 1.23 412 1.24 120 130 1.08 126 1.05
Set 11 68 5.00 362 378 1.04 383 1.06 110 135 1.23 135 1.23
Average 1.08 1.01 1.16 1.11
Standard deviation 0.14 0.12 0.17 0.13
235
Chapter 6
k
1
δ Lateral displacement
Axial displacement
236
Chapter 6
- 0.5Frs
+0.5Frs
A4 (1 x n)
A3 (12x12)
t/3
t/3
t/3
A1 (1 x 12)
+Frs
-Frs
A2 (12 x 80)
A5 (n x 1)
Effective Fully
Residual First Partially
centroid (c) x plastic
stress yielding yielded
element i c εi
yi Fy Fy Fy
e h + =
P or or
yc
ε
x
Strain distribution Stress distribution
Fy
Figure 6.2 Stress and strain distributions within the cross section of slender column
at mid-height.
237
Chapter 6
Loaded column
e’ δ L
L - ∆a
δ+e’
+
L
S
w
wo
L/2
z
z
400
350
∆L = L/15
Elastic
Itg 300
0.3 L
Load (kN)
250
Experiment
200
1 Segment
150 5 Segments
Specimen 9
100 10 Segments
Partially yielded
15 Segments
e’ = 7.04 mm
Iteff 50 Frs = 0.33 Fy
0.4 L
20 Segments
0
0 5 10 15 20 25 30 35 40 45
Lateral displacement δ (mm)
(b) Effect of number of segments on load-
lateral displacement response
Elastic
0.3 L
350
340
Peak load (kN)
330 convergence
320
310
(a) Variation of inertia 300
along column’s height 290
using 15 segments 280
0 5 10 15 20 25
Number of segments
(c) Variation of peak load with number of segments
Figure 6.4 Summary of the finite difference model and convergence study.
238
Chapter 6
300
∆a
Experiment
250
∆b Total (∆a + ∆b)
200
Load (kN)
150
100
50
Control Specimen 7
(e’ = 6.60 mm)
0
0 1 2 3 4 5 6
Axial displacement (mm)
1
CFRP plates
ultimate compressive strain to tensile strain
0.9
7 to 11
Sets
CFRP sheets
0.8 Average
0.7
Crushing
0.6
Best fit to average values
0.5
0.4
Set 6
ing
Set 4
0.3 nd
bo
kL/r = 75.6
De
Set 2
0.2
0.1
0
0 20 40 60 80 100 120
Slenderness ratio (kL/r)
Figure 6.6 Variation of ultimate compressive-to-tensile strain ratio of CFRP at
failure with slenderness ratio.
239
Chapter 6
700
Euler Load = 695 kN
650
Specimen 9 (e’ = 7.04 mm)
600
550
Case 1
500
450 Case 3
Load (kN)
400
Case 2
350
300
Case 4
250 Experiment
Failure of FRP
in compression
200
150 Case 1 Linear elastic HSS section (No yielding + No residual stress + No CFRP failure)
Case 2 Elasto-plastic HSS section (No residual stress + No CFRP failure)
100 Case 3 Same as case 2 + residual stresses (No CFRP failure)
50 Case 4 Same as case 3 + CFRP fails in compression
0
0 5 10 15 20 25 30 35 40 45
Lateral displacement δ (mm)
kL/r = 60
kL/r = 30
kL/r = 5
240
Chapter 6
900
GFRP
Stress (MPa)
600 CFRP (type C3)
Steel
(bi-linear) (a) Materials stress-strain curves
300
+ 0.5Frs
0
- 0.5Frs
0 15 30 45
-3
Strain x 10 (mm/mm)
t/3
Fy t/6
t/6
Stress (MPa)
t/3
+ Frs
- Frs
Used for the 2 inside layers
Frs - Frs
- 0.5Frs
+ 0.5Frs
Frs Strain x 10-3 (mm/mm) + Frs
y
o
z l
k
p
n
m
i j x
Top
Layer n z x
y
k
Layer 2
Layer 1
Bottom i
241
Chapter 6
291.7
89 x 89 x 3.2 mm
291.6
L = 2380 mm, e’=L/500
291
290.9
290.8
mesh 1 mesh 2 mesh 3
(a) One quarter of the
specimen (b) Mesh configurations
350
Exp. (kL/r=68)
Model 1
300
250 Model 2
Average stress (MPa)
200
Model 1
150
Model 2
Exp. (kL/r=98)
100
50
0
0 100 200 300 400 500 600
δ /L (x 10-5)
Figure 6.12 Verification of models 1 and 2 using test results on HSS 203 x 203 x 6.3
mm.
[Key and Hancock, 1985]
242
Chapter 6
350
Model 2
250
Average stress (MPa)
Model 2
200
150
Model 1
100
Exp. (kL/r=96)
50
0
0 100 200 300 400 500 600
δ /L (x 10-5)
Figure 6.13 Verification of models 1 and 2 using test results on HSS 152 x 152 x 4.9
mm.
[Key and Hancock, 1985]
250
kL/r = 46
e’ = 0.25 mm
200 Model 1
Frs = 0.49 Fy
150 P
Load (kN)
Experiment
100
e’ δ
50
0
0 5 10 15 20 25
243
Chapter 6
300
kL/r = 46
e’ = 0.26 mm
250
Frs = 0.49 Fy
Debonding of CFRP Model 1
200
Load (kN)
150 P Experiment
100
e’ δ
50
0
-1 4 9 14 19 24
180
kL/r = 70
160
Model 1 e’ = 0.36 mm
140 Frs = 0.49 Fy
120
P
Load (kN)
100
Experiment
80
60 e’ δ
40
20
0
0 5 10 15 20 25 30 35 40 45 50
244
Chapter 6
300
kL/r = 70
Debonding of CFRP
e’ = 0.32 mm
250
Model 1 Frs = 0.49 Fy
200
Load (kN)
150
P
100 Experiment
e’ δ
50
0
-1 4 9 14 19 24 29 34
120
kL/r = 93
Model 1 e’ = 0.96 mm
100
Frs = 0.49 Fy
80
Experiment
P
Load (kN)
60
40
e’ δ
20
0
0 10 20 30 40 50 60
245
Chapter 6
200
Crushing of CFRP on the kL/r = 93
180 inner compression flange
e’ = 0.84 mm
160 Frs = 0.49 Fy
140
Model 1
120
Load (kN)
100 P
80
Experiment
60
e’ δ
40
20
0
0 10 20 30 40 50 60
450
kL/r = 68
400
e’ = 6.6 mm
350 Frs = 0.33 Fy
Experiment
300
Load (kN)
250
P Model 2
200 Model 1
150
e’ δ
100
50
0
0 5 10 15 20 25 30 35 40
246
Chapter 6
450
kL/r = 68
400
Model 2 e’ = 0.92 mm
Model 1
350 Frs = 0.33 Fy
300
Crushing of CFRP on
Load (kN)
150
e’ δ
100
50
0
0 5 10 15 20 25 30 35 40
450
kL/r = 68
400
e’ = 7.04 mm
350 Experiment Frs = 0.33 Fy
300
Load (kN)
150
e’ δ
100
50
0
0 5 10 15 20 25 30 35 40
247
Chapter 6
450
Model 2 kL/r = 68
400
e’ = 2.04 mm
Model 1
350 Frs = 0.33 Fy
300
Experiment
Successive crushing of
Load (kN)
150
e’ δ
100
50
0
0 5 10 15 20 25 30 35 40
450
Crushing of CFRP kL/r = 68
Model 2 on the side walls
400 e’ = 5.00 mm
200
150
e’ δ
100
50
0
0 5 10 15 20 25 30 35 40
248
Chapter 6
300
kL/r = 46
250 Model 1
e’ = 0.25 mm
Frs = 0.49 Fy
200
Load (kN)
P
150
∆
Experiment
100
50
0
0 12 24 6
3 48 5
10 12
6 7
14
Axial Displacement ∆ (mm)
300
kL/r = 46
Model 1
e’ = 0.26 mm
250
Frs = 0.49 Fy
200
Load (kN)
Experiment
150 P
∆
100
50
0
0 21 42 3
6 4
8 5
10 6
12 7
14
Axial Displacement ∆ (mm)
249
Chapter 6
300
kL/r = 70
250 e’ = 0.36 mm
Frs = 0.49 Fy
200
Model 1
Load (kN)
150 P
Experiment
∆
100
50
0
0 21 42 63 4
8 5
10 6
12 7
14
Axial Displacement ∆ (mm)
300
kL/r = 70
e’ = 0.32 mm
250
Frs = 0.49 Fy
Model 1
200
Load (kN)
150 P
∆
100 Experiment
50
0
0 1
2 42 63 4
8 5
10 6
12 7
14
Axial Displacement ∆ (mm)
