Professional Documents
Culture Documents
by
September 2005
STATEMENT
STATEMENT
I hereby declare that this thesis contains no material accepted for any other
degree or diploma in any university. To the best of my knowledge, this thesis contains no
material previously written or published by another person, except where due
acknowledgment is made in the text.
___________________________
HUY BINH PHAM
iii
ACKNOWLEDGMENTS
ACKNOWLEDGMENTS
I wish to thank the staff at the Monash University Civil Engineering Laboratories.
The completion of the experimental aspect would not be possible without the hard work
and assistance of Graeme Rundle, Kevin Nievaart, Jeffrey Doddrell, Glenn Davis, Peter
Dunbar, Roy Goswell, Don McCarthy and Alan Taylor. I would also like to thank
Andrew Sarkady and MBT (Australia) Pty. Limited for providing partial sponsoring for
the materials required for this research.
My gratitude also goes to Jennifer Manson for her great help during my whole
time at Monash University. I am also grateful to Associate Professor Jay Sanjayan for his
advice on the reliability study. Thanks also to my colleagues and friends for their support
and friendship.
Finally, I would like to thank my family for their constant encouragement through
out the course of my life.
v
SUMMARY
SUMMARY
Literature reviews and assessment studies of the existing strength models of the
two member types, shear-lap specimens and retrofitted beams, were conducted. It is
shown that while the shear-lap strength models are able to give reasonable predictions,
the debonding strength models for retrofitted beams generally produce inaccurate and
scattered results and some models are not based on the actual mechanism and therefore
lack grounding.
The experimental and numerical work on the testing of RC beams bonded with
CFRP on their soffit showed two main debonding modes, end debond and intermediate
span debond. It was found that the behaviour of retrofitted beams depends a great deal on
the CFRP stiffness, the CFRP bond length and the steel reinforcement amount. The
investigation also demonstrated the effectiveness of steel clamps and U-straps for
anchorage.
vii
SUMMARY
and the maximum concrete shear stress along the tension rebar layer. The design method
was verified with the results from the experimental and numerical work carried out in this
research as well as a large database of tests available in the literature. It was found that
the method can predict the strength of a retrofitted beam with good accuracy.
viii
TABLE OF CONTENTS
TABLE OF CONTENTS
I
STATEMENT III
ACKNOWLEDGMENTS V
SUMMARY VII
TABLE OF CONTENTS IX
LIST OF FIGURES XV
CHAPTER 1 - INTRODUCTION 1
2.1 Introduction 7
2.2 Shear-lap specimens under tensile testing 7
2.2.1 Experimental work 7
2.2.2 Prediction formulae for bond strength 13
2.3 Retrofitted beams under bending 20
2.3.1 Experimental investigations 20
2.3.2 Existing theoretical studies for flexural failure 31
2.3.3 Existing theoretical studies for intermediate span debond 32
2.3.4 Existing theoretical studies for end debond 43
2.3.5 Design guidelines 56
ix
TABLE OF CONTENTS
3.1 Introduction 79
3.2 Assessment of analytical studies of shear-lap specimens 79
3.2.1 Database 79
3.2.2 Validation results 81
3.3 Assessment of analytical studies of beam specimens 85
3.3.1 Introduction 85
3.3.2 Existing database and new database 86
3.3.3 Validation of beam theory 89
3.3.4 Validation of existing models to predict intermediate span
debond 90
3.3.5 Validation of existing models to predict end debond 93
3.4 Validation of design guidelines 101
3.5 Summary of findings 101
x
TABLE OF CONTENTS
xi
TABLE OF CONTENTS
xii
TABLE OF CONTENTS
xiii
TABLE OF CONTENTS
REFERENCES 367
LIST OF PUBLICATIONS
xiv
LIST OF FIGURES
LIST OF FIGURES
Figure 2.1 Different set-ups for shear-lap tests: a) Double pull-pull test; b) Single pull-
push test; c) Bending test .............................................................................................8
Figure 2.2 Failure modes observed in shear-lap tests.........................................................9
Figure 2.3 FRP-concrete bond specimen (a) top view (b) strain distribution along FRP
and (c) shear stress distribution along FRP (Lee, 2003)............................................10
Figure 2.4 Measured bond-slip curves for a test series by Nakaba et al. (2001) ..............12
Figure 2.5 Bond-slip curves plotted by Savoia et al. (2003) ............................................13
Figure 2.6 Deformation and stresses of a bonded joint (Yuan et al., 2001) .....................16
Figure 2.7 Bond-slip relations: (a) Linear ascending; (b) bilinear relation; and (c) linear
descending .................................................................................................................16
Figure 2.8 A bond-slip curve following Popovics’ equation............................................19
Figure 2.9 Flexural strengthening using FRP or steel plates ............................................24
Figure 2.10 Flexural strengthening using FRP fabrics/sheets ..........................................24
Figure 2.11 Observed failure modes.................................................................................25
Figure 2.12 Shear strengthening using steel plates, FRP plates or FRP sheets ................26
Figure 2.13 FRP efficiency ratio (max. reported strain / strain at rupture) versus relative
stiffness (Bonacci and Maalej, 2001) ........................................................................30
Figure 2.14 Flexural crack debond (Sebastian, 2001) ......................................................32
Figure 2.15 Similarity between the shear-lap test and the situation near a flexural crack34
Figure 2.16 Forces acting on an element of a composite beam (Niu and Wu, 2001).......35
Figure 2.17 Debonding mechanism due to flexural cracks (Niu and Wu, 2001) .............35
Figure 2.18 Niedermeier’s methodology. .........................................................................37
Figure 2.19 Maximum possible increase in tensile stress between two subsequent cracks
(fib Bulletin 14, 2001)................................................................................................38
Figure 2.20 Displacements near the tip of a shear crack ..................................................39
Figure 2.21 Differential crack mouth opening displacements causing peeling (Neubauer
and Rostasy, 1999).....................................................................................................40
Figure 2.22 A peel test (Karbhari et al., 1997) .................................................................41
Figure 2.23 Crack sliding model (Mohamed Ali, 2000)...................................................42
Figure 2.24 Predicted shear and normal stress distributions near the plate end (Roberts,
1989) ..........................................................................................................................44
xv
LIST OF FIGURES
Figure 2.25 Stresses acting on a FRP laminate element near the cut-off point (Malek et
al, 1998) .....................................................................................................................46
Figure 2.26 Definition of bond development length for retrofitted beams by Nguyen et al.
(2001) .........................................................................................................................49
Figure 2.27 Stresses acting at the cut-off point of FRP (Saadatmanesh and Malek, 1998)
....................................................................................................................................49
Figure 2.28 Teeth scenario (Zhang and Raoof, 1995) ......................................................54
Figure 2.29 Shear-lap scenario for end debond ................................................................56
Figure 2.30 Finite element mesh adopted (a) and comparison of strain distributions
reported (b) by Maeda et al. (1997) ...........................................................................64
Figure 2.31 A typical FE mesh adopted by Chen et al. (2001).........................................65
Figure 2.32 The FE mesh (above) and the crack patterns for two example specimens with
bond lengths of 50 mm (below left) and 130 mm (below right) as reported in Lee
(2003) .........................................................................................................................66
Figure 2.33 Boundary conditions and mesh discretization by Camata et al. (2003) ........66
Figure 2.34 A FE mesh adopted (a) and the stress contours near the plate end reported (b)
by Teng et al. (2002) ..................................................................................................68
Figure 2.35 FE discretization model (a), loading curve (b) and FRP stress distributions
(c) as reported by Niu and Wu (2001) .......................................................................69
Figure 2.36 A crack pattern at failure reported by Ross et al. (1999)...............................70
Figure 2.37 A typical two-dimensional model of retrofitted beams adopted by Rahimi
and Hutchinson (2001)...............................................................................................71
Figure 2.38 Node and element displacements for RC beam model with slip in plate by
Aprile et al. (2001) .....................................................................................................73
Figure 2.39 Comparison of load-deflection curves for Zarnic et al.’s beams by Wong and
Vecchio (2003)...........................................................................................................73
Figure 2.40 The three-dimensional model of FRP plated beams adopted by Wong et al.
(2001) .........................................................................................................................74
Figure 2.41 Initial FE mesh (left) and crack pattern at a high load level (right) reported
by Yang et al. (2003)..................................................................................................75
Figure 3.1 Validation results for Van Gemert’s model (a) and ........................................81
Figure 3.2 Validation results for Maeda et al.’s model (a) and Izumo et al.’ model (b) ..82
Figure 3.3 Validation results for Niedermeier’s model (a), Neubauer and Rostasy’s
model (b) and Chen and Teng’s model (c) ................................................................83
xvi
LIST OF FIGURES
Figure 3.4 Validation results for Kanakubo et al.’s model (a) and Ulaga and Vogel’s
model (b)....................................................................................................................84
Figure 3.5 Validation of the beam theory .........................................................................90
Figure 3.6 Validation results for Arya and Farmer’s model (a) and Shehata et al.’s model
(b)...............................................................................................................................91
Figure 3.7 Validation results for Maruyama and Ueda’s model (a) and Teng et al.’s
model (b)....................................................................................................................92
Figure 3.8 Validation results for Blaschko (1997) model (a) and Mohamed Ali (2000)
model (b)....................................................................................................................92
Figure 3.9 Validation results for Niedermeier’s model ...................................................93
Figure 3.10 Validation results for Shehata et al.’s model.................................................94
Figure 3.11 Validation results for Nguyen et al.’s model (a) and Fanning and Kelly’s
model (b)....................................................................................................................94
Figure 3.12 Validation results for Saadatmanesh and Malek’s model (a), Tumialan et
al.’s model (b) and El-Mihilmy and Tedesco’s model (c) .........................................95
Figure 3.13 Validation results for Oehlers’ model (a) and Smith and Teng’s model (b) 96
Figure 3.14 Validation results for Jansze’s model (a) and Ahmed and Van Gemert’s
model (b)....................................................................................................................97
Figure 3.15 Validation results for Ziraba et al.’s model ..................................................97
Figure 3.16 Validation results for Raoof and Hassanen’s model – lower limit (a) and
Chaallal et al.’s model (b)..........................................................................................98
Figure 3.17 Validation results for Neubauer and Rostasy’s model (a) and Neubauer and
Rostasy’s model with Arya and Farmer’s limit (b) ...................................................99
Figure 3.18 Validation of the methods recommended in ACI 440.................................101
Figure 4.1 Specimen dimensions and test set-up............................................................104
Figure 4.2 MBrace FRP system components and main tools used for installation ........106
Figure 4.3 Locations of strain gauges on CFRP shear-lap specimens............................109
Figure 4.4 Loading and gripping method (a) and a specimen before testing (b)............110
Figure 4.5 Failure modes ................................................................................................112
Figure 4.6 Typical crack surfaces in interfacial debond failure .....................................113
Figure 4.7 A typical crack surface in shear-tension failure ...........................................113
Figure 4.8 Layers of the composite ................................................................................115
Figure 4.9 Strain distributions along bond length in specimen T1a (Lf = 60 mm).........116
Figure 4.10 Strain distributions along bond length in specimen T2a (Lf = 80 mm).......116
xvii
LIST OF FIGURES
Figure 4.11 Strain distributions along bond length in specimen T3a (Lf = 100 mm).....117
Figure 4.12 Strain distributions along bond length in specimen T4a (Lf = 140 mm).....117
Figure 4.13 Strain distributions along bond length in specimen T5a (Lf = 180 mm).....117
Figure 4.14 Strain distributions along bond length in specimen T6a (Lf = 220 mm).....118
Figure 4.15 Strain distributions along bond length in specimen T7a (H = 5 mm) .........118
Figure 4.16 Strain distributions along bond length in specimen T8a (H = 70 mm) .......119
Figure 4.17 Strain distributions along bond length in specimen T12a (bf = 50 mm) .....119
Figure 4.18 Strain distributions along bond length in specimen T13a (bf = 70 mm) .....120
Figure 4.19 Strain distributions along bond length in specimen T9a (Lf = 100 mm).....120
Figure 4.20 Strain distributions along bond length in specimen T10a (Lf = 140 mm)...121
Figure 4.21 Strain distributions along bond length in specimen T11a (Lf = 180 mm)...121
Figure 4.22 Average bond stress distributions along bond length in specimen T4a ......122
Figure 4.23 Average bond stress distributions along bond length in specimen T5a ......122
Figure 4.24 Average bond stress distributions along bond length in specimen T6a ......123
Figure 4.25 Bond stress versus slip development in specimen T4a (bf = 100 mm) .......124
Figure 4.26 Bond stress versus slip development in specimen T5a (bf = 100 mm) .......124
Figure 4.27 Bond stress versus slip development in specimen T6a (bf = 100 mm) .......125
Figure 4.28 Bond stress versus slip development in specimen T8a (bf = 100 mm) .......125
Figure 4.29 Fitted bond-slip relationships for specimens with two layers of CFRP of 100
mm width .................................................................................................................125
Figure 4.30 Bond stress versus slip development in specimen T12a (bf = 50 mm) .......126
Figure 4.31 Bond stress versus slip development in specimen T13a (bf = 70 mm) .......126
Figure 4.32 Load versus slip curves for specimens with two layers of CFRP and variable
bond lengths .............................................................................................................127
Figure 4.33 Load versus slip curves for specimens with six layers of CFRP and variable
bond lengths (subjected to alignment errors)...........................................................127
Figure 4.34 Effect of bond length on bond capacity.......................................................128
Figure 4.35 Load-slip curves for specimens with different CFRP thicknesses T3a (2
plies) and T9a (6 plies).............................................................................................128
Figure 4.36 Effect of CFRP thickness on bond capacity ................................................129
Figure 4.37 Load-slip curves for specimens with different CFRP widths T3a (100 mm),
T12a (50 mm) and T13a (75 mm)............................................................................129
Figure 4.38 Effect of CFRP width on bond capacity......................................................130
xviii
LIST OF FIGURES
Figure 4.39 Load-slip curves for specimens with different support clearances T3a (110
mm), T7a (5 mm) and T8a (70 mm) ........................................................................130
Figure 4.40 Effect of varying support clearance on bond capacity ................................131
Figure 5.1 Preliminary beam dimensions and test set-up ...............................................135
Figure 5.2 Load-displacement curves of two trial beams ...............................................135
Figure 5.3 Beams B1 and B2 after failure ......................................................................136
Figure 5.4 CFRP strain distributions ..............................................................................137
Figure 5.5 Average bond stress distributions in shear span............................................137
Figure 5.6 Photos of beams B1 (a) and B2 (b) near failure ............................................138
Figure 5.7 Applicability of theoretical models to predict IS debond loads for beam B1
.................................................................................................................................139
Figure 5.8 Applicability of theoretical models to predict end debond loads for beam B2
.................................................................................................................................139
Figure 5.9 Typical rectangular beam cross-sectional details (beams E1a and E1b).......141
Figure 5.10 Typical beam longitudinal details (beams E1a and E1b) ............................141
Figure 5.11 A steel clamp ...............................................................................................143
Figure 5.12 Surface texture after water jetting (a) and after priming (b) .......................144
Figure 5.13 Bonding operation (a) Applying a layer of resin; (b) Placing fibres on the
resin; (c) Pressing fibres into the resin with a ripped roller; (d) Finished surface...145
Figure 5.14 Strain gauge locations on beam E1a............................................................146
Figure 5.15 Load and deflection measurements .............................................................146
Figure 5.16 Test rig.........................................................................................................148
Figure 5.17 A support (a) and an anchorage of the reaction frame column (b)..............149
Figure 5.18 Load-deflection behaviour of the control beams.........................................150
Figure 5.19 Photos of beams C1a and C1b after failure .................................................150
Figure 5.20 Load-deflection behaviour of E beams .......................................................151
Figure 5.21 Load-deflection behaviour of beams E1a and E1b .....................................152
Figure 5.22 Photos of the debonded side of E1 beams ...................................................152
Figure 5.23 High-speed video captures for beam E1a at failure (end debond) (rate: 500
frames per second) ...................................................................................................153
Figure 5.24 End cover peeling failure surface in beam E1b...........................................154
Figure 5.25 Influence of stirrup extrusion on debond cracks in beam E1a ....................154
Figure 5.26 Load-deflection behaviour of beams E2a and E2b .....................................155
Figure 5.27 Photos of the debonded side of E2 beams ...................................................155
xix
LIST OF FIGURES
xx
LIST OF FIGURES
Figure 5.60 CFRP strain development in beam S2a on unclamped side ........................176
Figure 5.61 Tensile steel distributions in beam S2a on unclamped side ........................177
Figure 5.62 Stirrup strain development in beam S2a on unclamped side.......................177
Figure 5.63 Effect of CFRP and tension steel stiffness ratio..........................................179
Figure 5.64 Effect of CFRP bond length ........................................................................179
Figure 5.65 Crack patterns on the clamped side of E1a and S1a....................................181
Figure 5.66 Non-prestressed U strap system ..................................................................184
Figure 5.67 Prestressed U strap system ..........................................................................184
Figure 5.68 Non-prestressed U-strap system..................................................................186
Figure 5.69 Prestressed U-strap system..........................................................................186
Figure 5.70 Load and deflection measurement for beams tested in three-point bending
.................................................................................................................................187
Figure 5.71 Strain gauge locations on U-straps..............................................................187
Figure 5.72 Load-deflection behaviour of retested beams .............................................188
Figure 5.73 Load-deflection behaviour of beams A1a, A1b, A2a and A2b ...................188
Figure 5.74 Photo of the debonded side of beam E3b2 ..................................................189
Figure 5.75 Photos of the debonded side of beams anchored with a non-prestressed U
strap..........................................................................................................................190
Figure 5.76 Sliding of the longitudinal CFRP and local debonding and rupture of a strap
leg in beam E3a........................................................................................................190
Figure 5.77 Photos of the debonded side of beams anchored with a prestressed U strap
.................................................................................................................................191
Figure 5.78 Debonding of a leg of the prestressed strap in beam A1a ...........................191
Figure 5.79 Photos of beams anchored with three non-prestressed U straps after failure
.................................................................................................................................192
Figure 5.80 Photos of beams anchored with three prestressed U straps after failure .....193
Figure 5.81 Crack patterns in beam E3b2.......................................................................194
Figure 5.82 Crack patterns in beams with one U-strap...................................................194
Figure 5.83 Crack patterns in beams with three U-straps...............................................195
Figure 5.84 Longitudinal CFRP strain distributions in beam A1a .................................197
Figure 5.85 Longitudinal CFRP strain development in beam A1a.................................197
Figure 5.86 Strain development in the U-strap legs in beam A1a ..................................197
Figure 5.87 Longitudinal CFRP strain distributions in beam A1b .................................198
Figure 5.88 Longitudinal CFRP strain development in beam A1b ................................198
xxi
LIST OF FIGURES
Figure 5.89 Strain distributions along U strap height in beam A1b at different load levels
..................................................................................................................................198
Figure 5.90 Strain development in the U-strap legs in beam A1b..................................199
Figure 5.91 Strain development in the tension reinforcement in beam A1b ..................199
Figure 5.92 Longitudinal CFRP strain distributions in beam A2a .................................200
Figure 5.93 Longitudinal CFRP strain development in beam A2a.................................200
Figure 5.94 Strain development in the U-strap legs in beam A2a ..................................200
Figure 5.95 Strain development in the tension reinforcement in beam A2a...................201
Figure 5.96 Longitudinal CFRP strain distributions in beam A2b .................................202
Figure 5.97 Longitudinal CFRP strain development in beam A2b.................................202
Figure 5.98 Strain distributions along U strap height in beam A2b at different load levels
..................................................................................................................................202
Figure 5.99 Strain development in the U-strap legs in beam A2b..................................203
Figure 5.100 Strain development in the tension reinforcement in beam A2b ................203
Figure 6.1 Two-dimensional mesh of a typical specimen (T3) ......................................208
Figure 6.2 CQ12M element (a) and CQ16M element (b)...............................................209
Figure 6.3 Loading-unloading for total strain crack models (de Witte and Kikstra, 2003)
..................................................................................................................................210
Figure 6.4 Concrete stress-strain curve in tension (a) and compression (b) ...................212
Figure 6.5 Modified Newton-Raphson iteration (de Witte and Kikstra, 2003) ..............213
Figure 6.6 Deformed shapes of model T5 at different load levels (magnification factor =
60) ............................................................................................................................215
Figure 6.7 Deformed shapes of model T1 at different load levels (magnification factor =
60) ............................................................................................................................216
Figure 6.8 Crack patterns in models T1 to T6 ................................................................217
Figure 6.9 Crack patterns in models T7 and T8..............................................................217
Figure 6.10 Crack patterns in models T9, T10 and T11 .................................................217
Figure 6.11 Crack patterns in models T12 and T13........................................................217
Figure 6.12 Correlation of peak loads predicted by the FE model and measured in the
experiment................................................................................................................219
Figure 6.13 Effect of varying CFRP bond length (Experiment and FE) ........................220
Figure 6.14 Effect of varying support block height H (Experiment and FE) .................220
Figure 6.15 Effect of varying CFRP thickness (Experiment and FE) ............................221
Figure 6.16 Effect of varying CFRP width (Experiment and FE) ..................................221
xxii
LIST OF FIGURES
Figure 6.17 Comparison of CFRP strain distributions in specimens T1a, T2a and T3a 223
Figure 6.18 Comparison of CFRP strain distributions in specimens T4a, T5a and T6a 224
Figure 6.19 Comparison of CFRP strain distributions in specimens T7a and T8a ........225
Figure 6.20 Comparison of CFRP strain distributions in specimens T9a, T10a and T11a
.................................................................................................................................226
Figure 6.21 Comparison of CFRP strain distributions in specimens T12a and T13a ....227
Figure 6.22 Comparison of load-slip curves for specimens T1a to T6a.........................228
Figure 6.23 Comparison of load-slip curves for specimens T9a, T10a and T11a ..........229
Figure 6.24 Comparison of load-slip curves for specimens T7a, T8a, T12a and T13a..229
Figure 6.25 Comparison of local bond-slip relationships in specimen T5a (bf = 100 mm,
Lf = 180 mm) ...........................................................................................................230
Figure 6.26 Comparison of local bond-slip relationships in specimen T6a (bf = 100 mm,
Lf = 220 mm) ...........................................................................................................231
Figure 6.27 Comparison of local bond-slip relationships in specimen T12a (bf = 50 mm,
Lf = 100 mm) ...........................................................................................................231
Figure 6.28 Comparison of local bond-slip relationships in specimen T13a (bf = 70 mm,
Lf = 180 mm) ...........................................................................................................232
Figure 6.29 Gaussian integration points in the CFRP and adhesive elements ...............233
Figure 6.30 CFRP strain distributions in model T5........................................................234
Figure 6.31 CFRP strain distributions in model T11......................................................235
Figure 6.32 Prediction of CFRP strains near the loaded end at different load levels using
different mesh sizes .................................................................................................237
Figure 6.33 Comparison of crack patterns after peak in specimens T5 (left) and T1 (right)
predicted by the models with different mesh sizes ..................................................238
Figure 6.34 Comparison of load-slip curves predicted by the models with different mesh
sizes..........................................................................................................................238
Figure 6.35 Different meshes used .................................................................................239
Figure 6.36 Comparison of crack patterns after peak in specimens T5 (left) and T1 (right)
predicted by the models with different mesh types .................................................240
Figure 6.37 Comparison of load-slip curves predicted by the models with different mesh
types .........................................................................................................................240
Figure 6.38 Comparison of crack patterns after peak in specimens T5 (left) and T1 (right)
predicted by the models with different Gf values ....................................................242
xxiii
LIST OF FIGURES
Figure 6.39 Comparison of load-slip curves predicted by the models with different Gf
values .......................................................................................................................242
Figure 6.40 Comparison of crack patterns after peak in specimens T5 (left) and T1 (right)
predicted by the models with different fct values .....................................................243
Figure 6.41 Comparison of load-slip curves predicted by the models with different fct
values .......................................................................................................................243
Figure 6.42 Comparison of crack patterns after peak in specimens T5 (left) and T1 (right)
predicted by the models with different Ea values.....................................................245
Figure 6.43 Comparison of load-slip curves predicted by the models with different Ea
values .......................................................................................................................245
Figure 6.44 Two-dimensional mesh of a typical specimen (T3) ....................................246
Figure 6.45 L7BEN element ...........................................................................................247
Figure 6.46 3+3 node structural interface element (CL12I) ...........................................247
Figure 6.47 Idealised bond-slip relationships .................................................................247
Figure 6.48 Correlation of peak loads predicted by the bond-slip based model and
measured in the experiment .....................................................................................248
Figure 6.49 Comparison of the load-slip curves of specimens T5a and T1a..................249
Figure 6.50 Comparison of CFRP strain distributions of specimens T5a and T1a ........249
Figure 7.1 Two-dimensional mesh of beams loaded in 4-point bending........................251
Figure 7.2 Q8MEM element (a) and L7BEN element (b) ..............................................252
Figure 7.3 2+2 node structural interface element (L8IF)................................................252
Figure 7.4 Steel and CFRP reinforcement locations in model E1 ..................................253
Figure 7.5 Reduced concrete element width...................................................................253
Figure 7.6 Modelling of the steel clamp .........................................................................254
Figure 7.7 Concrete shear strength after cracking ..........................................................255
Figure 7.8 Bond-slip relationships adopted for steel bar/concrete interface (a) and
CFRP/concrete interface (b).....................................................................................257
Figure 7.9 Quasi-Newton iteration (de Witte and Kikstra, 2003)...................................258
Figure 7.10 Crack pattern in model C1...........................................................................259
Figure 7.11 Crack patterns in E1 (unclamped side)........................................................260
Figure 7.12 Crack patterns in model S2 (clamped side) .................................................261
Figure 7.13: Crack patterns in model S1 (unclamped side).............................................262
Figure 7.14 Deformed shapes of the unclamped side of model E1 near the peak load
(magnification factor = 20) ......................................................................................262
xxiv
LIST OF FIGURES
Figure 7.15 Deformed shapes of the clamped side of model S2 near the peak load
(magnification factor = 20) ......................................................................................263
Figure 7.16 Deformed shapes of the unclamped side of model E1 near the peak load
(magnification factor = 20) ......................................................................................263
Figure 7.17 Induced tooth-like crack..............................................................................264
Figure 7.18 Flexure-shear crack debond.........................................................................264
Figure 7.19 Crack patterns in S models at the peak loads ..............................................265
Figure 7.20 Crack patterns in E models (unclamped side) at the peak loads .................266
Figure 7.21 Correlation of peak loads as predicted by the FE model and measured in the
experiment ...............................................................................................................266
Figure 7.22 Comparison of load-displacement curves for S beams ...............................267
Figure 7.23 Comparison of load-displacement curves for E beams ...............................268
Figure 7.24 Gauss integration points in a CFRP beam element .....................................269
Figure 7.25 Variation of CFRP strain along the top and bottom halves in model S2N .269
Figure 7.26 Comparison of CFRP strain distributions in beam E1a ..............................270
Figure 7.27 Comparison of CFRP strain distributions in beam S2a...............................270
Figure 7.28 Comparison of CFRP strain distributions in beam S1a...............................271
Figure 7.29 Comparison of tensile steel strain distributions in beam E1a .....................272
Figure 7.30 Comparison of tensile steel strain distributions in beam S2a......................272
Figure 7.31 Comparison of tensile steel strain distributions in beam S1a......................273
Figure 7.32 Comparison of crack patterns at the peak loads in beams E1 (a) and S2C (b)
predicted by models with different Gf values ..........................................................275
Figure 7.33 Comparison of load-deflection curves predicted by models with different Gf
values .......................................................................................................................276
Figure 7.34 Comparison of crack patterns at the peak loads in beams E1 (a) and S2C (b)
predicted by models with different fct values...........................................................276
Figure 7.35 Comparison of load-deflection curves predicted by models with different fct
values .......................................................................................................................277
Figure 7.36 Comparison of crack patterns at the peak loads in beams E1 (a) and S2C (b)
predicted by models with different β values............................................................278
Figure 7.37 Comparison of load-deflection curves predicted by models with different β
values .......................................................................................................................278
xxv
LIST OF FIGURES
Figure 7.38 Behavioural trend of retrofitted beams with different CFRP cross sectional
areas .........................................................................................................................279
Figure 7.39 Behavioural trend of retrofitted beams with different tension reinforcement
cross sectional areas .................................................................................................280
Figure 7.40 Behavioural trend of retrofitted beams with different CFRP bond lengths.280
Figure 7.41 Behavioural trend of retrofitted beams with different concrete covers .......281
Figure 7.42 Mesh of a beam loaded in three-point bending ...........................................282
Figure 7.43 Modelling of reinforcements .......................................................................283
Figure 7.44 Location of interface elements in the FE mesh for model A2a ...................284
Figure 7.45 Crack patterns in retested beams at peak.....................................................285
Figure 7.46 Crack patterns in A beams at peak ..............................................................286
Figure 7.47 Deformed shapes after peak of beams with three straps (magnification factor
= 10) .........................................................................................................................286
Figure 7.48 Correlation of peak loads as predicted by the FE model and measured in the
experiment................................................................................................................287
Figure 7.49 Comparison of load-deflection curves for retested beams ..........................287
Figure 7.50 Comparison of load-deflection curves for A beams....................................288
Figure 7.51 Comparison of longitudinal CFRP strain distributions in beam A1a..........289
Figure 7.52 Comparison of strain development on U-strap sides in beam A1a .............289
Figure 8.1 Hillerborg’s Fictitious Crack Model (Saouma, 2002) ...................................293
Figure 8.2 Interface idealization and notations (Saouma, 2002) ....................................293
Figure 8.3 Failure surface (Saouma, 2002).....................................................................294
Figure 8.4 Bi-linear softening laws (Saouma, 2002) ......................................................294
Figure 8.5 Stiffness degradation in the equivalent uniaxial case (Saouma, 2002) .........295
Figure 8.6 Initial crack locations in specimens T1-T6....................................................297
Figure 8.7 Final transverse crack paths in specimens T1-T6..........................................297
Figure 8.8 Two typical meshes of the shear-lap specimens............................................298
Figure 8.9 Three-node plane stress element (a) and four-node interface element (b) ...298
Figure 8.10 Deformed shapes of model T5 at different load levels (magnification factor =
60) ............................................................................................................................302
Figure 8.11 Deformed shapes of model T1 at different load levels (magnification factor =
60) ............................................................................................................................303
Figure 8.12 Discrete crack opening in models T1 to T6 at the peak loads (magnification
factor = 60)...............................................................................................................304
xxvi
LIST OF FIGURES
Figure 8.13 Comparison of load-slip behaviours between the experiments, FEA using
smeared crack modelling and FEA using discrete/smeared crack modelling .........305
Figure 8.14 Comparison of CFRP strain distributions in specimens T1a, T2a and T3a
(exp. vs discrete/smeared crack model) ...................................................................306
Figure 8.15 Comparison of CFRP strain distributions in specimens T4a, T5a and T6a
(exp. vs discrete/smeared crack model) ...................................................................307
Figure 8.16 Comparison of load-slip curves predicted by the models with different GIf
values .......................................................................................................................308
Figure 8.17 Comparison of load-slip curves predicted by the models with different fct
values .......................................................................................................................308
Figure 8.18 Comparison of load-slip curves predicted by the models with different GIIf
values .......................................................................................................................309
Figure 8.19 Discrete crack locations in model E1 ..........................................................310
Figure 8.20 A typical mesh of retrofitted beams ............................................................311
Figure 8.21 Steel clamp location in model S2C .............................................................312
Figure 8.22 Deformed shape of the unclamped side of beam E1 at the peak load
(magnification factor = 40) ......................................................................................314
Figure 8.23 Deformed shape of the clamped side of beam S2C at the peak load
(magnification factor = 30) ......................................................................................315
Figure 8.24 Deformed shape of the unclamped side of beam S1N at the peak load
(magnification factor = 30) ......................................................................................316
Figure 8.25 Deformed shapes of the unclamped side of other E models at the peak loads
(magnification factor = 30) ......................................................................................317
Figure 8.26 Deformed shapes of other S models at the peak loads (magnification factor =
30) ............................................................................................................................318
Figure 8.27 Comparison of load-displacement curves for E beams ...............................320
Figure 8.28 Comparison of load-displacement curves for S beams ...............................321
Figure 8.29 Comparison of CFRP strain distributions in beam E1a (exp. vs
discrete/smeared crack model).................................................................................322
Figure 8.30 Comparison of CFRP strain distributions in beam S2a (exp. vs
discrete/smeared crack model).................................................................................322
Figure 8.31 Comparison of CFRP strain distributions in beam S1a (exp. vs
discrete/smeared crack model).................................................................................323
xxvii
LIST OF FIGURES
Figure 8.32 Comparison of tensile steel strain distributions in beam E1a (exp. vs
discrete/smeared crack model).................................................................................323
Figure 8.33 Comparison of tensile steel strain distributions in beam S2a (exp. vs
discrete/smeared crack model).................................................................................324
Figure 8.34 Comparison of tensile steel strain distributions in beam S1a (exp. vs
discrete/smeared crack model).................................................................................324
Figure 8.35 Comparison of load-displacement curves predicted by the models with
different GIf values ...................................................................................................325
Figure 8.36 Comparison of load-displacement curves predicted by the models with
different fct values ....................................................................................................326
Figure 8.37 Comparison of load-displacement curves predicted by the models with
different GIIf values ..................................................................................................327
Figure 9.1 Strain distribution before (a) and after installation (b) of FRP......................330
Figure 9.2 Flow chart for implementation of sectional analysis.....................................331
Figure 9.3 Local variations in steel strain and bond stress .............................................332
Figure 9.4 Calculation model for a RC member under pure flexure (Comite Euro-
International du Beton, 1985) ..................................................................................333
Figure 9.5 Sectional analysis results along beam E1a at maximum load level ..............334
Figure 9.6 Sectional analysis results along beam S1a at maximum load level...............335
Figure 9.7 Comparison of FRP strain distributions between experimental results and
predictions by beam theory in beam E1a .................................................................336
Figure 9.8: Comparison of FRP strain distributions between experimental results and
predictions by beam theory in beam S1a .................................................................336
Figure 9.9 Comparison of measured and calculated CFRP strains at the ultimate load
level for retrofitted beams without anchorage .........................................................337
Figure 9.10 Failure mechanisms .....................................................................................337
Figure 9.11 Average shear stress concept.......................................................................339
Figure 9.12 Variation of maximum average bond stress in beams failed by end debond
(a) and maximum tensile force in CFRP in beams failed by intermediate span
debond or mixed debond (b) ....................................................................................340
Figure 9.13 Prediction of end and intermediate span failure loads for the beams tested in
the present study ......................................................................................................342
Figure 9.14 Comparison between the theoretical predictions and experimental results.343
Figure 9.15 Comparison of the predicted trends with different CFRP thicknesses........344
xxviii
LIST OF FIGURES
Figure 9.16 Comparison of the predicted trends with different tensile steel amounts ...344
Figure 9.17 Comparison of the predicted trends with different CFRP bond lengths .....345
Figure 9.18 Comparison of the predicted trends with different concrete covers............345
Figure 9.19 Cross sections of T beams used...................................................................349
Figure 9.20 Side view of T beams and loading positions...............................................350
Figure 9.21 Design point and reliability index according to the first order reliability
method for normally distributed uncorrelated variables (Eurocode 2, 2002)..........356
xxix
LIST OF TABLES
LIST OF TABLES
Table 1.1 Typical properties of fibres (fib Bulletin 14, 2001) ............................................3
Table 2.1 Experimental studies on beams strengthened in flexure...................................22
Table 2.2 Different predictions for concrete tensile strength ...........................................61
Table 2.3 Different predictions for fracture energy..........................................................63
Table 3.1 Summary of the database of shear-lap tests......................................................80
Table 3.2 Variability of test parameters and results in the shear-lap test database ..........80
Table 3.3 Summary of validation results of strength models for shear-lap tests..............85
Table 3.4 Summary of the database of beam tests............................................................88
Table 3.5 Variability of some test parameters in the beam test database .........................89
Table 3.6 Summary of validation results of debond models for beams .........................100
Table 4.1 Variables in the experimental program...........................................................105
Table 4.2 Mechanical properties of MBrace FRP system as given by the manufacturer
.................................................................................................................................106
Table 4.3 Measured elastic modulus of the composite...................................................107
Table 4.4 Shear-lap test results .......................................................................................114
Table 5.1 Mechanical properties of materials used ........................................................134
Table 5.2 Variables in experimental program No. 1.......................................................142
Table 5.3 Mechanical properties of steel reinforcement.................................................144
Table 5.4 Summary of beam test results.........................................................................166
Table 5.5 Maximum CFRP strains and bond stresses ....................................................178
Table 5.6 Variables in experimental program No. 2.......................................................183
Table 5.7 Experimental results .......................................................................................194
Table 5.8 Maximum U-strap leg strains .........................................................................204
Table 6.1 Concrete properties adopted in the finite element model ...............................212
Table 6.2 Comparison between the experimental and the FEA results ..........................218
Table 6.3 Mesh size variations .......................................................................................236
Table 6.4 Mesh type variations.......................................................................................239
Table 6.5 Concrete tensile property variations ...............................................................241
Table 6.6 Adhesive modulus variations..........................................................................244
Table 6.7 The parameters defining Popovics’ bond-slip curve ......................................248
Table 7.1 Concrete properties used in the finite element model ....................................256
xxxi
LIST OF TABLES
Table 7.2 Steel properties used in the finite element model ..........................................256
Table 7.3 Concrete tensile property variations ...............................................................274
Table 7.4 Shear retention factor variations .....................................................................277
Table 8.1 CFRP and adhesive material model parameters ..............................................299
Table 8.2 Concrete smeared crack model parameters for shear-lap blocks....................299
Table 8.3 Interface crack model parameters for shear-lap blocks ..................................300
Table 8.4 Mode-II fracture energy variations .................................................................309
Table 8.5 Concrete smeared crack model parameters for S beams ................................312
Table 8.6 Interface crack model parameters for S beams ...............................................313
Table 8.7 Steel model parameters ...................................................................................313
Table 9.1 Design summary of T beam configurations....................................................350
Table 9.2 Calculation of model errors ............................................................................351
Table 9.3 Statistical parameters for some random variables from CONTECVET (2002)
..................................................................................................................................352
Table 9.4 Statistical parameters for FRP ultimate strength from Atadero et al. .............354
Table 9.5 Summary of variable statistical properties......................................................354
Table 9.6 Simulation results for the shear resistance R ..................................................355
Table 9.7 Capacity reduction factors corresponding to β = 3.25....................................358
Table 9.8 Capacity reduction factors recommended by AS5100....................................358
xxxii
NOTATION AND ABBREVIATION
xxxiii
NOTATION AND ABBREVIATION
xxxiv
NOTATION AND ABBREVIATION
xxxv
NOTATION AND ABBREVIATION
tl thickness of a laminate
uf FRP-to-concrete average bond strength
us steel-to-concrete average bond strength
*
V design shear force
V0 shear force at the cut-off point of a bonded laminate
Vcal calculated maximum shear force that a beam can support
Vexp actual maximum shear force that a beam can support
Vdebond actual debonding shear strength of a beam
Vmax, Vu,no debond calculated ultimate shear strength of a retrofitted beam failing by
concrete crushing or rupture of the laminate
Vu,end debond ultimate shear strength of a retrofitted beam failing by end debond
Vu,IS debond ultimate shear strength of a retrofitted beam failing by intermediate span
debond
yf distance from the neutral axis to the centroid of FRP reinforcement
α calibration factor for the maximum FRP tensile force in a retrofitted beam
failing by intermediate span debond; or
ratio of the FRP stiffness to the concrete stiffness (EfAf / EcAc); or
model error (reliability context)
β shear retention factor for concrete (finite element context); or
reliability index (reliability context)
∆L distance between two strain gauge positions
∆u0 displacement increment (finite element context)
δui ith iterative displacement increment (finite element context)
ε, εc concrete strain
ε0 concrete strain at which the concrete stress reaches its maximum value
(concrete stress-strain curve context)
εcu concrete crushing strain
εf FRP reinforcement strain
εfu FRP ultimate strain
εs tension steel reinforcement strain
εsy yield strain of tension steel reinforcement
ε’s compression steel reinforcement strain
ζ distribution coefficient
xxxvi
NOTATION AND ABBREVIATION
µ mean value
µR mean of the resistance (reliability context)
νc Poisson’s ratio of concrete
σ stress; or
standard deviation (reliability context)
σ0 normal stress at a laminate cut-off point
σR standard deviation of the resistance (reliability context)
ΣObar total perimeter of tension reinforcing bars
ρf FRP reinforcement ratio
ρs steel reinforcement ratio
τ average bond stress between two gauge positions
τ0 shear stress at a laminate cut-off point
τave average shear stress on the end peeling debond surface
τmax maximum average bond stress between two positions
τult interface shear strength
φ strength reduction factor
AFRP aramid fibre reinforced polymer composites
CFRP carbon fibre reinforced polymer composites
FCD flexural crack debond
FE finite element
FEA finite element analysis
FRP fibre reinforced polymer composites
FSCD flexure-shear crack debond
GFRP glass fibre reinforced polymer composites
IS debond intermediate span debond
RC reinforced concrete
SCD shear crack debond
xxxvii
INTRODUCTION
CHAPTER 1 - INTRODUCTION
Recently, a second type of laminate bonding has been developed, which is fibre
reinforced polymer composite bonding or FRP bonding. This system makes use of a new
excellent material for construction: fibre reinforced polymer composite (FRP). The new
system possesses many advantages over the steel plate system such as high strength of
laminates, easy installation, excellent durability and minimum maintenance requirement.