250
Chapter 6
120
P
100
∆
Model 1
80
Load (kN)
Experiment
60
40
kL/r = 93
20
e’ = 0.96 mm
Frs = 0.49 Fy
0
0 1 2 3 4 5 6 7
Axial Displacement ∆ (mm)
200
P kL/r = 93
180 e’ = 0.84 mm
∆
160 Frs = 0.49 Fy
140
Model 1
120
Load (kN)
100
80
Experiment
60
40
20
0
0 2 4 6 8 10
Axial Displacement ∆ (mm)
251
Chapter 6
450
400 P
350 ∆
Experiment
Model 1
300
Load (kN)
250
Model 2
200
150
100 kL/r = 68
e’ = 6.6 mm
50
Frs = 0.33 Fy
0
0 1 2 3 4 5 6
Axial Displacement ∆ (mm)
450
P
400
350 ∆
300
Model 2
Load (kN)
250 Model 1
200 Experiment
150
100 kL/r = 68
e’ = 0.92 mm
50
Frs = 0.33 Fy
0
0 1 2 3 4 5 6
Axial Displacement ∆ (mm)
252
Chapter 6
450
P
400
∆ Model 1
350
Experiment
300
250
Load (kN)
Model 2
200
150
100 kL/r = 68
e’ = 7.04 mm
50
Frs = 0.33 Fy
0
0 1 2 3 4 5 6
Axial Displacement ∆ (mm)
450
P Crushing of CFRP on
the concave side Model 2
400
∆
350 Experiment
300
Model 1
Load (kN)
250
200
150
100 kL/r = 68
e’ = 2.04 mm
50
Frs = 0.33 Fy
0
0 1 2 3 4 5 6
253
Chapter 6
450
P
400 Crushing of CFRP on Model 2
∆ the concave side
350
Experiment
300
Model 1
Load (kN)
250
200
150
100 kL/r = 68
e’ = 5.00 mm
50
Frs = 0.33 Fy
0
0 1 2 3 4 5 6
Axial Displacement ∆ (mm)
300
Experiment
200
Experiment
Load (kN)
150
100
S1 S2
kL/r = 46
50 e’ = 0.25 mm
Frs = 0.49 Fy
0
-8 -6 -4 -2 0 2
Axial strain (x 10-3)
254
Chapter 6
300
kL/r = 46
Model 1
250 e’ = 0.26 mm
Frs = 0.49 Fy
200
Experiment
Load (kN)
150
100
S1 S2
50
0
-8 -6 -4 -2 0 2
Axial strain (x 10-3)
300
kL/r = 70
e’ = 0.36 mm
250
Frs = 0.49 Fy
200
Model 1 Model 1
Load (kN)
Experiment
150
Experiment
100
S1 S2
50
0
-8 -6 -4 -2 0 2
Axial strain (x 10-3)
255
Chapter 6
300
kL/r = 70 Model 1
e’ = 0.32 mm
250
Frs = 0.49 Fy
200
Load (kN)
150 Experiment
100
S1 S2
50
0
-8 -6 -4 -2 0 2
Axial strain (x 10-3)
120
Model 1
Model 1
100
Experiment
80 Experiment
Load (kN)
60
40
S1 S2
kL/r = 93
20
e’ = 0.96 mm
Frs = 0.49 Fy
0
-4 -3 -2 -1 0 1 2
Axial strain (x 10-3)
256
Chapter 6
200
Experiment
Model 1
180
160
140 Experiment
120 Model 1
Load (kN)
100
80
60
kL/r = 93
40
S1 S2
e’ = 0.84 mm
20
Frs = 0.49 Fy
0
-4 -3 -2 -1 0 1 2
450
400
350
Experiment
Experiment
300
250
Load (kN)
Model 2 Model 1
200 Model 2
Model 1
150
100 kL/r = 68
S1 S2
e’ = 6.60 mm
50
Frs = 0.33 Fy
0
-2.5 -2 -1.5 -1 -0.5 0 0.5 1
Axial strain (x 10-3)
257
Chapter 6
450
400
350
Experiment
300
Load (kN)
200 Model 1
150 Experiment
100 kL/r = 68
S1 S2
e’ = 0.92 mm
50
Frs = 0.33 Fy
0
-2.5 -2 -1.5 -1 -0.5 0 0.5 1
Axial strain (x 10-3)
450
400
350
Experiment
300 Experiment
Model 1
Model 1
Load (kN)
250
Model 2
200 Model 2
150
100 kL/r = 68
S1 S2
e’ = 6.60 mm
50
Frs = 0.33 Fy
0
-2.5 -2 -1.5 -1 -0.5 0 0.5 1
Axial strain (x 10-3)
258
Chapter 6
450
Model 2
400 Model 1
350
Experiment
300
Model 2
250
Load (kN)
Model 1
Experiment
200
150
kL/r = 68
100 S1 S2
e’ = 2.04 mm
50 Frs = 0.33 Fy
0
-2.5 -2 -1.5 -1 -0.5 0 0.5 1
Axial strain (x 10-3)
Figure 6.45 Measured and predicted load-axial strain responses of specimen 10.
450
400 Model 2
Model 1
350 Experiment
Model 1
300 Model 2
Load (kN)
250
200 Experiment
150
kL/r = 68
100 S2
S1
e’ = 5.00 mm
50 Frs = 0.33 Fy
0
-2.5 -2 -1.5 -1 -0.5 0 0.5 1
Axial strain (x 10-3)
Figure 6.46 Measured and predicted load-axial strain responses of specimen 11.
259
Chapter 6
Min. Ux
x
y
z
Values are in mm
Max. Uz x
y
Undeformed
z
shape
Values are in mm
Max. Ry
x
y x
y
z z
Min. Ry
Figure 6.47 Comparison between the deformed shapes in experiments and FEM
(Model 2).
260
Chapter 6
450
ρ = 43 % (5 layers)
400
ρ = 25 % (3 layers)
350 ρ = 9 % (1 layer)
300
Load (kN)
250 Control
(no CFRP)
200
150
CFRP
100
50
kL/r = 68 e’ = L/500 Frs = 0.25 Fy
0
0 10 20 30 40 50
450
CFRP crushing CFRP crushing
in the concave in the convex
400 side ρ = 43 % (5 layers)
side
ρ = 25 % (3 layers)
350 ρ = 9 % (1 layer)
300
Load (kN)
250 Control
(no CFRP)
200
150
CFRP
100
50
kL/r = 68 e’ = L/600 Frs = 0.25 Fy
0
0 10 20 30 40 50
Lateral displacement δ (mm)
261
Chapter 6
450
350
300
Load (kN)
250 Control
ρ = 9 % (1 layer) (no CFRP)
200
150
CFRP
100
50
kL/r = 68 e’ = L/750 Frs = 0.25 Fy
0
0 10 20 30 40 50
Lateral displacement δ (mm)
450
400 ρ = 43 % (5 layers)
ρ = 25 % (3 layers)
350
ρ = 9 % (1 layer)
300
Load (kN)
Control
250 (no CFRP)
200
150
CFRP
100
50
kL/r = 68 e’ = L/1000 Frs = 0.25 Fy
0
0 10 20 30 40 50
Lateral displacement δ (mm)
262
Chapter 6
450
ρ = 43 % (5 layers)
400
ρ = 25 % (3 layers)
350
ρ = 9 % (1 layer)
300
Control
(no CFRP)
Load (kN)
250
200
150
CFRP
100
50
kL/r = 68 e’ = L/2000 Frs = 0.25 Fy
0
0 5 10 15 20 25 30 35 40 45 50
450
400 ρ = 43 % (5 layers)
ρ = 25 % (3 layers)
350 ρ = 9 % (1 layer)
300
Load (kN)
250 Control
(no CFRP)
200
150
CFRP
100
50
kL/r = 68 e’ = L/500 Frs = 0.50 Fy
0
0 10 20 30 40 50
Lateral displacement δ (mm)
Figure 6.53 Load-lateral displacement responses for specimens with Frs = 0.50 Fy.
263
Chapter 6
150
125
ρ = 43 % (5 layers)
100 ρ = 25 % (3 layers)
ρ = 9 % (1 layer)
Load (kN)
75
Control
(no CFRP)
50
CFRP
25
Figure 6.54 Load-lateral displacement responses for specimens with kL/r =160.
264
Chapter 6
40 80
Frs = 0.25 Fy Frs = 0.25 Fy
kL/r = 68 kL/r = 68
35 70 0
L/75
0
50
L/
% age Increase in axial strength
00
5 Layers
L/6
20 40
3 Layers
L/7 50
15 30
L/1 00 0
1 Layer
L/2 00 0 L/500
10 20
00
L/2 0
5 Layers
1 Layer
3 Layers
5 10 L/1 000
0 0
0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45
FRP reinforcement ratio ρf = Af /As x 100
40 80
Frs = 0.25 Fy e’ = L/500
kL/r = 68
35 70
ρ = 43% (5 Layers)
%
s . 3 Layers
% age Increase in axial strength
25
% age Increase in axial strength
=
30 60
ss
re
5 Layers
St
25 50
re
0,
16
=
20 40
/r
kL
5%
ρ = 25% (3 Layers) =2
15 ss
30 tre
1 Layer
S
ρ = 9% (1 Layer) s. 0%
8, re =5
6
10 20 /r = es s
kL S tr
es.
L/500
L/600
L/750
6 8, r
r=
L/1000
5 10 kL/
L/2000
0 0
0.0005 0.001 0.0015 0.002 0 5 10 15 20 25 30 35 40 45
265
Axial stiffness (kN/mm) Axial strength (kN)
0
50
100
150
200
250
300
350
400
450
500
0
20
40
60
80
100
120
140
160
88 268
Specimen 7 Specimen 7
85 267
102 360
Specimen 8 Specimen 8
98 359
109 343
Specimen 9 Specimen 9
109 326
130 407
Specimen 10 Specimen 10
126 412
135 378
Specimen 11 Specimen 11
266
135 383
Specimen identification
91 330
68-500-9-25 68-500-9-25
96 307
105 374
68-500-25-25 68-500-25-25
111 346
117 413
68-500-43-25 68-500-43-25
125 383
Model 2
Model 1
Chapter 6
Model 2
Model 1
Chapter 7
Chapter 7
Analytical Modeling of CFRP-Retrofitted
7.1 Introduction
This chapter discusses the analytical models developed to predict the flexural strength
the case of strengthening of intact girders, the model is based on establishing the
moment-curvature relationship of the cross section, which is then integrated along the
parametric study is also performed to examine the effects of elastic modulus and rupture
strain of CFRP, as well as the CFRP reinforcement ratio on the behaviour of the
girders having a cut in the steel flange, a simplified analytical approach is proposed to
267
Chapter 7
Figure 7.1(a) illustrates the strain and stress distributions over a typical cross section. An
incremental approach, similar to the one used for the fibre-element model of HSS
columns and described in chapter 6, is used. The concepts of equilibrium and strain
compatibility are satisfied at each loading step. The analytical procedure is executed
using a spread sheet-type programming. The following assumptions are considered in this
flexural model:
2. The following constitutive models are assumed to represent the behaviour of the
materials, as shown in Figure 7.1(b): (a) the stress-strain curve of steel is assumed
second degree parabola in compression (Collins and Mitchell, 1997), and (c) FRP
Figure 7.1(b) are based on the material properties used in test specimens G1 to
G3.
4. Perfect bond exists between the FRP bonded plates and steel.
5. Two different limit states may occur in the model, namely, concrete crushing or
268
Chapter 7
In order to establish the moment-curvature curve of a given cross section, the strain at the
top level of the concrete compression flange (εtop) is first assumed. The cross section of
the strengthened girder is divided into horizontal layers (elements), as shown in Figure
7.1(a), and the strain (εi) of each element i located at a distance di from the extreme top
fibres of the cross section can then be determined using the concept of similar triangles as
follows:
⎡d ⎤
ε i = ⎢ i − 1 ⎥ ε top (7.1)
⎣ c ⎦
The stress in each element can then be determined from strain using the corresponding
material stress-strain relationship, as given by Equations 7.2 to 7.4 for concrete, steel
⎡ ε ⎛ε ⎞
2 ⎤
' ⎢ ⎥
Fci = f c 2 ci' − ⎜ ci' ⎟ (7.2)
⎢ ε ⎜ε ⎟ ⎥
⎢⎣ c ⎝ c ⎠ ⎥⎦
Fs i = E s ε s i when ε s ≤ ε y
(7.3)
Fs i = F y when ε si > ε y
Ff = E f ε f i
when ε fi
≤ε fu
(7.4)
i
269
Chapter 7
For a given value of the neutral axis depth (c), the strain distribution over the entire cross
section can be determined and the internal force of each element can then be calculated.
Using Simpson’s Rule, the stress integration (i.e. internal forces) of the cross section can
(
R = ∑ ( Fci Aci ) + ∑ ( Frb Arb ) + ∑ ( F si A si ) + ∑ F f A f
i i
) (7.5)
concrete bars steel FRP
where Aci , Arb , A si , and A f are the areas of the concrete, reinforcing bar, steel, and
i
FRP elements, respectively, and Frb is the stress at the reinforcing bars.
An iterative approach is followed by varying the values of (c) until force equilibrium is
satisfied (i.e. R = 0). The moment capacity of the cross section can then be determined by
summing the bending moments of the element forces about the extreme top fibres of the
(
M = ∑ ( Fci Aci d i ) + ∑ ( Frb Arb d i ) + ∑ ( F si A si d i ) + ∑ F fi A f d i
i
) (7.6)
concrete bars steel FRP
Based on the assumption of strain compatibility and plane deformed sections, the
ε top
ψ = (7.7)
c
270
Chapter 7
A point (ψ, M) on the moment-curvature response has now been determined for the
concrete strain εtop. The procedure can be repeated for the next level of concrete
compressive strain (εtop). The process is continued until one of two limit states is reached,
either crushing of concrete, or rupture of the FRP material. Since both limit states are
strain-controlled, and the process is based on increasing the strains incrementally, the
Once the moment-curvature of the section is obtained, the load-deflection response of the
girder can be estimated for a given loading scheme. The deflection (y) is calculated by
integrating the curvatures (ψ) along the span using the moment-area method, as shown in
d2y
ψ = (7.8)
dx 2
y = ∫∫ψ ( x ) dx dx (7.9)
The deflection at any point is essentially the moment of the curvature diagram, which is
considered as elastic load acting on the conjugate beam (Ghali and Neville, 1989).
The model was verified using the test results of experimental Phase III, including a
control and two strengthened girders. Figure 7.3 to 7.5 show comparisons between the
271
Chapter 7
the top side of the concrete slab, and the bottom side of the tension steel flange (or the
CFRP plates for the strengthened girders). For girder G2, the strain gauge was attached to
the first layer of CFRP, which remained bonded to the steel flange throughout the test,
unlike the second CFRP short layer, which debonded at a tensile strain of 0.29 percent.
For girder G3, there was only one strain gauge attached to the outer CFRP plate, and was
lost after debonding of the outer CFRP plate at a tensile strain of 0.18 percent. Therefore,
and 7.5, the predictions are made for two cases. In one case, the two CFRP layers are
assumed fully bonded to the end and in the second case, only the first (inner) CFRP layer
is present from the onset of steel yielding. Figure 7.6 to 7.8 show comparisons between
the experimental and predicted responses of moment-strain of the tension flange, while
Figure 7.9 to 7.11 show the load-deflection responses of all the three girders. In the
predictions of all three girders, failure occurred at ultimate by crushing of concrete slab
before rupture of CFRP. Generally, the model shows good agreement with the
In this section, a parametric study is performed using the proposed model to study the
effects of the elastic modulus and the rupture strain (i.e. tensile strength) of CFRP
material, as well as the reinforcement ratio (i.e. amount of CFRP) on the flexural
behaviour of the strengthened girders. The same cross section of the girders tested in
experimental Phase III was selected for the analysis. Three commercially available CFRP
products with a wide range of elastic moduli are used in the analysis. CFRP types C4 and
272
Chapter 7
C5, which were previously used in the experimental program (EC4 = 152 GPa and EC5 =
313 GPa), as well as a third type, referred to here as C6 (EC6 = 457 GPa) are used in the
known as THM-450 (Schnerch, 2005). These types of CFRP are chosen to represent
Standard Modulus- (SM-), High Modulus- (HM-), and Ultra High Modulus- (UHM-)
CFRP. Three CFRP reinforcement ratios (ρ) of 1.3, 2.6, and 3.9 percent are considered
by having a constant plate thickness of 1.4 mm, and various plate widths of 30, 60, and
90 mm, respectively. Failure modes considered are either CFRP rupture or concrete
crushing. Table 7.1 summarizes the results of the parametric study. Figure 7.12 to 7.14
show the moment-curvature responses for all cases. The effect of each parameter is
Table 7.1 and Figure 7.12 to 7.14 clearly show the increase in flexural strength as a result
of increasing the reinforcement ratio for each type of CFRP. It is noted, however, that a
small reduction in strength and a large reduction in ductility are associated with the
increase in the elastic modulus of CFRP. Table 7.1 also indicates that for the same
reinforcement ratio (ρ = 3.9 percent), the stiffness of the strengthened girders is increased
by 6, 12, and 17 percent for CFRP elastic moduli of 152, 313, and 457 GPa, respectively.
It is also noted that the CFRP contribution and the effect of the elastic modulus on
flexural stiffness are more pronounced after yielding of steel, compared to those in elastic
range. Figure 7.15 shows the percentage increase in the elastic stiffness of the
strengthened girders versus the ratio of the CFRP-to-steel elastic moduli, for different
273
Chapter 7
reinforcement ratios. The figure shows linear relationships with a slope depending on the
reinforcement ratio.
Table 7.1 and Figure 7.16 show the relationships between the percentage increase in the
yielding moment of the strengthened girders and the ratio of the steel-to-CFRP elastic
moduli for different reinforcement ratios. The relationships are also linear with a slope
Figure 7.17 shows the effect of elastic modulus of CFRP on the percentage reduction
inductility. The figure shows that the rate of reduction is higher between SM-CFRP and
Table 7.1 suggests that using UHM-CFRP with a small reinforcement ratio may not be as
effective as using a SM-CFRP with a large reinforcement ratio in increasing the flexural
strength. In other words, using two different types of CFRP having the same (Ef Af)
product (i.e. product of modulus and area) will not provide the same increase in flexural
strength. This is attributed to the different tensile strengths of the different types of
CFRP.