FRP bonding can be used to strengthen not only reinforced concrete structures but also
masonry or timber structures (Bakis et al., 2002). Although it is a relatively new
1
CHAPTER 1
technique, FRP bonding has been used to strengthen numerous bridges and buildings
worldwide as well as in Australia in the last two decades. The Westgate Bridge
strengthening project in Melbourne Australia is one excellent example of this.
Fibres are polymeric materials. They are very small in diameter (for example
carbon fibres are about 8 µm in diameter). The most common types used in construction
are carbon, glass and aramid fibres. Carbon fibres, which can have very high modulus,
can be produced from two main sources: polyacrylonitrile fibres (PAN) and mesophase
pitch. Carbon fibres from the first source are used in construction because they have
higher strength and modulus. Glass fibres, which have lower modulus, are generally
cheaper. Three most popular types of glass fibres are E-glass, S-glass and C-glass.
Aramid fibres are synthesised from aromatic polyamides. Their modulus values lie
between those of carbon and glass fibres (Hull and Clyne, 1996).
The main mechanical properties of different fibres are summarised in Table 1.1.
Carbon fibres have high strength and elastic modulus, excellent fatigue properties, and
excellent moisture and chemical resistance. Glass fibres have lower strength and are
susceptible to alkaline attack, while aramid fibres can be prone to moisture, temperature,
2
INTRODUCTION
UV and alkaline attack (Rizkalla, 2002). Thus, for strengthening purposes, carbon fibres
are possibly the most suitable material.
The fibres are bonded to a structure using an epoxy resin by an adhesion process.
The epoxy resin normally contains a resin, a hardener and additives. The adhesion
process to a concrete surface generally involves several steps. At first, the epoxy flows
over and into the irregularities and wets the surface. It is then attached to the surface by
molecular forces and solidifies to form the joint. Due to the similar nature of fibres and
epoxy resins, good adhesion is normally assured. A good bond between concrete and
resins is more difficult. The bond mechanism between these two materials is more
complex involving not only mutual atomic/molecular attraction but also mechanical
interlocking (Hollaway and Leeming, 1999).
The adhesion can be implemented using two main methods: adhesive bonding or
wet lay-up. In the first method, the composite is prefabricated and then bonded onto the
concrete substrate using an adhesive. In the wet lay-up method, the composite and bond
are formed at the same time by impregnating layers of fabrics with the resin on the
concrete surface. It is a slow application and the composites can be subjected to non-
uniform wetting and waviness. However, it has great flexibility for field applications, i.e.
the composites can have different thicknesses and sizes, and can be bonded on curved
3
CHAPTER 1
surfaces. Wet lay-up is also generally the cheapest option (Karbhari and Seible, 1999;
Karbhari, 2001).
The general aim of this research is, therefore, to study the short-term behaviour of
concrete members bonded with CFRP composites using wet lay-up method. The specific
aims are:
• To summarise and assess various theoretical models for predicting the capacity of
a retrofitted concrete member
• To gain further understanding of the local bond characteristics between CFRP and
concrete
• To study different failure mechanisms and the effect of a number of important
parameters to a the capacity of a RC beam
• To develop and verify numerical and theoretical models for prediction of critical
failure loads
• To investigate mechanical anchorage methods for debond prevention.
The outcome from this research will be to contribute to the knowledge that
currently exists in this field so that greater confidence in the prediction of the load-
carrying capacity of retrofitted concrete members can be achieved.
4
INTRODUCTION
5
LITERATURE REVIEW
2.1 Introduction
Numerous experimental and analytical studies on concrete members bonded with
laminates have been conducted covering a number of important aspects of structural
performance and durability. In the present study, only short-term structural behaviour
under static loading of two member types, concrete shear-lap specimens and retrofitted
RC beams, is considered. The literature review presented below is devoted to
experimental, theoretical and numerical work on FRP strengthening of the two concrete
member types.
This chapter is organised into three main sections. The first presents a review of
existing experimental and theoretical studies on the characteristics of the bond between
concrete and FRP using shear-lap tests. The second is a review of experimental and
theoretical studies on retrofitted beams, while the third is a review of nonlinear finite
element modelling and its usage to simulate the behaviour of shear-lap specimens and
retrofitted beams under static loading.
7
CHAPTER 2
FRP
(a) steel bar concrete
FRP
(b) support concrete
block
FRP
Figure 2.1 Different set-ups for shear-lap tests: a) Double pull-pull test; b) Single
pull-push test; c) Bending test
The observed failure modes reported in the literature are shown in Figure 2.2 with
the colour lines representing the failure surfaces (Neubauer and Rostasy, 1997; Chen and
Teng; 2001 and Yao et al., 2002). Mode 1 is called interfacial debond and it is the most
8
LITERATURE REVIEW
commonly observed. The failure surface is in the concrete a few millimetres beneath the
concrete-adhesive interface. A concrete prism may also be pulled out near the loaded end.
Mode 2 is called shear-tension failure, in which the main crack propagates into the
concrete block from the FRP end. This mode tends to occur only when the laminate is
relatively thick. For modes 3, 4 and 5, the failure occurs in the laminate and adhesive
materials. Mode 3, FRP tensile rupture, can be observed if the FRP cross-sectional area is
very small. Mode 4, cohesion failure through the adhesive, has also been reported. The
composite can also be delaminated as in mode 5, FRP delamination, where the
delamination path can bridge across the adhesive layer and penetrate into the concrete.
These last three modes are not common especially for normal strength concrete, of which
the shear strength is much lower compared to that of the adhesive and FRP. The
following discussion applies for mode 1 only, which is the most widely observed failure
mode.
FRP Mode 5
Mode 3 Mode 4
adhesive
Mode 1
concrete Mode 2
9
CHAPTER 2
cracking initiates under the adhesive layer near the loaded end leading to a flattening
portion of the strain curve as shown in curve ‘3’. Further increase in the load extends the
crack and gradually shifts the area of active bonding towards the unloaded end until the
whole FRP is debonded from the concrete block. This is characterised by the shift in the
shear stress distribution curve as shown in Figure 2.3c.
concrete
FRP Lf
F
(a)
ε 4
2 3
1 adhesive
F
F
(b)
τ τ
crack
F F
F F
Prior to cracking After cracking
(c)
Figure 2.3 FRP-concrete bond specimen (a) top view (b) strain distribution along
FRP and (c) shear stress distribution along FRP (Lee, 2003)
Maeda et al. (1997) reported that the effective bond length decreased as the
stiffness of FRP increased. The stiffness was expressed as the product of the elastic
10
LITERATURE REVIEW
modulus and thickness of the FRP material. Täljsten (1997) reported that there was a
critical strain level, εcr, at which concrete began to fracture, and therefore a governing
factor for failure. It followed that for a given strain level, the higher the FRP stiffness,
the higher the load the bond could stand.
The effect of concrete strength was investigated by Chajes et al. (1996) and Xu et
al. (2001). Chajes et al. carried out bond tests with three different concrete strengths
ranging from 24 to 45 MPa and concluded that the ultimate bond strength was
proportional to f c . Xu et al. varied the concrete strength from 24 to 70 MPa and
reported that the ultimate bond strength was not always proportional to the concrete
tensile strength because different concrete strengths caused different effective bond
lengths.
The effects of concrete surface preparation and adhesive types on the bond
strength of the composite-concrete joint were investigated by Chajes et al. (1996),
Toutanji and Ortiz (2001) and Xu et al. (2001). Chajes et al. studied the bond on three
different surfaces: no surface preparation, a surface ground with a stone to give smooth
finish and a surface abraded mechanically with a wire wheel to leave the aggregate
slightly exposed. They reported that the concrete surface should be mechanically abraded
or sandblasted to achieve the best possible bond. Toutanji and Ortiz conducted a similar
study by comparing two surface preparations: water jet blasting and sanding. The
investigators reported that the surface treatment by a water jet produced higher bonding
strength than the surface treatment by a sander. Xu et al. showed from their experiments
that the bond formed by different adhesive resins had different ultimate strengths. They
also proved that the primer could improve the bond strength and the surface ground
heavily with a disk sander provides a better bond. The conclusions drawn from these
studies were similar. Surface treatments that produced a rough concrete surface produced
a better bond compared to surface treatments that resulted in a smooth concrete finish.
11
CHAPTER 2
called the bond-slip relation. This relation has been established experimentally by Nakaba
et al. (2001) and Savoia et al. (2003).
Nakaba et al. (2001) carried out a series of double pull-pull tests with the primary
test variables being the types of composites and concrete strength. The bond-slip curves
were plotted using the readings from the strain gauges placed at 15 mm spacing along the
bond length. The plot is given in Figure 2.4. The researchers concluded that composite
stiffness did not influence the shape of the bond-slip relation. The maximum local bond
stress increased as the concrete compressive strength increased. The relations were clearly
nonlinear and could be represented using Popovics’ equation (Popovics, 1973) given by
τ s a
= (2.1)
τ max s1 (a − 1) + (s / s1 ) a
Nakaba et al. recommended that the constant, a, could be taken to be 3 and the slip at
maximum bond stress, s1, could be taken as 0.065 mm.
Figure 2.4 Measured bond-slip curves for a test series by Nakaba et al. (2001)
Savoia et al. (2003) used the test results from Chajes et al. (1996) to plot the bond-
slip relationships (Figure 2.5). They also recommended that Popovics’s equation could be
used to describe the interface law. The maximum bond stress was found to be 3.5 fc0.19 at
a slip of 0.051 mm. The constant, a, was taken to be 2.86.
12
LITERATURE REVIEW
In summary, these past studies have highlighted the complexity of the failure
mechanisms of a concrete-composite joint and the factors affecting the bond’s behaviour
and strength. It is clear that some of the important issues are not yet conclusive. For
example, the effect of FRP width and thickness on the bond-slip curve and therefore the
bond performance needs to be investigated further. Also, the bond behaviour when
transverse shear-tension cracks formed within the bond line is not yet clear. Further
experimental study is, therefore, still required.
13
CHAPTER 2
obtained by calibrating the FRP strain at failure and the concrete strength. The term f c2 / 3
is related to the tensile strength of concrete as reported in Japanese Standards.
Calibrations were carried out for specimens with the bond length, Lf, larger or equal to
100 mm and the following expressions were recommended:
P = (3.8fc2/3 + 15.2) Ef Lf bf tf × 10-6 for CFRP (2.6a)
P = (3.8fc2/3 + 69.0) Ef Lf bf tf × 10-6 for AFRP (2.6b)
14
LITERATURE REVIEW
bond strength between steel plates and concrete. It was modified by Niedermeier (1996)
for the bond between CFRP plates and concrete and given as:
⎧ 0.78k f b f E f t f f ct if L f ≥ L e
⎪
P=⎨ Lf ⎛ Lf ⎞ (2.7a)
⎪0.78k f b f E f t f f ct L ⎜⎜ 2 − L ⎟⎟ if L f < L e
⎩ e ⎝ e ⎠
where the effective bond length is given by:
Ef t f
Le = (2.7b)
4f ct
Another modification was made by Neubauer and Rostasy (1999). In their paper, a
slightly different formulation was presented for P and Le, as follows
⎧0.64k f b f E f t f f ct if L f ≥ L e
⎪
P=⎨ Lf ⎛ Lf ⎞ (2.8a)
⎪0.64k f b f E f t f f ct L ⎜⎜ 2 − L ⎟⎟ if L f < L e
⎩ e ⎝ e ⎠
where
Ef t f
Le = (2.8b)
2f ct
In general, these expressions were derived based on the assumption that the
fracture energy can be calculated as a function of the concrete tensile strength, fct, as
follows
G f = k f2 c f f ct (2.9a)
where kf is a geometrical factor related to the width of the plate, bf, and the width of the
concrete member, bc, and can be calculated using the following expression:
2 − bf / bc
k f = 1.125 (2.9b)
1 + b f / 400
where cf is an empirical constant containing all secondary effects. This constant was
determined in a linear regression analysis using the results of double shear or similar
tests. Neubauer and Rostasy reported that cf equals 0.202 mm.
A more adequately reported derivation was by Yuan et al. (2001). They analysed
the interfacial stress state of externally bonded laminates for single pull-push situation.
The stress state is illustrated in the figure below.
15
CHAPTER 2
FRP
σ1 σ1 + dσ/dx dx
τ
τ
concrete
σ2 σ2 + dσ/dx dx
dx
Figure 2.6 Deformation and stresses of a bonded joint (Yuan et al., 2001)
By assuming that the shear stress is a function of the relative displacement or slip, they
derived a differential equation relating the deformation, δ, with the interfacial fracture
energy, Gf as shown below
d 2 δ 2G f 2
− λ f ( δ) = 0 (2.10)
dx 2 τ 2max
where
τ 2max ⎛ 1 bf ⎞
λ2 = ⎜⎜ + ⎟⎟ (2.11)
2G f ⎝ Ef t f bcEc t c ⎠
Two parameters are introduced in the above expressions: the local bond strength τmax and
the interfacial fracture energy Gf. The solutions were obtained by solving the differential
equation using different assumed relations between bond shear stress and slip (bond-slip
or shear-slip relations). Three bond-slip relations considered by Yuan et al. are illustrated
in Figure 2.7.
Gf Gf
Gf slip slip slip
s1 s1 s2 s2
The solution for Case (a), linear ascending with a sudden stress drop, was found to give:
16
LITERATURE REVIEW
τ max b f
P= tanh(λL f ) (2.12)
λ
The solution for Case (c), linear descending, gave:
⎧ τ max b f
⎪ if L f ≥ L e
P=⎨ λ (2.13)
τ b
⎪ max f sin (λL f ) if L f < L e
⎩ λ
π
where L e = and Gf = 0.5τmaxs2.
2λ
The solution for Case (b), bilinear shear-slip, yielded a relatively complicated implicit
expression.
A more complete solution was reported later in Wu et al. (2002). In this paper, the
solutions for both pull-pull and pull-push situations were reported. They found that when
the bilinear relationship was used and the bond length Lf is sufficiently large, the ultimate
load only depends on the interfacial fracture energy Gf or the area under the bond-slip
curve. In such cases, the following equations were derived for pull-push (single shear)
and pull-pull (double shear):
2E f t f G f
P = bf for pull-push (2.14a)
1+ α
where α = Ef Af / Ec Ac.
The equations reported in Yuan et al. and Wu et al. were, however, not verified
with experimental data. A more practical formula was proposed by Chen and Teng
(2001). The researchers noticed that s1 is small compared to s2 and used the solution for
Case (c) in Figure 2.7 (Equation 2.13). They also used the findings from other
experimental studies to derive the formula for the bond strength, as follows
P = 0.427β f β L f c b f L e (2.15a)
where
Ef tf 2 − bf / bc ⎧ 1 if L f ≥ L e
Le = ;βf = ; βL = ⎨ (2.15b)
fc 1 + bf / bc ⎩sin( πL / 2L e ) if L f < L e
17
CHAPTER 2
The constant 0.427 in Equation 2.15a was obtained by calibrating with a database of
shear-lap tests. Chen and Teng showed that the expression agreed well with the test data.
The ratio of the experimental to predicted ultimate bond strength was found to have an
average of 1.00 with the corresponding coefficient of variation of 0.159.
Dai and Ueda (2003) proposed different expressions for the bond-slip curve. The
expressions used are:
λ
⎛s⎞
τ = τ max ⎜⎜ ⎟⎟ s ≤ s1 (2.16a)
⎝ s1 ⎠
τ = τ max e −β ( s −s1 ) s > s1 (2.16b)
A different bond-slip relation was used by Kanakubo et al. (2003). The relation
was based on Popovics’ models (Equation 2.1). The shape of the bond-slip plot is shown
in Figure 2.8. By calibrating with their experiments, Kanakubo et al. recommended that
a = 3, τmax = 3.5 fc0.19 and s1 = 0.065 mm. To solve for the ultimate bond strength, they
introduced a concept of ‘equivalent bond stress block’ and derived the following
formulae:
P = k e τ max b f L e (Lf ≥ Le) (2.17a)
18
LITERATURE REVIEW
2λ f s e
Le = (2.17c)
ke
1− ke ⎛L ⎞ 1+ ke
k= cos⎜⎜ f π ⎟⎟ + (2.17d)
2 ⎝ Le ⎠ 2
where the bond length index λf = tfEf / τmax, se = 0.354 mm and ke = 0.428. It is also worth
noting that the researchers had assumed that the concrete stiffness was significantly
higher than that of the FRP. Therefore, concrete dimensions were not present in the
formulae. Kanakubo et al. also verified their model against a number of tests done in
Japan. They showed that the ratios of the experimental to predicted ultimate strength had
an average of 0.99 and a coefficient of variation of 17 %.
Bond stress
τmax
Gf
slip
s1
Ulaga and Vogel (2003) attempted to derive the shape of the bond-slip
relationship using micromechanical considerations. They stated that at a low load, shear
deformation is concentrated in the adhesive layer; whereas at a high load, the crack faces
are in the concrete layer and contain aggregate particles which partly protrude from the
cement matrix. Using ‘aggregate interlock theory’, they derived a bond-slip relationship
with nonlinear softening. To simplify the rather complicated relationship, they adopted a
linear softening curve with τmax = 0.4(fc)2/3. The formulae derived to predict the ultimate
capacity are:
2E f t f G f
P = bf when Lf > Le (2.18a)
(1 + α )
2
(1 + α)τ 2max L f 2E f t f G f
P = sin bf when Lf ≤ Le (2.18b)
2E f t f G f (1 + α )
where
19
CHAPTER 2
π 2E f t f G f
Le = (2.18c)
2 (1 + α)τ 2max
G f ≈ 0.045 f c2 / 3 (2.18d)
Equation 2.18a is the same as Equation 2.14a. However, Ulaga and Vogel (2003) also
derived formula for the case of a short bond length (Equation 2.18b).
In summary, there have been numerous analytical studies on the bond behaviour
between FRP and concrete. There are two main methodologies: empirical and semi-
empirical. The empirical methods are based mainly on the result of a certain set of
experiments and tend to yield simpler equations. The semi-empirical methods are based
on fracture mechanics using an idealized local bond-slip relationship. The shape of the
bond-slip curves and therefore the prediction formulae were generally calibrated with a
limited set of tests. Since, both methods relied on a certain set of experiments which
usually had similar configurations, the accuracy of these models still needs to be further
investigated.
In the listed experimental studies, most of the beams are rectangular and simply
supported. The laminates are bonded in various ways as illustrated in Figure 2.9 and
Figure 2.10. Different anchorage methods are developed using bolts, clamps, angles or
wraps.
20
LITERATURE REVIEW
A variety of failure modes have been observed (Figure 2.11). They can be
grouped as follows.
Flexural failure: This failure includes yielding of longitudinal steel bars followed by
secondary compression failure of the concrete or tensile rupture of the laminate.
Shear failure: It can occur in beams with insufficient shear capacity. The typical critical
shear crack extends from the end of the laminate to the point of loading. The crack
initiated close to the support has also been reported.
Debonding failure: Debonding is the most frequently observed mode. About two-thirds
of the specimens tested for FRP flexural strengthening failed due to laminate debonding.
More than one debonding mode exist. They can be classified into two main categories
based on the initial starting point. End debond is the failure that originates near the plate
end and propagates in the concrete either along the tension steel reinforcement (end cover
peeling) or near the bond line (end interfacial debond). Intermediate span debond is the
failure that originates either from a wide flexural crack (flexure crack debond), a wide
flexure-shear crack (flexure-shear crack debond) or a wide diagonal shear crack (shear
crack debond). The failure then propagates from the crack tip to the laminate end parallel
to the adhesive/concrete interface. Flexure-shear crack debond is also often accompanied
by cracking along the tension reinforcement layer.
In addition to these modes, failures in the adhesive or FRP layer have also been
reported. Failure can also result from the unevenness of the concrete surface. However,
these are not common and can be prevented by using proper adhesion and suitable FRP
materials and by carrying out correct surface preparation.
21
CHAPTER 2
22
LITERATURE REVIEW
23
CHAPTER 2
grooves nails
plate
clamp
I jacket
side plate
U jacket
confining plate
FRP wrap
anchorage U anchor
fabric
anchorage wrap
24
LITERATURE REVIEW
1. Concrete crushing
3. End shear
Shear strengthening
In practice, flexural strengthening is often combined with shear strengthening by
bonding laminates on both the soffit and the sides of the beam. The laminates on the sides
act like steel stirrups to cross diagonal cracks. They can be narrow strips placed at certain
spacing or wide plates covering the entire shear span. The main direction of the fibres can
be vertical or inclined to intersect shear cracks. The shear laminates can also be anchored
by being wrapped around the beam or by mechanical anchorage (Figure 2.12).
25
CHAPTER 2
FRP bolt
anchorage hole
U-anchor
U-anchor
with FRP
without
rod
FRP rod
lap
Figure 2.12 Shear strengthening using steel plates, FRP plates or FRP sheets
26
LITERATURE REVIEW
27
CHAPTER 2
higher level of both shear and normal stress. A similar observation was also made by Cha
et al. (1999) for strengthened prestressed concrete beams.
Instead of varying the shear span and the laminate length, Garden et al. (1997) and
Garden et al. (1998) changed the shear span and the beam depth. In their studies, the
shear span to depth ratio was varied from 3.00 to 7.72. A cantilever set-up was utilised to
obtain high ratios. It was found that as the ratio increased, the failure modes shifted from
end cover peeling to flexure-shear crack debond and then to shear crack debond. There
was inconsistency with the trend and it was claimed to be the result of the small amount
of shear reinforcement. Garden et al. (1998) also showed that the ratio of the shear span
to the beam depth did not significantly affect the ultimate shear capacity but affected the
ultimate moment capacity. Furthermore, they suggested that beyond a shear span to depth
ratio of 5.9, there was no further increase in the ultimate moment capacity.
Several researchers varied the laminate bonded length solely to study its effect on
the behaviour of retrofitted beams. Hearing (2000) fabricated nine rectangular beams
retrofitted with CFRP and tested them in four-point bending with a shear span of 600 and
a total span of 1800 mm. The laminated lengths varied from 600 to 1600 mm. Initial
delamination cracks were introduced near the laminate ends. This method forced all the
beams to fail by end interfacial debond. He reported that the beams with a longer
laminated length had a higher delamination initiation load and a higher ultimate failure
load, and also demonstrated a more brittle failure. Fanning and Kelly (2001) tested six
beams spanning 2800 mm and loaded in four-point bending with a shear span of 1100
28
LITERATURE REVIEW
mm. The CFRP bonded lengths were 715, 638 and 550 mm. All of these beams failed due
to end cover peeling with average ultimate loads of 102, 81 and 72 kN, respectively.
Fanning and Kelly also found that for the beams of shorter bonded length, there was
significant inconsistency in the results. Nguyen et al. (2001) also investigated the effect of
the laminate bond length in their experiments. They noticed that the ultimate loads only
slightly increased with longer bond lengths. The CFRP lengths were 950, 1000 and 1150
mm and the ultimate loads measured were 56, 57 and 59 kN, respectively.
29
CHAPTER 2
the ratio reduces. There is only one experimental study by Kishi et al. (2001), in which
the effect of the relative stiffness was investigated directly. They carried out testing on
RC beams strengthened with AFRP sheets and varied both the tension reinforcement and
AFRP area. They found that the beams with the area ratio Af / As ranging from 1.3 to 1.7
failed by debonding, whereas the beams with this ratio ranging from 0.6 to 0.8 failed in
flexure.
0.6
0.4
0.2
0.0
0.00 0.25 0.50 0.75 1.00 1.25
AfEf / AsEs
Figure 2.13 FRP efficiency ratio (max. reported strain / strain at rupture) versus
relative stiffness (Bonacci and Maalej, 2001)
30
LITERATURE REVIEW
2.3.1.2.7 Anchorage
Various anchorage methods have been used (Figures 2.9 and 2.10). Basically,
there are two main anchorage types corresponding to two main debonding failures: end
anchorage and intermediate span anchorage. Some forms of end anchorage are bolts, side
angles, I jackets or U anchors. It has been shown that end anchorage improves the beam
ductility since premature debond is delayed (Hollaway and Leeming, 1999). The beam
strength can also be enhanced by end anchorage (Hussain et al., 1995; Hollaway and
Leeming, 1995). Intermediate span anchorage is usually provided by U-shaped transverse
straps (Swamy and Mukhopadhyaya, 1999). These U-straps have been found to be able to
reduce the opening of diagonal shear cracks, prevent spalling of concrete cover and limit
the propagation of debonding cracks to the portion between the straps. They can therefore
increase deflection and load carrying capacities for beams susceptible to intermediate
span debond.
31
CHAPTER 2
32
LITERATURE REVIEW
The simplest method to design against this failure is to limit the FRP strain in the
vicinity of flexural cracks. Arya and Farmer (2001) suggested two strain limits: 0.008 if
the applied load was uniformly distributed and 0.006 if both high shear forces and
bending moments were present. They also proposed a limit on the steel reinforcement
strain of five times the yield strain. The failure mode referred to was called ‘debonding’.
Similar recommendation was made by Shehata et al. (2001), in which the limiting strain
was 0.005. This value was found from testing of two RC beams. They also proposed a
limit for ‘tearing off the concrete cover’, which will be discussed later in Section 2.3.4.
Some researchers believed that flexural crack debond mechanism is similar to that
in a shear-lap test (Figure 2.15). Consequently, the maximum force Ff in the FRP can be
calculated using the results from a shear-lap test. Maruyama and Ueda (2001) reported
the design recommendation by the Japan Society of Civil Engineers, which states that to
neglect peeling in beams strengthened by FRP sheet, the tensile stress at the location of
the flexural cracking should be limited to
2G f E f
ff ≤ (2.19)
n f t f ,0
where nf is the number of attached FRP layers and tf,0 is the thickness of a layer. The
interfacial fracture energy, Gf, is introduced in this formulation, which is to be obtained
from a double shear-lap test or taken as 0.5 N/mm.
Instead of limiting the FRP stress, Teng et al. (2001) proposed to limit the tensile
force in the FRP. To calculate the limiting force, they used Chen and Teng’s formulation
for a shear-lap test (Equation 2.15). A modification factor, α, was introduced to account
for the differences to the beam bending situation. The limiting force is calculated as
follows
P = αβ f β L f c b f L e (2.20)
33
CHAPTER 2
FRP
Ff
Ff
Shear-lap test
Peeling crack
Figure 2.15 Similarity between the shear-lap test and the situation near a flexural
crack
For most RC beams, there exist more than just one wide flexural crack. The
cracks are generally spaced at a certain distance over which the shear stress is transferred.
The crack spacing, therefore, might affect debonding failure. To account for that, analyses
were carried out by Malek et al. (1998) and Niu and Wu (2001b).
Malek et al. derived a formula to calculate the interfacial shear stress distribution
near a flexural crack using a closed-form solution. They indicated that the shear stress
near the tip of a flexural crack should be compared with the interfacial shear strength of
the adhesive.
Niu and Wu analysed the forces acting on an infinitesimal element (Figure 2.16)
and used a bilinear shear bond-slip model to describe the bond between FRP and
concrete. They derived expressions for the shear stress distribution near a flexural crack
or between two flexural cracks. The expressions were verified with a finite element
model and an experiment. A stress-slip model with τmax = 8 MPa, s1 = 0.05 mm and s2 =
0.3 mm was used. Close results were found. Niu and Wu also explained the debonding
mechanism in details (Figure 2.17) for plain concrete and RC beams. They found that for
the latter case, crack spacing played an important role in the stress distribution and
therefore debond propagation.
34
LITERATURE REVIEW
Load
M1 M1 + dM1
N1 RC concrete N1 + dN1
V1 V1 + dV1
τ τ
N2 FRP N2 + dN2
dx
Figure 2.16 Forces acting on an element of a composite beam (Niu and Wu, 2001)
Debonding propagation
Flexural crack
Figure 2.17 Debonding mechanism due to flexural cracks (Niu and Wu, 2001)
Both the expressions by Malek et al. and Niu and Wu are, however, cumbersome.
The proposed formulae were also based on a simplification that all materials had linear
elastic properties.
35
CHAPTER 2
FRP end and at flexural cracks was treated with the same model. To calculate the
debonding load, two main steps were involved.
In the first step, the most unfavourable spacing of flexural cracks was determined
assuming that the mean bond strengths along the steel-concrete and FRP-concrete
interfaces are constant. The spacing was given as
M cr 1
s rm = 2
z m (τ fm b f + ∑ τ sm d b π )
(2.21a)
The bond strength of steel, τsm, was taken as 1.85 fct. The bond strength of FRP, τfm, was
taken as 0.44 fct. The cracking moment was calculated as
kf ct bh 2
M cr = (2.21b)
6
In above equation, the factor k was taken as 2 to take into account the difference between
flexural and direct tensile strength. The mean lever arm, zm, was calculated using the
following formula:
(hE f A f + d s E s A s )
z m = 0.85 (2.21c)
(E f A f + E s A s )
The second step involved checking the tensile force transfer between two
subsequent cracks. The increase in the FRP tensile stress, ∆σf, should be less than the
maximum allowable increase, max∆σfd, between two cracked sections. The researcher
stated that max∆σfd reduced as σf increased. The envelope for max∆σfd is illustrated in
Figure 2.19. Point A corresponds to the end of FRP, where the maximum tensile stress
was calculated as the following:
c1 E f f c f ct
σ fad ,max = (2.21d)
γc tf
If the bond length lb was smaller than lb,max, the maximum tensile stress was calculated as
36
LITERATURE REVIEW
lb ⎛ l ⎞
σ fad = σ fad ,max × ⎜2 − b ⎟ (2.21f)
l b ,max ⎜⎝ l b ,max ⎟
⎠
To find point B, where the linear decrease ends, the following equations were provided:
c3E f s
σ fB = − c 4 f c f ct rm (2.21g)
s rm 4t f
1 ⎡⎢ c1 E f f c f ct ⎤
2
max ∆σ = B
fd + (σ f ) − σ f ⎥
B 2 B
(2.21h)
γc ⎢ tf ⎥
⎣ ⎦
Point C is defined by the stress limit reaching the tensile strength of FRP.
L0
Ff,1
Ff,i+1 - Ff,i
Ff envelope
37
CHAPTER 2
Figure 2.19 Maximum possible increase in tensile stress between two subsequent
cracks (fib Bulletin 14, 2001).
A simple design method was proposed by Matthys (2000). The shear stress at the
FRP-concrete surface was limited to the bond strength of concrete, fcv, which was
calculated as:
f cv = 1.8f ct (2.22a)
Matthys adopted a number of simplifications and proposed the following formulae to
calculate the final design shear force:
⎛ A E ⎞
Vd = f cv 0.95db f ⎜⎜1 + s s ⎟⎟ if tensile steel does not yield (2.22b)
⎝ Af Ef ⎠
Vd = f cv 0.95db f if tensile steel yields (2.22c)
38
LITERATURE REVIEW
steel
concrete
FRP
There are only a few studies that identified and analysed this mechanism. The first
attempt was by Triantafillou and Plevris (1992). For simplification, bending action was
ignored. The ultimate shear force was believed to be proportional to the total shear
stiffness of the steel reinforcement and the FRP, as follows
⎛v⎞
V ~ ⎜ ⎟ (G s A s + G f A f ) (2.23)
⎝ w ⎠ cr
where v and w are the vertical and horizontal movements at a crack tip. The researchers
indicated that peeling occurred when the ratio of vertical to horizontal displacement
reached a critical value. However, no detailed analysis and verification were reported.
Neubauer and Rostasy (1999) presented a detailed analysis of these shear crack
mouth opening displacements using a truss model with shear crack friction (Figure 2.21).
The mouth displacements were given as
w = 2s cr (ε x − ε cw cos 2 θ) (2.24a)
where scr is the crack spacing, βcr is the angle of inclined shear cracks and θ is the angle
of the compression strut inclination. εx, εz and εcw are the strains in the longitudinal,
perpendicular and compression strut directions given by
ηB − 1 ⎛M τb v d ⎞
εx = ⎜⎜ + ⎟⎟ (2.24c)
2η B E f A f ⎝ Z l 2 tan θ ⎠
τ tan θ
εz = (2.24d)
E s µ sw
39
CHAPTER 2
τ(1 + tan 2 θ)
ε cw =− (2.24e)
E c tan θ
V
where τ = and ηB is the degree of strengthening in bending. However, many
bvd
approximations were needed and a clear design implication for shear crack debond was
not available.
The vertical and horizontal movement of the concrete tip cause peeling of the FRP
from the concrete surface possibly in a similar manner as in the peel tests described by
Karbhari and Engineer (1996) and Karbhari et al. (1997) and illustrated in Figure 2.22. In
the first paper, different specimens with a number of peeling angles, α, were tested to
determine the interfacial fracture energy for mode I (tension) and mode II (shear).
Karbhari and Engineer found that as the angle decreased, GfI increased and GfII decreased.
The total interfacial fracture energy for the bonding between CFRP and concrete was
found to be around 0.56 N/mm. In the second paper, the specimens were exposed to
various conditions (ambient temperature, immersion in fresh water, immersion in sea
water, low temperature and freeze-thaw cycles) to study the environmental effects on the
bond. The authors also listed four generic mechanisms of adhesion with the emphasis on
the dominance of mechanical interlocking.
40
LITERATURE REVIEW
δy
FRP
α δx
roller
concrete
To design for shear crack debond, a simple design method was described in
Blaschko (1997), in which the acting shear force was limited to the modified concrete
shear capacity without shear reinforcement. The proposed formulae are
VR = k × b v d (1.2 + 40ρ l ) τ Rk (2.25a)
Ef
ρ l = (A s + A f ) /( b v d ) (2.25b)
Es
The factor k was to take size effect into account and given as k = 1.6 − d ≥ 1 , where d is
in meters. The shear strength, τRk, was found by calibrating with a number of test results
and given by:
τ Rk = 0.18(f c )1 / 3 (2.25c)
Another design model was developed by Mohamed Ali (2000). This study was
based on Zhang’s model (Zhang, 1997), which used the theory of plasticity to calculate
shear strength of reinforced concrete beams. For plain concrete beams, Zhang’s model
postulates that shear cracks generally extend from the load point at the top of the beam to
the position near the support. The closer the crack to the support, the larger the cracking
load, Vcr, (due to a larger crack area) and the smaller the shear strength, Vu, are (due to a
larger sliding angle, α). The critical diagonal crack is at a position where Vcr equals Vu.
For beams bonded with laminates, Mohamed Ali made modifications for Vcr (to account
for the additional resistance due to the tensile force in the laminates) and for Vu (by
41
CHAPTER 2
adding an additional term for the longitudinal reinforcement) and proposed the following
formulae:
Lb
Pb
V Vu
Vuc Vcr
(Ld)crit Ld
(a) Unplated (b) Plated
For beams with laminate bonded on the tension face, the cracking load was given as
⎛ x2 + h2 ⎞⎛ f tef b c mf ct b f t f (h + 0.5t f ) ⎞
Vcr = ⎜⎜ ⎟⎜
⎟ 2 + ⎟ (2.26a)
⎝ a ⎠⎝ h2 ⎠
where x is the distance from the crack tip to the load point, h is the total depth of the
beam, fct is the direct tensile strength of concrete and m is the ratio of FRP to concrete
stiffness Ef / Ec. The tensile strength, ftef, was calculated as the following:
−0.3
⎛ h ⎞
f tef = 0.156f c
2/3
⎜ ⎟ (2.26b)
⎝ 100 ⎠
The shear strength was given as
1 ⎛ ⎛x⎞
2
x⎞
Vu = γ s γ 0, tfp f c 1 + ⎜ ⎟ − ⎟b c h
⎜ (2.26c)
2 ⎜ ⎝h⎠ h⎟
⎝ ⎠
where
γs = 0.5; γ0,tfp = λ.f1(fc).f2(h).f3(ρf); λ = 1.6 (2.26d)
3.5
f 1 (f c ) = 5 MPa < fc < 60 MPa (2.26e)
fc
⎛ 1 ⎞
f 2 (h ) = 0.27⎜⎜1 + ⎟⎟ 0.08 m < h < 0.7 m (2.26f)
⎝ h⎠
42
LITERATURE REVIEW
As A tfp ,eff
ρs = 100 ; ρ f = 100 (2.26h)
bch bch
The effective area, Atfp,eff, was to account to the fact that the FRP might delaminate from
the concrete. Mohamed Ali suggested the following equations, which were developed for
a shear test between a steel plate and concrete:
Atfp, eff = Pb / ffu (2.26i)
Pb = 1.09 Lb fct bf ≤ 71ft tf bf (2.26j)
where Lb = (a – x – L0).
In this review, prediction models for end debond are classified into four main
approaches. They are interfacial localised failure based models, shear capacity based
models, concrete tooth scenario based models and shear-lap scenario based models. Some
models are the combination of several approaches. It is important to emphasize that
although some models were developed for steel plated beams, they are expected to be
applicable for FRP plated beams with or without modification. In the formulae presented
here, the units used are Newton (N) and millimetre (mm), unless noted otherwise. For
comparison, some equations are put in a slightly different form from the originals. Since a
number of models are dependent on the studies on the interfacial stress concentration near
the FRP end, a summary of those studies is presented first in the next section.
43
CHAPTER 2
end of the laminate. Several approximate closed-form solutions were derived to predict
the peak stresses near the laminate ends.
Shear
Stress
Normal
Figure 2.24 Predicted shear and normal stress distributions near the plate end
(Roberts, 1989)
σ0 = C R 2 τ0 (2.27b)
where
⎡ ⎛ K ⎞
1/ 2
M0 ⎤ bf t f
C R1 = ⎢1 + ⎜⎜ s
⎟⎟ ⎥ (d f − d n ) (2.27c)
⎢⎣ ⎝ f f t f
E b ⎠ V0 ⎥ I tr ,f b a
⎦
1/ 4
⎛ Kn ⎞
CR2 = t f ⎜⎜ ⎟⎟ (2.27d)
⎝ 4E f I f ⎠
Ga ba Ea ba
Ks = Kn = (2.27e)
ta ta
M0 and V0 are the bending moment and shear force at the FRP end, respectively. By
comparison with more rigorous solutions and test data, Roberts recommended that M0
should be replaced by M*, which is the bending moment evaluated at a distance (h + tf) / 2
from the end of the laminate.
44
LITERATURE REVIEW
Several modifications of Roberts’ solution have also been proposed. Ziraba et al.
(1994) calibrated Roberts’ equations with their finite element analyses. They proposed
the following formulae to calculate the maximum shear and normal stresses:
5/ 4
⎛C V ⎞
τ 0 = 35f ct ⎜⎜ R1 0 ⎟
⎟ (2.28a)
⎝ fc ⎠
σ 0 = 1.1C R 2 τ0 (2.28b)
El-Mihilmy and Tedesco (2001) carried out a calibration study with nine experimental
programs and proposed additional empirical coefficients for Roberts’ equations. Their
formulae to calculate the peak shear and normal stress near the cut-off point are:
Ea tf
τ 0 = τ + c1 σx (2.29a)
Ef
1/ 4
⎛ 3E t ⎞
σ 0 = ⎜⎜ a f ⎟⎟ τ0 (2.29b)
⎝ Ef ta ⎠
where τ and σx are the elastic shear and longitudinal laminate stress, respectively. They
are calculated as the following:
V0 t f y f
τ= (2.30a)
I tr .f
M0 yf
σx = (2.30b)
I tr ,f
Using a linear regression, the researchers found that the constant c1 could be taken to be
0.28.
K n ⎛ Vf V0* + β*M 0 ⎞ qE f I f
⎜
σ 0 = *3 ⎜ − ⎟⎟ + (2.31c)
2β ⎝ E f I f E c Ic ⎠ bf E c Ic
45
CHAPTER 2
where
Ga
A= (2.31d)
t a t f Ef
y f a 1E f
b1 = (2.31e)
I tr ,c E c
yf Ef
b2 = (2a 1 L 0 + a 2 ) (2.31f)
I tr ,c E c
⎡ yf t t ⎤
b3 = Ef ⎢ (a 1 L20 + a 2 L 0 + a 3 ) + 2b1 a f ⎥ (2.31g)
⎣ I tr , cE c Ga ⎦
Vf = −0.5b f t f2 (b 3 A + b 2 ) (2.31i)
1/ 4
⎛K b ⎞
β = ⎜⎜ n f
*
⎟⎟ (2.31j)
⎝ 4E f I f ⎠
q is the external distributed load applied on the beam, yf is the distance from the neutral
axis to the centroid of FRP plate, Itr,c is the moment of inertia of the transformed section
and Kn = Ea / ta.
fn
σ FRP σ + dσ/dx dx
dx
Figure 2.25 Stresses acting on a FRP laminate element near the cut-off point (Malek
et al, 1998)
In a recent study by Smith and Teng (2001), a more rigorous analysis was carried
out to derive the formulae for shear and normal stresses. The solution was later verified
with a linear elastic finite element model using a very dense mesh. However, their
formulation is rather complicated, which makes it difficult to be used in design.
46
LITERATURE REVIEW
the retrofitted steel plates in the beams tested by Jones et al. (1988) with his theory. No
verification of the end peak stresses has been made for beams bonded with FRP
laminates. It might be due to the difficulty in measuring the stress state close to the FRP
ends since the peak stress region there is much less clear for these beams (Aprile et al.,
2001; Rahimi and Hutchinson, 2001)
Jones et al. (1988) used the interfacial shear capacity as the failure criteria. Based
on their experimental work on steel plated beams, the researchers stated that anchorage
failure occurred when the peak shear stress exceeded 2 times the concrete tensile
strength. The peak shear stress was estimated to be about twice the elastic shear stress.
The elastic shear stress was given by
V0 A f ( E f / E c ) y f
τ= (2.32)
I tr ,c b f
Sharif et al. (1994), Ziraba et al. (1994) and Chaallal et al. (1998) suggested that
the interfacial shear stress in the adhesive should be limited to avoid debonding failure.
The peak shear stress was calculated using Roberts’ analytical solution. Sharif et al.
(1994) checked for end debond by comparing the interfacial shear stress with the shear
capacity of the adhesive. They recommended a shear strength of 3.5 MPa. Ziraba et al.
(1994) and Chaallal et al. (1998) differentiated between end interfacial debond and end
cover peeling. The latter mode was checked using another approach and will be described
later. The former mode was checked using Mohr-Coulomb criterion for the stress state in
the adhesive. Failure was deemed to occur when τ0 + σ0tanφ exceeded the allowable shear
cohesion, c, for the epoxy. φ is the internal friction angle. Ziraba et al. (1994) calculated
the peak shear stress at the laminate end using their modified expressions of Roberts’
equations. The cohesion, c, and internal friction angle, φ, were found to be 5.36 MPa and
47
CHAPTER 2
28 degrees, respectively. Chaallal et al. (1998) calculated the shear stress using Roberts’
solution without modification. The cohesion and internal friction angle recommended
were 5.4 MPa and 33 degrees, respectively.
Shehata et al. (2001) suggested that the average interfacial shear stress along the
shear span should be limited to
τlim = 0.3fct (2.33a)
to avoid end debond. The corresponding limit on the plate stress was found using an
equilibrium condition for the plate and given as
σlim tf = τlim a (2.33b)
This recommendation was based on the observation of tearing-off of the CFRP strips used
for shear reinforcement in their experiments. No verification for end debond of
composites bonded on a beam soffit was carried out.
where Gc is the concrete cracked shear modulus taken as 0.4Ec / 2(1 + νc). They
postulated that the limiting strain depended mainly on the interfacial shear strength,
which could be found from a shear test. In their tests, the limiting strain varied from
0.0017 to 0.0024.
48
LITERATURE REVIEW
P/2
ldev
lb
CFRP strain
(1) (2) (3)
transition point
Figure 2.26 Definition of bond development length for retrofitted beams by Nguyen
et al. (2001)
In another study, Fanning and Kelly (2001) suggested that the limiting strain
gradient controlled the ultimate loads. This was based on the results of six specimens
strengthened with CFRP composites, in which the CFRP strain gradient varying from 4.2
to 5.5 (strain/mm × 10-6). They reported that the strain dropped from the maximum value
under the load point to zero at the plate end.
In the second approach, the stress state at the cut-off point is used to check for end
debond failure. End cover peeling is deemed to happen when the maximum principle
stress in the concrete just above the cut-off point exceeds the concrete strength. The
concrete at this location undergoes a biaxial stress state with three stress components
present: interfacial shear stress, interfacial normal stress and flexural longitudinal stress
(Figure 2.27). There are differences among the researchers in the calculation of these
stresses and the selection of a concrete strength model.
Figure 2.27 Stresses acting at the cut-off point of FRP (Saadatmanesh and Malek,
1998)
49
CHAPTER 2
Saadatmanesh and Malek (1998) proposed that the interfacial shear and normal
stresses were calculated using their formulation presented in Section 2.3.4.1. To find the
flexural longitudinal stress, the researchers suggested that the applied bending moment at
the plate curtailment location must be increased to an amount given by the moment Mm,
as follows
M m = L 0 t f b f y c (b 3 A + b 2 ) (2.35)
where y c is the distance between the centroidal axis and the bottom of concrete beam.
The strength model for concrete under bi-axial stresses presented by Kupfer and Grestle
(1973) was used. According to this model, the strength of concrete was approximated by:
σ2 σ
= 1 + 0 .8 1 biaxial compression-tension (2.36a)
f ct fc
Tumialan et al. (1999) developed a similar model to predict end debond failure
loads. However, the interfacial shear and normal stress were calculated using Roberts’
formulae. The longitudinal stress was also determined from a bending analysis but the
additional moment was not included. End cover peeling was deemed to occur when the
maximum principle stress reached the concrete tensile strength, which was recommended
as
f ct = 0.689 f c (2.37)
El-Mihilmy and Tedesco (2001) used their formulae (Section 2.3.4.1) to find the
interfacial shear and normal stress. To find the longitudinal stress, they proposed to
replace the bending moment M0 at the plate curtailment with M* given by
⎛ L ⎞ L
M * = ⎜1.35 − 12.5 0 ⎟M 0 ; 0 ≤ 0.1 (2.38)
⎝ L ⎠ L
where L is the total span length. They also suggested that the longitudinal stress may be
ignored since it is quite small near the support. The strength model used for concrete
under bi-axial state was based on Tasuji and Slate’s criterion (Tasuji and Slate, 1978).
50
LITERATURE REVIEW
This criterion stated that the failure occurred if the principle tensile stress was greater
than the tensile strength, ftu, given by:
⎛ σ ⎞
f tu = f ct ⎜⎜1 + 2 ⎟⎟ biaxial compression-tension (2.39a)
⎝ fc ⎠
Different values for the moment of inertia, Itr, were used in the above three
models. Saadatmanesh and Malek, and Tumialan et al. used the moment of inertia of a
uncracked section; whereas El-Mihilmy and Tedesco used the moment of inertia of a
cracked section.
Oehlers (1992) separated two cases of peeling in beams bonded with steel plates.
When the moment to shear ratio near the laminate end, M0/V0, is high, flexural peeling
occurs. When this ratio is low, shear peeling occurs. For the first case, peeling was
assumed to occur when the sum of the normal stresses due to the beam curvature and due
to the axial strain near the plate end exceeded the concrete tensile strength. The
corresponding moment at which flexural peeling occurs, Mdb,f, was then calculated and
calibrated with the test results reported in Oehlers and Moran (1990). For the second case,
Oehlers observed that the amount of stirrups did not influence the shear peeling strength
and proposed a limitation on the shear capacity of the concrete RC beam without stirrups,
Vbd,s. For a general case, peeling was deemed to occur when the sum of V0/Vbd,s and
moment M0/Mdb,f at the cut-off point was greater than 1.17. The proposed formulae are:
M0 V
+ 0 ≤ 1.17 (2.40a)
M db ,f Vbd ,s
1/ 3
⎛A f ⎞
Vbd ,s = β1β 2 β 3 b v d 0 ⎜⎜ st c ⎟
⎟ (2.40b)
⎝ bvd0 ⎠
E c I tr ,c f ct
M db ,f = (2.40c)
0.474E f t f
51
CHAPTER 2
Smith and Teng (2003) modified these equations slightly based on their tests.
Their formulae are:
M0 V
0.4 + 0 ≤ 1.0 if Vdb,end ≥ 0.6Vdb,s (2.41a)
M db ,f Vbd ,s
Jansze (1997) used the ‘fictitious shear span’ concept and modified the model
described in Kim and White (1991) to compute the beam shear resistance for RC beams
bonded with steel plate. In the study by Kim and White for RC beams, the shear stress
distributions in a cracked beam and the reduction of internal moment arm length due to
arch action were considered to compute the shear force Vcr,s required for the initiation of
an inclined-shear crack. Another expression for the shear force Vcr,f at which a crack was
initiated for crack was derived based on cracking moment of the section. Where these two
curves intersected was the location of the critical inclined shear crack. Noticing that for a
retrofitted beam, the critical location was at the end of the steel plate, Jansze, by analogy,
derived an expression for the fictitious shear span, aL, and calculated the shear force
capacity, VRd, as the following:
VRd = τ Rd b c d s (2.42a)
ds ⎛ ⎞
τ Rd = 0.183 3 ⎜1 + 200 ⎟3 100ρ s f c (2.42b)
aL ⎜ d s ⎟⎠
⎝
aL = 4
(1 − ρs )2
a > L + ds (2.42d)
Jansze stated that this design method may be used for both plate-end shear and end cover
peeling.
Ahmed and Van Gemert (1999) modified the study by Jansze (1997) to account
for the difference between the shear stress generated due to shear force when replacing a
52
LITERATURE REVIEW
steel plate with an equivalent FRP laminate. The expressions to calculate the maximum
shear capacity, VRd, was reported as
VRd = τ FES b c d s (2.43a)
⎛ S S ⎞ ⎛ τ − 4.121 ⎞
τ FES = τ PES + τ PES b c d s ⎜⎜ s − f ⎟⎟ + 6188.5⎜⎜ ⎟⎟ (2.43b)
⎝ Is bf If ba ⎠ ⎝ b d
c s ⎠
⎛ 17.1366ρ s d s ⎞ A sv f sy
τ = ⎜ 0.15776 f c' + ⎟ + 0.9 (2.43c)
⎝ a ⎠ sb c
ds ⎛ 200 ⎞3
τ PES = 0.183 3 ⎜1 + ⎟ 100ρ s f cm (2.43d)
aL ⎜ d ⎟
⎝ ⎠
(1 − ρ s ) 2
aL = 4 dL3 (2.43e)
ρs
Ss and Is are the first and second moment of area of the cracked section transformed to
concrete with an equivalent steel plate with the thickness of tf ×(ffu /fsy). It is not clear
how the reported formulae were derived.
Ziraba et al. (1994) checked for end cover peeling based on the modified shear
capacity Vuc + kVus where Vuc and Vus are the shear contribution of plain concrete and
stirrups, respectively, and k is the empirical modification coefficient used to account for
the stress concentration near the cut-off point. From a regression analysis of a set of
experimental data containing several plated beams, they obtained an expression for k, as
follows
6
k = 2.4e −0.08×10 C R 1C R 2
(2.44a)
where CR1 and CR2 are given in Equations 2.27a and 2.27b, respectively. The shear
capacities, Vuc and Vus, were given in Ziraba et al.’s paper as
1
Vuc = (f c + 100ρ s )b c d s (2.44b)
6
A sv f sy ,f d s
Vus = (2.44c)
s
53
CHAPTER 2
exceeds the concrete tensile strength. To calculate the crack width and the interfacial
shear stress, different methods were used among the researchers.
Tension bar
lmin
A model based on concrete tooth scenario was developed by M. Raoof and his
associates (Zhang and Raoof, 1995; Raoof and Zhang, 1997; Raoof and Hassanen, 2000).
They assumed that the crack patterns for plated and unplated beams were similar. The
crack spacing was calculated from the classical theory of cracking by Watstein and
Parsons (1943) as the following:
A e f ct
l min = (2.45a)
u s ∑ O bar + u f b f
where Ae is the area of concrete in tension, ∑O bar is the total perimeter of the tension
reinforcing bars, us and uf are the steel-to-concrete and FRP-to-concrete average bond
strengths, respectively. Raoof and Hassanen (2000) used the following formulae to
calculate these bond strengths:
u s = 0.28 f cu (2.45c)
With a known Lp, the limit axial stress σ in the plate was calculated using the equilibrium
condition for the plate (σ = τLp / tf). The lower bound of FRP tensile stress was found to
be
54
LITERATURE REVIEW
A e f ct2 L p bc
σ f (min) = (2.45g)
6ct f (u s ∑ O bar + u f bf ) bf
To find the tensile strength of concrete, the researchers used the following formula:
f ct = 0.36 f cu (2.45i)
Another concrete tooth scenario based model was proposed by Chaallal et al.
(1998). The crack spacing was simply taken as the stirrups spacing. Only the last tooth
near the FRP end was considered and the shear stress acting on the tooth was calculated
using Robert’s expressions. The ultimate shear peeling stress was calculated as
f ct s b c
τ peeling
ult = (2.46a)
6c b f
The researchers recommended that the tensile strength of concrete was taken as
f ct = 0.53 f c (2.46b)
55
CHAPTER 2
FRP
lb
2.3.5.1 fib Bulletin 14: Externally bonded FRP reinforcement for RC structures
This report has the most complete treatment for debonding or loss of composite
action. The failure modes are divided into: ‘peeling-off caused at shear cracks’, ‘peeling-
off at the end anchorage and at flexural cracks’, ‘end shear failure’ and ‘peeling-off
caused by the unevenness of the concrete surface’. For the first mode, the design method
recommended is by Blaschko (1997). To check for the second modes, three approaches
can be used. Approach 1 is based on Ayer and Farmer (2001) and Neubauer and Rostasy
(1999) models. Approach 2 is based on Neidermeier (2000). Approach 3 is based on
Matthys (2000). For end shear failure, the model developed by Jansze (1997) is
recommended.
56
LITERATURE REVIEW
⎧ 4 − 500ε c
⎪⎪ 6 − 1000ε for ε c ≤ 0.002
γ=⎨ c
(2.47b)
1000ε c (3000ε c − 4) + 2
⎪ for 0.002 < ε c ≤ 0.003
⎪⎩ 1000ε c (3000ε c − 2)
57
CHAPTER 2
k is the proportionality factor relating the mean bond strength to the concrete strength. k
depends on the type of FRP reinforcement, the modulus of rupture of the concrete and the
type of application. k can be taken to be 0.184 as found from the study by Bizindavyi and
Neale (1999).
58
LITERATURE REVIEW
Hence, it is usually necessary to have some test results against which the output from the
finite element analysis may be compared.
59
CHAPTER 2
characterised by the specific fracture energy, Gf, dissipated during complete crack
formation. It is the area under the complete stress-displacement curve (discrete crack) or
stress-strain curve (crack band) in pure tension. Consequently, concrete cracking is
influenced primarily by two properties: the tensile strength, fct, and the specific fracture
energy, Gf.
where both strengths are in MPa. An estimate of the characteristic flexural tensile
strength can be made using this expression:
60
LITERATURE REVIEW
The table below summarises the predictions of the tensile strength of the concrete
with an average compressive strength of 53.7 MPa, which is mainly used in the present
research.
The fracture energy of concrete, Gf, can also be variable. It is relatively difficult to
be predicted and therefore there are very limited formulations recommended in the
current codes. The fracture energy can be found using testing. There have been a large
number of tests from different laboratories, based on different fracture specimens and
different testing methods. A wide range of values for Gf has been reported (Bazant and
Becq-Giraudon, 2002; Wittmann, 2002). Gf has been found to be dependant on several
parameters, including but not limited to the aggregate sizes and properties, the water-
cement ratio, the test method and the specimen size (RILEM Technical Committee 90-
FMA, 1989; Wittmann et al., 1990; Bazant, 2002; Wittmann, 2002). To estimate the
fracture energy from the basic characteristics of concrete, a number of empirical formulae
have been proposed. In the following expressions, the concrete strength and modulus are
in MPa, the maximum aggregate size, da, is in mm and the fracture energy, Gf, is in
N/mm.
A simple formula for the mean fracture energy was developed by Bazant and Oh
(1983) on the basis of the data of notched specimens. The energy is given by
( )
G f = 2.5 2.72 + 0.0214f ' ct f ' ct
2 da
Ec
(2.53)
where f’ct is the concrete tensile strength and Ec is the elastic modulus of the concrete.
61
CHAPTER 2
Trunk and Wittmann (1998) studied the relationship between the fracture energy
and the maximum aggregate size. They proposed a power function to calculate Gf, as
follows
Gf = a . dan (2.56)
From their data, they found that a = 80.6 and n = 0.32.
In a recent study by Bazant and Becq-Giraudon (2002), the fracture energy was
stated to be influenced by the maximum aggregate size, the water-cement ratio, w/c, the
concrete elastic modulus, the unit weight of the concrete and the concrete strength. From
a relatively large database, the researchers derived a statistical formula to compute the
concrete fracture energy as the following:
0.46 0.22 −0.30
⎛ f' ⎞ ⎛ d ⎞ ⎛w⎞
G f = 2.5α 0 ⎜⎜ c ⎟⎟ ⎜1 + a ⎟ ⎜ ⎟ (2.57)
⎝ 0.051 ⎠ ⎝ 11.27 ⎠ ⎝c⎠
where the coefficient α0 equals 1.12 for crushed aggregates. The coefficient of variation
was found to be 29.9%.
These formulae produce different values for Gf. For example, the table below
summarises the predictions for the concrete with the compressive strength of 53.7MPa
with the maximum aggregate size of 14 mm and the water/cement ratio of 0.59. These are
the properties of the concrete used mainly in the present research.
62
LITERATURE REVIEW
63
CHAPTER 2
(a)
3kN (FEM)
5kN (FEM)
9kN (FEM)
4000
2000
0 50 100 150
Location (mm)
(b)
Figure 2.30 Finite element mesh adopted (a) and comparison of strain distributions
reported (b) by Maeda et al. (1997)
Chen et al. (2001) carried out a study on the stress distributions in several
common test arrangements using a linear elastic finite element analysis. A very fine mesh
was used near the pulling end and the plate end (Figure 2.31). The stress concentrations in
the plate-adhesive interface were found showing the peaks within a very short distance
(0.5 mm) near the plate ends. The stress variation across and along the composite plate
64
LITERATURE REVIEW
were also investigated. It was found that the stress difference on the upper and lower
surfaces gradually decreased and became insignificant at a certain distance from the
pulling end. The effect of the support block clearance in the pushing tests was also
studied. The maximum shear stress near the pulling end was not significantly affected
once the clearance was greater than 37.5 mm. Chen et al. also found that the bond
strength could be dependant on the test methods. The bending tests gave higher bond
strengths than those obtained using shear tests. Little difference was found between the
corresponding double and single tests. However, these findings were only for the linear
elastic range and cracking in the concrete was not accounted for in the analyses.
Plate
Adhesive
Concrete
A recent investigation was by Lee (2003), where concrete cracking was taken into
account by smeared cracks modelling. Lee showed similar normal cracks in the concrete
between his FE model and experiment and was able to predict the ultimate failure loads
for his series of specimens. However, interfacial cracks along the interface were not
observed even though it was the dominant failure surface observed in his experiments.
65
CHAPTER 2
X CL
- Nodes free to move in X-direction, fixed in Y-direction
- Uniform displacement applied along the nodes
Figure 2.32 The FE mesh (above) and the crack patterns for two example specimens
with bond lengths of 50 mm (below left) and 130 mm (below right) as reported in
Lee (2003)
Camata et al. modelled Lee’s tests using a different technique. The researchers
used a combination of smeared and discrete crack approaches. The analyses were carried
out using the program MERLIN (MERLIN User's Manual, 2002). An auto-remeshing
technique was used to find the crack paths in the concrete blocks. The researchers were
able to simulate shear-tension failure and reported good agreement between the numerical
analysis and the experimental results for a number of specimens.
Figure 2.33 Boundary conditions and mesh discretization by Camata et al. (2003)
66
LITERATURE REVIEW
nonlinear analyses were carried out. To simulate debond failures, both discrete and
smeared crack modelling were used.
Malek et al. (1998) and Teng et al. (2002) built linear elastic models to investigate
the stress distributions in RC beams strengthened with a bonded soffit plate. Malek et al.
used the FE analysis to verify their theoretical predictions of the local interfacial stress
concentrations near the FRP cut-off point and near the tip of a flexural crack. Relatively
fine meshes were used at those locations. Different meshes with different element types
and different mesh sizes were chosen. The interfacial shear and normal stress, and the
tensile stress of the plate predicted by the FEA were in good agreement with the
theoretical solution. Teng et al. adopted a much finer mesh (Figure 2.34a). They were
able to show that the stresses vary strongly in the adhesive layer. In particular, near the
end of the plate, the interfacial normal stress is tensile along the adhesive-concrete
interface but compressive along the FRP-adhesive interface (Figure 2.34b). Teng et al.
also found that the effect of a spew fillet at the end of the plate formed from excess
adhesive can reduce the interfacial stresses. They also pointed out that a closed-form
solution based on the assumption of uniform stresses in the thickness direction is
incapable of predicting accurately such stress concentrations but can provide reasonably
close predictions of stresses along the middle-thickness section of the adhesive layer.
67
CHAPTER 2
(a)
Niu and Wu (2001a) and Niu and Wu (2001b) built FE models based on discrete
cracks to study interfacial peeling of FRP from flexural and flexure-shear cracks. The
main cracks were inserted manually. In the first paper, the commercial FE package
ABAQUS was used and the main focus was on the interfacial stress distributions in
beams with flexural cracks. Concrete was assumed to be linear elastic. The cracks were
simulated using a tension-softening model. The researchers compared the FE results with
their theoretical solutions and found good agreement. No comparison with experimental
results was however made. In the second paper, the FE package DIANA version 7.2 was
used to simulate the behaviour of a retrofitted beam under testing. Both flexural and
flexure-shear cracks were inserted in the model as shown in Figure 2.35a. Plane stress
and truss elements were used to model concrete and FRP, respectively. Concrete material
was modelled as nonlinear in compression. The interfacial fracture behaviour was
simulated using bilinear relationships between the bond stress and the slip and between
the normal stress and the normal displacement. The researchers found that the
68
LITERATURE REVIEW
propagation of peeling depends mainly on the interfacial facture energy consumed for
mode II. Comparison of the load-displacement curves was done for two beams as shown
in Figure 2.35b. The predicted curves were found to have a larger stiffness than those of
the experiments. That was attributed to the fact that the model did not take into account
distributed cracks in concrete.
(a)
(b) (c)
Figure 2.35 FE discretization model (a), loading curve (b) and FRP stress
distributions (c) as reported by Niu and Wu (2001)
A number of nonlinear analyses using smeared cracks have also been carried out.
Arduini et al. (1997) modelled RC beams strengthened with CFRP plates and sheets using
ABAQUS. A two-dimensional mesh was used for the beams strengthened with CFRP
plates, while the beams strengthened with CFRP sheets were modelled in three
dimensions. The concrete elements used were eight-node and twenty-node, respectively.
The meshes were relatively coarse (50 and 100 mm mesh size). A perfect bond was
assumed between the FRP composite and concrete. The researchers reported that the
numerical models showed good accordance to the experimental results in terms of the
load-deflection response, the load-FRP strain response and the evolution of cracking.
However, the assumption of a perfect bond and the course mesh used led to an
overestimation of the beam stiffness. Debond cracks were not observed. In addition,
convergence difficulty was encountered and therefore the models fell short in the
identification of the maximum load carrying capacity of the beams.
69
CHAPTER 2
David et al. (1999) used a FE code developed in France to model their tests on
beams retrofitted with CFRP. These beams failed by end cover peeling. Three-node
triangular elements and two-node linear elements were used to model the concrete and the
CFRP, respectively. An elasto-plastic model with a softening post-peak phase was used
for the concrete in compression. In tension, the concrete was modelled using a smeared
fixed crack approach. The CFRP-concrete interface was assumed to have a perfect elastic
behaviour. The author reported that the model was able to predict correctly the flexural
behaviour of the strengthened beams. However, comparison of only one loading curve
was showed. The crack patterns and predicted failure modes were not presented.
Ross et al. (1999) modelled their beams using another commercial package
(ADINA), which is capable of nonlinear modelling. Eight-node plane stress elements and
three-node truss elements were used to represent the concrete and the FRP composite,
respectively. Good agreement with the experiments was observed in terms of the beam
stiffness but the failure mode and therefore the ultimate load were not predicted by the
model. The crack pattern at failure of a beam is illustrated in Figure 2.36, showing no
delamination crack even though it was the dominant failure mode of the beams.