Table 7.1 and Figure 7.15 and 7.16 show that, within the elastic range, the same gain in
either the stiffness or the yielding moment could be achieved by increasing either the
CFRP elastic modulus Ef or the reinforcement ratio ρ. It should be noted that, these
274
Chapter 7
results are based on increasing the CFRP reinforcement ratio by increasing the width of
the plate (i.e. a corresponding increase in contact surface area is also achieved, and hence
the bond strength is maintained). However, had the thickness of the CFRP material been
increased instead of the width, a debonding type of failure, rather than rupture could
occur, which may lead to a reduction in the yielding moment. Moreover, increasing the
thickness of the CFRP material could magnify the shear lag effect, which may affect the
In general, increasing the CFRP elastic modulus is usually associated with reduction of
its ultimate tensile strength (Figure 3.8). Therefore, it was decided to compare three
actual CFRP products in the parametric study. Table 7.1 lists the ultimate strain at rupture
for each type of CFRP. Figure 7.18 clearly shows that the strength of the CFRP
CFRP Materials
This section presents a simplified model used to predict the ultimate moment and
deflection at service load of damaged and the repaired steel-concrete composite girders
using FRP materials. The damage is assumed to be a complete loss of the tension flange
at one cross section. The following sections provide a detailed description of the model.
275
Chapter 7
A similar approach to the one described previously in Section 7.2 is used to predict the
ultimate moment. However, a few additional assumptions are made to account for the
stress concentration in the vicinity of the damaged flange as well as the crack propagation
from the cut flange into the web, in a simplified manner. The model covers the following
cases:
Additionally, a calibration for the assumptions made in the model is introduced. The
control girder. The calculations of flexural strength are based on first principles of strain
compatibility and force equilibrium, exactly as discussed in section 7.27.2.1. The cross
sectional moment capacity M will be calculated using Equation (7.6), at one point only
when the uppermost concrete fibres reach the crushing strain (εc top = εcr = 0.35 percent),
276
Chapter 7
The introduced cut in the steel tension flange creates a zone of stress concentration,
which eventually leads to crack propagation through the steel web, as observed
section, a simplified approach is proposed to account for this phenomenon. This approach
ignores the lower flange as well as a part of the web of a height (ηd), as shown in Figure
7.19(b), where d is the total depth of the section. The value of (η) is assumed and then
calibrated. It should be noted that linear strain distribution is assumed within the
remaining part of the steel section and the concrete slab for simplicity, which may not be
quite the case in reality. The cross sectional moment capacity M is calculated using
Equation (7.6), when concret crushes (εc top = εcr = 0.35 percent).
The value of the elastic modulus of the CFRP used in the repair of damaged girders plays
an important role in determining the failure mode of the repaired girder. Based on the
results of Phase III of the experimental program, it was shown that CFRP rupture occurs
when using HM-CFRP, due to the very high modulus and consequently the small strain at
rupture of CFRP. This rupture strain is usually close to or slightly higher than the yield
strain of steel. On the other hand, debonding typically occurs when using SM-CFRP
since the ultimate (rupture) strain of CFRP is very high, relative to the yielding strain of
steel. It was also observed experimentally that debonding happens at different strain
levels, depending on the bonded surface area (i.e. length and width of the CFRP). In the
277
Chapter 7
following sections, the procedure used to account for the type of CFRP and the associated
In the case of HM-CFRP, failure may be due to rupture of the CFRP material. Since this
type of tension failure occurs at a very small tensile strain, crushing of the concrete slab
is highly unlikely to occur. Also, most of the intact part of the steel section would still be
elastic at failure. Therefore, the flexural strength calculations would be very similar to
that described in section 7.2.1, except that the ultimate moment capacity M is calculated
when the CFRP strain reaches the rupture strain (εf = εr), instead of assuming a
compressive crushing strain in the extreme fibre of the concrete slab. The concrete strain
can then be checked at ultimate to ensure that it has not reached the crushing strain.
In the case of SM-CFRP, failure may either be due to debonding of CFRP or crushing of
the concrete slab. The flexural strength can also be calculated using the same procedure
described in section 7.2.1. The main difference would be the governing strain of CFRP at
debonding. It has been shown in Figure 5.40 that the axial strain in CFRP at debonding
depends on the bonded length. Also, the effect of bond length on strain at ultimate varies
in the shear span from the constant moment region. Therefore, the following two
equations are proposed to obtain the strain εmax in CFRP at debonding of SM-CFRP,
278
Chapter 7
ε max
= 0.0055 L CFRP (No shear) (7.10)
ε ult
ε max
= 0.0003 L CFRP + 0.6324 (Bending and shear) (7.11)
ε ult
where εult is the reported rupture strain of CFRP and LCFRP is the bonded length of CFRP
Debonding may be assumed first by using εmax and then check that the top concrete strain
is below crushing strain. If it is not, the analysis should be revised assuming concrete
7.4.1.4 Calibration of parameter η for the neglected part of the steel web
A calibration study is carried out to assess the length of the lower part of the web (ηd),
which is considered inefficient due to crack propagation, and can then be neglected in
section analysis. Three values of η (0.25, 0.33, and 0.50) were assumed in this study. The
moment capacities of the beams tested in Phase III, except for B5, were calculated using
the different values of η. Table 7.2 shows the maximum strain of CFRP at ultimate, based
on Equations 7.10 and 7.11, the measured and predicted ultimate moment capacities for
each beam and η value as well as their ratio. Also given in Table 7.2, is the average and
standard deviation values of the (predicted moment / measured moment) ratios for each
value of η. The same results are plotted in Figure 7.20 for beams B3 to B11 (except B5)
as well as for B2. Generally, Table 7.2 and Figure 7.20 show that η has a relatively small
yet significant effect on the predicted moment. It is noted that the model tends to
279
Chapter 7
underestimate the moment capacity of the damaged and unrepaired specimen (B2) for all
the three values of η. On the other hand, the model tends to overestimate the moment
capacity of the damaged and repaired specimens (B3, B4, and B6 to B11) for all values of
η. It is also noted that the value of η has less effect on the repaired specimens. This is
attributed to the fact that the unrepaired section is reduced to a T-section, where its
moment of inertia becomes much more sensitive to the intact height of the web. CFRP
section, for which the moment of inertia is not greatly affected by the intact height of the
web. Figure 7.20 suggests that the optimum values of η, which result in an average
moment ratio of 1.00, are 0.57 and 0.23 for the repaired and unrepaired beams,
respectively. It should be noted that this study investigated the effect of η on specimens
with several variables, namely, different types of CFRP, different force equivalence
indices (ω), and different bonded lengths and areas. The approach adopted in this model
to deal with the cut in the flange, the crack propagation, and the associated stress
This is suited for engineers and designers. A more rigorous approach could possibly
This section provides a simplified procedure for calculating the mid-span deflection at
service load. The service load is defined as the unfactored dead plus live loads. In order
to estimate the equivalent service load of the test beams at which deflection will be
280
Chapter 7
M f = αD M D +αL M L (7.12)
where M D , M L , and M f are the dead and live service moments, and the total factored
moment, respectively. α D and α L are the dead and live load factors, and are taken as
1.25 and 1.5, respectively, (NBCC, 2005). Equation (7.12) is then rearranged to isolate
ML as follows:
M f −αD M D
ML = (7.13)
αL
By setting αL ML equal to the experimental measured maximum moment of the beam and
MD equal to the self weight of the beam, the moment at service load (Mservice = MD + ML)
can be obtained.
The deflection at service load can be calculated using various methods, including the
virtual work method (Ghali and Neville, 1989). The general virtual work equation used to
calculate the vertical deflection at any point along the beam span (L) can be formulated as
follows:
L mM service L vV service
δ =∫ dx + ∫ dx (7.14)
0
Es It 0
G s Aw
where m and v are the internal virtual bending moment and shear force, respectively, due
to a virtual unit load at the deflection point of interest along the beam. M and V are the
actual bending moment and shear force, respectively, acting on the beam due to service
loads. Gs is the elastic shear modulus of steel, and is taken as 77 GPa (Kulak and
Grondin, 2002). It is the transformed moment of inertia, and Aw is the area of the steel
web. It should be noted that for girders with relatively large shear span-to-depth ratio, the
281
Chapter 7
second term of Equation (7.14) becomes insignificant and can be neglected. This,
however, is not the case for the beams tested in Phase III of this study, since the spans
The transformed moment of inertia (It) is calculated after transforming the concrete slab
breadth (bc) and the FRP breadth (bf) to equivalent steel breadths beqc and beqf, according
E ct
b eq = bc (7.15)
c Es
Ef
b eq = bf (7.16)
f Es
where Ect is the initial tangent modulus of concrete. For the assumed parabolic stress-
f c'
E ct = 2 (7.17)
ε c'
Alternatively, the product (Es It) of the cross section can easily be calculated using the
M service
Es I t = (7.18)
ψ
For concrete-steel composite girders with shear connectors, the effective moment of
inertia (Ieff) is used in Equation (7.14) in lieu of It to account for the increase in deflection
that may result from the interfacial slip between the concrete slab and the steel girder. Ieff
282
Chapter 7
where Is is moment of inertia of the steel cross section alone (i.e. without the concrete
n'
ρ= (7.20)
n
where n’ is the actual number of shear connectors used in the girder, and n is the number
of shear connectors required for full composite action. For the beams tested in Phase III,
[( )
min A s F y or ( 0.85 b c t c f ' c )]
n= (7.21)
[(
min 0.5 A sc )
f ' c E ct or ( A sc F u )]
where Asc is the cross sectional area of one shear stud, As is the area of the steel cross
section, Fy is the yield stress of steel, and Fu is the ultimate stress of the shear stud.