Hearing (2000) used the same FE package (ADINA) and built different two-
dimensional models with different mesh types and mesh sizes to simulate debonding
failure. The nonlinear FE models were able to produce the global behaviour, i.e. the load-
deflection stiffness, of the specimens tested by the researcher. However, the analyses
were not able to capture local debond mechanisms.
Another FE study using a smeared crack approach was done by Rahimi and
Hutchinson (2001). The cracking model used was incorporated with an isotropic damage
model to simulate the nonlinear behaviour of the concrete. Four- or eight-node
70
LITERATURE REVIEW
quadrilateral isoparametric elements were used for concrete. Three- or six-node elements
were also needed in the transition zones. The internal rebars were modelled with two- or
three-node bar elements, while the adhesive layer and external reinforcement were
modelled with a single row of four- or eight-node elements (Figure 2.37). The adhesive
was assumed to be elastic. The FE analysis was used mainly to supplement the
experimental measurement. The FE predictions showed two peak interfacial shear
stresses near the FRP end and under the load point. However, some discrepancy with the
experiments was observed. The FE analyses were also found to be sensitive to the
concrete tensile strength. The beam strengths predicted were found to be within 20 % of
the test results. No crack pattern was reported.
Fanning and Kelly (2000) used the smeared crack model incorporated in the three-
dimensional eight-node solid isoparametric element, Solid65, in the commercial package
ANSYS to model both the concrete and the adhesive layer. Four-node isoparemetric
71
CHAPTER 2
linear elastic shell elements were utilised to represent the CFRP plates. The researchers
reported that their model was able to simulate the strain distributions in the CFRP
composites and the load-deflection curves for a number of beams in their test series. A
crack pattern was compared for a beam failed in shear. However, the cracking patterns for
the beams failed in end cover peeling were not reported and the deflection and plate strain
were not showed for most of the beams failing by end cover peeling.
72
LITERATURE REVIEW
stresses normal to the rebars and to the external plates were however neglected. The
researchers modelled the same beams reported in Zarnic et al. (1999). Two failure modes,
midspan and end debond, were indicated in the simulation by looking at the bond stress
distributions. However, due to the nature of the models, the failure surfaces were assumed
to be the bond surfaces and the capacities were determined by an assumed strength of the
epoxy resin. Close match was found for the loading curves of all specimens.
y
U31 U22
U21 vB (x) U32
uB (x)
x z
U11 U12
u (x)
U41 U42
Figure 2.38 Node and element displacements for RC beam model with slip in plate
by Aprile et al. (2001)
Wong and Vecchio (2003) also modelled Zarnic et al.’s beams. The researchers
used a smeared rotating crack model to analyse the beams and bond elements to simulate
debonding. Good agreement with the experiments was achieved when the maximum bond
stress was taken as 3 MPa (Figure 2.39).
CFRP (FE)
(perfect bond)
CFRP (FE)
(τmax = 3 MPa)
Total load (kN)
CFRP (Exp.)
Control (Exp.)
A similar approach was adopted by Wong et al. (2001). This research group built
a three dimensional FE model of FRP plated beams tested by Rahimi and Hutchinson
73
CHAPTER 2
Figure 2.40 The three-dimensional model of FRP plated beams adopted by Wong et
al. (2001)
Yang et al. (2002) used a discrete crack model based FEA. Linear elastic fracture
mechanics (LEFM) was used in conjunction with a remeshing algorithm. A computer
program based on AUTOFRAP was specially developed to incorporate these features. In
the LEFM approach, all the materials involved in the plated RC beams were assumed to
be linear elastic in both tension and compression. Since LEFM is meant to be applicable
to large concrete structures only, the researchers argued that acceptable accuracy may still
be achieved for normal-sized structures if energy based criteria are properly applied.
In Yang et al.’s model, the bond between the concrete and internal reinforcements
was modelled using interface elements. A perfect bond was assumed initially. Once a
crack intersected an internal rebar, a full loss of bond over the length of one
74
LITERATURE REVIEW
reinforcement element was assumed. The bond-slip behaviour between the concrete-
adhesive and adhesive-plate was modelled using four-node contact elements. A maximum
energy release rate criterion was adopted for the crack propagation condition. To allow
multiple crack propagation, Yang et al. used a combination of two loading algorithms, the
G-scaling procedure and the displacement control procedure, to avoid the problems
associated with each individual method. The G-scaling procedure was used at first to
allow only one crack to initiate at each step. In this procedure, the load level and the
corresponding structural responses were scaled from a linear elastic analysis to meet the
crack propagation condition. When multiple cracks were initiated at a load step, the
displacement control procedure was used thereafter to allow multiple cracks to propagate
within a single step. It was achieved by scaling the load and the structural responses so
that the displacement satisfied a pre-set value.
Yang et al. modelled an example RC beam retrofitted with a CFRP plate and
failing by end cover peeling. The concrete cover on the tension face was modelled using
four rows of elements. Both the adhesive and the CFRP plate were modelled using one
layer of elements. The crack pattern was simulated successfully (Figure 2.41). The effect
of the plate length on the failure mode was also investigated. Shifting from flexural to end
cover peeling was observed as the plate length was reduced. However, very limited
comparison with the experimental measurements was made. Only the load deflection was
compared but the peak load was not captured accurately.
Figure 2.41 Initial FE mesh (left) and crack pattern at a high load level (right)
reported by Yang et al. (2003)
In summary, the FE studies were able to predict the global behaviour of retrofitted
beams. The load versus deflection curves predicted by most FE models matched the
corresponding experimental ones reasonably well. However, most models were not able
75
CHAPTER 2
to capture local debonding failures. Cracking patterns were rarely reported. Several
studies showed the general crack bands but the debond cracks were not observed. There
were attempts to simulate debonding using a predefined interface behaviour. However, in
this model delamination surface location was assumed to be the bond line between the
FRP composite and concrete. Last but not least, most of the reported FE models lack
experimental verification. Further investigation using nonlinear FEA is therefore still
needed.
• Investigation on shear-tension failure and effect of vertical cracks and concrete prism
formation
• Further investigation on the effect of FRP dimensions and the bond-slip relations
• Verification of existing prediction models.
• Further investigation on the debond failure mechanisms and the effects of a number of
parameters
• Further experimental data to verify existing theoretical models
• Development or recommendation of a reliable design methods.
76
LITERATURE REVIEW
The final section of the review was about FE modelling of retrofitted concrete
members. An outline of concrete mechanical properties and modelling techniques was
presented. It was followed by a critical review of FE modelling of shear-lap and beam
tests. The following tasks were highlighted:
• Further investigation using FE on the bond between the FRP composite and concrete
• Further FE modelling of retrofitted beams to capture different debonding
mechanisms.
77
ASSESSMENT OF PREDICTION MODELS
3.1 Introduction
In Chapter 2, existing experimental and theoretical studies on shear-lap specimens
and retrofitted beams are presented. It was found that there are a number of theoretical
models to predict the load capacity of a joint between FRP and concrete. These models
are based on different experimental data and/or different theoretical assumptions.
Therefore, there is a need to investigate the accuracy of the models. Similarly, there are
many experimental and theoretical studies on the retrofitted beams. Different failure
modes have been described and different approaches have been adopted to predict the
peak loads of retrofitted beams. It was found that not only these prediction methods need
to be validated, but the failure mechanisms also require further investigation.
In this chapter, existing prediction models are assessed. The assessment of the
models for shear-lap tests was done based on a relatively large database of shear-lap tests
collected from past experiments. The assessment of the models for retrofitted beams was
based on a large database of beam tests from past experiments. For completeness, the data
obtained from the experimental programs undertaken by the author was also included in
the databases. The details of these experiments will be reported in Chapters 4 and 5.
79
CHAPTER 3
Table 3.2 Variability of test parameters and results in the shear-lap test database
Parameter Minimum Maximum Average
Concrete strength, fc (MPa) 18.9 70 35.5
Concrete width, bc (mm) 100 228.6 138.0
FRP width, bf (mm) 15 100 51.8
FRP modulus, Ef (MPa) 22500 380000 216432
FRP thickness, tf (mm) 0.11 2 0.39
Bond length, Lf (mm) 50 700 137
Relative stiffness, Ef Af / Ec Ac 0.0097 1.754 0.366
Ultimate capacity, P (kN) 3.81 42.8 13.7
As shown in Table 3.1 and Table 3.2, the most common setup is single shear-lap
test (143 specimens) and the most common bonded composite is CFRP (152 specimens).
The most frequently observed failure is mode 1, interfacial debond (138 specimens).
Sixteen specimens failed by mode 2, shear-tension failure. Only one specimen failed by
FRP tensile rupture and one failed by cohesion failure of the adhesive.
In a number of theoretical models used to predict the debonding failure load, the
elastic modulus and thickness of the adhesive are needed. However, in the experiments,
these parameters were often not measured. In those cases, the adhesive thickness was
80
ASSESSMENT OF PREDICTION MODELS
taken as 1.0 mm as observed in the experiments by the author, and the adhesive elastic
modulus was taken as 8500 MPa, as recommended by Smith and Teng (2002).
The validation results for the first two simple formulae by Van Gemert (1980) and
Chajes et al. (1996) are shown in Figure 3.1. It is clear that the two models do not provide
accurate predictions. Van Gemert’s model is based on an incorrect distribution of the
bond stress. The average of Pexp / Pcal is 1.98 and the coefficient of variation is 53 %. In
Chajes et al.’s model, the ultimate load is dependent on two empirical factors, which were
found by calibrating with their limited experimental data. The model predicts non-
conservative results with an average Pexp / Pcal of 0.62 and a coefficient of variation of
35 %.
40 40
30 30
20 20
10 10
0 0
0 10 20 30 40 50 0 10 20 30 40 50
Measured max. load (kN) Measured max. load (kN)
(a) (b)
Figure 3.1 Validation results for Van Gemert’s model (a) and
Chajes et al.’s model (b)
Figure 3.2 shows the validation results for the models by Maeda et al. (1997) and
Izumo et al. (1998). Izumo et al.’s model was validated using the specimens bonded with
CFRP only. Both of these models are purely empirical, based on a number of tests done
81
CHAPTER 3
in Japan. However, they give relatively accurate results for most specimens in the
database described above. The average values of Pexp / Pcal are 0.95 and 0.99 for Maeda et
al.’s and Izumo et al.’s models, respectively. The validation results for the model by
Izumo et al. have a high degree of scatter. Their coefficient of variation is 39 % compared
to 21 % for Maeda et al.’s model. Both models also tend to be non-conservative for the
specimens with high bond strengths.
40 40
30 30
20 20
10 10
0 0
0 10 20 30 40 50 0 10 20 30 40 50
Measured max. load (kN) Measured max. load (kN)
(a) (b)
Figure 3.2 Validation results for Maeda et al.’s model (a) and Izumo et al.’ model
(b)
Figure 3.3 presents the evaluation results for three models based on fracture
mechanics. All three models provide good correlations between Pexp and Pcal. The
coefficients of variation are less than 20 % for all models. The model by Chen and Teng
gives the best result in terms of the accuracy with the average Pexp / Pcal of 1.07; whereas
Niedermeier’s model is the least conservative with the average Pexp / Pcal of 0.73.
82
ASSESSMENT OF PREDICTION MODELS
30 30
20 20
10 10
0 0
0 10 20 30 40 50 0 10 20 30 40 50
Measured max. load (kN) Measured max. load (kN)
(a) (b)
Mode 1 Mode 2
Mode 3 Mode 4
50
Predicted max. load (kN)
40
30
20
10
0
0 10 20 30 40 50
Measured max. load (kN)
(c)
Figure 3.3 Validation results for Niedermeier’s model (a), Neubauer and Rostasy’s
model (b) and Chen and Teng’s model (c)
The validation results for the other two models by Kanakubo et al. (2003) and
Ulaga and Vogel (2003) are presented in Figure 3.4. These models also produce relatively
good predictions. The former tends to have non-conservative results with the average
Pexp / Pcal of 0.83; whereas the latter is conservative with the average Pexp / Pcal of 1.20.
83
CHAPTER 3
40 40
30 30
20 20
10 10
0 0
0 10 20 30 40 50 0 10 20 30 40 50
Measured max. load (kN) Measured max. load (kN)
(a) (b)
Figure 3.4 Validation results for Kanakubo et al.’s model (a) and Ulaga and Vogel’s
model (b)
84
ASSESSMENT OF PREDICTION MODELS
Table 3.3 Summary of validation results of strength models for shear-lap tests
Tests failed by mode 1
Verified against All tests
only
Coefficient Coefficient
Average of Average of
Statistical parameters of variation of variation
Pexp / Pcal Pexp / Pcal
(%) (%)
Van Gemert (1980) 1.98 53 2.02 54
Chajes et al. (1996) 0.62 35 0.63 34
Maeda et al. (1997) 0.95 21 0.98 19
Izumo et al. (1997) 0.99 39 1.04 35
Niedermeier (1996) 0.73 19 0.74 17
Neubauer and Rostasy
0.92 18 0.94 17
(1999)
Chen and Teng (2001) 1.07 16 1.08 16
Kanakubo et al. (2003) 0.83 20 0.86 18
Ulaga et al. (2003) 1.20 23 1.25 20
Therefore, this section presents an assessment of all those models together. The
assessment was based on a large database of the results from testing of simply supported
retrofitted RC beams done by the author and several other research groups. In the
following sections, some description on the database is presented first. It is followed by
the validation and assessment results of the strength models for three main failure
categories: flexure failure, intermediate span debond and end debond.
85
CHAPTER 3
There are a number of reasons to build a new database. Firstly, even though a
number of studies have showed that the beam theory can predict the ultimate loads for
flexure failures, i.e. concrete crushing or FRP rupture, the conclusion made was based on
a very limited number of test specimens and therefore it needs to be consolidated with a
larger number of tests. Secondly, the main focus of the existing databases is either on the
general beam behaviour or on end debond. There have been very limited assessment
studies on intermediate span debond. Lastly, a large number of new tests have become
available recently for end cover peeling (including tests by the author). Further validation
for this mode is therefore possible to confirm the previous findings.
To build the database, an extensive literature survey was carried out. For inclusion
in the database, the beam chosen must satisfy following requirements. It was of a
rectangular section, simply supported and loaded in three or four-point bending. No
preloading was applied. The laminate material was FRP composite only and not
86
ASSESSMENT OF PREDICTION MODELS
prestressed. For the beam failing by debonding, no form of corresponding anchorage was
present.
A list of the experimental studies included in the database is shown in Table 3.4
together with the number of beams tested and the failure modes observed. The
classification of the failure modes was based mainly on the photos and/or descriptions
provided. More details of those beam tests can be found in Appendix A.3. In total, 181
beams were found. 36 beams failed due to flexural failures, in which 19 beams failed by
concrete crushing and 17 failed by FRP rupture. 58 beams failed by intermediate span
debond, in which 43 beams failed due to flexure-shear crack debond, 14 beams due to
flexural crack debond and one beam due to shear crack debond. 82 beams failed by end
debond, in which the only mode observed was end cover peeling. The variability in some
of the test parameters is shown in Table 3.5.
87
CHAPTER 3
88
ASSESSMENT OF PREDICTION MODELS
Table 3.5 Variability of some test parameters in the beam test database
Parameter Minimum Maximum Average
Beam span, L (mm) 812 6000 1972
Beam aspect ratio, bc/h 0.45 1.67 0.78
Ratio of FRP length to beam depth, Lf/h 0.50 5.40 2.99
Relative stiffness, EfAf / EsAs 0.03 1.06 0.27
Concrete strength, fc (MPa) 25 80 45.48
FRP modulus, Ef (MPa) 10343 400000 150754
In the theoretical models used to predict debonding modes, there were a few
parameters, which were not measured or mentioned in a number of experimental studies.
In those cases, the parameters were assumed as follows: The elastic modulus of steel, Es,
was taken as 200000 MPa. The concrete elastic modulus was taken as 5050 f c
according to AS3600 (Standards Australia, 2001). The average concrete tensile was
calculated according to CEB-FIB Model Code 1990 (Comite Euro-International du
Beton, 1991) The adhesive thickness was taken as 1 mm as observed in the experiments
by the author. The elastic modulus of the adhesive was taken as 8500 MPa as
recommended by Smith and Teng (2002).
Similar to the validation study of the shear-lap tests, model accuracy is determined
based on two statistical parameters: the average and the coefficient of variation of the
ratios of the capacity measured in the experiments to the predicted ultimate capacity
Mexp / Mcal or Vexp / Vcal.
89
CHAPTER 3
120
100
80
Mcal (kN.m)
60
40
20
0
0 20 40 60 80 100 120
Mexp (kN.m)
To assess the debonding strength models, sectional analysis was carried out using
the beam theory (described in Appendix A.1). For the cases when the predicted
debonding failure load exceeded the ultimate load based on the sectional moment
capacity, the latter was used. To find the failure load, a backward substitution method was
used. A procedure was set up using VBA in Microsoft Excel 2002 to vary the applied
shear load until a failure ratio was close to 1. The failure ratio was defined as the ratio of
failure criteria to the corresponding capacity. The allowable error was 1 %. Checks were
also carried out for the parameter limits and the results were ignored if a requirement was
not satisfied.
The validation results of the models by Arya and Farmer (2001) and Shehata et al.
(2001) are shown in Figure 3.6. It was found that the FRP strain limit was always critical
90
ASSESSMENT OF PREDICTION MODELS
in Arya and Farmer’s model. The steel strain did not exceed 0.5εsy for any specimen.
Despite the simplification made, the predicted failure loads agree well with the
experimental results. The average of the ratios Vexp/Vcal is 1.08 with the coefficient of
variation being 15 %. However, for some beams, the calculated load is not conservative.
For Shehata et al.’s model, since the allowable strain limit was reduced from 0.006 to
0.005, the average of Vexp/Vcal increases to 1.15 and the predictions are conservative for
most cases.
250 250
Vcal (kN)
Vcal (kN)
200 200
150 150
100 100
50 50
0 0
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350
Vexp (kN) Vexp (kN)
Figure 3.6 Validation results for Arya and Farmer’s model (a) and Shehata et al.’s
model (b)
The validation results of the models by Maruyama and Ueda (2001) and Teng et
al. (2001) are illustrated in Figure 3.7. Validation of Maruyama and Ueda’s model was
done assuming that the fracture energy was 0.5 N/mm. The model is generally
conservative with the average of the ratios Vexp/Vcal being 1.47 and the coefficient of
variation being 20 %. Very similar results are found for Teng et al.’s model.
91
CHAPTER 3
Vcal (kN)
200 200
150 150
100 100
50 50
0 0
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350
Vexp (kN) Vexp (kN)
Validation results for the models by Blaschko (1997) and Mohamed Ali (2000)
are plotted in Figure 3.8. These models produce relatively scattered results with the
coefficient of variation being 23 % for both cases. Mohamed Ali’s model is very
conservative. The average of Vexp/Vcal for this model is 2.70.
250 250
Vcal (kN)
Vcal (kN)
200 200
150 150
100 100
50 50
0 0
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350
Vexp (kN) Vexp (kN)
92
ASSESSMENT OF PREDICTION MODELS
350 FSCD
FCD
SCD
300
250
Vcal (kN)
200
150
100
50
0
0 50 100 150 200 250 300 350
Vexp (kN)
The simple model by Jones et al. for steel plated beams does not work well for
FRP plated beams. As the predicted failure loads are higher than the ultimate shear
capacities assuming no debond for most cases, the validation results are invalid and not
presented here. Shehata et al.’s model produces relatively good results (Figure 3.10). The
average of Vexp/Vcal for this model is 1.15 with 22 % variation.
93
CHAPTER 3
160
140
120
100
Vcal (kN)
80
60
40
20
0
0 20 40 60 80 100 120 140 160
Vexp (kN)
To verify Nguyen et al.’s model, the limiting strain was taken as the lowest value
found by the researchers, i.e. 0.0017. Similarly for Fanning and Kelly’s model, the limit
strain slope was taken as 4.2 strain/mm × 10-6. The validation results for these two models
are shown in Figure 3.11. Both models produce slightly scattered results with the
coefficient of variation being 25 and 26 %, respectively. Nguyen et al.’s model provides
relatively more accurate predictions with the average of Vexp/Vcal being 1.02.
200 200
160 160
Vcal (kN)
Vcal (kN)
120 120
80 80
40 40
0 0
0 40 80 120 160 200 0 40 80 120 160 200
Vexp (kN) Vexp (kN)
94
ASSESSMENT OF PREDICTION MODELS
Figure 3.12 shows the comparison of the predictions of end debond strengths
using the models by Saadatmanesh and Malek (1998), Tumialan et al. (1999) and El-
Mihilmy and Tedesco (2001) with the experimental data. It is clear that these models do
not provide accurate predictions. The average of Vexp/Vcal ranges from 1.77 to 4.27 with
the coefficient of variation ranging from 26 to 35 %.
160 160
140 140
120 120
100 100
Vcal (kN)
Vcal (kN)
80 80
60 60
40 40
20 20
0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
Vexp (kN) Vexp (kN)
160
140
120
100
Vcal (kN)
80
60
40
20
0
0 20 40 60 80 100 120 140 160
Vexp (kN)
Validation results for the shear capacity based strength models are shown in
Figure 3.13 to Figure 3.15. These models tend to give less scattered results with the
variability of around 15 %. Oehlers’ and Smith and Teng’s model base on the shear
95
CHAPTER 3
bending moment interaction diagrams and produce similar results. Both models provide
moderately conservative predictions. The averages of Vexp/Vcal are 1.82 and 1.84,
respectively. The models by Jansze (1997) and Ahmed and Van Gemert (1999) are less
conservative with the corresponding averages of Vexp/Vcal being 1.44 and 1.21,
respectively. In Ziraba et al.’s model, the beam shear strength comprises of the shear
contribution of plain concrete and of the modified shear contribution of stirrups. The
latter was found to be insignificant for most cases. Therefore, the results for this model
were found to be similar to those of other shear based models, where the shear
contribution of stirrups was not considered.
160 160
140 140
120 120
100 100
Vcal (kN)
Vcal (kN)
80 80
60 60
40 40
20 20
0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
Vexp (kN) Vexp (kN)
96
ASSESSMENT OF PREDICTION MODELS
160 160
140 140
120 120
100 100
Vcal (kN)
Vcal (kN)
80 80
60 60
40 40
20 20
0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
Vexp (kN) Vexp (kN)
160
140
120
100
Vcal (kN)
80
60
40
20
0
0 20 40 60 80 100 120 140 160
Vexp (kN)
Validation results of concrete tooth scenario based models are shown in Figure
3.16. Predictions using the lower limit by Raoof and Hassanen are presented only since
the upper limit was not critical for most cases. The average of Vexp/Vcal for this model is
1.26. A scattered distribution is also observed. Chaallal et al.’s model gives very
inaccurate predictions. The validation results for this model are also very scattered.
97
CHAPTER 3
160 160
140 140
120 120
100
Vcal (kN)
100
Vcal (kN)
80 80
60 60
40 40
20 20
0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
Vexp (kN) Vexp (kN)
(a) Raoof and Hassanen (2000): lower bound (b) Chaallal et al. (1998)
Figure 3.16 Validation results for Raoof and Hassanen’s model – lower limit (a)
and Chaallal et al.’s model (b)
98
ASSESSMENT OF PREDICTION MODELS
200 200
160 160
Vcal (kN)
Vcal (kN)
120 120
80 80
40 40
0 0
0 40 80 120 160 200 0 40 80 120 160 200
Vexp (kN) Vexp (kN)
(a) Neubauer and Rostasy (1997) (b) Arya and Farmer (2001)
Figure 3.17 Validation results for Neubauer and Rostasy’s model (a) and Neubauer
and Rostasy’s model with Arya and Farmer’s limit (b)
Table 3.6 summarises the validation results for the models discussed above. For
intermediate span debond, the simple models by limiting FRP strain level (Arya and
Farmer’s and Shehata et al.’s) tend to perform best despite the fact that they assume a
constant FRP tensile strain at peeling-off. Recent test results have demonstrated that the
FRP tensile strain at failure depends on a broad range of parameters, such as the
properties of FRP and concrete (fib Bulletin 14, 2001). These models can, however, be
used for a quick check. For end debond, most models tend to give scattered and/or
inaccurate results. Even though Nguyen et al.’s model has the average of Vexp/Vcal closest
to 1, the assumption of a constant FRP strain at failure for all cases has not been proved
by experiments.
99
CHAPTER 3
100
ASSESSMENT OF PREDICTION MODELS
FSCD 160
350 FCD
SCD 140
300
120
250
Vcal (kN)
100
Vcal (kN)
200
80
150
60
100
40
50
20
0
0
0 50 100 150 200 250 300 350
Vexp (kN) 0 20 40 60 80 100 120 140 160
Vexp (kN)
101
CHAPTER 3
The second study involved an assessment of the strength models for three main
failure modes: flexural failures, intermediate span debond and end debond. 26 models
were verified using a database of 181 beams. The main conclusions drawn from this study
are:
• Beam theory is able to predict the full composite action of beams strengthened with
FRP composites.
• Intermediate span debond capacity is closely related to the CFRP tensile strain. Even
though the models based purely on limiting FRP strain yield good correlation with
current database, the effect of other important parameters such as FRP and concrete
properties need to be taken into account. The empirical formulae recommended by
ACI provide good correlation.
• End debond capacities are not predicted well by existing models. Most models fail to
recognise that the failure plane is in the concrete layer at tension reinforcement level.
The ACI guideline provides a safe design limit for this kind of failure.
102
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
4.1 Introduction
The mechanical performance and durability of FRP composites in strengthening
applications depend primarily on their bond with the main structure. However, FRP
composites tend to delaminate or debond from concrete due to the local shear stress
concentration and the low shear strength of concrete. To be able to use FRP composites
for strengthening, an understanding of bonding and debonding mechanisms is therefore
essential.
In order that the results of these shear-lap tests could be used in the study of the
beam tests reported later in Chapter 5, the concrete block width and CFRP cross sectional
dimensions were chosen to be the same as in those beams, and the same fibre, epoxy and
bonding procedures were selected. The blocks were cast together with the beams so that
the concrete strength would be similar. To ensure that the bond length was sufficient and
that the specimens could be easily fitted in the steel cage used to fix the blocks to the bed
of the loading machine, the concrete block length was chosen to be 300 mm. The gripping
103
CHAPTER 4
steel plate length was also chosen to be 250 mm in total, which was long enough such
that the stress was transferred to the CFRP relatively uniformly.
Steel cage H
CFRP
Lf
300
bf
140
140
104
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
4.3.2 Materials
4.3.2.1 Concrete
The concrete material used was supplied separately, pre-mixed from a local
supplier. Compressive strength tests were carried out on concrete cylinders for each batch
in accordance with AS1012.9 (Standards Australia, 1999). The concrete strength after 67
days was 53.7 MPa. Further information on the concrete strength development with time
is presented in Appendix B.1.
105
CHAPTER 4
Fibres, which are also known as S&P C-sheet 240, are high tensile strength carbon fibres
supplied in unidirectional tow sheets of 300 mm width. The fibres were held together
with a backing grid and rolled together with a plastic sheet. The nominal thickness, based
on the total thickness of the fibres in a unit width, was reported to be 0.176 mm. The fibre
weight was 300 g/m2. MBrace Saturant is a two-part epoxy with 100 % solids content
used to impregnate the fibres to form a composite and to provide bonding to the primed
surface. Its compressive strength was reported to be greater than 80 MPa. The pot life of
the saturant was 30 minutes at 25°C and it reached the design strength after 7 days.
Figure 4.2 MBrace FRP system components and main tools used for installation
106
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
The bonding operation was carried out about 30 minutes to 1 hour after the
application of the primer. The operation included resin under-coating, carbon fibres
placement and resin over-coating. MBrace Saturant was first applied on the primed
surface using a steel scraper to form a coat approximately 2 mm thick, which proved to be
sufficient to achieve a wet-out of the FRP fabrics. Then, a piece of MBrace CF 130 fibre
sheet, which had been cut beforehand into prescribed sizes using scissors, was placed
107
CHAPTER 4
with the fibre side down onto the coating and generally smoothed down by hand. After
that, the surface of the sheet was rolled over along the longitudinal direction of the fibres
using a ribbed roller to impregnate resin into the fibres and remove any air bubbles.
Rolling was continued until the resin was squeezed out between the fibres. A thin layer of
resin was then placed on the fibres and pressed down with the scraper to smooth out any
remaining imperfections. To bond another ply of fibres, the same steps were followed. To
allow for epoxy impregnation, a period of five minutes was allowed between subsequent
applications. The specimens were left to cure at room temperature for at least seven days
before testing.
The specimens were instrumented to record strain, load and extension readings.
Out of two identical specimens manufactured for each configuration, one block (block
‘a’) had more instrumentation than the other (block ‘b’). Strain gauges were used to
measure the strain levels in the composite along its centre line (Figure 4.3). Loads and
extensions were measured using a load cell and a linear variable displacement transducer
(LVDT) in the loading machine. LVDTs were also mounted on the sides of the block in
an attempt to measure the slip of the CFRP. However, it was found that their readings
were not reliable since the slips were too small to be picked up by these LVTDs.
108
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
15
15
15
2@18
3@18
4@18
Lf = 60 mm Lf = 80 mm Lf = 100 mm
6@18 15
15
15
8@18
10@18
Lf = 140 mm Lf = 180 mm Lf = 220 mm
The data acquisition equipment made by dataTaker, ‘DT505’, was used in the
experimental program with a channel expander module ‘CEM-AD’. During testing,
readings were recorded once every second using the computer programs ‘Defriend’ and
‘Detransfer’. Crack propagation was captured by a digital video recorder and a digital
high-speed video recorder. The high-speed video recorded at 500 frames per second and
it was used during two tests only.
109
CHAPTER 4
the bonding of CFRP on concrete. The plates were connected to the jaw through a
connection rod. The plate thickness was chosen to be 3 mm to ensure that yielding would
not occur near the connection hole. During installation of the specimen, good alignment
was controlled using a plumb-line. The tension bolts were tightened slightly before the
test.
Connection
Tension rod
bolts CFRP limit
Support
block
Stiffener
Steel Bottom
clamp bed
(a) (b)
Figure 4.4 Loading and gripping method (a) and a specimen before testing (b)
110
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
The specimens bonded with two plies of CFRP with a very short bond length (60
mm) or with six plies of CFRP of 100 mm bond length failed by shear-tension failure. A
crack initiated from the unloaded CFRP end, propagated deep into the concrete block and
finally terminated near the support block location. The failure was very brittle. The crack
propagated through both the cement paste and coarse aggregate (Figure 4.7).
Some specimens with two plies of CFRP bonded over a very short length or with
six plies of CFRP bonded at a long bond length failed either due to shear-tension failure
or interfacial debond. One reason for the inconsistency was that, in those cases, the stress
transferred well to the unloaded CFRP end, which resulted in a high stress concentration
at both ends of the composite and therefore both failure modes were likely. Another
possible reason for the inconsistency was the difficulty in keeping the CFRP perfectly
aligned with the concrete block. This led to bending of the composites and therefore a
different stress distribution. This could have been exacerbated under high loads due to the
rotation of the concrete block.
111
CHAPTER 4
The observed failure modes and recorded ultimate capacities are summarised in
Table 4.4. The failure loads varied from 18.8 kN to 42.8 kN with an average of 28 kN.
The averages of the average bond stress and the maximum FRP stress were 2.7 MPa and
704.5 MPa, respectively. The table also shows that the model by Chen and Teng (2001)
predicts the bond strengths of the specimens relatively well. The average of the ratios of
the experimental failure load to the prediction is 0.98 with the coefficient of variation
being 13 %.
The photos of all specimens after failure are shown in Appendix C.1. The high-
speed video captures for a specimen are also illustrated in Appendix C.2.
112
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
Carbon fibre
Coarse
Transverse crack aggregate
initiation location Cement paste
Surface cracks
Force direction
Force direction
Coarse
aggregate
113
CHAPTER 4
114
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
than the fibre/saturant layers because of the pressure applied during bonding. The
approximate thicknesses of the layers are illustrated in Figure 4.8b.
1.0
Concrete
Concrete
All dimensions are in mm
(a) A cut through a 6-ply composite (dark (b) Layers of a 2-ply CFRP composite bonded on
layers contain mainly carbon fibres) concrete
In the experiments, the CFRP strain was measured near the composite top surface
(Figure 4.8b). The top strain was expected to be different to the average composite strain
as the strain could vary across the thickness of the composite. This variation was very
difficult to measure experimentally. However, as shown by the finite element analysis
presented in Chapter 6, the difference between the top strain and the average value was
only significant near the ends of the CFRP and near the tips of widely open transverse
cracks. Hence, to describe the global behaviour of the bond, it was reasonable to assume
that the strain measured can represent the average CFRP strain.
The strain distributions measured in the specimens with varying bond lengths are
plotted in Figure 4.9 to Figure 4.14. At a low load level, the distributions show a gradual
decline from the peak near the loaded edge to the other end. As the load increases but still
at a relative low level (less than 10 kN), the strain at the first gauge increases most
rapidly. As a result, the distributions have the steepest slope near the loaded edge
(between the first and second gauges from the loaded edge). As the load increases further
to around 15 kN, the strain at the second gauge starts to increase faster than that at the
first gauge. Consequently, the steepest slope moves to the right of the second gauge.
115
CHAPTER 4
With addition of more loads, the steepest portion shifts further to the right. The
distributions start to flatten out near the loaded edge indicating delamination underneath
the CFRP there. This is observed most clearly in the specimens with long bond lengths
(T4, T5 and T6). The effective bond length can be approximated from the strain
distributions of those specimens. It is the distance over which the maximum strain
decreases to near zero. From the plots, the effective bond length is found to be
approximately 100 mm. The maximum CFRP strain is approximately 3000 microstrain.
1500 18.0
20.0
1000
500
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 4.9 Strain distributions along bond length in specimen T1a (Lf = 60 mm)
20.0
1500 22.0
24.0
1000 25.8
500
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 4.10 Strain distributions along bond length in specimen T2a (Lf = 80 mm)
116
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
Microstrain
2000 20.0
22.0
1500 25.7
1000
500
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 4.11 Strain distributions along bond length in specimen T3a (Lf = 100 mm)
2000 20.0
22.0
1500
24.0
1000 26.0
500 26.7
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 4.12 Strain distributions along bond length in specimen T4a (Lf = 140 mm)
3000 22.0
2500 24.0
2000 26.0
1500 27.0
1000
27.8
500
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 4.13 Strain distributions along bond length in specimen T5a (Lf = 180 mm)
117
CHAPTER 4
Figure 4.14 Strain distributions along bond length in specimen T6a (Lf = 220 mm)
The strain distributions measured in the specimens with varying support block
heights are plotted in Figure 4.15 and Figure 4.16. There is little difference in the
distributions.
2500 20.0
2000 25.0
30.0
1500
32.0
1000
33.0
500
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 4.15 Strain distributions along bond length in specimen T7a (H = 5 mm)
118
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
3500 5.0
10.0
3000 15.0
Microstrain
2500 20.0
2000 25.0
26.0
1500 27.0
1000 28.0
500 28.5
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 4.16 Strain distributions along bond length in specimen T8a (H = 70 mm)
The strain distributions measured in the specimens with varying CFRP widths are
plotted in Figure 4.17 and Figure 4.18. It can be seen that the maximum composite strain
increases as the CFRP width decreases. The maximum CFRP strains increase to
approximately 3500 and 4500 microstrain for the specimens with the CFRP width of
70 mm and 50 mm, respectively.