The following sections provide a proposed methodology to account for the effects of
section loss and the resulting stress distribution in the vicinity of the crack as well as the
It has been established in section 7.4.1.4, that the lower flange and 57 percent of the web
are ignored at the mid-span section, to account for the crack propagation at ultimate. For
simplicity and as a conservative approach, the same same level of section loss is assumed
at service. The upward propagation of the crack within the web affects the stress flow in
the region around the crack. The stress flow within the web and flange is idealized, as
shown in Figure 7.21(a). The inclined line starts from point “a” at 0.57d of the web and
283
Chapter 7
follows a slope of (z:1) until it reaches the flange at point “b”. The line then spreads
within the flange in the transverse direction with a slope (1:z), until it reaches the edges
of the flange at line “c-c”. The slope value (z) is determined through a calibration study.
For the damaged unrepaired girders [Figure 7.19(b)], the moment of inertia of the intact
section (Ieff1) [Figure 7.19(a)] beyond line “c-c” can be assumed and is calculated using
Equation 7.19. Equation 7.19 is also used to calculate (Ieff2) at mid-span accounting for
section loss. Between the mid-span section and line “c-c”, the moment of inertia is
For the CFRP-repaired girders, the moment of inertia at mid-span is referred to as Ieff4,
based on the cross section shown in Figure 7.19(c or d). For sections beyond line “c-c”,
and if CFRP extends beyond line “c-c”, the moment of inertia is referred to as Ieff3, which
is based on intact cross section strengthened with CFRP. The moment of inertia is
assumed to vary linearly between Ieff3 to Ieff4 as shown in Figure 7.21(b, left side). Both
Ieff3 and Ieff4 are also calculated using Equation 7.19, where Is in this case is the moment of
inertia of the cross section at mid-span without the concrete slab (i.e. Is = IT-section + ICFRP).
Based on the conclusions drawn by Miller (2000), Colombi and Poggi (2006), and the
experimental observations of Phase III of this study, the development length of CFRP
sheets can reasonably be assumed 100 mm from the free edge of the sheets. As shown in
Figure 7.21(b, left side), a linear transition is assumed between from Ieff1 and Ieff3 within
this length.
284
Chapter 7
If the CFRP is terminated before line “c-c” [Figure 7.21(b, right side)], then Ieff4 is
assumed at mid-span. A fictitious Ieff3 value is assumed at line “c-c” and linearly
connected to Ieff4 at mid-span. The diagram is then corrected with a 100 mm transfer
Three slopes (1:1, 1.5:1, and 2:1) were examined in this study. The deflections were
calculated using Equation 7.14 and compared with the experimental results of the beams
tested in Phase III. Table 7.3 shows the service load and moment at which the deflection
is calculated, the measured deflection at service load, the predicted deflection at service
load, and the ratio of the two values for each beam. The values are calculated for each of
the three assumed slopes. Also given in Table 7.3, are the average and standard deviation
values of the (predicted deflection / measured deflection) ratios for each slope. The same
results are plotted in Figure 7.22 for beams B3 to B11 (except B5) as well as for B2. The
figure suggests that deflections are highly sensitive to the z value. Also, extending the
average lines of the repaired and unrepaired beams to intercept the unity ratio leads to a
values of z equal to 0.48 and 2.85, respectively. It is noted from Figure 7.22 that this
approach has resulted in overestimating the deflections for the repaired beams and
285
Chapter 7
maximum load, ψ
Flexural stiffness
Moment capacity
Yielding moment
Elastic modulus,
%age reduction
Curvature @
(1/m x 10-3)
%age gain
%age gain
%age gain
(kN/mm)
E (GPa)
(kN.m)
(kN.m)
x 10-3
ρ
control 17.6 132 180 63.1
152 12.57 1.3 18.0 2 135 2 201 12 47.3 25
2.6 18.3 4 138 5 223 24 47.7 24
3.9 18.7 6 141 7 244 36 48.0 24
313 4.71 1.3 18.3 4 138 5 182 1 18.7 70
2.6 19.0 8 144 9 199 11 19.0 70
3.9 19.7 12 151 14 216 20 19.1 70
457 3.35 1.3 18.7 6 141 7 180 0 13.8 78
2.6 19.7 12 150 13 192 7 14.0 78
3.9 20.7 17 160 21 210 17 14.3 77
Table 7.2 Calibration of the neglected part of the steel web in repair applications.
η
Specimen
286
Chapter 7
Specimen
Service 1:1 1.5:1 2:1
I.D.
Service moment
load, P M, δexp. δpred. δpred. δpred. δpred. δpred. δpred.
(kN) (k.Nm) (mm) (mm) δexp. (mm) δexp. (mm) δexp.
B1 212 83 6.33 5.91 0.93 5.91 0.93 5.91 0.93
B2 85 33 5.92 4.45 0.75 4.63 0.78 4.84 0.82
B3 97 38 2.55 2.66 1.05 2.70 1.06 2.74 1.08
B4 235 92 5.49 5.47 1.00 5.51 1.00 5.37 0.98
B6 185 72 6.40 7.71 1.21 7.98 1.25 8.31 1.30
B7 182 71 5.65 5.99 1.06 6.10 1.08 6.26 1.11
B8 247 96 7.58 7.83 1.03 7.97 1.05 8.15 1.08
B9 210 82 6.24 6.75 1.08 6.88 1.10 6.99 1.12
B10 190 74 6.06 6.67 1.10 6.82 1.13 7.07 1.17
B11 152 59 4.87 5.69 1.17 5.85 1.20 6.14 1.26
Average 1.04 1.06 1.08
Standard deviation 0.13 0.13 0.15
400
1960 Dims are in mm.
The effect of shear deformation is considered
in the predicted deflection calculations.
Gs = 77 GPa
287
Chapter 7
bc
εtop Fc
tc/2
εrb Rc
Reinforcing bars c
ψ Rs1
di
N.A. Bending
moment
element i “M”
εi Rs2
Rs3
εbottom Rf
εf Ff
FRP (Strains) (Stresses)
εc’ = 0.0025
εcr’ = 0.0035
εs = 0.0017
εs εc εf
Steel Concrete CFRP
Figure 7.1 Steel-concrete composite girder strengthened with FRP and subjected to
pure bending.
P/2 P/2
Moment
(y-diagram)
y(x)
x
M(x)
y = ∫∫ψ ( x)dx dx
ψ (x)
(M-diagram) Curvature
M(x)
Load
P
(ψ-diagram)
ψ(x)
Get y(x) as the moment of the area under the ψ - diagram y(x) Deflection
288
Chapter 7
275
250
225
200 Concrete
crushing
Model
175
Moment (kN.m)
150
Experiment
125
100
75
50
25
0
0 10 20 30 40 50 60 70 80
Curvature (1/m) x 10-3
275
Model
250 (Two CFRP layers) CFRP
rupture
225
200 Debonding of
outer layer Model
175 (One CFRP layer)
Moment (kN.m)
150 Experiment
125
100
75
50
25
0
0 10 20 30 40 50 60 70 80
Curvature (1/m) x 10-3
289
Chapter 7
275
250
Model CFRP rupture
225 (Two CFRP layers)
200 Model
Moment (kN.m)
275
250
225
200
Model
175
Moment (kN.m)
150
Experiment
125
100
75
50
25
0
0 5 10 15 20
Strain in tension flange x 10-3
290
Chapter 7
275
Model
250 (Two CFRP layers)
225
Model
200 Debonding of (One CFRP layer)
outer layer
175
Moment (kN.m)
Experiment
150
125
100
75
50
25
0
0 5 10 15 20
Strain in tension flange x 10-3
275
250
Model CFRP rupture
225 CFRP rupture
(Two CFRP layers)
Model
200 (One CFRP layer)
175
Moment (kN.m)
150 Debonding of
outer layer
125
Experiment
100
75
50
25
0
0 5 10 15 20
291
Chapter 7
220
200
180
Concrete
160 crushing
Model
140
Load (kN)
120
Experiment
100
80
60
40
20
0
0 25 50 75 100 125 150 175
Mid-span deflection (mm)
220
Model
200 Experiment CFRP rupture
180 Model
(Two CFRP layers)
Model
160 (One CFRP layer)
140
Load (kN)
120 Debonding of
outer layer
100
80
60
40
20
0
0 25 50 75 100 125 150 175
Mid-span deflection (mm)
292
Chapter 7
220
Experiment
200
CFRP rupture CFRP rupture
180 Model
(Two CFRP layers) Model
160 (One CFRP layer)
140
Load (kN)
120
100
80
60
40
20
0
0 25 50 75 100 125 150 175
Mid-span deflection (mm)
250
ρ = 3.9 %
225 ρ = 2.6 %
ρ = 1.3 %
200
175 control
Moment (kN.m)
150
125
100
(E)FRP = 152 GPa
75
(Fy)steel = 345 MPa
50
fc’ = 40 MPa
25 tc = 65 mm
0
0 10 20 30 40 50 60 70 80
Curvature (1/m) x 10-3
Figure 7.12 Moment-curvature responses of composite sections strengthened using
SM-CFRP.
293
Chapter 7
250
225 ρ = 3.9 %
ρ = 2.6 %
200
ρ = 1.3 %
175
Moment (kN.m)
control
150
125
100
250
225
ρ = 3.9 %
200 ρ = 2.6 %
ρ = 1.3 %
175
Moment (kN.m)
150
control
125
100
294
Chapter 7
18
(Fy)steel = 345 MPa
16 1.96 ρ
fc’ = 40 MPa
1
14 tc = 65 mm
% increase in stiffness
12
1.98 ρ
10 .9 %
ρ =3 1
8
.6 %
ρ =2
6
2.00 ρ
1
4
ρ = 1.3 %
0
0.5 1 1.5 2 2.5
Ef / Es
Figure 7.15 Effect of the modulus of CFRP on percentage increase in stiffness for
different reinforcement ratios.
25
(Fy)steel = 345 MPa
fc’ = 40 MPa
20 tc = 65 mm 2.34 ρ
1
% increase in Myield
15
2.25 ρ
3. 9%
ρ = 1
10
.6 %
ρ =2
2.26 ρ
5 1
ρ = 1.3 %
0
0.5 1 1.5 2 2.5
Ef / Es
295
Chapter 7
90
(Fy)steel = 345 MPa
80
fc’ = 40 MPa
70 tc = 65 mm
% reduction in ductility
60
E = 457 GPa
E = 313 GPa
50
E = 152 GPa
40
ρ = 1.3, 2.6, and 3.9 %
30
20
10
0
0.5 1 1.5 2 2.5
Ef / Es
Figure 7.17 Effect of the modulus of CFRP on percentage reduction in ductility.
40
(Fy)steel = 345 MPa
35 fc’ = 40 MPa
E = 313 GPa
tc = 65 mm
30
9%
% increase in strength
3.
ρ =
25
E = 457 GPa
E = 152 GPa
20
%
2. 6
ρ =
15
10
1.3 %
ρ =
5
0
0 2 4 6 8 10 12 14
Rupture strain of CFRP x 10-3
Figure 7.18 Effect of CFRP rupture strain on percentage increase in strength for
different reinforcement ratios.
296
Chapter 7
bc x (Ect/Es) bc
Fc εc top = εcr
tc Rc C
N.A. tc
εs top
Y
d yi
Rs d
Ieff1
fi εi
b Fy εs bot >> εy b
Rs
d
Ieff2
Fy εs bot >> εy ηd
b b
tc Rc C tc
N.A.
(1-η) d
Rs d
Ieff4 Fs bot εs bot
ηd
Rf
Ff = Fr εf = εmax bf
bf x (Ef /Es)
b
(c) Cross section repaired using HM-CFRP
bc x (Ect/Es) bc
Fc εc top = εcr
Rc
tc C tc
N.A.
(1-η) d
Rs
d
Fs bot εs bot
ηd
Rf
Ff εf bf
bf x (Ef /Es) b
(d) Cross section repaired using SM-CFRP
Figure 7.19 Effective cross sections and corresponding stress and strain
distributions in intact, damaged, and repaired girders.
297
Chapter 7
1.4
1.2
Repaired beams
(B3, B4, B6-B11)
1
Damag
unrepair ed and
Mpred / Mexp
0.8 ed beam
(B2)
0.6
0.4
η = 0.57
η = 0.23
0.2
0
0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
η
298
Chapter 7
Ieff1 Ieff2
a d
z:1
e ηd
S lop
b b
2zηd
L
c c
Sl
op :z
e1 e1
:z
S lop
b
c c
2zηd + bz
Ieff2
Ieff1 Ieff1
LCFRP/2 LCFRP/2
Ieff4 Ieff2
Ieff1 Ieff1 Ieff4 Ieff1
Ieff3 Ieff4 Ieff3
100 mm
(L-LCFRP)/2100 mm zηd + bz/2 zηd + bz/2
LCFRP/2 LCFRP/2
299
Chapter 7
1.60
1.40
s
1.20 d beam
Repaire 6-B11)
B 4, B
(B3,
1.00
δpred / δexp
d and
Damage m (B2)
0.80 ed ea
b
unrepair
0.60
0.40
z = 2.85
z = 0.48
0.20
0.00
0 0.5 1 1.5 2 2.5 3
300
Chapter 8
Chapter 8
8.1 Summary
The main objective of this study was to evaluate the use of CFRP composite materials in
members to increase their load carrying capacity and stiffness as well as repair of
damaged members to restore their original capacities. Both experimental and theoretical
investigations were carried out. In principle, the study demonstrated the great success of
this retrofitting technique. Also, the findings of this research program will enable
engineers to make more informative decisions regarding retrofit of steel structures using
adhesively bonded composites and can assist in developing reliable design guides.
columns with slenderness ratios ranging from 4 to 93. The columns were strengthened
using different configurations of CFRP composites and were tested under axial
compression. The study considered the effects of number of CFRP layers, type of CFRP,
fibre orientation, and slenderness ratio of the columns, on their axial behaviour. Phases II
301
Chapter 8
and III of the experimental program were focused on strengthening and repair,
respectively, of steel W-sections acting compositely with concrete slabs, using different
CFRP materials. In Phase II, three large-scale intact girders were strengthened using
CFRP plates. In Phase III, 11 girders, most of which were artificially damaged, have been
repaired using CFRP sheets. All girders were tested in four-point bending. The
parameters considered were the effect of CFRP type, number of layers, number of
retrofitted sides of the tension flange, and the length of CFRP repair patch.
Four analytical and numerical models have been developed for the columns and girders.
The first and second models are an analytical fibre-element model and a non-linear
numerical finite element model, developed for FRP-strengthened HSS slender columns.
complete cut in their tension flanges. The models were verified using experimental
8.2 Conclusions
1. Both the axial strength and stiffness of HSS slender columns are increased using
302
Chapter 8
the columns against lateral deflection is improved, and hence, overall buckling occurs
at higher loads.
2. The effectiveness of the CFRP system in increasing the axial strength of slender
columns increases greatly as the slenderness ratios become higher. Its effectiveness in
increasing axial stiffness, on the other hand, is not much affected by slenderness ratio.
The strength of columns with slenderness ratios ranging from 46 to 93 was increased
percent.
3. The axial strength of a column reduces as slenderness ratio increases. However, the
4. The effectiveness of the CFRP system increases in columns with larger out-of-
becomes constant. The gain in stiffness, on the other hand, is not affected much by
were successfully developed and predicted reasonably well the complete behaviour of
higher axial strength and a slightly lower axial stiffness than Model 2, and its
303
Chapter 8
significantly.
has little effect on the gain in axial strength when using CFRP. In general, the lower
8. Slender columns fail by excessive overall buckling. In thin-walled sections, this may
up to kL/r = 76, the CFRP plate on the extreme compression face completely debonds
from the steel surface. For higher slenderness ratio (kL/r = 93), the CFRP plate
crushes at mid-height. For columns strengthened using CFRP sheets, the secondary
local buckling is associated with local debonding and crushing of CFRP. The CFRP
9. Based on the experimental study, simple empirical equations are proposed to establish
the maximum strain that CFRP material can reach at ultimate as a function of
1. CFRP wraps with fibres oriented in the transverse direction appear to be more
efficient in increasing axial strength of HSS short columns than those with fibres in
the longitudinal direction. This is evident by achieving the highest gain among all
configurations considered in this study. However, the maximum gain in axial stiffness
304
Chapter 8
on the stiffness of the CFRP jacket (Ef Af). In this study, the SM-CFRP resulted in
better strengthening than the high modulus- (HM-) CFRP because it has 10 percent
higher stiffness. Also, the stiff nature of fibres in the HM-CFRP wraps resulted in
symmetric local buckling, where two opposite sides buckled inwards and the other
two sides buckled outwards. Transverse CFRP wraps are effective in bracing the
outwards buckling but they tend to debond from the sides that buckle inwards.
4. In all short columns strengthened with longitudinal CFRP layers, debonding occurred
between CFRP and steel at one end, even in the columns with additional outer CFRP
transverse wrap.
1. The SM-CFRP plates have indeed increased the flexural strength and stiffness of
experimental study.
2. The control girder failed by crushing of the concrete slab after excessive yielding of
the steel section. In the strengthened girders, the CFRP plate bonded to steel over a
length equals to 67 percent of the span remained fully bonded to the section till
buckling. At this stage, tensile rupture of the CFRP plate was quite imminent. The
305
Chapter 8
second CFRP plate, which was significantly shorter (25 percent of the span), was
3. Increasing the elastic modulus of CFRP leads to a reduction in flexural strength gain,
as a result of the reduced tensile strength of CFRP. It also leads to a reduction in the
4. Similar gains in the elastic flexural stiffness and yielding moment are achieved by
either increasing the elastic modulus or the cross sectional area of CFRP. The effect
5. The percentage increases in both the elastic stiffness and yielding moment of the
girders are linearly proportional to the ratio of CFRP-to-steel elastic moduli. The rate
between the CFRP plate and the steel substrate increase rapidly near the end of the
CFRP plates.
1. The flexural strength and stiffness of steel-concrete composite girders tested in this
2. The girders repaired using HM-CFRP sheets of a 152 percent force equivalence index
(ω) recovered the original flexural strength and stiffness and even exceeded them by
306
Chapter 8
10 and 26 percent, respectively. On the other hand, the girder repaired using SM-
CFRP sheets required a higher ω of 210 percent to just recover the original stiffness
but exceeded the original strength by 16 percent. ω is the cross sectional area ratio of
CFRP and steel flange, normalized to the ratio of CFRP tensile strength and steel
yield strength.