3000
15.0
2500
2000 16.0
1500 17.0
1000 18.0
500
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 4.17 Strain distributions along bond length in specimen T12a (bf = 50 mm)
119
CHAPTER 4
Figure 4.18 Strain distributions along bond length in specimen T13a (bf = 70 mm)
For the specimens bonded with six plies of CFRP, i.e. specimens T9, T10 and
T11, the maximum applied loads were higher and the measured strains were much lower
than those of the specimens bonded with two plies. Therefore, the measurements on these
specimens were much more sensitive to the alignment errors. The errors were difficult to
be avoided and consequently led to bending of the composite plate during loading. This
especially affected the readings of the gauges near the loaded edge. As a result, the strain
measurements for these specimens were not reliable near the edge. As illustrated in
Figure 4.19 to Figure 4.21, the gauges closest to the loaded edge measured low and even
negative strains indicating that the composite was bended there.
500
0
0 50 100 150 200
Microstrain
-500
Load (kN)
5.1
-1000 10.0
15.0
-1500 20.0
25.0
29.8
-2000
Distance from loaded edge (mm)
Figure 4.19 Strain distributions along bond length in specimen T9a (Lf = 100 mm)
120
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
Microstrain
20.0
300 25.0
200 30.0
35.0
100 36.0
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 4.20 Strain distributions along bond length in specimen T10a (Lf = 140 mm)
600 35.0
400 40.0
200 42.8
0
-200 0 50 100 150 200
-400
Distance from loaded edge (mm)
Figure 4.21 Strain distributions along bond length in specimen T11a (Lf = 180 mm)
The average bond stress between two adjacent strain gauges can be calculated by
dividing the force difference by the bond area and given by the following equation:
E f (ε f ,i +1 − ε f ,i ) t f
τ= (4.1)
∆L
where Ef and tf are the CFRP elastic modulus and thickness, respectively; εf,i+1 and εf,i are
the CFRP strains; and ∆L is the distance between strain gauges. The bond stress
distributions at discrete data points for three specimens with relatively long bond lengths
are plotted in Figure 4.22 to Figure 4.24. The distributions of other specimens are
included in Appendix C.3.
The area under the bond stress curve is proportional to the amount of tension
transferred by the bond. It can be observed from these plots that the tension is transferred
121
CHAPTER 4
via a certain length only, called the ‘active length’. When the shear stress at the loaded
end exceeded the shear capacity, concrete cracking occurred and the ‘active length’
displaces toward the unloaded end of the specimen.
5 Load (kN)
5.0
4 10.0
Bond stress (MPa)
15.0
3 20.0
22.0
24.0
2
26.0
26.7
1
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 4.22 Average bond stress distributions along bond length in specimen T4a
8 Load (kN)
5.0
7
10.0
6
Bond stress (MPa)
15.0
5 20.0
4 22.0
24.0
3 26.0
2 27.0
1 27.8
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 4.23 Average bond stress distributions along bond length in specimen T5a
122
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
8 Load (kN)
7 5.0
10.0
6 15.0
Figure 4.24 Average bond stress distributions along bond length in specimen T6a
The average slip can be calculated as the incremental sum of the CFRP extension,
as follows
ε f ,i +1 + ε f ,i
si = ∆L + s i −1 (4.2)
2
In Equation 4.2, since the concrete block is much stiffer than the CFRP, concrete
elongation has been ignored.
The plots of the average bond stress versus the average slip at several load levels
for some specimens are presented below. The distributions for the other specimens are
shown in Appendix C.4. In these plots, the average bond stresses and slips are calculated
at several locations along the bond length. For example, the bond stress and slip at the
location 24 mm from the loaded edge are calculated using the readings from the strain
gauges at 15 and 33 mm from the loaded edge.
The plots for the specimens bonded with the CFRP composite of 100 mm width
bonded over a relatively long bond length (specimens T4a, T5a, and T6a) are shown in
Figure 4.25 to Figure 4.27. The curves vary significantly, possibly due to the presence of
transverse cracks along the joint. However, it can be seen that the peak shear stress was
around 4 MPa at a slip of approximately 0.05 mm. Similar observation can also be made
from the bond stress slip distributions at several load levels for the specimens with the
support block height equal to 70 mm (Figure 4.28). However, all of these curves appear
to have a nonlinear ascending and descending trends. It was found that these trends can
123
CHAPTER 4
The maximum bond stresses of the local bond-slip curves seem to depend
significantly on the composite width. The peak bond stresses are approximately 6 and 5
MPa for the specimens with the composite widths of 50 and 70 mm (specimens T12a and
T13a), respectively.
0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25
Slip (mm) Slip (mm)
(a) (b)
Figure 4.25 Bond stress versus slip development in specimen T4a (bf = 100 mm)
15.0 20.0
Bond stress (MPa)
5 6
96 22.0 24.0
4 5
26.0 27.0
4 27.8
3
3
2
2
1 1
0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25
Slip (mm) Slip (mm)
(a) (b)
Figure 4.26 Bond stress versus slip development in specimen T5a (bf = 100 mm)
124
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
(a) (b)
Figure 4.27 Bond stress versus slip development in specimen T6a (bf = 100 mm)
6 6
60 25.0 26.0
5 5 27.0 28.0
4 4 28.5
3 3
2 2
1 1
0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25
Slip (mm) Slip (mm)
(a) (b)
Figure 4.28 Bond stress versus slip development in specimen T8a (bf = 100 mm)
8
7
6
Bond stress (MPa)
Popovics'
5
equation
4
3
2
1
0
0.00 0.05 0.10 0.15 0.20 0.25
Slip (mm)
Figure 4.29 Fitted bond-slip relationships for specimens with two layers of CFRP of
100 mm width
125
CHAPTER 4
6
60 78 16.0 17.0
5 5
18.0
4 4
3 3
2 2
1 1
0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25
Slip (mm) Slip (mm)
(a) (b)
Figure 4.30 Bond stress versus slip development in specimen T12a (bf = 50 mm)
7 24 42 7 5.0 10.0
15.0 18.0
Bond stress (MPa)
6
Bond stress (MPa)
6
60 78 19.0 20.0
5 5
20.5 21.0
4 4
3 3
2 2
1 1
0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25
Slip (mm) Slip (mm)
(a) (b)
Figure 4.31 Bond stress versus slip development in specimen T13a (bf = 70 mm)
126
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
It can be observed that the initial stiffness is very similar for the specimens with
two plies of CFRP. For the specimens with six plies, the stiffness varies due to the
alignment errors but the general trend is similar. As expected, the specimens with longer
bond lengths show more ductile behaviour.
35
30
Total load (kN)
25
T1a - Exp.
20 T2a - Exp.
15 T3a - Exp.
T4a - Exp.
10 T5a - Exp.
5 T6a - Exp.
0
0.00 0.10 0.20 0.30 0.40 0.50
Slip at 42 mm from loaded edge (mm)
Figure 4.32 Load versus slip curves for specimens with two layers of CFRP and
variable bond lengths
45
40
35
Total load (kN)
30
25
20 T9a - Exp.
15 T10a - Exp.
10 T11a - Exp.
5
0
0.00 0.05 0.10 0.15
Slip at 42 mm from loaded edge (mm)
Figure 4.33 Load versus slip curves for specimens with six layers of CFRP and
variable bond lengths (subjected to alignment errors)
The effect of the bond length, Lf, on the ultimate load capacity and average
maximum bond stress is illustrated in Figure 4.34. The average bond stress was calculated
by simply dividing the total load by the total bond area. It is clear that for the specimens
with two plies of carbon fibres, the ultimate load does not increase significantly once the
bond length exceeds 100 mm. The average bond stress therefore decreases greatly as Lf
increases beyond that limit. A slight increase in the ultimate load is still observed for the
127
CHAPTER 4
specimens with a very long bond length possibly due to some friction after debonding.
When the failure is dominantly shear-tension (specimens T9a to T11a), an increase in the
bond length results in a significant increase in the bond capacity.
30 2.0
20 1.5
1.0
10 0.5
0 0.0
0 50 100 150 200 250 0 50 100 150 200 250
Lf (mm) Lf (mm)
(a) (b)
Figure 4.34 Effect of bond length on bond capacity
A thicker composite can allow better stress transfer along the bond area. However,
the improvement usually does not outweigh the waste in CFRP materials. Figure 4.36b
demonstrates this point showing that the maximum CFRP stress is reduced significantly
as the nominal fibre thickness increases from 0.352 to 1.056 mm. The specimens with a
thicker composite also have much stiffer behaviour.
30
25
Total load (kN)
20
15 T3a - Exp.
10 T9a - Exp.
5
0
0.00 0.05 0.10 0.15 0.20
Slip at 42 mm from loaded edge (mm)
Figure 4.35 Load-slip curves for specimens with different CFRP thicknesses T3a (2
plies) and T9a (6 plies)
128
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
35 620
30 420
25 220
20 20
0 0.5 tf (mm) 1 1.5 0 0.5 1 1.5
tf (mm)
(a) (b)
Figure 4.36 Effect of CFRP thickness on bond capacity
Figure 4.37 shows that the composite width has a significant effect on the ultimate
capacity. The specimen with a narrower CFRP width has a lower bond capacity.
However, if the average bond stress is considered, the narrower strip proves to be more
effective (Figure 4.38b).
30
25
Total load (kN)
20
15
T3a - Exp.
10 T12a - Exp.
T13a - Exp.
5
0
0.00 0.05 0.10 0.15 0.20
Slip at 42 mm from loading edge (mm)
Figure 4.37 Load-slip curves for specimens with different CFRP widths T3a (100
mm), T12a (50 mm) and T13a (75 mm)
129
CHAPTER 4
30 5
3
15
2
10
1
5
0 0
20 40 60 80 100 120 20 40 60 80 100 120
b f (mm) b f (mm)
(a) (b)
Figure 4.38 Effect of CFRP width on bond capacity
Figure 4.39 and Figure 4.40 indicate that the support block clearance has an
insignificant effect on the bond except when the clearance is very small. Specimen T3a
with only 5 mm clearance failed at a slightly higher load. However, the variation between
two identical specimens T3a and T3b is also significant (20 %). Therefore, it is not
possible to draw conclusions about the effect of a small clearance on the bond capacity.
35
30
Total load (kN)
25
20
15 T3a - Exp.
10 T7a - Exp.
T8a - Exp.
5
0
0.00 0.05 0.10 0.15 0.20
Slip at 42 mm from loading edge (mm)
Figure 4.39 Load-slip curves for specimens with different support clearances T3a
(110 mm), T7a (5 mm) and T8a (70 mm)
130
EXPERIMENTAL INVESTIGATION OF SHEAR-LAP SPECIMENS UNDER TENSILE TESTING
35
30
25
Load (kN)
20
15
10
5
0
0 25 50 75 100 125
H (mm)
• Two types of failure were observed. Interfacial debond was dominant in the
specimens with two plies of CFRP bonded over a sufficient length. Shear-tension
failure was likely for the ones with six plies of CFRP or with two plies of CFRP but
bonded over a very short length. The failure surface was in the concrete through both
the cement matrix and aggregates.
• The failure loads varied from 18.8 kN to 42.8 kN with an average of 28.0 kN. The
means of the average bond stress and the maximum FRP stress were 2.7 MPa and
704.5 MPa, respectively.
• A typical CFRP strain distribution showed an exponential curve at early stages of
loading with the maximum value located near the loaded edge. The curve gradually
flattened as the load approached the ultimate capacity.
• The strain distributions also indicated an effective stress transfer length of around 100
mm for specimens with 2 plies of CFRP of 100 mm width. The local bond-slip curves
were nonlinear and varied significantly possibly due to the presence of inclined cracks
intersecting in the bond. The average maximum bond stress was around 4 MPa at an
average slip of approximately 0.05 mm for the specimens with 2 plies of CFRP of 100
131
CHAPTER 4
132
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
5.1 Introduction
There have been a number of studies on debonding failures in RC beams
strengthened using FRP composites as reported in Chapter 2. However, due to the
complexity of the failures, the mechanisms are still not fully understood and the
influences of several parameters are still not clear.
Therefore, in this study, two experimental programs are carried out to further
investigate the behaviour of RC beams retrofitted with CFRP fabrics using a wet lay-up
method. The aim of the first program is to study the failure mechanisms and the influence
of a number of parameters. In this program, a total of eighteen RC beams are constructed
and tested under four-point bending. Two are control beams and sixteen are retrofitted
with CFRP fabrics. The variables are the CFRP bond length, CFRP thickness, steel
tension reinforcement amount, concrete cover and stirrup spacing. The aim of the second
experimental program is to investigate the effectiveness of anchorage systems using
composite U-straps. In the program, the beams are tested under three-point bending. A
total of eight tests are carried out. One is not anchored and seven are anchored with either
non-prestressed or prestressed CFRP U-straps.
133
CHAPTER 5
experiments are also to help identify critical failure modes and establish design
parameters for the main experimental programs.
The beams were loaded in four-point loading with a total span of 2300 mm and a
shear span of 700 mm. A steel clamp was used at one end to force the failure at the other
end. A high-speed video camera was utilised to capture the failure sequences. Strain
gauges were also installed on the CFRP and tension reinforcement.
134
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Beam B1 failed due to intermediate span debond (flexure-shear crack debond) and
beam B2 failed by end debond (end cover peeling) (Figure 5.3). Both beams experienced
a very brittle failure. As shown in Figure 5.2, three regions can be clearly identified
according to the changes in the slope when the concrete cracked and when the tension
reinforcement yielded. As expected, beam B1 had lower stiffness after the yielding of
steel as the bonded composite was thinner. The ultimate loads were however very close
for beams B1 and B2 (149 kN and 155 kN, respectively).
100
80
Shear load (kN)
60
B1
40
B2
20
0
0 10 20 30 40
Midspan deflection (mm)
135
CHAPTER 5
(B1)
(B2)
The CFRP strain distributions are shown in Figure 5.4. The maximum strain level
in beam B1 is much higher than that in beam B2. The average bond stress distributions
were derived from the CFRP strain distributions (using Equation 4.1) and are plotted in
Figure 5.5. Beam B1 has a peak value near its midspan whereas B2 has a peak value near
the composite end. This agrees with the failure modes observed.
Examination of the crack patterns reveals the existence of more flexural cracks
near the midspan and less inclined cracks near the CFRP end in beam B1 than in beam
B2. The clamp proved to be very effective in preventing the formation of inclined cracks
leaving no crack in beam B1 and only two thin cracks in beam B2.
The images from the high-speed video illustrated the failure mechanism of end
cover peeling. At first, when the steel had yielded, visible inclined shear cracks were
observed near FRP end. Figure 5.6a shows a sudden crack propagation originating from
those inclined cracks. The concrete teeth failed simultaneously. Flexure-shear crack
debond had a different mechanism. Inclined cracks appeared closer to the load point.
Those cracks became more inclined as the load increased. Vertical shear deformation was
observed clearly before failure (Figure 5.6b). Cracks propagated towards the CFRP end
136
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
along the concrete layer located approximately 1 to 2 mm above the bond surface. The
concrete cover in the bending region was also sheared off.
Load Load
(kN) 8000 (kN)
10000
10 10
8000 20 6000 20
30 30
Microstrain
Microstrain
6000 40 40
4000
50 50
4000
60 60
65 2000
2000 70
70 75
0 74 0 78
0 500 1000 0 500 1000
Distance from FRP end Distance from FRP end
(mm) (mm)
(a) B1 (b) B2
Figure 5.4 CFRP strain distributions
Load Load
7 (kN) 10 (kN)
10 9 10
6
20 8 20
Bond stress (MPa)
Bond stress (MPa)
5 7
30 30
4 6
40 40
5
3 50 4 50
2 60 3 60
65 2 70
1
70 1 75
0 0
74 78
0 275 550 0 275 550
Distance from FRP end Distance from FRP end
(mm) (mm)
(a) B1 (b) B2
Figure 5.5 Average bond stress distributions in shear span
137
CHAPTER 5
CFRP
end
(a) End cover peeling
(high-speed video
capture)
Load
point
(b) Flexure-shear crack
debond
138
End debond failure shear load (kN)
0
10
20
30
40
50
60
70
80
90
100
68
Shehata et al. (2001)
Intermediate span debond failure shear load (kN)
88
Nguyen et al. (2001) 0
10
20
30
40
50
60
70
80
90
100
59
Fanning and Kelly (2001)
21
Saadatmanesh and Malek (1998) Arya and Farmer (2001)
62 60
40
Tumialan et al. (1999) Shehata et al. (2001)
15
El-Mihilmy and Tedesco (2001) Maruyama and Ueda (2001)
58 56
failure load
B2
B1
Experimental
47
38 38
Smith and Teng (2003) Blaschko et al. (1997)
25
46
Jansze (1997) Mohamed Ali (2000)
failure load
Experimental
75
59
Ahmed and Van Gemert (1999) Niedermeier (2000)
39
Ziraba et al. (1994)
77
Raoof and Hassanen (2000): Lower
13
Chaallal et al. (1998)
Neubauer and Rostasy (1997)
90 90
Figure 5.8 Applicability of theoretical models to predict end debond loads for beam
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
139
CHAPTER 5
A total of eighteen RC beams were constructed. Two were control beams and
sixteen were retrofitted with carbon fibre reinforced fabrics using a wet lay-up method.
The variables for the experiment program are listed in Table 5.2. The retrofitted beams
were divided into two main groups. The E group had ten beams retrofitted with a
140
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
relatively thick layer of CFRP (six or nine plies). The S group had six beams retrofitted
with two plies of CFRP. Two identical beams were manufactured for each configuration
and denoted as ‘a’ and ‘b’. For example, E1 configuration has two beams: E1a and E1b.
The test variables included the CFRP bond length, the area of tension reinforcement, the
concrete cover, the number of CFRP plies and the amount of shear reinforcement.
2N12
46
220
R10*
260*
3N12*
100 CFRP*
140
Figure 5.9 Typical rectangular beam cross-sectional details (beams E1a and E1b)
Longitudinal
Support CFRP Support
reinforcement
150* 150*
* varying parameter S N
All dimensions are in mm.
Figure 5.10 Typical beam longitudinal details (beams E1a and E1b)
141
CHAPTER 5
On one side of the retrofitted beams, the longitudinal CFRP was anchored with a
steel clamp as shown in Figure 5.11. The clamp was used to force end failure (if it
occurred) at the other side. This allowed the camera to be focused on one side of the
beam. The clamp comprised of two steel plates 20 mm in thickness connected by two
bars 14 mm in diameter. Four nuts were used to tighten the clamp. To monitor the tension
in the bars, a torque wrench was used together with a strain gauge installed on one bar.
The wrench was set to a torque level so that a strain level of approximately 400
microstrain was introduced to the bolts.
142
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Concrete beam
Strain gauge
location on bolt
CFRP
5.4.2 Materials
5.4.2.1 Concrete
The concrete material used was supplied separately, pre-mixed from a local
supplier. The concrete was poured in two batches. The first batch was for S beams. The
second batch was for E beams. Compressive tests were carried out in accordance with
AS1012.9-1999 (Standards Australia, 1999). The concrete strengths for S beams and E
beams at the day of the first test in the series were 47.7 MPa (82 days) and 53.7 MPa (67
days), respectively. Further information on the concrete strength development with time
is presented in Appendix B.1.
143
CHAPTER 5
(a) (b)
Figure 5.12 Surface texture after water jetting (a) and after priming (b)
144
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
(a) (b)
(c) (d)
Figure 5.13 Bonding operation (a) Applying a layer of resin; (b) Placing fibres on
the resin; (c) Pressing fibres into the resin with a ripped roller; (d) Finished surface
The beams were instrumented to record strain, load and deflection readings. Strain
gauges were used to measure strain levels in the steel and CFRP reinforcement. Out of
two identical beams manufactured for each configuration, one beam (beam ‘a’) had more
instrumentation than the other. The strain gauge locations for a typical beam are
illustrated in Figure 5.14. The strain gauges were bonded on the horizontal tension
reinforcement on the bottom side of the middle bar. They are labelled as ‘Hi’ in Figure
5.14, where ‘i’ is the gauge number as counted from right to left. The strain gauges on the
stirrups were located at the middle height of the legs. They are labelled as ‘Via’ or ‘Vib’,
where ‘a’ and ‘b’ correspond to the gauges on the front and back legs, respectively. For
the retrofitted beams, the strain gauges were also placed on the composite surface and
labelled as ‘Ci’. All gauges were protected from the environment by a layer of M-Coat.
For the gauges on the steel reinforcement, wax coating was also applied before pouring
concrete to provide a water resistance layer. The deflections of the beams were measured
at three locations using linear variable displacement transducers (LVDTs) as shown in
145
CHAPTER 5
Figure 5.15. The loads and reactions were measured using load cells placed in the
actuator and the supports.
H5 H4 H3 H2 H1
C7 C6 C5 C4 C3 C2 C1
Two gauges on two 450 450 125 125 125 125
stirrup legs
S N
All dimensions are in mm.
Figure 5.14 Strain gauge locations on beam E1a
S N
All dimensions are in mm.
Figure 5.15 Load and deflection measurements
A digital video recorder and a high-speed video recorder were used to capture the
crack propagation. The high-speed video recorded at 500 frames per second.
The loading set-up and testing machine details are shown in Figure 5.16. The
beam was placed on two steel support blocks, which seated on two low friction bearing
146
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
strips to allow horizontal movements (Figure 5.17a). Load cells of 100 kN capacity were
placed underneath the bearing steel blocks to measure the reaction at the supports. The
load from the actuator was transferred to the beam through an I-beam and two rollers.
The rollers seated on two steel bearing plates. The I-beam stiffness was chosen to ensure
that its maximum deflection under the maximum applied load was negligible (less than
0.25 mm). The bearing plates were 100 mm wide and 20 mm thick. The actuator was
mounted on a loading frame, which consisted of two steel I crossbeams of 540 mm depth
bolted to two steel I-columns of 253 mm depth and 4500 mm high. Each column was
secured in place using two anchor bolts pre-tensioned to 50 kN against the strong floor as
illustrated in Figure 5.17b.
147
148
CHAPTER 5
Reaction frame
Loading jack
Concrete beam
Supports
(a) (b)
Figure 5.17 A support (a) and an anchorage of the reaction frame column (b)
The two control beams failed by typical steel yielding followed by secondary
compression failure of concrete. The load and crack development were very similar for
both beams C1a and C1b. At approximately 15 kN, fine flexural cracks became visible in
the middle of the beams. At around 35 kN, very fine flexure-shear cracks were observed
in the shear spans. As the load increased, more flexural cracks appeared and propagated
deeper into the beams. When the load reached a value of approximately 54 kN, the tensile
steel yielded and the loading curve levelled off. As loading was continued, the beams
showed wide flexural cracks at their midspan. These cracks extended to the compressive
area and the concrete crushed underneath the load point.
In beam C1a, local crushing and excessive flexural cracking occurred near the
right load point when the deflection was 43 mm. When the deflection reached 63 mm,
149
CHAPTER 5
similar crushing and cracking were also seen under the left load point. As the load
increased further, the most severe cracks were observed under the two load points (Figure
5.19a). Beam C1a showed very ductile behaviour. It could still hold a significant load
when the midspan deflection reached 142 mm. In beam C1b, concrete crushed first near
the left load point when the midspan deflection reached 40 mm. As the beam deflection
increased to 55 mm, more crushing between the left load point and the beam middle was
observed (Figure 5.19b). The beam also experienced ductile behaviour.
60
50
Shear load (kN)
40
Loading
30 terminated C1a
20 C1b
10
0
0 25 50 75 100 125 150
Midspan deflection (mm)
150
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
5.4.5.1.2 E beams
All of the beams in E series failed by end debond on the unclamped side. The
loading curves of these beams are shown together in Figure 5.20. The curves of the
control beams are also shown for reference. The stiffness of the retrofitted beams was
increased significantly compared to that of the control beams. However, the loading
curves of the retrofitted beams also show a more brittle failure. A detailed description of
each beam responses is presented in the following paragraphs.
80
70
60 C1a C1b
Shear load (kN)
50 E1a E1b
40 E2a E2b
E3a E3b
30
E4a E4b
20
E5a E5b
10
0
0 5 10 15 20 25 30
Midspan deflection (mm)
The load-deflection responses for E1 beams are plotted in Figure 5.21 and the
crack patterns after failure are shown in Figure 5.22. The crack development of beam E1a
is typical for all beams failing by end debond. Vertical flexural cracks appeared first in
the pure bending region. As the load increased, diagonal flexure-shear cracks became
visible in the shear span. A transverse crack originating from the CFRP end was observed
at 39 kN shear load level. As the load increased further, the cracks in the shear span
widened progressively. At the same time, the transverse crack propagated diagonally into
the concrete cover and then parallel along the tension reinforcement level. It finally
joined the adjacent diagonal crack. At 59 kN, this crack opened further and a portion of
about 90 mm length from the CFRP end was widely open. Since the tensile steel
reinforcement was not yet yielded, further application of load led to gradual transferring
of the tensile force in CFRP to the intact portion and the debond crack propagated into the
151
CHAPTER 5
shear span until the load reached a peak of 70.7 kN. The propagation sequences are
illustrated in the high-speed video captures shown in Figure 5.23. Beam E1b behaved in a
very similar fashion as beam E1a. The ultimate load of beam E1b was 74.5 kN.
80
70
Shear load (kN)
60
50
40 C1a
30 C1b
20 E1a
10 E1b
0
0 5 10 15 20 25 30
Midspan deflection (mm)
Right load
point
(E1b)
152
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Right load
point
(1) (2)
(3) (4)
(5) (6)
Figure 5.23 High-speed video captures for beam E1a at failure (end debond) (rate:
500 frames per second)
The debond crack surface was approximately at the mid-height of the rebars
(Figure 5.24). The debonded cover had several tooth-like transverse cracks, which were
spaced at the stirrup spacing. This was due to the interference of the stirrup extrusions to
the cracking surface (Figure 5.25). As the debond crack was propagating along the weak
plane at the tension reinforcement level, its path was intersected by the stirrup extrusions,
which induced a transverse crack that originated from the tip of the extrusions. This
153
CHAPTER 5
mechanism might have helped to increase the delamination strength and therefore the
load was still able to increase when the debond crack at the composite end had already
opened.
Rebar locations
Cracking surface
Stirrup extrusions
Induced cracks
The load deflection responses of beams E2a and E2b are plotted in Figure 5.26.
Their crack patterns after failure are shown in Figure 5.27. These beams had a much
shorter bond length. A debond crack near the cut-off point was observed very early. The
crack propagated more gradually since the load level was not high. In both beams, the
debond crack opened wide at the load level of around 49 kN. After that, the load still
increased since the tension steel reinforcement could take more loading. The cracks along
the tension reinforcement level finally caused total separation of the composite and the
concrete cover.
154
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
60
50
Right load
point
(E2a)
Right load
point
(E2b)
Beams E3a and E3b had less tension reinforcement. The load deflection curves
and crack patterns are shown in Figure 5.28 and Figure 5.29, respectively. The failure
occurred in a similar way to that of E1 beams.
155
CHAPTER 5
70
60
Right load
point
(E3a)
Right load
point
(E3b)
The concrete covers in beams E4a and E4b were deeper than the other beams.
Their load deflection responses and crack patterns are plotted in Figure 5.30 and Figure
5.31, respectively. The ultimate capacities were not consistent. The maximum loads
recorded for E4a and E4b were 79kN and 61kN, respectively. The reason for this
inconsistency is unknown.
156
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
80
70
Right load
point
(E4a)
Right load
point
(E4b)
Beams E5a and E5b were bonded with a very thick layer of CFRP. As shown in
Figure 5.32, the stiffness of the beams after cracking was much higher that that of the
control beams. Less cracking was seen in the beam body due to a high amount of tension
reinforcement (Figure 5.33). These two beams failed at slightly lower shear loads
compared to those of E1 beams.
157
CHAPTER 5
70
60
Right load
point
(E5a)
Right load
point
(E5b)
5.4.5.1.3 S beams
The load-deflection responses for S beams (the beams retrofitted with two plies of
fibres) are plotted in Figure 5.34. All of S beams failed by yielding of steel followed by
debonding of CFRP. Along the loading curves, three regions can be clearly identified
corresponding to when the concrete was not cracked, when the concrete had cracked and
when the steel reinforcement had yielded. After the concrete had cracked, the retrofitted
158
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
beams showed slightly stiffer behaviour compared to the control beams. After the
yielding of steel, the major increase in tension was transferred to the CFRP and the load
still increased further with increasing deflection.
90
C1a
80
70 C1b
Beams S1a and S1b failed by a combination of intermediate span debond and end
debond. The debond sequences in beam S1a are illustrated in Figure 5.35. The photos of
the cracking patterns after failure are shown in Figure 5.36. The two beams experienced
very similar cracking behaviour prior to failure. Visible fine flexural cracks were seen at
the load level of approximately 30 kN. These cracks propagated gradually as the load
increased. They were spaced more evenly and closer compared to those in the control
beams. At around 40 kN load level, diagonal flexure-shear cracks became visible. After
the yielding of the steel (at around 61 kN), these cracks appeared more clearly in the
shear spans. They widened as the load increased. Cracking noises could be heard before
failure. Concrete fractured simultaneously from the unclamped end of the CFRP and from
the tip of a wide flexure-shear crack under the right load point. Stress transferred and
cracking propagated toward the beam centre along the tension reinforcement level. The
failure was very brittle.
159
CHAPTER 5
Right load
point
(1) (2)
(3) (4)
(5) (6)
Figure 5.35 High-speed video captures of beam S1a at failure (combination of end
and intermediate span debond) (rate: 500 frames per second)
160
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Right load
point Initiation locations
(S1a)
Right load
point
Initiation locations
(S1b)
Figure 5.36 Photos of the debonded side of beams S1a and S1b
Beams S2a and S2b failed by intermediate span debond. The full loading curves
of these two beams are plotted in Figure 5.37. The failures were slightly different in the
two beams. In beam S2a, debonding occurred on the clamped side. A wide flexure-shear
crack was observed under the load point and delamination of the laminate from concrete
was initiated at its tip as the shear load reached 78 kN. This was followed by gradual
delamination of the composite from the concrete beam along the bond surface toward the
clamp location (crack 1 in Figure 5.39). The crack path was within a thin layer of
concrete near the bond surface (Figure 5.40). The load continued to increase to a peak of
80.4 kN until the concrete cover near the tip of the flexure-shear crack broke from the
beam leading to complete separation of the composite from the crack location to the steel
clamp. The load level stayed at around 70 kN as the CFRP was still held by the clamp. As
the prescribed displacement increased further, severe crushing of concrete occurred under
the load points. A crack along the tension reinforcement layer also formed (crack 2 in
Figure 5.39). As the load increased further, more cracks formed in the concrete cover
(cracks 3 in Figure 5.39). Eventually, the entire concrete cover on the right side was
sheared off and the load decreased to approximately 50 kN. In beam S2b, delamination
occurred on the unclamped side. The debonding sequences of this beam were captured
161
CHAPTER 5
with the high-speed camera and are illustrated in Figure 5.38. As the composite was
peeled from the beam towards the right side of a flexure-shear crack, the unbalanced
force in the CFRP transferred to the concrete cover on the left and led to the fracture of
the cover. The failure was brittle. The ultimate load of beam S2b was slightly less than
that of beam S2a.
Intermediate Total
100 span debond separation
of CFRP
80
Total load (kN)
60
40 S2a
S2b
20
0
0 20 40 60 80 100
Midspan deflection (mm)
162
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Right load
point
(1) (2)
(3) (4)
(5) (6)
Figure 5.38 High-speed video captures of beam S2b at failure (intermediate span
debond) (rate: 500 frames per second)
163
CHAPTER 5
Left load
point
(S2a) Debond sequence
1 2 3
Right load
point
(S2b)
Initiation location
Figure 5.39 Photos of the debonded side of beams S2a and S2b
Initiation location
Beams S3a and S3b had less tension reinforcement, which was indicated by their
lower stiffness as seen in Figure 5.34. Beam S3a failed by intermediate span debond,
which first appeared on the clamped side. Beam S3b failed due to a combination of
intermediate span and end debond on the unclamped side (Figure 5.41).
164
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Left load
Initiation point point
(S3a)
Right
load point
Initiation points
(S3b)
Figure 5.41 Photos of the debonded side of beams S3a and S3b
C1b
50 E2a (retested)
40 E2b
30 E4a
20 E4b
10 E5b
0
0 5 10 15 20 25 30
Midspan deflection (mm)
Figure 5.42 Loading curves for all beams tested until concrete crushed
5.4.5.1.5 Summary
A summary of the failure modes and load capacities of the beams is presented in
Table 5.4. The crack patterns of all beams are compared in Figure 5.43 to Figure 5.45.
165
CHAPTER 5
Except for the beams strengthened with a very short length of CFRP, the ultimate
capacities of retrofitted beams were increased compared to the control beams. For E
beams, which were strengthened with 6 or more layers of CFRP, the ultimate load
capacities were increased up to 38 %. For S beams, which were strengthened with 2
layers of FRP, the maximum increase in the capacity was 49 %.
* Load level where the strain level at 50mm from the CFRP end reached its peak.
166
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
It is visually clear that the flexural crack widths were reduced when the CFRP
composite were bonded to a beam. The thicker CFRP, the narrower the crack widths
were. The crack spacing in constant moment region was also more uniformly distributed
in retrofitted beams.
C1a
C1b
S1a S1b
S2a S2b
S3a S3b
Figure 5.44 Crack patterns in S beams
167
CHAPTER 5
E1a E1b
E2a E2b
E3a E3b
E4a E4b
E5a E5b
Figure 5.45 Crack patterns in E beams
168
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
on the stirrup indicate no yielding of steel. The strains at different stirrup locations along
the shear span also have comparably similar values.
The CFRP strain in all retrofitted beams is found to decrease from a maximum
value under the load point to a minimum value at the end of the composite (Figures 5.48,
5.53 and 5.58). The descending slopes in the shear span are close to being linear for
beams S1a and S2a. However, for beam E1a, the curves have a nonlinear distribution
with the slope reducing gradually from the CFRP end to around zero under the load point,
especially at higher loads. As expected, the maximum strain level in beams S1a and S2a
is much higher than that in beam E1a.
With the exception at the peak load, the average bond stress distributions of
beams S1a and S2a appear to have two peaks: one near the CFRP end and one under the
load point; whereas the distributions of beam E1a show only one clear peak near the FRP
end (Figures 5.49, 5.54 and 5.59). These are the indication of the mixed mode of end and
intermediate span debond in S beams and end debond in E beams. For beam S2a, since
the failure happened to be on the clamped side, where the end debond was prevented, this
beam failed by intermediate span debond.