CFRP sheets, before developing their full tensile strength. On the other hand, girders
repaired using HM-CFRP sheets failed by rupture of the sheets, without any sign of
debonding. This is attributed to the higher modulus and smaller rupture strain of the
HM-CFRP.
4. The bonded length of the SM-CFRP sheets has an insignificant effect on the elastic
5. The longer the bonded length of SM-CFRP sheets, the higher the maximum strain
reached in the sheets at failure. This strain is also affected by the state of stress in the
girder (i.e. pure bending or combined bending and shear). Although no SM-CFRP
rupture was observed in this study, it has been estimated that bonded lengths of at
least 180 and 1175 mm, from each side of the crack, are required to achieve rupture
and avoid debonding, in the cases of pure bending and combined bending and shear,
6. Bonding SM-CFRP sheets to both sides of the steel flange, instead of the bottom
surface only, enhances flexural stiffness of the repaired girders, but has no effect on
307
Chapter 8
flexural strength. Curvature of the girders could induce peeling stresses and triggers
7. Increasing the force equivalence index (ω) increases the percentage recovery of
strength in CFRP-repaired girders. This rate of increase follows similar trends for
8. Shear stresses at the interface between the CFRP sheets and the steel substrate, in the
vicinity of the flange cut, increase rapidly with loading until they reach the
characteristic strength of the adhesive, and then drop to zero as debonding progresses.
near the cut at mid-span, than at the ends. Although not investigated in this study, it is
believed that providing an anchorage system such as transverse wraps near the flange
9. An analytical approach has been developed to calculate flexural strength and service
load deflection of CFRP-repaired girders with a complete flange cut. It deals with the
steel girder of a height of 0.57d and base length of (0.27 d + 0.24 b), at both sides of
the cut, where b and d are the flange breadth and steel section depth, respectively.
10. Late application of CFRP sheets (i.e. several months after sandblasting the surface)
substantially lower flexural strength was achieved. This occurred despite the fact that
308
Chapter 8
steel surface was protected by a coating of oil, which was completely removed and
then the surface was thoroughly wiped with acetone, just prior to FRP installation.
The research work carried out in this study on retrofit of steel structures using CFRP
materials covered a wide range of applications and parameters, and indeed demonstrated
modeling. Future research in related areas, however, still needs to be carried out on the
following topics:
1. Optimizing the length of CFRP plates in steel columns with high slenderness ratios,
2. A study focused on the effect of CFRP wraps on delaying local buckling of short HSS
columns (or box sections) with very thin walls, which are vulnerable to local buckling
prior to yielding. This may also be extended to include very thin webs in girders
subjected to shear.
3. Developing analytical and numerical models to predict the behaviour and axial
load and bending moment as well as slender columns with different end conditions.
repaired using CFRP. The effect of CFRP on arresting fatigue cracks would be of
great interest.
309
Chapter 8
6. Studying the effect of using prestressed CFRP plates on the behaviour of fatigue-
exposures.
9. Developing techniques to use the FRP material in strengthening (or repair) of bolted
10. Establishing comprehensive design guidelines for steel girders and columns
310
References
1. ACI (2002). ACI 440. 2R-02: Guide for the design and construction of externally
Technology 59(7):1127-1137.
4. Al-Emrani, M., Linghoff, D., and Kliger, R. (2005) “Bonding Strength and
D.C.
10. American Society for Testing and Materials (ASTM), (2000), “Designation D
International.
11. American Society for Testing and Materials (ASTM), (2000), “Designation D
3039/D 3039M –Standard test method for tensile properties of polymer matrix
12. American Society for Testing and Materials (ASTM), (2000), “Designation
D5868-01, Standard test method for lap shear adhesion for Fiber Reinforced
13. American Society for Testing and Materials (ASTM), (2000), “Designation E
8M-04, Standard test methods for Tension Testing of Metallic Materials [Metric]”
312
References
16. Bakht, B., Sweeney, R. A. P., Saunders, W. W., Yule, R. B., Byers, W. G.,
Mosese, F., McKeel, W. T., Higgins, M., Kahn, L., and Heins, C. P. (1979)
17. Bassetti, A., Liechti, P., and Nussbaumer, A. (1998) “Fatigue Resistance and
Repairs of Riveted Bridge Members.” Fatigue Design ’98, Espoo, Finland, pp.
535-546.
18. Bassetti, A., Nussbaumer, A., and Manfred, A.H. (2000) “Crack Repair and
19. Beedle, L. S., and Tall, L. (1960) “Basic Column Strength.” Journal of Structures.
21. Bjorhovde, R., and Birkemoe, P. C. (1979). “Limit states design of HSS
22. Boyd, G. M. (1970). Brittle Fracture in Steel Structures, Butterworth & Co,
Printed in Great Britain by Hazell Watson & Viney Ltd, Aylesbury, Bucks: 117
pp.
313
References
24. Bruneau, M., Uang, C., and Whittaker, A. (1998). Ductile Design of Steel
25. Buyukozturk O., Gunes O., and Karaca, E. (2003) “Progress on Understanding
Exhibition.
26. Buyukozturk, O., Gunes, O., and Karaca, E. (2004) “Progress on Understanding
27. Cadei, J.M.C., Stratford, T.J., Hollaway, L.C. & Duckett, W.G. (2004).
Publications, Canada.
314
References
33. Colombi, P. and Poggi, C. (2006) “ Strengthening of Tensile Steel Members and
Bolted Joints using Adhesively Bonded CFRP Plates.” Construction and Building
35. Dawe, J.L., Elgabry, A.A. and Grondin, G.Y. (1985) “Local Buckling of Hollow
1112.
Materials.” Master’s thesis, North Carolina State University, Raleigh (NC). 213p.
37. Deng, J., and Lee, M. M. K. (2007) “ Behaviour under Static Loading of Metallic
78(2):232-242.
38. Deng, J., Lee, M. M. K., and Moy, S. S. J. (2004) “ Stress Analysis of Steel
65(2):205-215.
39. Edberg, W., Mertz, D. and Gillespie Jr., J. (1996) “Rehabilitation of Steel Beams
Conference, Materials for the New Millenium, ASCE, New York, NY, Nov 10-
315
References
40. El Damatty, A. A., Abushagur, M., and Youssef, M. A. (2003) “Experimental and
41. El Damatty, A. A., and Abushagur, M. (2003) “Testing and Modeling of Shear
41(11): 987-1003.
42. Fam, A., Witt, S., and Rizkalla, S. (2006) “Repair of Damaged Aluminum Truss
43. Fawzia, S., Al-Mahaidi, R., Zhao, X., Rizkalla, S. (2007) “ Strengthening of
Circular Hollow Steel Tubular Sections using High Modulus CFRP Sheets”
45. Garden, H. N., Quantrill, R. J., Hollaway, L. C., Thorne, A. M., and Parke, G. A.
46. Ghali, A. and Neville, A. M. (1989). Structural analysis: A unified classical and
matrix approach, 3rd Ed., Chapman and Hall, London and New York, 870pp.
47. Gillespie, J. W., Mertz, D. R., Edberg, W. M., Ammar, N., Kasai, K., and
316
References
48. Gillespie, J. W., Mertz, D. R., Kasai, K., Edberg, W. M., Demitz, J. R., and
49. Grabovac, I., Bartholomeusz, R. A., and Baker, A. A. (1991) “Fibre Composite
50. Haedir, J., Bambach, M. R., Zhao, X. L., and Grzebieta, R. (2006) “Bending
31.
52. Hashim, S. A. (1999) “Adhesive Bonding of Thick Steel Adherents for Marine
53. He, W. (1992) “Punching Behaviour of Composite Bridge Decks with Transverse
317
References
54. Hollaway, L.C. and Cadei, J. (2002), “Progress in the Technique of Upgrading
56. Jiao, H and Zhao, X. -L. (2004) “CFRP Strengthened Butt-Welded Very High
ASCE, 7(4):311-322.
58. Karbhari, V. M., and Shulley, S. B. (1995) “Use of Composites for Rehabilitation
column behaviour of cold formed square hollow sections. Res. Rep. No. R493,
Australia.
60. Key, P.W. and Hancock, G. J. (1993). “A theoretical investigation of the column
4), 31-64.
61. Key, P.W., Hasan, S.W. and Hancock, G.J. (1988) “Column Behavior of Cold-
407.
318
References
62. Kulak, G.L., and Grondin, G.Y. (2002). Limit states design in structural steel,
63. Lam, C.C.A., Cheng, J.J.R. and Yam, C.H.M. (2004) “Study of the Tensile
Alberta, CD-ROM.
64. Linghoff, D., Al-Emrani, M., and Kliger, R. (2006) “Strengthening Steel Beams
65. Liu, X., Silva, P. F., and Nanni, A. (2001) “Rehabilitation of Steel Bridge
613-617.
66. Luke, S. (2001) “The Use of Carbon Fiber Plates for the Strengthening of Two
Metallic Bridges of a Historic Nature in the UK.” Proceedings of the CICE 2001,
FRP Composites in Civil Engineering, Vol. II J.-G. Teng (Ed.), pp. 975-983.
68. Miller, T. C. (2000) “The Rehabilitation of Steel Bridge Girders Using Advanced
319
References
69. Miller, T. C., Chajes, M. J., Mertz, D. R., and Hastings, J. N. (2001)
Repair of Steel Structures.” 43rd International SAMPE Symposium May 31- June
4, pp. 1826-1837.
71. Moy, S.S.J. and Nikoukar, F. (2002), “Flexural Behaviour of Steel Beams
73. Nozaka, K., Shield, C. K., and Hajjar, J. F. (2005a) “Effective bond length of
74. Nozaka, K., Shield, C. K., and Hajjar, J. F. (2005b) “Design of test specimen to
9(4), 304-312.
75. Ontario Ministry of Transportation (1983) “Ontario Highway Bridge Design Code
Canada.
320
References
76. Patnaik, A.K. & Bauer, C.L. (2004). “Strengthening of steel beams with carbon
Canada. CD ROM.
79. Salmon, C. G., and Johnson, J. E. (1980). Steel structures. Design and behavior,
Adhesive Bonded Struct., U.S. Army Armament Res. And Devel. Command,
Dover, N. J.
82. Schnerch, D. (2005). Strengthening of steel structures with high modulus carbon
321
References
83. Schnerch, D., Dawood, M., Rizkalla, S., and Sumner, E. (2007) “Proposed Design
84. Schnerch, D., Dawood, M., Rizkalla, S., Sumner, E., and Stanford, K. (2006)
85. Schnerch, D., Stanford, K., Sumner, E., and Rizkalla, S., (2004) “Strengthening of
Steel Structures and Bridges with High Modulus Carbon Fiber Reinforced
86. Seica, M. V., Packer, J. A., Guevara Ramirez, P., Bell, S. A. H., and Zhao, X. L.
Structures. August 31- September 2, 2006, Quebec City, Quebec, Canada: 365-
373.
87. Sen, R., Liby, L., and Mullins, G. (2001) “Strengthening Steel Bridge Sections
88. Shaat, A. and Fam, A. (2006) “Axial Loading Tests on Short and Long Hollow
322
References
89. Shaat, A. and Fam, A. (2007a) “Fiber-Element Model for Slender HSS Columns
90. Shaat, A. and Fam, A. (2007b) “Finite Element Analysis of Slender HSS Columns
7(1): 19-34.
91. Shulley, S. B., Huang, X., Karbhari, V. M., and Gillespie, J. W. (1994)
Rehabilitation Schemes for Steel Girders with Web Distress.” Proceedings of the
16: 1187-1194.
92. Smith, S. T., and Teng, J. G. (2001) “Interfacial Stresses in Plated Beams” Journal
Girders Using CFRP Sheets.” Proc. ISEC-01, Hawaii, Jan. 24-27: 461-466
323
References
99. Tebedge, N., Alpsten, G. A., and Tall, L. (1973) “Residual Stress Measurment by
100. Vatovec, M., Kelley, P. L., Brainerd, M. L., and Kivela, J. B. (2002) “Post
Strengthening of Steel Members with CFRP.” Simpson Gumpertz & Heger Inc.
SAMPE 2002.
102. Weng, C. C. (1984). “Cold-bending of thick steel plates at low R/t ratios.”
103. Weng, C. C., and Pekoz, T. (1990). “Residual stresses in cold-formed steel
324
References
106. Zhao, X., Fernando, D., and Al-Mahaidi, R. (2006) “CFRP Strengthened RHS
1555-1565.
325
Appendix A
Measurements of Out-of-Straightness Profiles for
Column Sets 1 to 6
A.1 General
perpendicular sides (a and b) using an ILD1400 laser optical displacement sensor. Figure
A.1 to A.17 show the out-of-straightness profiles of all specimens in column sets 1 to 6,
except those of specimen 6-3, which are previously shown in Figure 3.13. It should be
noted that all specimens were oriented in the test setup such that the side of the highest
0.3
a
Side (a)
0.25
b
0.2
Out-of-Straightness (mm)
0.15
0.1
Side (b)
0.05
0
0 100 200 300 400 500 600 700 800
-0.05
-0.1
Length (mm)
0.3
a
0.25
b
Side (a)
0.2
Out-of-Straightness (mm)
0.15
0.1
Side (b)
0.05
0
0 100 200 300 400 500 600 700 800
-0.05
-0.1
Length (mm)
327
Appendix A
0.3
a
Side (a)
0.25
b
0.2
Out-of-Straightness (mm)
0.15
0.1
0.05
Side (b)
0
0 100 200 300 400 500 600 700 800
-0.05
-0.1
Length (mm)
0.3
Side (a) a
0.25
b
0.2
Out-of-Straightness (mm)
0.15
0.1
0.05
Side (b)
0
0 100 200 300 400 500 600 700 800
-0.05
-0.1
Length (mm)
328
Appendix A
0.3
a
0.25
Out-of-Straightness (mm) Side (a) b
0.2
0.15
0.1
Side (b)
0.05
0
0 100 200 300 400 500 600 700 800
-0.05
-0.1
Length (mm)
0.3
a
0.25
b
0.2
Side (a)
Out-of-Straightness (mm)
0.15
0.1
Side (b)
0.05
0
0 100 200 300 400 500 600 700 800
-0.05
-0.1
Length (mm)
329
Appendix A
0.4
a
0.35
b
0.3
Side (a)
Out-of-Straightness (mm)
0.25
0.2
0.15
Side (b)
0.1
0.05
0
0 200 400 600 800 1000 1200
-0.05
-0.1
Length (mm)
0.4
0.35
Side (a)
0.3
a
Out-of-Straightness (mm)
0.25
b
0.2
0.15
0.1
0.05
Side (b)
0
0 200 400 600 800 1000 1200
-0.05
-0.1
Length (mm)
330
Appendix A
0.4
0.35 a
0.3 b
Out-of-Straightness (mm)
0.25
Side (a)
0.2
0.15
0.1
0.05
0
0 200 400 600 800 1000 1200
-0.05
Side (b)
-0.1
Length (mm)
0.4
a
0.35
b
0.3
Side (a)
Out-of-Straightness (mm)
0.25
0.2
0.15
Side (b)
0.1
0.05
0
0 200 400 600 800 1000 1200
-0.05
-0.1
Length (mm)
331
Appendix A
0.4
a
0.35
Side (a) b
0.3
Out-of-Straightness (mm)
0.25
0.2
Side (b)
0.15
0.1
0.05
0
0 200 400 600 800 1000 1200
-0.05
-0.1
Length (mm)
0.4
a
0.35
b
0.3
Out-of-Straightness (mm)
0.25
Side (a)
0.2
Side (b)
0.15
0.1
0.05
0
0 200 400 600 800 1000 1200
-0.05
-0.1
Length (mm)
332
Appendix A
1
a
0.8 b
Out-of-Straightness (mm)
0.6 Side (a)
0.4
Side (b)
0.2
0
0 200 400 600 800 1000 1200 1400 1600
-0.2
Length (mm)
1
a
0.8 b
Out-of-Straightness (mm)
0.6
Side (a)
0.4
0.2
Side (b)
0
0 200 400 600 800 1000 1200 1400 1600
-0.2
Length (mm)
333
Appendix A
1
a Side (a)
0.8 b
Out-of-Straightness (mm) Side (b)
0.6
0.4
0.2
0
0 200 400 600 800 1000 1200 1400 1600
-0.2
Length (mm)
1
a
Side (a)
0.8 b
Out-of-Straightness (mm)
0.6
0.4
Side (b)
0.2
0
0 200 400 600 800 1000 1200 1400 1600
-0.2
Length (mm)
334
Appendix A
1
a
0.6
Side (a)
0.4
0.2
Side (b)
0
0 200 400 600 800 1000 1200 1400 1600
-0.2
Length (mm)
335
Appendix B
Appendix B
B.1 General
mid-height (e’) for column sets 3 to 11 (plots of column sets 1 and 2 were shown earlier
in Figure 4.24). The imperfection values at mid-height are calculated based on Equation
4.1 and plotted versus the applied load in Figure B.1 to Figure B.3. It is noted that e’
varies with the applied load P and also differs slightly when calculated from both sides of
the columns. The initial imperfection is then estimated as the y-intercept (i.e. the value at
336
Appendix B
1.8
1.6
Specime
n 4-1, e
Mid-height Imperfection, e’ (mm) P=0 = 1.63
1.4
1.2
1
Specimen 4-2, eP=0= 0.77
0.8
Specimen 3-2, eP=0= 0.71
0.6
Figure B.1 Mid-height imperfections of column sets 3 and 4 versus the applied load.
1.8
1.6
1.4
Mid-height Imperfection, e’ (mm)
Specimen 6-1, e =
1.2 P=0 1.22
0.6
Specimen
Specimen 5-2, 5-1, e
P=0= 0.57
e P=0= 0.41
0.4
Specimen 5-3,
e P=0= 0.37
0.2
Specimen 6-3, eP=0= 0.17
0
0 10 20 30 40 50
Applied load, P (kN)
Figure B.2 Mid-height imperfections of column sets 5 and 6 versus the applied load.
337
Appendix B
7 Specimen 9, e = 7.0
P=0 4
Mid-height Imperfection, e’ (mm)
6 Specimen 7, e
P=0= 6.60
5
Specimen 11, e
4 P=0= 5.00
2
Specimen 8, eP=0= 0.92
1
0
-1 0 10 20 30 40 50 60 70
Specimen 10, eP=0= 2.04
-2
-3
Applied load, P (kN)
338