The effects of concrete cracking, steel yielding and debonding are illustrated
clearly in the CFRP strain development plots (Figures 5.50, 5.55 and 5.60). Concrete
cracking and steel yielding influence the tension in the composite noticeably. Before
concrete cracking, the strain level in the composite is very small. In beams S1a and S2a,
after concrete cracking at approximately 10 kN, the CFRP strain gradually increases with
the load. After yielding of steel at around 65 kN, there is a steep increase in the CFRP
169
CHAPTER 5
strain indicating that the major increase in tension was taken by the composite. In beam
E1a, except at the ultimate load, the longitudinal steel did not yield and the CFRP strain
increased gradually until the ultimate load was reached. The CFRP strain development
also reveals early debonding near the CFRP end at a load level much lower than the
ultimate load. Early sign of debonding is indicated by the change of the curve slope
recorded by the gauge at 50 mm from the composite end. This change occurred at 47 kN
for both beam S1a and beam S2a, and at 41 kN for beam E1a. The composite is debonded
when the slope becomes negative. The load levels when that occurred for all beams are
listed in Table 5.4
The strain levels in the stirrup closer to the composite end also appear to be
slightly higher than those under the load point. However, this is not clearly demonstrated
considering the variation in the measurement and the fact that the strain depends largely
on the position of major cracks. Except for the stirrup near the CFRP end in beam S1a,
yielding of the steel was not seen.
170
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Microstrain
3000 25.0
30.0
2000 35.1
40.0
1000 45.1
50.0
0 53.6
1000 750 500 250 0
Distance from left support (mm)
1400 Location
1200 from left
support (mm)
1000
Microstrain
800 250 F
600 250 B
500 F
400
500 B
200
625 F
0 625 B
0 10 20 30 40 50
Shear load (kN)
171
CHAPTER 5
30.0
0.4 40.0
0.2 50.1
0 60.0
-0.21000 750 500 250 0 65.1
-0.4 70.7
Distance from end
of FRP (mm)
2500
175
2000
1500 300
1000 425
500 550
0 1000
0 20 40 60 80
Shear load (kN)
172
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Microstrain
2000 40.0
1500 50.1
55.1
1000
60.0
500 65.1
0 70.7
1000 750 500 250 0
Distance from end of FRP (mm)
Figure 5.51 Tensile steel strain distributions in beam E1a on unclamped side
800 100 B
600 350 F
400 350 B
200 475 F
0 475 B
0 20 40 60 80
Shear load (kN)
173
CHAPTER 5
Microstrain
6000 40.1
50.0
4000
60.0
2000 65.0
70.0
0 73.7
1000 750 500 250 0
Distance from end of FRP (mm)
1.5 30.0
40.1
1 50.0
0.5 60.0
65.0
0
70.0
-0.51000 750 500 250 0
73.7
Distance from end
of FRP (mm)
6000 175
4000 300
425
2000
550
0 1000
0 20 40 60 80
Shear load (kN)
174
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Microstrain
6000 40.1
50.0
4000
60.0
2000 65.0
70.0
0 73.7
1000 750 500 250 0
Distance from end of FRP (mm)
Figure 5.56 Tensile steel strain distributions in beam S1a on unclamped side
100 B
2500
2000 350 F
1500 350 B
1000
475 F
500
0 475 B
0 20 40 60 80
Shear load (kN)
175
CHAPTER 5
1 40.0
50.0
0.5 60.0
70.0
0
75.0
1000 750 500 250 0
-0.5 80.2
Distance from end
of FRP (mm)
6000 175
4000 300
425
2000
550
0 1000
0 20 40 60 80
Shear load (kN)
176
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Microstrain
30.0
6000 40.0
50.0
4000
60.0
2000 70.0
75.0
0 80.2
1000 750 500 250 0
Distance from end of FRP (mm)
85 B
1000 445 F
445B
500 535 F
535 B
0
0 20 40 60 80
Shear load (kN)
The maximum CFRP strain and maximum bond stress for all retrofitted beams are
listed in Table 5.5. Since the bond stress was calculated based on the strain reading at two
locations, the calculated result was not accurate once debonding had occurred at a strain
gauge location. For those cases, the maximum bond stress values were taken from the
previous load level. The bond stress values are only available for the beams instrumented
with enough strain gauges.
It can be seen that for S beams, the maximum CFRP strains varied around an
average value of 8615 microstrain with the variation being less than 12 % for all cases.
For E beams, it can be seen that end debond occurred at shear bond stress levels varying
around 1 MPa. However, the calculated bond stresses near the CFRP end were less
177
CHAPTER 5
accurate due to the fact that they depended on the CFRP strains near the CFRP end where
the readings were more susceptible to noise.
178
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
As shown in Figure 5.63a, the amount of the CFRP and tension steel
reinforcement clearly affects the beam behaviour. It is apparent that retrofitting RC beams
with a thicker layer of CFRP does not always lead to higher capacity. The average
ultimate load for the beams retrofitted with 2, 6 and 9 layers of FRP were 148.3 kN, 145.3
kN and 126.4 kN respectively. Figure 5.63b clearly shows that the effectiveness of the
CFRP is greatly reduced as the relative stiffness of CFRP to steel increases.
6 plies 2 plies
3 bars 3 bars
80 9 plies 1
70 3 bars
0.8
60
Shear load (kN)
Vdebond / Vmax
50 0.6
40
2 plies 0.4
30
2 bars
20 0.2
6 plies
10 2 bars
0 0
0 5 10 15 20 25 30 35 0 0.1 0.2 0.3 0.4 0.5 0.6
Midspan deflection (mm) A f Ef / A s Es
(a) (b)
Figure 5.63 Effect of CFRP and tension steel stiffness ratio
The bond length in the shear span, Lf, also has a significant influence on the
beam’s capacity. As illustrated in Figure 5.64, the beam capacity reduces by 39 % as the
bond length is shortened by 36 %.
80 1
70 Lf = 550 mm
0.8
60
Shear load (kN)
Vdebond / Vmax
50 0.6
40 Lf = 350 mm
30 0.4
20 0.2
10
0 0
0 5 10 15 20 25 30 0 0.25 0.5 0.75 1
Midspan deflection (mm) Lf / a
(a) (b)
Figure 5.64 Effect of CFRP bond length
179
CHAPTER 5
Concrete cover and stirrup spacing seem to have an insignificant effect on the beam
capacity.
These remarks will be investigated further in Chapters 7 and 9 by comparing with the
results from numerical and theoretical modelling.
End steel clamping was not able to prevent intermediate span debond. Two of S
beams, S2a and S3a, failed by intermediate span debond on the clamped side as described
previously. However, end steel clamping improved ductility of the beams significantly.
After intermediate span debond occurred, the CFRP was still tightened to concrete by the
clamp and the beam continued to carry a significant portion of the peak load. This was
clearly observed in beam S2a as described previously.
180
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Left load
(a) S1a: point
Clamped
side
Left load
point
(b) E1a:
Clamped
side
Figure 5.65 Crack patterns on the clamped side of E1a and S1a
From the above observation, it is clear that steel clamping is a simple and
effective anchorage method. However, steel poses corrosion problems. Another possible
set back of steel clamping is that it needs to connect to the top of the beam. It means that
the top flange of T-beams and the top slab need to be drilled through. This might not be
practical for some cases. A better anchorage method is using composite U-straps. This is
investigated in the next section.
181
CHAPTER 5
Out of these eight tests, one was carried out on the beam without anchorage. In the
other seven tests, the beams were anchored with non-prestressed and prestressed CFRP
U-straps placed either at the CFRP end only or at a spacing of 180 mm in the shear span.
In the non-prestressed anchorage system, two plies of CFRP fabrics of 50 mm width were
wrapped and bonded around the sides and the soffit of the concrete beam near the end of
the longitudinal CFRP (Figure 5.66). In the prestressed system, a gap was introduced
between the strap and concrete soffit. Prestressing was introduced into the sides of the
CFRP strap by inserting a wedge into the gap (Figure 5.67). More description of this new
prestressed system is followed.
The strap had at a slope of 1:20 to the vertical, which was also the wedge surface
slope. This slope was chosen so that a prestressing strain of 500×10-6 could be introduced
to the strap sides. The prestressing strain was monitored by two strain gauges placed on
the two sides of the strap over the un-bonded area close to the beam soffit. To avoid
slipping of the wedge over the longitudinal composite, grooves were introduced to the
wedge bond surface. The grooves were designed for one-directional slippage only. To
further prevent slippage, epoxy resin was also injected between the grooves the wedge.
182
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Slipping was however allowed between the wedge and the bottom side of the U strap.
This method of prestressing was designed specifically for CFRP fabrics bonded using wet
lay-up method since the fibres can not be stressed prior the formation of the composite
due to breaking of the individual strands. Another prestressing method is also available
for CFRP thermoplastic tape and has been reported elsewhere (Lees et al., 2002).
Since no stirrup was provided between the load points, additional steel clamps
were used there to improve the beam shear strength. The steel clamps should have an
insignificant effect on the failure in the shear span.
1 1
Cover1
Label ns x db db,sv – s nf x tf,01 L0 (mm) 1 U-strap
(mm)
E3b2 2 - 12 10 - 125 25 6 - 0.176 150 None
E3a2 2 - 12 10 - 125 25 6 - 0.176 150 1-N2
E1a2 3 - 12 10 - 125 25 6 - 0.176 150 1-P26
E5a2 3 - 12 10 - 125 25 9 - 0.176 150 3-P5
A1a 3 - 12 10 - 125 25 6 - 0.176 150 1-N2
A1b 3 - 12 10 - 125 25 6 - 0.176 150 3-N3
A2a 3 - 12 10 - 125 25 6 - 0.176 150 1-P4
A2b 3 - 12 10 - 125 25 6 - 0.176 150 3-P5
183
CHAPTER 5
Load point
Support Support
U-strap
Longitudinal
CFRP
Load point
Support Support
CFRP concrete
U-strap
steel reinforcement
duct tape
CFRP
PERSPEX grooves
Longitudinal wedge no bonding
CFRP
184
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
5.5.2 Materials
The material properties are described in Section 5.4.2.
For the prestressed straps, the installation involved surface preparation, bonding
of the U-strap and prestressing of the U-strap. To create a gap between the U-strap and
concrete soffit, a PERSPEX ‘forming’ wedge was attached to the soffit. The wedge was
wrapped with a plastic sheet to prevent it from bonding to the fibres. Two peaces of duct
tape were also adhered on the sides of the concrete beam within the concrete cover
portion in order to leave two un-bonded areas for monitoring prestressed strain level. The
fibre strips were applied on the concrete as for the non-prestressed system. After that, it
was left to cure. During that time, strain gauges were installed. After seven days of
curing, the ‘forming’ wedge was then removed and a prestressing wedge was inserted to
the gap. The prestressing wedge was forced into the gap using a plastic hammer until the
strain level at the control area reached the desire level (using the monitoring strain gauges
on the two legs). Epoxy was also injected to the contact surface between the wedge and
the longitudinal CFRP after the wedge had been positioned. Together with the grooves on
the wedge surface, a strong bond was formed between the wedge and the longitudinal
CFRP. Photos of this system are presented in Figure 5.69.
185
CHAPTER 5
The location of LVDTs and load cells are shown in Figure 5.70. The strain gauges
were placed on the steel and the longitudinal CFRP at similar locations to the beams
tested in four-point bending. A number of strain gauges were also installed on the U-
straps. The locations of those are shown in Figure 5.71. A digital camera was also used to
capture crack propagation. The beams were loaded at a rate of 1 mm per minute for most
of the time.
186
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
900 700
450 725 425
Load cell
LVDT LVDT
LVDT
Figure 5.70 Load and deflection measurement for beams tested in three-point
bending
30 30
80 80
80 80
30 30
187
CHAPTER 5
significantly compared with those of the beams without U-straps. Detailed descriptions
for each beam are presented in the following paragraphs.
120
E5a2
100 E1a2 E1a load capacity
E3a2
Shear force (kN)
80
E3a load capacity
E3b load capacity
60
E5a load capacity
40
E3b2
20
0
0 5 10 15 20 25 30 35
Deflection under load (mm)
A2b
120
A1b
100
A2a E1a load capacity
A1a
Shear force (kN)
80
60
40
20
0
0 5 10 15 20 25 30 35
Deflection under load (mm)
Figure 5.73 Load-deflection behaviour of beams A1a, A1b, A2a and A2b
Beam E3b2 was a retest of beam E3b in three-point bending. It had no U-strap. It
failed by end debond in a similar fashion as beam E3b (Figure 5.74). Its ultimate load
carrying capacity was however slightly less than that of beam E3b.
188
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Load point
Beams E3a2 and A1a were anchored with one end non-prestressed strap. The end
strap shifted the failure mode to intermediate span debond, which occurred at a higher
load (Figure 5.75). Similar flexural and flexure-shear cracks were observed as in the
beams without anchorage (beams E3a and E1a). The end debond crack was also seen but
its propagation was restricted significantly by the U-strap. As the load increased beyond
the capacity of the beams without anchorage, more inclined cracks formed in the concrete
cover and intermediate span debond was initiated from the tip of one of those cracks. As
this occurred, the load dropped. However, the strap held the longitudinal composite and
the beam was still able to hold a significant portion of the peak load. As the tension in the
CFRP increased further, sliding of the longitudinal plate underneath the U-strap was
observed visually. The sliding action caused bending of the strap legs near the beam soffit
and led to local debonding and rupture of one of the legs before the longitudinal CFRP
was complete separated from the beam (Figure 5.76).
189
CHAPTER 5
Load point
(E3a2)
Load point
(A1a)
Figure 5.75 Photos of the debonded side of beams anchored with a non-prestressed
U strap
rupture
end debond
crack
debonding
sliding
Figure 5.76 Sliding of the longitudinal CFRP and local debonding and rupture of a
strap leg in beam E3a
Beams E1a and E2a were anchored with one end prestressed strap and they failed
by intermediate span debond in a similar way as in beams E3a2 and A1a (Figure 5.77).
However, no evidence of end debond crack or sliding of the longitudinal composite near
190
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
its end was found. This was due to the higher concrete shear capacity resulted from the
compressive stress field caused by the prestressing force. A leg of the strap was
eventually delaminated from concrete surface (Figure 5.78).
Load point
(E1b2)
Load point
(A2a)
Figure 5.77 Photos of the debonded side of beams anchored with a prestressed U
strap
total debonding
No sign of end
debond crack
191
CHAPTER 5
The beams with three U-straps (E5a2, A1b and A2b) failed by debonding of the
straps followed by intermediate span debond (Figure 5.79 and Figure 5.80). The straps in
the shear span crossed the flexure-shear cracks and limited their opening. They also
prevented the formation of inclined cracks in the concrete cover and increased the bond
strength between the concrete and longitudinal CFRP. As a result, both end debond and
intermediate span debond were limited. Debonding of the vertical strap nearest to the load
point occurred first. It was followed by intermediate span debond occurring near the
debonded strap location. As the tension transferred to other two straps, they were
subsequently subjected to either local debonding and rupture (in the beam bonded with
non-prestressed straps) or total peeling off (in the beams bonded with prestressed straps).
Load point
(A1b)
Figure 5.79 Photos of beams anchored with three non-prestressed U straps after
failure
192
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Load point
(E5a2)
Load point
(A2b)
Figure 5.80 Photos of beams anchored with three prestressed U straps after failure
193
CHAPTER 5
E3a2 E1b2
A1a A2a
Figure 5.82 Crack patterns in beams with one U-strap
194
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
E5a2
A1b A2b
Figure 5.83 Crack patterns in beams with three U-straps
The CFRP strain distributions (Figures 5.84, 5.87, 5.92 and 5.96) show high strain
values extending deep into the shear span. This was due to the fact that more severe
cracking was present in the cover, which led to more stress redistribution in the
composite. The CFRP strain development curves also indicate three clear regions
corresponding to when the concrete was uncracked, when the concrete was cracked and
when the steel yielded. Once the CFRP was debonded, its strain decreased rapidly as
expected.
The tensile steel strain development in the beams with anchorage (Figures 5.91,
5.95 and 5.100) shows clear yielding of steel. After debonding of the CFRP, the steel
rebar strain increased promptly as all tension transferred to the rebar.
The strain development in the strap legs (Figures 5.86, 5.90, 5.94 and 5.99) shows
that the straps only became effective after the load reached approximately 60 kN. The U-
strap strain increased significantly after the peak as the longitudinal CFRP was separated
from the beam. As it occurred, the tension was resisted by the end friction, which was
mainly due to the interlocking mechanism as a result of the confinement from the U-strap.
195
CHAPTER 5
The crack frictional faces dilated and therefore imposed more strain in the strap legs until
the legs ruptured or debonded.
The strain distributions along the strap height (Figures 5.89 and 5.98) near the
ultimate load levels were affected by the intersected cracks. In beam A1b, the crack
located at 100 mm height along the leftmost strap; whereas in beam A2b, the crack was at
approximately 200 mm height.
196
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Load point
5000 Load (kN)
20.0
4000
40.0
Microstrain
3000 50.0
60.1
2000
70.0
1000 80.0
0 90.0
500 400 300 200 100 0 94.5
5000
3000 137.5
2000 275
412.5
1000
550
0
0 20 40 60 80 100
Shear load (kN)
3500
3000 Gauge
2500 location
Microstrain
2000 Front
1500 Back
1000
500
0
-500 0 20 40 60 80 100
Shear load (kN)
197
CHAPTER 5
Microstrain
4000 60.1
3000 70.1
2000 80.0
90.0
1000
100.0
0 108.6
500 400 300 200 100 0
Distance from end of FRP (mm)
6000
5000 Location
from CFRP
4000
Microstrain
end (mm)
3000 137.5
2000 412.5
1000 550
0
0 20 40 60 80 100
Shear load (kN)
50 50 50
0 0 0
-500 500 1500 2500 -500 500 1500 2500 3500 -500 500 1500 2500
Microstrain Microstrain Microstrain
Load (kN)
Front 20.0 40.0 59.9 80.0 90.0 100.0 108.7
Back 20.0 40.0 59.9 80.0 90.0 100.0 108.7
Figure 5.89 Strain distributions along U strap height in beam A1b at different load
levels
198
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Front Back
5000
4000
Microstrain
3000
2000
1000
0
0 20 40 60 80 100 120
Shear load (kN)
199
CHAPTER 5
4000 40.0
Microstrain 3000
60.0
70.1
2000
80.0
1000 89.9
0 101.7
500 400 300 200 100 0
Distance from end of FRP (mm)
5000
3000 137.5
2000 275
412.5
1000
550
0
0 20 40 60 80 100
Shear load (kN)
5000
Gauge
4000
location
Microstrain
3000 Front
Back
2000
1000
0
0 20 40 60 80 100
Shear load (kN)
200
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
10000
8000
Microstrain
6000
4000
2000
0
0 20 40 60 80 100 120
Shear load (kN)
201
CHAPTER 5
Microstrain
4000 60.0
3000 80.0
90.1
2000
100.0
1000
109.9
0 114.5
500 400 300 200 100 0
Distance from end of FRP (mm)
6000
5000 Location
from CFRP
Microstrain
50 50 50
0 0 0
-500 500 1500 2500 -500 500 1500 2500 -500 500 1500 2500 3500
Microstrain Microstrain Microstrain
Load (kN)
Front 20.1 40.0 60.0 80.0 90.0 100.0 109.9 114.5
Back 20.1 40.0 60.0 80.0 90.0 100.0 109.9 114.5
Figure 5.98 Strain distributions along U strap height in beam A2b at different load
levels
202
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
Front Back
10000
8000
Microstrain
6000
4000
2000
0
0 20 40 60 80 100 120
Shear load (kN)
Table 5.8 summarises the maximum strains recorded for the U-strap legs and the
failure types. It can be seen from the table that the non-prestressed systems failed by
CFRP rupture for most cases. The average maximum strain was 3270 microstrain, which
was much lower than the tensile rupture strain. This was due to bending of the strips and
therefore non-uniform stress distributions. The prestressed systems failed by CFRP
debonding for all cases with the average debonding strain in the straps of 4540
microstrain. This strain is approximately 90 % of the average debonding strain measured
in the shear-lap tests with the same composite width.
203
CHAPTER 5
Notes: 1) Local failure: the strap was not ruptured or completely debonded but some localised debonding
and minor damage were present.
2) Not measured.
• Two failure modes were observed: intermediate span and end debond. They were
the result of the high shear stress level in concrete. The former mode appeared to occur at
a fixed CFRP strain level of approximately 8600 microstrain. The later appeared to occur
when the shear bond stress reached approximately 1 MPa.
204
EXPERIMENTAL INVESTIGATION OF RETROFITTED BEAMS UNDER BENDING
• U-straps placed over longitudinal composite ends prevented end debond. In these
cases, the failure mode shifted to intermediate span debond.
• U-straps placed in shear span limited intermediate span debond.
• Prestressed U-strap anchorage system only improved the beam stiffness and ultimate
capacity slightly compared with the system with non-prestressed straps.
• Non-prestressed systems failed by rupture of the side strips at a strain of
approximately 3200 microstrain.
• Prestressed systems failed by debonding of side strips at a strain of approximately
4500 microstrain.
205
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
Two methods of finite element modelling are used in the study. The first method
is based on smeared cracking, whereas the second method uses a combination of smeared
and discrete cracking.
In this chapter, the smeared crack modelling of the shear-lap specimens under
testing is presented. The simulation was implemented using a commercially available
non-linear finite element code, DIsplacement method ANAlyser or DIANA (version
8.1.2), developed by TNO Building and Construction Research in the Netherlands (de
Witte and Kikstra, 2003). The chapter reports the finite element idealisation, material
models, solution procedure and verification with the shear-lap test results. A study of the
model sensitivity to mesh size, mesh type, concrete tensile properties and adhesive
stiffness is also included. In Chapter 7, modelling of the beam tests using smeared cracks
is presented, whereas modelling of the shear-lap and beam tests using discrete/smeared
cracks is reported in Chapter 8.
207
CHAPTER 6
A different approach was used in the finite element model presented in this
chapter. Instead of lumping the bond failure into an interface slip, the failure was
simulated as cracking in concrete near the bond surface using a fine mesh near the bond
line. This approach was successfully used to model the bond between steel reinforcement
and concrete and has the potential of explaining the fundamentals of the bond (Rots,
1989).
A typical finite element mesh and boundary conditions of the model are shown in
Figure 6.1. Since a fine mesh was required near the bond line, the concrete block was
modelled using three layers: a fine layer with an element size of approximately 2 mm, a
transition layer with transition elements and a coarse layer with an element size of
approximately 10 mm. The CFRP was connected to the top concrete surface via an
adhesive layer. The adhesive layer was assumed to be 1 mm thick and the CFRP
composite formed by one ply of fibres was taken to be 0.7 mm thick (Figure 4.8b).
Horizontal and vertical roller supports were placed appropriately to simulate the actual
physical restraints. A prescribed displacement was applied on the CFRP end.
Lf Horizontal
Adhesive CFRP
roller
H Fine layer
Y
Transition
layer
X
Coarse
Vertical Horizontal roller layer
roller
The concrete in the fine and transition layers was modelled using six-node
quadrilateral isoparametric plane stress elements (CQ12M) (Figure 6.2a). Each element
has twelve degrees of freedom (dof) with two displacements, ux and uy, at each node. A
three-point Gaussian integration scheme was used. The CFRP, the adhesive and the
208
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
5
6 5
7
6 η 4
η
8 4
ξ
1 3 ξ
2
1 3
2
(a) (b)
Figure 6.2 CQ12M element (a) and CQ16M element (b)
1) The stress is evaluated as a function of the strain in the directions of the cracks
and this function follows an explicit curve as in the following formula:
t + ∆t
i +1 σ nst = σ( t +i +∆1t ε nst ) (6.1)
where nst is the crack coordinate system. The strain and stress in the element
coordinate system are related to the strain and stress in the crack direction by a
strain transformation matrix T, as follows
t + ∆t
t + ∆t
i +1 ε nst =T t +i ∆+1t ε xyz ; i +1 σ nst = T T t +i +∆1t σ nst (6.2)
209
CHAPTER 6
variables are for tension and the last three are for compression). Since no damage
recovery is assumed, to ensure that these internal damaged variables are not
decreasing in case of unloading, the unloading constraints rk are used. They are
active (rk = 1) if unloading occurs and inactive (rk = 0) otherwise (Figure 6.3). The
internal variables are updated as
t + ∆t
i +1 α = t α + (1 − rk )∆ε (6.4)
3) The stress is related to the strain with the following equation:
σ = f (α, ε).g(α, ε nst ) (6.5)
where the reduction due to the loading-unloading effect is taken into account by
the following function:
αk − ε j
g =1− (6.6)
αk
σj
αj+3
εj
αj
rj = 1 rj = 0
rj+3 = 1 rj+3 = 0
Figure 6.3 Loading-unloading for total strain crack models (de Witte and Kikstra,
2003)
4) The tensile behaviour of concrete is described with a softening curve. For the
fixed crack model, the shear behaviour after cracking can also be reduced by a
shear retention factor, β.
5) The stress strain curve for the compressive behaviour of concrete depends on the
effect of lateral confinement (increasing strength and ductility) or lateral cracking
(reducing strength and ductility), which can be modelled using Thorenfeldt’s
210
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
curve (Thorenfeldt et al., 1987) and Vecchio and Collin’s reduction factors
(Vecchio and Collin, 1986), respectively.
Two total strain crack models are available in DIANA: a fixed crack model and a
rotating crack model. In the former, the stress-strain relationship is evaluated in a
coordinate system which is fixed upon cracking. For the latter, the coaxial stress-strain
concept is used, in which the stress-strain relationship is evaluated in the principal
directions of the strain vector.
To avoid stress-locking problems associated with the fixed crack model when
changes in crack orientation occur during loading (Rots, 1989), the rotating crack model
was used in this study to model the concrete block.
The input for the model comprises the elastic and nonlinear properties in tension,
shear and compression. The elastic properties consist of the elastic modulus and Poisson’s
ratio. In this study, the elastic modulus was determined using the expression given by
ACI Committee 363 (1992), as follows
This expression can be applied to concrete of all strengths (Warner, 1998). The Poisson’s
ratio, νc, was taken to be 0.2.
211
CHAPTER 6
The behaviour of the concrete in compression was modelled using the function
proposed by Thorenfeldt et al. (1987), of which the general shape is illustrated in Figure
6.4b. The function described both the hardening and softening behaviour of concrete.
σ σ
fct ε
Gf/h
fc
ε
(a) (b)
Figure 6.4 Concrete stress-strain curve in tension (a) and compression (b)
212
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
δu iT (f int,i +1 + f int,i )
Energy Norm Ratio = (6.8)
∆u T0 (f int,1 + f int,0 )
where δui is the ith iterative displacement increment, ∆u0 is the displacement increment
and fint is the internal force. A new reference norm was determined at the start of each
step. In all of the models, the tolerance for convergence was set to 0.0005 to ensure
accurate and good results. The maximum number of iterations for each load step or
increment was set to 100.
Figure 6.5 Modified Newton-Raphson iteration (de Witte and Kikstra, 2003)
213
CHAPTER 6
To improve the convergence, the line search algorithm was also used in the study.
The algorithm scaled the prediction of the iterative displacement increment, δu, by a
value to minimize the energy potential.
Model T5 predicted an interfacial debond failure. As the load was applied, the
tensile force transferred to the concrete near the loaded end. The first clear transverse
shear-tension crack formed normal to the bond at approximately 10 mm from the loaded
end and propagated at an inclined angle into the concrete block. As the load increased, the
second transverse crack formed at 20 mm from the loaded end. When this crack opened,
the stress was released around the first crack and the crack opening was reduced there. As
the load increased further, a debond crack was initiated at the tip of the second normal
crack and propagated parallel to the bond surface. The shear stress transferred along and
subsequent cracking occurred. The transverse cracks formed at an equal interval of
approximately 10 mm with the two dominant cracks located at 20 and 60 mm from the
loaded edge.
214
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
24 kN
27 kN
Transverse cracks Debond cracks
28 kN (peak)
Figure 6.6 Deformed shapes of model T5 at different load levels (magnification
factor = 60)
215
CHAPTER 6
10 kN
15 kN
Shear-tension crack
18 kN (peak)
Figure 6.7 Deformed shapes of model T1 at different load levels (magnification
factor = 60)
The crack patterns at the peak load were illustrated in Figure 6.8 to Figure 6.11 for
all specimens. For clarity, only the cracks, of which the crack strain is large enough so
that the crack tensile stress has reduced by 50 % from the maximum tensile strength, are
shown. The predicted peak loads and failure modes are compared in Table 6.2
It can also be seen that for some specimens the number of cracks, the pulled-out
concrete prism sizes and the failure modes predicted by the FEA were different to those
observed in the experiments. A possible reason for the discrepancy was that in the
experiment, stress transferring occurred in a more brittle manner. The propagation, which
led to total debonding of the CFRP composite, might occur from the tip of a transverse
crack closer to the loaded edge as a result of the dynamic effect from a sudden release of
energy once the crack opened. It was not captured with the static FE model.
216
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
(T1) (T2)
(T3) (T4)
(T5) (T6)
Figure 6.8 Crack patterns in models T1 to T6
(T7) (T8)
Figure 6.9 Crack patterns in models T7 and T8
(T9) (T10)
(T11)
Figure 6.10 Crack patterns in models T9, T10 and T11
(T12) (T13)
Figure 6.11 Crack patterns in models T12 and T13
217
CHAPTER 6
Table 6.2 Comparison between the experimental and the FEA results
Experiment Finite element analysis
Spec- Var- Peak Vari- Peak Devia-
imen iable load ation1 Failure mode load tion2 Failure mode
(kN) (%) (kN) (%)
T1a 20.0 Shear-tension 18.3 8.3 Shear-tension
6.2
T1b 18.8 Interfacial debond 18.3 2.5 Shear-tension
T2a 25.8 Interfacial debond 22.1 14.3 Shear-tension
2.4
T2b 25.2 Interfacial debond 22.1 12.2 Shear-tension
Shear-tension/
T3a 25.8 Interfacial debond 26.9 4.1
Interfacial debond
5.6
Shear-tension/
T3b Lf 27.3 Interfacial debond 26.9 1.6
Interfacial debond
T4a 26.7 Interfacial debond 28.7 14.0 Interfacial debond
3.0
T4b 25.9 Interfacial debond 28.7 17.5 Interfacial debond
T5a 27.8 Interfacial debond 29.4 5.9 Interfacial debond
13.1
T5b 31.7 Interfacial debond 29.4 7.1 Interfacial debond
T6a 31.7 Interfacial debond 29.4 7.3 Interfacial debond
10.3
T6b 28.6 Interfacial debond 29.4 2.8 Interfacial debond
T7a 33.0 Interfacial debond 26.7 19.2 Interfacial debond
20.4
T7b 26.9 Interfacial debond 26.7 0.9 Interfacial debond
H
T8a 28.5 Interfacial debond 27.3 4.1 Shear-tension
4.5
T8b 29.8 Interfacial debond 27.3 8.3 Shear-tension
T9a 28.4 Shear-tension 27.3 4.0 Shear-tension
4.8
T9b 29.8 Shear-tension 27.3 8.5 Shear-tension
T10a 37.4 Interfacial debond 36.7 1.9 Shear-tension
11.6
T10b 33.3 Shear-tension 36.7 10.2 Shear-tension
tf
Shear-tension/
T11a 42.8 Interfacial debond 46.0 7.5
Interfacial debond
9.3
Shear-tension/
T11b 39.0 Shear-tension 46.0 18.0
Interfacial debond
T12a 18.8 Interfacial debond 20.5 9.2 Interfacial debond
2.6
T12b 19.3 Interfacial debond 20.5 6.4 Interfacial debond
bf
T13a 21.2 Interfacial debond 24.0 13.1 Interfacial debond
13.2
T13b 24.2 Interfacial debond 24.0 0.9 Interfacial debond
Notes: 1) Variation = |Fa – Fb| / (Fa/2+Fb/2), where Fa and Fb are the peak loads for specimens ‘a’ and ‘b’,
respectively.
2) Deviation = |Fexp-FFE| / Fexp, where Fexp and FFE are the experimental and predicted load
capacities of the bond.
218
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
50 Varying H tf bf
Lf
45 parameter
40
Maximum load (kN
35
30
25
20
15
10 Experiment FEA
5
0
T10a
T11a
T12a
T13a
T10b
T11b
T12b
T13b
T1a
T2a
T3a
T4a
T5a
T6a
T7a
T8a
T9a
T1b
T2b
T3b
T4b
T5b
T6b
T7b
T8b
T9b
(a)
50
40
Predicted max. load (kN
30
20
10
0
0 10 20 30 40 50
Measured max. load (kN)
(b)
Figure 6.12 Correlation of peak loads predicted by the FE model and measured in
the experiment
The behavioural trends of the specimens with different CFRP bond lengths,
support block clearances, CFRP thicknesses and CFRP widths as observed in the
experiments have been previously presented in section 4.4.3. The trends are depicted
219
CHAPTER 6
again in Figure 6.13 to Figure 6.16 where the predictions by the FEA are also included. It
is clear that the trends are captured well by the FE model. For the specimens bonded with
2 plies of CFRP, the effective bond length is found to be 100 mm. The effective bond
stress decreases almost linearly with the CFRP width. It is also found that the support
clearance has an insignificant effect on the bond capacity.
30 2.0
20 1.5
1.0
10 0.5
0 0.0
0 50 100 150 200 250 0 50 100 150 200 250
Lf (mm) Lf (mm)
Figure 6.13 Effect of varying CFRP bond length (Experiment and FE)
Experiment FEA
35
30
25
Load (kN)
20
15
10
5
0
0 25 50 75 100 125
H (mm)
Figure 6.14 Effect of varying support block height H (Experiment and FE)
220
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
40
620
35
30 420
25 220
20
20
0 0.5 1 1.5
0 0.5 1 1.5
tf (mm) tf (mm)
20 3
15
2
10
5 1
0 0
20 40 60 80 100 120 20 40 60 80 100 120
b f (mm) b f (mm)
In general, the numerical models simulated the strain distributions well. There are
exceptions for specimens T9, T10 and T11, for which, as discussed in Chapter 4, the
experimental strain readings were not reliable. Comparable trends can be observed in the
distribution plots for most specimens. At low load levels, the predicted strain distributions
221
CHAPTER 6
exhibit an exponential trend with the highest strain near the loaded end. As the load
increases, the strain distribution curves displace further along the joint similar to the
observation in the experiments. This phenomenon is most clearly seen in the specimens
with a long bond length, for example specimens T5a and T6a. When the failure mode is
shear-tension, the shifting of the high stress region is not observed clearly because once
the end crack opens, the CFRP strain level drops and the peak load is reached. This is
typically observed in the specimens with a very short bond length or with a very thick
CFRP layer, for example specimens T1a, T9a and T10a.
The strain readings were utilised to plot load-slip curves for all specimens in the
same way as for the experiments. They are compared in Figure 6.22 and Figure 6.24. The
load-slip plots from the experiments and the numerical models are comparable. For most
cases, the FE results show slightly more brittle behaviour. As expected, the initial
stiffness of the specimens with different bond lengths is similar but higher peak loads and
more ductile behaviour are observed in those with longer bond lengths.
222
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
Microstrain
2500 18.0 18.0
T1a 2000
1500
1000
500
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 6.17 Comparison of CFRP strain distributions in specimens T1a, T2a and
T3a
223
CHAPTER 6
T4a 2000
1000
0
0 50 100 150 200
Distance from loaded edge (mm)
5000
10.0 9.9
4000 15.0 15.3
20.0 20.2
Microstrain
1000
0
0 50 100 150 200
Distance from loaded edge (mm)
24.0 24.1
3000 26.0 26.1
28.0 28.0
2000 26.9 26.9
T6a
1000
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 6.18 Comparison of CFRP strain distributions in specimens T4a, T5a and
T6a
224
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
Microstrain
2000 25.0 25.0
1500
T7a
1000
500
0
0 50 100 150 200
Distance from loaded edge (mm)
25.0 25.0
2000 26.0 27.1
1500
T8a
1000
500
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 6.19 Comparison of CFRP strain distributions in specimens T7a and T8a
225
CHAPTER 6
-1000
-1500
Distance from loaded edge (mm)
30.0 29.9
40.0 40.0
1000
42.8 46.0
T11a
500
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 6.20 Comparison of CFRP strain distributions in specimens T9a, T10a and
T11a
226
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
Microstrain
3000 18.0 18.3
T12a 2000
1000
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 6.21 Comparison of CFRP strain distributions in specimens T12a and T13a
227
CHAPTER 6
25 30
20 25
20
15 T1a (Exp.) T2a (Exp.)
15
10 10
T1 (FE) T2 (FE)
5 5
0 0
0.00 0.01 0.02 0.03 0.00 0.02 0.04 0.06 0.08
Slip at 42 mm from loaded edge Slip at 42 mm from loaded edge
T1a T2a
30 40
25
30
20
15 T3a (Exp.) 20
T4a (Exp.)
10 T3 (FE) 10 T4 (FE)
5
0 0
0.00 0.05 0.10 0.15 0.20 0.00 0.10 0.20 0.30
Slip at 42 mm from loaded edge Slip at 42 mm from loaded edge
T3a T4a
40 40
Total load (kN)
30 30
20 T5a (Exp.) 20
T6a (Exp.)
10 T5(FE) 10
T6 (FE)
0 0
0.00 0.20 0.40 0.60 0.00 0.10 0.20 0.30 0.40
Slip at 42 mm from loaded edge Slip at 42 mm from loaded edge
T5a T6a
228
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
30 40
25
T9a T10a
50
40
Total load (kN)
30
20 T11a
10 T11 (FE)
0
0.00 0.05 0.10 0.15
Slip at 42 mm from loaded edge
T11a
Figure 6.23 Comparison of load-slip curves for specimens T9a, T10a and T11a
40 30
Total load (kN)
25
30
20
20 T7a (Exp.) 15 T8a (Exp.)
10
10 T7 (FE) T8 (FE)
5
0 0
0.00 0.05 0.10 0.15 0.20 0.00 0.05 0.10 0.15 0.20
Slip at 42 mm from loaded edge Slip at 42 mm from loaded edge
T7a T8a
25 30
Total load (kN)
Total load (kN)
20 25
20
15
T12a (Exp.) 15
10 T13a (Exp.)
10
5 T12 (FE) 5 T13 (FE)
0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20
Slip at 42 mm from loaded edge Slip at 42 mm from loaded edge
T12a T13a
Figure 6.24 Comparison of load-slip curves for specimens T7a, T8a, T12a and T13a
229
CHAPTER 6
The local bond-slip curves are compared in Figure 6.25 and Figure 6.26 for two
specimens with a 100 mm wide CFRP composite bonded over a relatively long distance
(180 and 220 mm). The bond-slip curves for two specimens bonded with narrower
composites (50 and 70 mm wide), are also plotted in Figure 6.27 and Figure 6.28. The
plotted stresses and slips are the average over the distance between strain gauges and
calculated in the same way as in the experiments. In the plots, the locations are measured
from the loaded edge.
The bond-slip curves vary significantly from one location to the other along the
CFRP length. This was also observed in the experiments. The main reason for the
variation is due to the influence from the transverse cracks. The bond-slip curves at
locations further from the loaded edge appear to have stiffer behaviour because they are
less influenced by transverse cracks formed near the edge. It is not possible to identify a
curve which would describe the bond behaviour accurately for all specimens at any
position along the joint. However, generally, the bond-slip curves have a non-linear
ascending and descending portion. The maximum bond stress is reached at a slip of
approximately 0.05 mm. The values for the maximum bond stress are approximately 4
MPa for specimens with CFRP width of 100 mm, 5 MPa for specimens with CFRP width
of 70 mm and 6 MPa for specimens with CFRP width of 50 mm. The maximum slip is
approximately 0.2 mm. These values are close to those observed in the experiments.
5 60 78 5
96 96
4 4
3 3
2 2
1 1
0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25
Slip (mm) Slip (mm)
230
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
2 2
1 1
0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25
Slip (mm) Slip (mm)
24 42 8
10 24 42
7
Bond stress (MPa)
8 60 78 6
60 78
5
6
4
4 3
2
2
1
0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25
Slip (mm) Slip (mm)
231
CHAPTER 6
(a) (b)
Figure 6.28 Comparison of local bond-slip relationships in specimen T13a (bf = 70
mm, Lf = 180 mm)
The analysis is an approximation of the actual situation and there are limitations
inherent in the model. Firstly, the composite itself consists of a number of different
layers. Its prosperities could depend on the thickness and the properties of the layers, and
the bond between the layers. However, in the analysis, the composite was assumed to be
homogenous and have uniform properties across its cross section. Secondly, the spew
fillets at the composite ends formed from excess adhesive were not avoidable during
preparation of the specimens. These fillets could affect the local stress distributions near
the CFRP end. Nevertheless, the spew fillets were not considered in the analysis. Thirdly,
there are stress singularities at two-material wedges in the problem. The stress will never
232
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
converge at the singular points on the bond lines near the composite ends and therefore
the stress magnitude near the composite ends might be sensitive to the mesh size. In this
analysis, the mesh size of 2 mm was used. It will be proved later (in section 6.7.1) that
this size is adequate to estimate the stresses near the composite ends.
It is also worth noting that, in this analysis, the stress and strain outputs were
taken not at the nodes but rather at the Gaussian integration points in the element (Figure
6.29). As a result, the finite element outputs were only the indication of the stress/strain
levels in the upper and lower halves of the layer.
Points representing
the top half
CFRP elements
Points representing
the bottom half
Adhesive elements
Figure 6.29 Gaussian integration points in the CFRP and adhesive elements
Figure 6.30 shows the CFRP strain distributions in the top and bottom halves of
the composite at three load levels for specimen T5. This specimen was bonded with two
plies of CFRP over a length of 180 mm. It can be seen clearly that at the loaded end, the
CFRP composite was bended and the strain in the top half is much lower than that in the
bottom half. The difference diminishes further away from the loaded edge. At or below
service load levels, the difference is insignificant beyond a distance of 10 mm from the
loaded edge. At the unloaded end, similar bending action also existed but the difference is
less noticeable than that at the loaded edge. It diminishes beyond a distance of around
7 mm from the end. Near the ultimate load, the influence of the major normal cracks on
the CFRP strain distribution can also be seen. Over the tip of these cracks, the strain on
the CFRP top reduces from a high crest to a low trough.
233
CHAPTER 6
CFRP microstrain
4000 20.2(Top)
20.2(Bottom)
(a) The whole 3000 29.4(Top)
29.4(Bottom)
CFRP strain 2000
1000
distribution
0
-1000 0 50 100 150 200
4000 20.2(Top)
20.2(Bottom)
(b) CFRP strain 3000 29.4(Top)
29.4(Bottom)
distributions near 2000
1000
the loaded end
0
-1000 0 10 20 30 40
Load (kN)
200
10.4(Top)
10.4(Bottom)
150
CFRP microstrain
20.2(Top)
20.2(Bottom)
(b) CFRP strain 100 29.4(Top)
29.4(Bottom)
distributions near 50
the un-loaded end
0
160 165 170 175 180
-50
Distance from loaded end (mm)
The CFRP strain distributions in the specimen T11 bonded with six plies of CFRP
over the same bond length of 180 mm are shown in Figure 6.31. The distributions are
similar to those of model T5. However, the lengths, over which significant differences
between the top and bottom layers exit, are longer compared to those in model T5.
234
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
CFRP microstrain
2500 23.4(Bottom layer)
2000 46.0(Top layer)
(a) The whole 46.0(Bottom layer)
1500
CFRP strain 1000
distribution 500
0
-500 0 50 100 150 200
Distance from loaded end (mm)
23.4(Bottom layer)
2000 46.0(Top layer)
(b) CFRP strain 46.0(Bottom layer)
1500
distributions near 1000
the loaded end 500
0
-500 0 10 20 30 40
Distance from loaded end (mm)
23.4(Bottom layer)
46.0(Top layer)
(b) CFRP strain 100 46.0(Bottom layer)
distributions near 50
the un-loaded end
0
160 165 170 175 180
-50
Distance from loaded end (mm)
235
CHAPTER 6
6.7.1.2 Comparison
The predictions of the top and bottom halves of the CFRP composite near the
loaded edge are illustrated in Figure 6.32. Due to the singularity near the edge, the
smaller the mesh size, the more accurate prediction can be obtained. It can be seen that
the differences between the strain distributions near the loaded edge are not significant
once the mesh size is equal to or lower than 2 mm. The predictions of the CFRP strain are
however very close beyond a distance of 10 mm from the loaded edge.
236
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
CFRP microstrain
CFRP microstrain
Bottom - Mesh2
100 Top - Mesh1 1500 Top - Mesh1
Bottom - Mesh1
80 Bottom - Mesh1
1000
60
40 500
20
0
0
0 5 10 15 20
-20 0 5 10 15 20 -500
Distance from loaded end (mm) Distance from loaded end (mm)
6000
5000
Top - Mesh5
CFRP microstrain
(c) At ultimate
Figure 6.32 Prediction of CFRP strains near the loaded end at different load levels
using different mesh sizes
The crack patterns after peak in the models of T5 and T1 are compared in Figure
6.33. All models of T5 indicate interfacial debond failure mode. While the models Mesh2
and Mesh1 show similar crack patterns, model Mesh5 predicts debonding initiated from
an inclined crack closer to the loaded edge. The crack patterns are very similar for the
models of T1.
The load-slip curves are compared in Figure 6.34. There is little difference
between the curves.
237
CHAPTER 6
Mesh5
Mesh2
Mesh1
Figure 6.33 Comparison of crack patterns after peak in specimens T5 (left) and T1
(right) predicted by the models with different mesh sizes
30 20
Total load (kN)
25 15
20
15 10
10 5
5
0
0
0.00 0.01 0.02 0.03
0.00 0.10 0.20 0.30 0.40 0.50
Slip at 42 mm from loaded edge
Slip at 42 mm from loaded edge
T5 T1
Figure 6.34 Comparison of load-slip curves predicted by the models with different
mesh sizes
238
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
6.7.2.2 Comparison
The crack patterns after peak in the models of T5 and T1 are compared in Figure
6.36. All models of T5 show two major inclined cracks at around 20 and 60 mm from the
loaded edge. There is a slight difference in the inclined crack orientation between model
D6 and the two models M6 and M8. The interfacial cracks appear to propagate more
easily in the last two models due to their more regular mesh orientation. In the models of
T1, the crack patterns are very similar, showing a shear-tension failure.
239
CHAPTER 6
The load-slip curves are compared in Figure 6.37. The models with a triangular
mesh predict slightly higher ultimate load capacities than the ones with a rectangular
mesh.
(D6)
(M6)
(M8)
Figure 6.36 Comparison of crack patterns after peak in specimens T5 (left) and T1
(right) predicted by the models with different mesh types
20
30
15
20
10
10 5
0 0
0.00 0.10 0.20 0.30 0.40 0.50 0.00 0.01 0.02 0.03
Slip at 42 mm from loaded edge Slip at 42 mm from loaded edge
(a) T5 (b) T1
Figure 6.37 Comparison of load-slip curves predicted by the models with different
mesh types
240
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
6.7.3.2 Comparison
The crack patterns after peak for the specimens with different Gf values are
compared in Figure 6.38. The load-slip curves are compared in Figure 6.39. It is found
that Gf has a significant influence in the behaviour of the shear-lap tests. In T5 models,
the location, at which the debond crack starts to propagate, shifts closer to the loaded
edge as Gf reduces. This is due to the fact that with a lower value of fracture energy, the
interfacial debond crack propagates more easily and the transverse cracks can open wider.
As a result, the stress does not transfer further and the predicted ultimate load drops. In
T1 models, there is little difference in the crack patterns. The predicted ultimate load also
reduces as the fracture energy value decreases.
241
CHAPTER 6
The crack patterns after peak for specimens with different fct values are compared
in Figure 6.40. The load-slip curves are compared in Figure 6.41. The crack patterns in
T5 models indicate that with a lower tensile strength, the transverse cracks tend to be
more inclined towards the loaded edge. The transverse cracks also form at lower load
levels and therefore the load-slip curves have lower stiffness. The predicted ultimate
loads are however only slightly different. For T1 models, the crack propagation is very
rapid and no stress redistribution is possible. Therefore, the predicted ultimate load drops
as fct decreases.
(F3.86 –
G0.188)
(F3.86 –
G0.124)
(F3.86 –
G0.091)
Figure 6.38 Comparison of crack patterns after peak in specimens T5 (left) and T1
(right) predicted by the models with different Gf values
30
Total load (kN)
20
25
15
20
15 10
10 5
5
0 0
0.00 0.10 0.20 0.30 0.40 0.50 0.00 0.01 0.02 0.03
Slip at 42 mm from loaded edge Slip at 42 mm from loaded edge
(a) T5 (b) T1
Figure 6.39 Comparison of load-slip curves predicted by the models with different
Gf values
242
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
(F4.40 –
G0.188)
(F3.86 –
G0.188)
(F2.93 –
G0.188)
Figure 6.40 Comparison of crack patterns after peak in specimens T5 (left) and T1
(right) predicted by the models with different fct values
30
15
20 10
10 5
0
0
0.00 0.01 0.02 0.03
0.00 0.10 0.20 0.30 0.40 0.50
Slip at 42 mm from loaded edge
Slip at 42 mm from loaded edge
(a) T5 (b) T1
Figure 6.41 Comparison of load-slip curves predicted by the models with different
fct values
243
CHAPTER 6
6.7.4.2 Comparison
The crack patterns after peak for specimens with different Ea values are compared
in Figure 6.42. The load-slip curves are compared in Figure 6.43. There is little difference
in the behaviour of T1 models except the fact that the load-slip curve stiffness is slightly
higher with a higher adhesive modulus. For T5 models, the models with the adhesive
stiffness of 8500 and 12800 MPa behave very similarly in terms of crack patterns. In
model A3500, the major inclined crack, from which the interfacial debond crack initiates,
shifts slightly closer to the loaded edge. The ultimate loads predicted by all models are
however similar even through model A3500 predicts slightly more ductile behaviour.
244
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
(A12800)
(A8500)
(A3500)
Figure 6.42 Comparison of crack patterns after peak in specimens T5 (left) and T1
(right) predicted by the models with different Ea values
30
15
20 10
10 5
0 0
0.00 0.10 0.20 0.30 0.40 0.50 0.00 0.01 0.02 0.03
Slip at 42 mm from loaded edge Slip at 42 mm from loaded edge
(a) T5 (b) T1
Figure 6.43 Comparison of load-slip curves predicted by the models with different
Ea values
245
CHAPTER 6
In this section, an approximate FE analysis of the bond between the CFRP and
concrete is presented. A relatively coarse mesh was used with the bond behaviour
idealised by a nonlinear bond-slip relationship. The analysis was carried out on the
specimens with only two plies of CFRP. The typical finite element mesh and boundary
conditions are shown in Figure 6.44. The element size was maintained at approximately 9
mm.
Horizontal
Lf CFRP
roller
X
Vertical
roller Horizontal roller
246
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
η
1
2
ξ 3
Interface element
CFRP element
Concrete element
(a) Location
5 uy tn
4 6 ux
y
1 x 3 tt
z 2
τmax
s1
247
CHAPTER 6
6.8.3 Results
The predicted ultimate load capacities are compared with the actual values in
Figure 6.48. The predicted load-slip curves and the CFRP strain distributions in two
example specimens, T5 and T1, are compared with the experimental ones in Figure 6.18
and Figure 6.20, respectively.
The load capacities predicted are quite close to the experimental values. The
bond-slip based models have lower initial stiffness compared with the testing. However,
in general, the curves are reasonably similar.
40 Varying H bf
parameters L
35
Maximum load (kN
30
25
20
15
10
Exp. Bond-slip
5
0
T12a
T13a
T12b
T13b
T1a
T2a
T3a
T4a
T5a
T6a
T7a
T8a
T1b
T2b
T3b
T4b
T5b
T6b
T7b
T8b
Figure 6.48 Correlation of peak loads predicted by the bond-slip based model and
measured in the experiment
248
SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS
30 25
Total load (kN)
(a) T5 (b) T1
Figure 6.49 Comparison of the load-slip curves of specimens T5a and T1a
(a) T5 2000
1000
0
0 50 100 150 200
Distance from loaded edge (mm)
(b) T1 1000
500
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 6.50 Comparison of CFRP strain distributions of specimens T5a and T1a
249
CHAPTER 6
and inputs for concrete tensile properties and adhesive stiffness was also discussed. The
findings from this work are:
• The nonlinear finite element model was able to simulate the test behaviour relatively
well by predicting the peak load, strain distributions and load-slip curves with reasonable
accuracy. The predicted crack patterns were also quite comparable to those observed in
the experiments indicating two typical failure modes: interfacial debond and shear-
tension failure. The behavioural trends were predicted well when the CFRP bond length,
CFRP thickness, CFRP width and support block height were varied.
• The numerical results were slightly sensitive to mesh sizes, mesh types and mesh
orientation. Adhesive stiffness had little influence on the model behaviour. The major
influence was from the concrete tensile properties. The effect was different in the
specimens failing by interfacial debond and the ones failing by shear-tension.
• The local bond-slips, which were often assumed to be constant along the bond, were
found to be variable due to the presence of intersecting inclined cracks. This finding was
consistent with the experimental observations.
• An approximate solution to the bond between the CFRP and concrete could be
obtained using a bond-slip relationship following Popovics’ equation.
250
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
7.1 Introduction
In this chapter, nonlinear finite element modelling of the retrofitted RC beams
under bending using smeared cracks is presented. The beams are modelled using DIANA
(version 8.1.2). The modelling is carried out for beams under four-point and three-point
bending. The models are verified using the data reported in Chapter 5. A parametric study
is also presented to investigate the effect of several parameters.
Horizontal X
Vertical roller CFRP Concrete below tension roller
reinforcement
Figure 7.1 Two-dimensional mesh of beams loaded in 4-point bending
251
CHAPTER 7
integration scheme was used. The steel and CFRP reinforcements in the beams were
modelled using beam elements (L7BEN). This element type is a two-node, two-
dimensional beam element with the basic variables including the translations ux and uy
and the rotation φz in the nodes (Figure 7.2b). The element was numerically integrated
over the cross-section and along their axis. Two-point and three-point Gauss integration
schemes were used along the element axis and in the element cross-section, respectively.
The reinforcement elements were connected to concrete elements through 2+2 node
structural interface elements (L8IF). The normal and shear tractions, and the normal and
shear relative displacements across the interface are illustrated in Figure 7.3. Three-point
Newton-Cotes integration scheme and linear interpolation were used. The nodes for the
steel reinforcement were superimposed on top of the concrete nodes, whereas the nodes
for the CFRP reinforcement were offset from the concrete beam soffit by a distance equal
to the sum of the adhesive thickness and half of the composite thickness.
3
4 η
η 1
ξ
ξ 2
1
2
(a) (b)
Figure 7.2 Q8MEM element (a) and L7BEN element (b)
Concrete element
Interface element
Reinforcement
element
uy tn
4 3 ux
y
1 x 2 tt
z
252
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
45
177
38
Figure 7.4 plots the location of the reinforcement elements. Since anchorage
failure of steel reinforcement was not observed, the bended portions of the tension bars
were not modelled. The longitudinal bar locations were simplified as straight lines
extending from one end of the beam to the other end. At the longitudinal tension
reinforcement level, the thickness of the concrete was reduced to account for the
reduction of concrete volume due to the presence of the bars. This was achieved by
reducing the thickness of the concrete elements just below the tensile beam elements such
that the concrete area lost was equivalent to the area of tension reinforcement. By doing
so, the weakest shear plane was assumed to be in the concrete elements just below the
tension reinforcement. In the experiments, a portion of the strength of the failure plane
was contributed from the bond between the steel bars and concrete. However, this
contribution was small and it was ignored in the present analysis.
Tension
reinforcement
(typ.)
To account for the influence of the stirrup extrusions when they intersected the
debond crack, the beam elements representing the vertical stirrups were extended
downward passing the longitudinal tension reinforcement by an element size. The beam
element extrusions crossed the weak layer of concrete elements.
253
CHAPTER 7
In the experiments, a steel plate was placed under the actuator to distribute the
load over a small area. The supports also had a certain contact area with the beam to
provide an even load distribution. As a result, localised crushing in the region around the
load points and supports were not observed in the experiments. In the finite element
models, 20-mm thick steel plates, modelled using Q8MEM elements, were added at the
load location. A layer of interface elements were placed between the plate and concrete to
model the friction between the two materials. The steel plate elements were assumed to
have linear elastic properties. To further overcome the problem of localised crushing in
the finite element models, selected concrete elements around the load point and support
were assumed to have linear elastic properties.
The beams strengthened with two plies of CFRP failed either by intermediate span
debond or combination of intermediate span debond and end debond. In these beams,
intermediate span debond occurred either on the clamped or unclamped side of the
beams. Therefore, for each of these beams, two FE models were built. The behaviour of
the unclamped side was studied using the model described above. To model the failure on
the clamped side of the beam, a prestressed steel clamp was added (Figure 7.6). The
clamp consisted of two steel plates connected by a steel bolt. The steel plates were
modelled using Q8MEM elements and the bolt was idealised by a beam element L7BEN.
A frictional slip was assumed between the plate and the concrete and between the plate
and the CFRP. An initial strain of 500×10-6 was applied to the bolt before loading.
Concrete element
Bolt element
Interface element
Common node at
CFRP element connection of bolt and
Steel plate steel plate
254
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
G cr = β G (7.1)
βG
γ
Figure 7.7 Concrete shear strength after cracking
255
CHAPTER 7
7.2.2.2 Steel
The main flexural and shear steel reinforcements in the finite element models
were assumed to be an isotropic linear elastic material up until the yield point. Yielding
of the reinforcement was based on the yield criterion of Von Mises with strain hardening.
The elastic moduli, yield stresses and hardening diagrams used in the models were
determined from the tensile testing as described in Appendix B.2. The moduli and yield
stresses are summarised in Table 7.2. The steel plates were assumed to be a linear elastic
material.
7.2.2.3 Interfaces
In reinforced concrete, the interaction between the reinforcement and the concrete
is highly complex. The interaction is governed by secondary transverse and longitudinal
cracks in the vicinity of the reinforcement. This behaviour can be modelled with a bond-
slip mechanism (de Witte and Kikstra, 2003). The bond-slip constitutive law in DIANA is
based on a ‘total deformation’ theory, which expresses the tractions as a function of the
total relative displacements. The relationship between the normal traction and the normal
256
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
relative displacement is assumed to be linear elastic, whereas the relationship between the
shear traction and the slip is assumed to follow a nonlinear function. In the present
analysis, a cubic function proposed by Dorr (1980) was used to describe the bond-slip
relation between the reinforcement rebars and concrete (Figure 7.8a). The normal
stiffness was assumed to be 1000 MPa.
A different bond-slip relation was used to describe the bond between the CFRP
and concrete. The bond-slip relation followed a nonlinear curve established previously
from the shear-lap tests (Figure 7.8b). The normal stiffness was calculated using the
following equation.
D11 = E / t (7.2)
where the modulus, E, was taken as 8500 MPa and the bond thickness, t, was taken to be
1 mm.
6 4
3
4 2
2 1
0 0
0 0.05 0.1 0.15 0.2 0.25 0 0.05 0.1 0.15 0.2 0.25
Slip (mm) Slip (mm)
(a) (b)
Figure 7.8 Bond-slip relationships adopted for steel bar/concrete interface (a) and
CFRP/concrete interface (b)
257
CHAPTER 7
suitable for the beam models since it converged quickly for most steps. The convergence
criterion adopted was based on the energy norm composed of internal forces and relative
displacements. A new reference norm was determined at the start of each step. In all of
the models, the tolerance for convergence was set to 0.0005 to ensure accurate and good
results. The maximum number of iterations for each load step or increment was set to
100. To improve the convergence, the line search algorithm was also used in the study.
7.2.4 Verification
To verify that the finite element models simulated the behaviour of the beams
properly, four outputs from the experiments and numerical simulations were compared.
They were the crack patterns at failure, the load-displacement curves and the strain
distributions in the steel and CFRP reinforcements. For clarity, only the cracks, of which
the crack strain is large enough so that the crack tensile stress has reduced by 50 % from
the maximum tensile strength, are shown in the crack patterns.
258
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
The FE models of the retrofitted beams indicated three failure modes similar to
those observed in the experiments. The modes were end debond, intermediate span
debond, and combination of end span debond and intermediate span debond. Three
respective example models are models E1, S2, and S1. Their crack patterns at different
load levels are illustrated in Figure 7.11 to Figure 7.13. The deformed shapes of the
example models at or near the peak load are plotted in Figure 7.14 to Figure 7.16.
259
CHAPTER 7
40.4 kN
50.3 kN
60.6 kN
70.0 kN
76.0 kN
(peak)
67.3 kN
(post
peak)
260
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
40.3 kN
60.3 kN
70.0 kN
75.0 kN
77.8 kN
(peak)
68.9 kN
(post
peak)
261
CHAPTER 7
40.2 kN
50.6 kN
60.5 kN
70.0 kN
76.1 kN
(peak)
72.7 kN
(post
peak)
262
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
The models of the retrofitted beams showed more closely spaced and uniformly
distributed flexural cracks compared to those in model C1. More diagonal cracks were
also observed in the shear span as the load increased beyond the capacity of the control
beams.
In model E1, a crack originated from the CFRP end and propagated upward until
it reached the tensile rebar. The crack then branched into two cracks: a crack propagating
up into the concrete beam and a crack propagating horizontally along the rebar. As the
load increased further, the horizontal crack opened wider, propagated towards the beam
middle and joined other vertical cracks in the concrete cover. When the propagation path
was intersected by a stirrup extrusion, it was forced to go around the stirrup. The tensile
stress near the stirrup extrusion induced tooth-like cracks in the concrete cover (Figure
7.17). The peak load was reached when debonding of the concrete cover near the CFRP
end was clearly observed. As more displacement was applied, the horizontal debond
crack propagated deeper toward the beam middle.
263
CHAPTER 7
CFRP
Propagation direction
In model S2, since the CFRP layer was thinner, the flexural cracks opened wider
compared to those in model E1. The tensile steel yielded at around 60 kN. After yielding,
more diagonal cracks became visible, while the flexural cracks continued to widen. The
widest flexural crack located in the shear span near the load point. Upon additional
application of loading, the strain level in the steel rebar bridging the crack increased much
beyond the yield strain. A horizontal crack was seen along the steel rebar level. (Figure
7.18). This crack propagated to a location where the steel strain reduced to a lower value
of around the yield strain. The crack then propagated across the concrete cover towards
the bond surface and eventually along the bond surface just above the CFRP. The final
propagation path was not in the interface elements but in the concrete elements above the
joint.
The situation in model S1 was similar to that in model S2. The only exception was
that as the steel clamp was not present, an end debond crack was also seen together with
the horizontal crack near the bond line. The failure mode indicated was a combination of
intermediate span and end debond.
Wide
Region of high flexural
steel strain crack Stirrup
Steel strain
around
yielding level
Tensile steel
CFRP Tensile
force
Propagation direction
264
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
The crack patterns after the peak load in the models of S beams (S models) are
compared in Figure 7.19. Both unclamped and clamped side were shown. As expected, if
the failure occurred on the clamped side, the failure mode was clearly intermediate span
debond. Meanwhile, the models of the unclamped side indicated a combination of
intermediate span and end debond. In general, the patterns matched reasonably well with
the observations from the experimental study.
The models of E beams (E models) are compared in Figure 7.20. All E models
predicted end debond failure. The crack patterns predicted are also close to those
observed in the experimental study.
(S1U)
(S1C)
(S2U)
(S2C)
(S3U)
(S3C)
265
CHAPTER 7
(E1)
(E2)
(E3)
(E4)
(E5)
Figure 7.20 Crack patterns in E models (unclamped side) at the peak loads
120
C beams S beams E beams
100
Maximum load (kN)
80
60
40
Experiment
20
FE
0
C1a
E1a
E2a
E3a
E4a
E5a
C1b
E1b
E2b
E3b
E4b
E5b
S1a
S2a
S3a
S1b
S2b
S3b
Figure 7.21 Correlation of peak loads as predicted by the FE model and measured
in the experiment
266
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
The loading curves are compared in Figure 7.22 and Figure 7.23. The curves
generated by the finite element analysis and measured in the experiments are in good
agreement, showing similar stiffness before and after yielding of the steel.
100 100
80 80
Total load (kN)
(a) S1 (b) S2
70
60
Total load (kN)
50
S3a
40
30 S3b
20 S3N (FE)
10 S3C (FE)
0
0 10 20 30
Midspan deflection (mm)
(c) S3
Figure 7.22 Comparison of load-displacement curves for S beams
267
CHAPTER 7
80 70
70 60
60 50
50 E1a (Exp.) 40 E2a
40 30 E2b
E1b (Exp.)
30 20
20 E2 (FE)
E1 (FE) 10
10
0
0
0 5 10 15 20
0 5 10 15 20
Midspan deflection (mm)
Midspan deflection (mm)
(a) E1 (b) E2
70 80
60 70
Shear load (kN) 60
Shear load (kN)
50
50
40 40 E4a
30 E3a 30
E4b
20 E3b 20
10 E4 (FE)
10 E3 (FE)
0
0
0 5 10 15 20
0 5 10 15 20
Midspan deflection (mm)
Midspan deflection (mm)
(c) E3 (d) E4
80
70
Shear load (kN)
60
50
40 E5a
30
E5b
20
10 E5 (FE)
0
0 5 10 15 20
Midspan deflection (mm)
(e) E5
Figure 7.23 Comparison of load-displacement curves for E beams
268
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
selected to be as close as possible to each other. In the FE models, the CFRP strains were
taken at the average of the values at Gaussian integration points 1 and 4 (Figure 7.24).
These strains represent the average strain in the bottom half of the composite. It was
found that the strains were not significantly different between the top and bottom layers
of the composite (Figure 7.25).
The simulated CFRP strain distributions are, in general, close to those in the
experimental program at both high and low load levels for all beams. The fluctuation in
the simulated CFRP strains at high load levels is due to the nature of the smeared crack
model, where the composite strains are sensitive to the presence of distributed concrete
cracks transverse to the bond.
Concrete element
Interface element
3 6
2 5 FRP element
1 4
Node 1 Node 2
Gauss integration point
12000
Load (kN)
10000 Top Bottom
20.3 20.3
8000
Microstrain
60.3 60.3
6000 70.0 70.0
2000
0
1000 800 600 400 200 0
Distance from end of FRP (mm)
Figure 7.25 Variation of CFRP strain along the top and bottom halves in model S2N
269
CHAPTER 7
Load point
4000 Load (kN)
40.0 40.4
2000
50.1 50.3
1500 60.0 60.6
70.7 76.0
1000
500
0
1000 750 500 250 0
Distance from end of CFRP (mm)
Load point
12000
Load (kN)
10000
Exp. FEA
2000
0
1000 750 500 250 0
Distance from end of CFRP (mm)
270
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
Load point
10000
9000
40.1 40.2
5000 60.0 60.5
4000 65.0 65.2
1000
0
1000 750 500 250 0
Distance from end of CFRP (mm)
The strain distributions in the tensile steel reinforcement are plotted in Figure 7.29
to Figure 7.31 for three representative beams E1a, S2a and S1a, respectively, at several
load levels. Comparisons for other beams are shown in Appendix D.4.
In general, the numerical models simulate the tensile steel strain distributions
reasonably well before yielding. After yielding, the steel strain depends greatly on the
location of the main flexural cracks and therefore it is not possible to compare the
experimental measurements with the numerical results.
271
CHAPTER 7
Load point
3500
Load (kN)
3000 Exp. FEA
20.1 20.3
2500
30.0 30.3
Microstrain
500
0
1000 750 500 250 0
Distance from end of CFRP (mm)
Load point
10000
9000
Load (kN)
8000 Exp. FEA
7000 20.1 20.3
40.0 40.3
6000
Microstrain
60.0 60.3
5000
70.0 70.0
4000
75.0 75.0
3000
80.2 77.2
2000
1000
0
1000 750 500 250 0
Distance from end of CFRP (mm)
272
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
Load point
10000
Load (kN)
9000
Exp. FEA
8000
20.0 19.8
7000 40.1 40.2
6000
Microstrain
60.0 61.1
5000 65.0 64.8
2000
1000
0
1000 750 500 250 0
Distance from end of CFRP (mm)
273
CHAPTER 7
274
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
The crack patterns at the peak loads for the models with different Gf values are
compared in Figure 7.32. The load-displacement curves are compared in Figure 7.33. It
can be seen that Gf has a moderate influence on the behaviour of the beams. The models
with a lower Gf tend to experience more cracking in the concrete cover prior to failure.
The peak loads are influenced most in E1 models, where lowering Gf results in a lower
ultimate capacity.
The crack patterns at the peak loads for the models with different fct values are
compared in Figure 7.34. The load-displacement curves are compared in Figure 7.35. fct
has a minor influence on the behaviour of the beams. The peak loads predicted for beams
E1 and S2C are similar for three different values of fct.
275
CHAPTER 7
E1a S2a
E1b S2b
E1 (F3.86-G0.188) S2C (F3.51-G0.188)
E1 (F3.86-G0.124) S2C (F3.51-G0.117)
80 E1 (F3.86-G0.091) 100 S2C (F3.51-G0.086)
70
80
Total load (kN)
276
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
E1a S2a
E1b S2b
E1 (F4.40-G0.188) S2C (F4.14-G0.188)
E1 (F3.86-G0.188) S2C (F3.51-G0.188)
100 E1 (F2.93-G0.188) 100 S2C (F2.76-G0.188)
80 80
Total load (kN)
40 40
20 20
0 0
0 5 10 15 0 10 20 30 40
Midspan deflection (mm) Midspan deflection (mm)
Comparisons of crack patterns and loading curves for two representative cases are
included in Figure 7.36 to Figure 7.37. The crack patterns are similar even though the
models with a higher shear retention capacity seem to have more cracks in the shear
spans. These models also show stiffer behaviour as expected.
277
CHAPTER 7
Beta0.001 Beta0.001
Beta0.05 Beta0.05
Beta0.10 Beta0.10
(a) (b)
Figure 7.36 Comparison of crack patterns at the peak loads in beams E1 (a) and
S2C (b) predicted by models with different β values
E1a S2a
E1b S2b
E1 (Beta0.001) S2C (Beta0.001)
E1 (Beta0.05) S2C (Beta0.05)
100 E1 (Beta0.10) 100 S2C (Beta0.10)
80 80
Total load (kN)
60 60
40 40
20 20
0 0
0 5 10 15 0 10 20 30
Midspan deflection (mm) Midspan deflection (mm)
(a) E1 (b) S2
Figure 7.37 Comparison of load-deflection curves predicted by models with
different β values
278
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
results are also included. The study is carried out on a base specimen with the dimensions
and material properties similar to those of beam E1a. The mode intermediate span debond
is abbreviated as IS debond in the following paragraphs.
Figure 7.38 shows the behavioural trend of retrofitted RC beams bonded with
different composite cross sectional areas, Af. As the area increases, the failure mode shifts
in the following order:
concrete crushing => IS debond => mixed IS and end debond => to end debond
It is clear that increasing the CFRP amount does not always lead to an increase in the
beam capacity. The optimal CFRP cross sectional area, when the capacity of the beam
reaches the maximum value, is around 50 mm2 (or 3 plies of CFRP of 100 mm width).
The corresponding failure mode is mixed debond. When the failure mode is concrete
crushing or FSCD, the beam capacity increases with Af. However, when the mode is end
debond, the beam capacity reduces gradually with Af.
IS debond Mixed
Crush End debond
debond
Ultimate shear load (kN)
100
80
60
40
Exp.
FE
20
0
0 50 100 150 200
2
A f (mm )
Figure 7.38 Behavioural trend of retrofitted beams with different CFRP cross
sectional areas
Figure 7.39 shows the behavioural trend of retrofitted beams reinforced with
different amounts of tensile rebars. When the tensile rebar cross-sectional area, As, is
small, flexural and flexural-shear cracks in the intermediate span can open wider.
Therefore, the failure mode is IS debond. As As increases, the failure mode shifts to a
mixed mode of IS and end debond, and then to end debond.
279
CHAPTER 7
100
Ultimate shear load (kN)
80
60
40 Exp
FE
20
0
0 50 100 150 200 250 300 350 400
2
A s (mm )
Figure 7.40 presents the relationship between the ultimate shear load and the ratio
of the CFRP bond length, Lf, to the shear span, a. The failure mode is end debond with
the exception of the beam with the longest bonded length, where the mode is IS debond.
The general trend is that the beam capacity increases moderately as the bond length
increases.
100
80
60
40
Exp.
FE
20
0
0.5 0.6 0.7 0.8 0.9 1.0
Lf / a
Figure 7.40 Behavioural trend of retrofitted beams with different CFRP bond
lengths
280
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
Figure 7.41 illustrates the behavioural trend of retrofitted beams with different
concrete covers. All of the beams investigated fail by end debond. It is clear that the
cover has an insignificant influence on the beam capacity.
100
80
60
40
Exp.
FE
20
0
10 20 30 40 50
c (mm)
Figure 7.41 Behavioural trend of retrofitted beams with different concrete covers
The same material models as described in Section 7.2.2 were used. Orthotropic
material properties were used for the vertical CFRP straps. The CFRP stiffness in the
transverse direction was assumed to be the same as that of the adhesive. To simulate the
bond behaviour between the concrete and CFRP composite, the nonlinear bond-slip
model derived previously from the shear-lap tests was used. In those shear-lap tests, the
stress distribution in the CFRP was close to uniform. However, in the vertical straps,
bending action resulted in non-uniform stress distribution, which could lead to premature
rupture or debonding. To model the distribution accurately, a much finer mesh is
required. This was however undesirable since the computational time would increase
281
CHAPTER 7
significantly. Therefore, the current models are not expected to be able to pick up
debonding and rupture of the composite.
In the experiments, after the peak load was reached and a sudden failure was
observed, loading was continued until complete debonding of the composite. The
dynamic nature of the failure caused severe damage to the straps, concrete and bonding
between them. This action is not picked up by the current static model. Therefore, the
simulation was done only up to the peak load and the behaviour after peak was not
considered.
For the models of the beams with prestressed straps, i.e. models A2a and A2b, the
concrete just above the wedge was subjected to a compressive field and therefore behaved
more liked reinforced concrete. Therefore, the concrete portion above the wedges was
modelled using the fixed smeared cracks.
282
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
(a) E3b2
concrete
strap elements
interface elements
main CFRP
(c) A1b
concrete
strap elements
interface elements
main CFRP
(d) A2a and E1b2 (for E1b2, end strap moved to the left by 60 mm)
283
CHAPTER 7
Interface
elements Concrete
layers
Interface
Longitudinal
elements
CFRP
Figure 7.44 Location of interface elements in the FE mesh for model A2a
7.3.2 Verification
7.3.2.1 Crack patterns
The crack patterns at the peak load of retested beams and A beams are illustrated
in Figure 7.45 and Figure 7.46, respectively. The beams with one strap failed by
intermediate span debond; whereas the beams with three straps failed by debonding of the
straps in the concrete cover followed by debonding of the longitudinal CFRP between the
straps (Figure 7.47). Slight crushing of concrete under the load point was also observed.
The patterns are similar to those observed in the experiments. However, complete
debonding of the straps was not seen.
284
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
(E3b2)
(E3a2)
(E1a2)
(E5a2)
Figure 7.45 Crack patterns in retested beams at peak
285
CHAPTER 7
(A1a)
(A1b)
(A2a)
(A2b)
Figure 7.46 Crack patterns in A beams at peak
debond debond
286
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
predictions do not match as well as for the beams tested in four-point bending. However,
the predictions and the measurement are still in relatively good agreement.
140
120
E3a2
E1a2
E5a2
A1a
A2a
A1b
A2b
Figure 7.48 Correlation of peak loads as predicted by the FE model and measured
in the experiment
140 140
120 120
Shear load (kN)
100 100
80 80
E3b2 (Exp.)
60 60 E3a2 (Exp.)
E3b2 (FE)
40 40
E3a2 (FE)
20 20
0 0
0 2.5 5 7.5 10 12.5 15 0 2.5 5 7.5 10 12.5 15
Midspan deflection (mm) Midspan deflection (mm)
140 140
120 120
Shear load (kN)
Shear load (kN)
100 100
80 80
60 60 E5a2 (Exp.)
E1a2 (Exp.)
40 40
E1a2 (FE) E5a2 (FE)
20 20
0 0
0 2.5 5 7.5 10 12.5 15 0 2.5 5 7.5 10 12.5 15
Midspan deflection (mm) Midspan deflection (mm)
287
CHAPTER 7
140 140
120 120
Shear load (kN)
140 140
120 120
Shear load (kN)
Shear load (kN)
100 100
80 80
60 A2a (Exp.) 60 A2b (Exp.)
40 A2a (FE) 40 A2b (FE)
20 20
0 0
0 2.5 5 7.5 10 12.5 15 0 2.5 5 7.5 10 12.5 15
Midspan deflection (mm) Midspan deflection (mm)
288
SMEARED CRACK MODELLING OF RETROFITTED BEAMS
Load point
6000
Load (kN)
5000 Exp. FE
20.0 18.2
4000
40.0 40.1
Microstrain
80.0 80.6
2000
94.5 93.6
1000
0
500 400 300 200 100 0
Distance from end of CFRP (mm)
3500
3000
2500
Exp. (Front)
Microstrain
2000
Exp. (Back)
1500
FE
1000
500
0
-500 0 25 50 75 100
Shear load (kN)
• The nonlinear finite element analysis using smeared cracks was able to simulate the
test behaviour reasonably well, predicting similar peak loads, load-deflection curves
and strain distributions.
289
CHAPTER 7
• The failure mechanisms were also captured showing two debonding modes, end and
intermediate span debond, similar to those observed in the experiments.
• The finite element analysis indicated that the behaviour of retrofitted RC beam
depended a great deal on the amount of CFRP and steel reinforcement. The failure
mode changed from intermediate span debond to end debond as the CFRP amount
increased. The beams with a low amount of tension reinforcement were more likely to
fail by IS debond.
• The CFRP bonded length also had a moderate influence on the beam end debond
capacity. The longer the bond length, the higher the capacity was. By extending the
CFRP close to the support, end debond could be suppressed and the failure mode
shifted to intermediate span debond.
• Despite some discrepancy, the effect of the anchorage straps was simulated generally
well. The straps proved to be able to prevent or delay debonding. However,
debonding and rupture of the straps were not captured with the two-dimensional
model.
290
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
One problem with the use of discrete crack modelling is the difficulties in
simulating structures of which the behaviour is influenced by a large number of cracks.
Debonding failure of CFRP in shear-lap specimens or retrofitted beams is one of those
cases, where the structural behaviour is influenced by many cracks including both local
and global cracks. The local cracks are in the proximity of the bond area whereas the
global cracks are the distributed cracks in other parts of the structure. Therefore, to
simulate the behaviour of retrofitted members, a combination of discrete and smeared
crack modelling approaches can be used, where the major local cracks are simulated by
discrete cracks and the minor distributed cracks are modelled using smeared cracks.
This chapter reports the results from a numerical investigation of the behaviour of
the shear lap and beam specimens under testing using discrete/smeared cracks. Modelling
is implemented using the computer program MERLIN II. The program is developed by
Dr Jan Cervenka and Ron Reich at the University of Colorado where Professor Victor
291
CHAPTER 8
Saouma is the principle investigator. The model of discrete cracks is based on nonlinear
fracture mechanics.
The goal of the study presented in this chapter is to investigate the possibility of
modelling debonding failures in retrofitted structures using discrete/smeared crack
modelling based on nonlinear fracture mechanics. The discrete/smeared model is also to
further validate the failure mechanisms, which are used to develop the prediction models
presented in Chapter 9. For those purposes, only six shear-lap configurations, T1 to T6,
and eight retrofitted beams (E and S beams) are modelled in this study.
The Fictitious Crack Model proposed by Hillerborg et al. (1976) is probably the
most widely used in nonlinear fracture mechanics finite element analysis (Saouma, 2002).
Its basic concept is illustrated in Figure 8.1. It is related well to the true physical nature of
cracks in concrete. The model is based on the softening curve of the tensile stress versus
crack opening displacement or COD.
292
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
COD
Microcracks Spalling
Fracture Process Zone True crack
Effective crack
f’t
s1 GF Thickness
w1 w2
where φf is the angle of friction; τ is the interface traction vector; and σ is the normal
traction component (Figure 8.2). The shape of the failure function in a two-dimensional
case is shown in Figure 8.3.
Material 1
Material 1
ux, τ
uy, σ
Material 2 Material 2
293
CHAPTER 8
1
tan (φf)
Initial failure
1 fucntion
tan (φf)
c
Final failure
fucntion
σt σ
The cohesion, c, and the tensile strength, σt, are reduced by softening laws based
on a softening parameter uieff (as opposed to COD in Hillerborg’s model). The bi-linear
softening laws available in MERLIN are illustrated in Figure 8.4. As uieff increases, the
cohesion and tensile strength soften and the failure surface changes from its initial state to
its final state as shown in Figure 8.3.
σt c
σt0 c0
GIF GIIF
s1σ s1c
w1σ wσ w1c wc
uieff uieff
To obtain uieff, the displacement vector u is decomposed into an elastic part ue and
an inelastic part ui:
u = ue + ui (8.2a)
uieff is defined as the norm of the inelastic displacement and is given by:
2 2
u ieff = u ix + u iy (8.2b)
ui is composed of the plastic up and fracturing displacement uf, and is given by:
ui = up + uf (8.2c)
294
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
up = γui (8.2d)
For concrete, γ is usually assumed to be equal to 0.2 or 0.3 (Saouma, 2002).
The stress at the interface depends on the displacement and is given as:
σ = αE(u – up) (8.2e)
where
σ = {τ, σ}T (8.2f)
⎡K 0 ⎤
E = ⎢ to (8.2g)
⎣ 0 K no ⎥⎦
Kno and Kto are the initial normal and tangential stiffness.
σ σ
σi σi
Kno
Kns GIF
u ui = uieff
up ui up ui
Figure 8.5 Stiffness degradation in the equivalent uniaxial case (Saouma, 2002)
In the elastic range, no dilatancy is considered and the off-diagonal terms of the
matrix E are all equal to zero. The dilatancy is introduced after the failure limit has been
reached. The dilatancy angle is also assumed to be a function of uieff.
⎛ u ieff ⎞
φ d (u ieff ) = φ d 0 ⎜⎜1 − ⎟⎟ when u ieff ≤ u dil
⎝ u dil ⎠ (8.2h)
φ d (u ieff ) = 0 when u ieff > u dil
where φd0 is the initial dilatancy angle and udil is the maximum displacement after which
the dilatancy vanishes.
295
CHAPTER 8
296
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
Concrete
The typical final finite element mesh and boundary conditions are shown in
Figure 8.8. The concrete, CFRP and adhesive were modelled using three-node triangle
isoparametric plane stress elements. Line-to-line four-node interface elements were used
to model the crack surface (Figure 8.9). The thicknesses of CFRP and adhesive were
assumed to be the same as those used in the smeared crack model (0.7 mm per ply of
CFRP and 1 mm for adhesive).
297
CHAPTER 8
(a) T1 (b) T5
Figure 8.8 Two typical meshes of the shear-lap specimens
4 3
1 2
1 2
(a) (b)
Figure 8.9 Three-node plane stress element (a) and four-node interface element (b)
The material parameters were chosen to be as close to those used for the smeared
crack model presented in Chapter 6 as possible. For the interface crack models, there are
a number of extra parameters which were not required in the smeared crack model. The
values used for these parameters were obtained through calibration with experimental
results. The normal and tangential stiffness were taken to be 1.5 times of concrete elastic
modulus. The friction and dilatancy angles were taken to be 45 degrees. The cohesion
was assumed to be 10 % higher the tensile strength.
298
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
by Karbhari and Engineer (1996), in which the interfacial facture energy of the bond was
investigated. However, the failure surface observed was within both the concrete and
bonded materials. In MERLIN User’s Manual, it is suggested that GfII can be taken as 10
times of GfI. In Niu and Wu (2001), GfII was varied from 0.5GfI to 1.0GfI. In this study,
the specific mode II fracture energy was taken to be twice the specific mode-I fracture
energy.
Table 8.2 Concrete smeared crack model parameters for shear-lap blocks
Description Unit Value
Thickness mm 140
Mass density kg/mm3 2.40E-06
Elastic modulus MPa 29344
Poisson's ratio 0.2
Tensile strength MPa 3.86
Fracture energy N/mm 0.188
Compressive strength MPa -53.7
Compressive critical displacement mm -0.5
Factor for return direction in Haigh-Westergaard space 0
Factor for roundness of Menetrey-Westergaard failure surface 0.55
Onset of nonlinearity in compression MPa -43.0
Plastic strain at compressive strength -1.05E-03
299
CHAPTER 8
300
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
301
CHAPTER 8
24.0 kN
27.2 kN
28.29 kN
Figure 8.10 Deformed shapes of model T5 at different load levels (magnification
factor = 60)
302
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
10.1 kN
15.0 kN
17.7 kN
Figure 8.11 Deformed shapes of model T1 at different load levels (magnification
factor = 60)
The deformed shapes of all six models T1 to T6 are illustrated in Figure 8.12. It
can be seen that as the bond length increased beyond 80 mm (specimen T2), the failure
mode changed from shear-tension to interfacial debond. This trend is similar to the
observation from the experimental study and numerical investigation using smeared
cracks.
303
CHAPTER 8
(T1) (T2)
(T3) (T4)
(T5) (T6)
Figure 8.12 Discrete crack opening in models T1 to T6 at the peak loads
(magnification factor = 60)
The CFRP strain distributions are compared in Figure 8.14 and Figure 8.15
between the experimental and the discrete/smeared crack modelling results. Since the
numerical model behaviour tends to be stiffer than the actual behaviour, the CFRP strains
304
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
are generally lower compared with the measurements. However, the shapes of the
distributions are similar.
25 30
25
Total load (kN)
(T1) (T2)
30 35
25 30
Total load (kN)
(T3) (T4)
35 35
30 30
Total load (kN)
25 25
20 20
15 T5a (Exp.) 15 T6a (Exp.)
10 T5 (Smeared) 10 T6 (Smeared)
5 T5 (Discrete/smeared) 5 T6 (Discrete/smeared)
0 0
0.00 0.10 0.20 0.30 0.40 0.50 0.00 0.10 0.20 0.30 0.40 0.50
Slip at 42 mm from loaded edge Slip at 42 mm from loaded edge
(T5) (T6)
Figure 8.13 Comparison of load-slip behaviours between the experiments, FEA
using smeared crack modelling and FEA using discrete/smeared crack modelling
305
CHAPTER 8
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 8.14 Comparison of CFRP strain distributions in specimens T1a, T2a and
T3a (exp. vs discrete/smeared crack model)
306
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
Microstrain
24.0 24.1
2000 26.0 27.8
1500
(T4a)
1000
500
0
0 50 100 150 200
Distance from loaded edge (mm)
1000
0
0 50 100 150 200
Distance from loaded edge (mm)
0
0 50 100 150 200
Distance from loaded edge (mm)
Figure 8.15 Comparison of CFRP strain distributions in specimens T4a, T5a and
T6a (exp. vs discrete/smeared crack model)
307
CHAPTER 8
‘F3.86-G0.188’. The crack patterns of these models are similar and therefore not shown
herein.
The load-slip curves are compared in Figure 8.16 and Figure 8.17. Similar to the
smeared crack models of the shear-lap tests, the mode-I fracture energy has a significant
influence on the specimen behaviour. The predicted ultimate load capacity reduces as the
fracture energy decreases for both models T5 and T1. The concrete tensile strength seems
to have a less significant effect on the predicted ultimate load model T5 than on model
T1. However, the model with a lower tensile strength shows less stiff behaviour.
20
30
15
20
10
10 5
0 0
0.00 0.10 0.20 0.30 0.40 0.50 0.00 0.01 0.02 0.03
Slip at 42 mm from loaded edge Slip at 42 mm from loaded edge
(a) T5 (b) T1
Figure 8.16 Comparison of load-slip curves predicted by the models with different
GIf values
30 20
15
20
10
10 5
0 0
0.00 0.10 0.20 0.30 0.40 0.50 0.00 0.01 0.02 0.03
Slip at 42 mm from loaded edge Slip at 42 mm from loaded edge
(c) T5 (d) T1
Figure 8.17 Comparison of load-slip curves predicted by the models with different
fct values
308
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
The load-slip curves are compared in Figure 8.18. It is clear that since model T1 is
subjected to a shear-tension failure where mode-I is possibly more dominant, the mode-II
facture energy, GIIf, does not influence the load-slip curve significantly. Meanwhile, GIIf
has a significant effect in model T5, especially after the interfacial debond crack starts to
propagate. The model with a lower value for GIIf shows less stiff behaviour.
30 20
15
20
10
10 5
0 0
0.00 0.10 0.20 0.30 0.40 0.50 0.00 0.01 0.01 0.02 0.02 0.03
Slip at 42 mm from loaded edge Slip at 42 mm from loaded edge
(a) T5 (b) T1
Figure 8.18 Comparison of load-slip curves predicted by the models with different
GIIf values
309
CHAPTER 8
Discrete cracks
The typical finite element mesh and boundary conditions are shown in Figure
8.20. The element size varied from 25x25 mm to 5x5 mm. Due to symmetry, only half a
beam was modelled. The model was supported vertically at the base and horizontally at
the beam’s centreline. The load was applied to a single node on the top of the beam.
310
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
Tensile
rebar
Discrete
cracks
Adhesive
CFRP
near the flexure-shear crack tip near the CFRP end
Figure 8.20 A typical mesh of retrofitted beams
The concrete, CFRP, adhesive and steel load plates were modelled using three-
node triangle isoparametric plane stress elements (Element Type 8). The top longitudinal
rebar and stirrups were modelled as embedded reinforcement (ReinfRods). MERLIN
internally determines which continuum elements are crossed by the reinforcement and
properly adjusts the stiffness matrix. The tensile steel reinforcement was modelled using
two-node bar elements (Element Type 2). This allowed the reinforcement to cross the
flexure-shear crack. A perfect bond was assumed between the reinforcement and the
concrete
311
CHAPTER 8
312
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
313
CHAPTER 8
Figure 8.22 Deformed shape of the unclamped side of beam E1 at the peak load
(magnification factor = 40)
314
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
Figure 8.23 Deformed shape of the clamped side of beam S2C at the peak load
(magnification factor = 30)
315
CHAPTER 8
Figure 8.24 Deformed shape of the unclamped side of beam S1N at the peak load
(magnification factor = 30)
For completeness, the deformed shapes of other modelled beams are plotted in
Figure 8.25 and Figure 8.26.
316
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
(E2)
(E3)
(E4)
(E5)
Figure 8.25 Deformed shapes of the unclamped side of other E models at the peak
loads (magnification factor = 30)
317
CHAPTER 8
(S1C)
(S2N)
(S3N)
(S3C)
Figure 8.26 Deformed shapes of other S models at the peak loads (magnification
factor = 30)
318
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
319
CHAPTER 8
80 70
70 60
Shear load (kN)
(a) E1 (b) E2
70 80
60 70
Shear load (kN)
50 60
50
40
40
30 E3a (Exp.) E4a (Exp.)
30
20 E3b (Exp.) E4b (Exp.)
E3 (Smeared) 20
10 E4 (Smeared)
E3 (Discrete/smeared) 10
E4 (Discrete/smeared)
0 0
0 5 10 15 20 0 5 10 15 20
Midspan deflection (mm) Midspan deflection (mm)
(c) E3 (d) E4
80
70
Shear load (kN)
60
50
40
30 E5a (Exp.)
E5b (Exp.)
20
E5 (Smeared)
10 E5 (Discrete/smeared)
0
0 5 10 15 20
Midspan deflection (mm)
(e) E5
Figure 8.27 Comparison of load-displacement curves for E beams
320
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
90 90
80 80
70 70
Total load (kN)
(a) S1 (b) S2
70
60
Total load (kN)
50
40
S3a (Exp.)
30 S3b (Exp.)
S3N (FE)
20 S3C (FE)
10 S3N (Discrete/smeared)
S3C (Discrete/smeared)
0
0 10 20 30
Midspan deflection (mm)
(c) S3
Figure 8.28 Comparison of load-displacement curves for S beams
321
CHAPTER 8
Load point
4000
30.0 30.5
2000 40.0 40.5
0
1000 750 500 250 0
Distance from end of CFRP (mm)
Load (kN)
10000
Exp. FEA
20.1 20.9
8000
40.0 39.8
Microstrain
0
1000 750 500 250 0
Distance from end of CFRP (mm)
322
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
Load point
10000
9000
Load (kN)
8000 Exp. FEA
7000 20.0 20.9
Microstrain
6000 40.1 39.7
5000 60.0 60.5
4000 65.0 70.2
3000 70.0 75.6
2000 73.7 80.6
1000
0
1000 750 500 250 0
Distance from end of CFRP (mm)
Load point
3500
Load (kN)
3000
Exp. FEA
2500 20.1 20.4
Microstrain
30.0 30.5
2000
40.0 40.5
1500
50.1 50.0
1000 60.0 60.0
0
1000 750 500 250 0
Distance from end of CFRP (mm)
Figure 8.32 Comparison of tensile steel strain distributions in beam E1a (exp. vs
discrete/smeared crack model)
323
CHAPTER 8
Load point
10000
9000
Load (kN)
8000 Exp. FEA
7000 20.1 20.9
Microstrain
Figure 8.33 Comparison of tensile steel strain distributions in beam S2a (exp. vs
discrete/smeared crack model)
Load point
10000 Load (kN)
9000 Exp. FEA
8000 20.0 20.9
7000 40.1 39.7
Microstrain
Figure 8.34 Comparison of tensile steel strain distributions in beam S1a (exp. vs
discrete/smeared crack model)
324
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
The load-displacement curves are compared in Figure 8.35 and Figure 8.36. The
mode-I fracture energy has a more significant effect on model E1 than model S2C. As
expected, the higher the energy, the higher the predicted load capacity. A similar
observation can also be made for the effect of concrete tensile strength, ft.
60
50 60
40 40
30
20 20
10
0 0
0 5 10 15 0 10 20 30 40
Midspan deflection (mm) Midspan deflection (mm)
325
CHAPTER 8
60
50 60
40
40
30
20 20
10
0 0
0 5 10 15 0 10 20 30 40
Midspan deflection (mm) Midspan deflection (mm)
The load-displacement curves are compared in Figure 8.37. It is clear that the
mode-II fracture energy has a more significant influence on the behaviour of model S2C
than on model E1. The predicted beam capacity increases with the mode-II fracture
energy.
326
DISCRETE/SMEARED CRACK MODELLING OF SHEAR-LAP SPECIMENS AND RETROFITTED BEAMS
60
50 60
40 40
30
20 20
10
0 0
0 5 10 15 0 10 20 30 40
Midspan deflection (mm) Midspan deflection (mm)
• The discrete/smeared crack models were able to simulate the behaviour of shear-lap
tests reasonably well. The predicted peak load, load-slip curves and CFRP strain
distributions were similar to those obtained from the smeared crack models. The
failure mechanisms were simulated relatively well showing two main failure modes:
interfacial debond and shear-tension failure.
• The discrete/smeared crack models were also able to simulate the behaviour of beam
tests relatively well. The simulation results were similar to those from the
corresponding smeared crack models. The modelling results confirmed the failure
mechanism of end and intermediate span debond. End debond was the shearing of
concrete along the tension reinforcement level; whereas intermediate span debond
was the shearing of concrete along the bond surface from the tip of a diagonal crack.
327
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
Therefore, in this chapter, an attempt is carried out to develop and verify design
models to calculate the strength of retrofitted beams. Three most frequently observed
failure modes are to be addressed: flexure failure, intermediate span debond and end
debond. The models are developed based on the observations from the main experimental
investigations and from the numerical analyses. Even though the models are derived
based on the data of beams retrofitted with CFRP, they are expected to be valid for beams
retrofitted with other available composites. A reliability study is also carried out for these
models to recommend appropriate capacity reduction factors for design.
329
CHAPTER 9
εc0 εc
ε’s0 ε’s
εs0 εs
εfrp
(a) (b)
Figure 9.1 Strain distribution before (a) and after installation (b) of FRP
The analysis algorithm is illustrated in Figure 9.2. Two loops were used to
estimate the location of the neutral axis, d, and the concrete strain at the top fibre, εc. The
implementation was executed using VBA in Microsoft Excel 2002.
330
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
START
No
M > Mcr
Yes
Initial estimate of εc
Initial estimate of d
No
∆F < ∆F tolerance
Yes
Calculate unbalanced moment ∆M
Estimate of εc to minimise ∆M
No
∆M < ∆M tolerance
Yes
STOP
331
CHAPTER 9
average
actual steel εs2 steel strain
strain εsm
bond stress
Figure 9.4 illustrates the calculation model for a member subjected to a constant
moment M. To find the average strain, the actual element is replaced by a model
composed of two parts. Part I works as an uncracked section. Part II works as a cracked
section where only the reinforcement and the concrete in compression are taken into
account. The average strain is defined by:
332
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
∆l ∆l1 + ∆l 2 l1 ε s1 + l 2 ε s2
ε sm = = = = (1 − ζ)ε s1 + ζε s2 (9.1)
l l l
where εs1 is the strain in the reinforcement calculated assuming the section is uncracked.
εs2 is the strain in the reinforcement calculated assuming the section is cracked and
neglecting the contribution of concrete in tension. ζ is a distribution coefficient which can
be calculated using the following formula:
2
⎛M ⎞
ζ = 1 − β1β 2 ⎜ cr ⎟ (9.2)
⎝ M ⎠
II II
I
M M
εs2 εs1 εs2
l2/2 l1 l2/2
l
Figure 9.4 Calculation model for a RC member under pure flexure (Comite Euro-
International du Beton, 1985)
333
CHAPTER 9
failure. This verification study was based on the measured strain values in the
experiments described in Chapter 5.
To plot the strain distributions along the beam, sectional analyses were carried out
at 50 locations along the beams at several load levels. The average strain levels were
calculated according to Equation 9.1.
The calculation results for two typical beams, E1a and S1a, at the ultimate load
are shown in Figure 9.5 and Figure 9.6, respectively. In these figures, M is the applied
moment, Fs is the force in the tension reinforcement, Ff is the force in the FRP and dn is
the neutral axis depth from the section top. The figures clearly demonstrate the influence
of steel yielding on the Ff distribution. In beam S1a, once the steel has yielded, there is a
steep increase in the Ff slope and a gradual decrease in the neutral axis depth.
50
M (kN.m)
0
1000 750 500 250 0
Distance from left support (mm)
Figure 9.5 Sectional analysis results along beam E1a at maximum load level
334
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
Start of
steel Start of
yielding End of FRP concrete
Load point cracking
200
Fs (kN)
150
100 dn (mm)
Ff (kN)
50
M (kN.m)
0
1000 750 500 250 0
Distance from left support (mm)
Figure 9.6 Sectional analysis results along beam S1a at maximum load level
The CFRP strain distributions predicted by the beam theory at different load
levels along the whole beam for E1a and S1a are compared with the corresponding
experimental curves in Figure 9.7 and Figure 9.8, respectively. Comparisons for other
beams are presented in Appendix D.4. The plots show that the beam theory can predict
the peak CFRP strain level (in the constant moment region) with good accuracy at both
low and high load levels. However, the predicted CFRP strains in the shear span deviate
significantly from the measured values at high loads. Deviation occurs most clearly when
the steel reinforcement undergoes yielding. For those cases, the actual CFRP strain shows
a more gradual decline in the shear span (Figure 9.8). This finding was also observed in
the numerical models presented previously in Chapters 7 and 8. This was due to presence
of diagonal cracks which led to strain redistribution in the concrete cover.
335
CHAPTER 9
Load point
4000
3500 Load (kN)
Exp. Theory
3000
20.1 20.0
2500
Microstrain
40.0 40.0
2000 60.0 60.0
1500 70.7 70.7
1000
500
0
1000 750 500 250 0
Distance from end of FRP (mm)
Load point
10000 Load (kN)
9000 Exp. Theory
8000
20.0 20.0
7000
40.1 40.0
Microstrain
6000
5000 60.0 60.0
4000 65.0 65.0
3000 70.0 70.0
2000
73.7 73.8
1000
0
1000 750 500 250 0
Distance from end of FRP (mm)
Since the applied moment, M, under the load point is significantly higher than the
cracking moment, Mr, at ultimate, the distribution coefficient, ζ, could be taken to be
unity at ultimate and the maximum CFRP strain could be calculated assuming a cracked
section. As illustrated in Figure 9.9, the CFRP strain levels at ultimate are predicted
accurately using this method.
336
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
10000
Theory (no reduction)
CFRP microstrain
8000 Exp. (at midspan)
Exp. (under load point)
6000
Exp. (between midspan and load point)
4000
2000
E1a
E2a
E3a
E4a
E5a
E1b
E2b
E3b
E4b
E5b
S1a
S2a
S3a
S1b
S2b
S3b
Figure 9.9 Comparison of measured and calculated CFRP strains at the ultimate
load level for retrofitted beams without anchorage
337
CHAPTER 9
Strictly speaking, the tensile force in the composite is transferred through the
shear stress on the concrete area at the weakest plane (Figure 9.11a, surface A) and the
bond area between concrete and steel rebars (Figure 9.11a, surface B). The shear stress is
also not uniform over the failure surface. However, surface B is generally small compared
to surface A and the bond strength between the steel rebars and the concrete cover
(unconfined concrete) is also possibly small compared to the concrete shear strength.
Therefore, for simplification, the average shear stress is assumed to be distributed over
the failure surface A only. End debond can be assumed to occur when the following
inequality is satisfied:
where τave is the average shear stress on the assumed debond surface (surface A) and fcv,d
is the shear strength of the surface. To calculate τave, the force distribution in the
composite plate can be assumed to be linear over the shear span based on the observation
from the experiments described previously (Figure 9.11b). As a result, the average shear
stress can be estimated to be:
where Ff,max is the tensile force in the FRP under the load point; Lf is the FRP bond length
in the shear span; and bf is the FRP width.
338
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
Surface A Surface B
(a)
Lf
FRP
(b)
Figure 9.11 Average shear stress concept
339
CHAPTER 9
1.6 80
1.4 70
1.2 60
1.0 50
MPa
0.8
kN
40
0.6 Average bond stress 30
0.4 Average value 20 Tensile force in CFRP
0.2 10 Average value
0.0 0
E1a
E2a
E3a
E4a
E5a
E1b
E2b
E3b
E4b
E5b
S1a
S2a
S3a
S1b
S2b
S3b
(a) (b)
Figure 9.12 Variation of maximum average bond stress in beams failed by end
debond (a) and maximum tensile force in CFRP in beams failed by intermediate
span debond or mixed debond (b)
P = αβ f β L f c b f L e (9.5)
where the factor, α, is a calibration factor to account for any difference between a beam
failing by intermediate span debond and a shear-lap specimen failing by interfacial
debond. Despite many similarities, the situation in a retrofitted beam is much more
340
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
complex due to the presence of several factors such as bending deformation, shear
deformation and the presence of tensile steel reinforcement. Teng et al. calibrated a with 8
beams and 9 slabs and suggested a value of 0.4 for α. A similar calibration study was
carried out using the experimental results reported in Chapter 5. The study showed that
the maximum CFRP tensile force was approximately 64.4 kN, and the corresponding
value for the factor α was 1.04. This is demonstrated in Figure 9.12b.
It was also found that the factor α can be taken to be 1.0 for simplification without
altering the accuracy significantly.
With these modifications, the first verification results are shown in Figure 9.13.
As indicated by the predicted curves, beams E1a to B2 fail by end debond and beams S1
to B1 fail by intermediate span debond. These predictions are consistent with the actual
failure modes observed in the experiment. For beams S1a to S3b, the two predicted
curves are relatively close suggesting a mixed mode of end and intermediate span
debond, which was also seen the experiment. For the last 4 beams, A2a to E3a, the
anchorage improved the strength significantly by preventing end debond and therefore
forcing intermediate span debond. For these beams, the predicted failure loads are slightly
higher than those observed in the experiments.
341
CHAPTER 9
160
140
Failure shear load (kN)
120
100
80
60
40 Exp.
Theory - No debond
20 Theory - End debond
Theory - IS debond
0
E3b2
E1a2
E3a2
E1a
E2a
E3a
E4a
E5a
E1b
E2b
E3b
E4b
E5b
B2
B1
S1a
S2a
S3a
S1b
S2b
S3b
A2a
A1a
Beam label
Figure 9.13 Prediction of end and intermediate span failure loads for the beams
tested in the present study
The results from the second verification study using the beam database are shown
in Figure 9.14, which also indicates good agreement. The averages of the ratios of
predicted failure loads on the experimental results Vexp/Vcal are 1.01 and 1.05 for
intermediate span and end debond, respectively. The coefficients of variation are 14 and
20 %, respectively. For the end debond strength model (Figure 9.14a), the predicted
values are quite conservative for a number of beams. The reason is that these beams were
reinforced with a relatively large amount of tension reinforcement (ku ≈ 0.4) and therefore
the bond area of the tensile rebars could be significant.
342
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
Vcal (kN)
80 200
150
40 100
50
0
0 40 80 120 160 0
0 50 100 150 200 250 300 350
Vexp (kN) Vexp (kN)
Note: FSCD: flexure-shear crack debond; FCD: flexural crack debond; SCD: shear crack debond
Figure 9.15 compares the behavioural trends as the CFRP thickness varies. The
theoretical results include predictions for intermediate span debond (IS debond), end
debond, flexural failure assuming no debond (or full composite action) and flexural
failure of the original beams without CFRP. There predicted trends are very similar to
those observed from the FEA. As the CFRP area increases beyond an optimal value of
approximately 50 mm2, the predicted failure mode changes from intermediate span
debond to end debond (as the intermediate span debond capacity increases to a value
higher than the end debond capacity).
343
CHAPTER 9
Failure mode
predicted by IS debond Mixed
FEA Crush End debond
Ultimate shear load (kN) debond
100
80
60
Exp.
40 FE
Theory - IS debond
20 Theory - End debond
Theory - No debond
Theory - No FRP
0
0 50 100 150 200
2
A f (mm )
Figure 9.15 Comparison of the predicted trends with different CFRP thicknesses
Figure 9.16 compares the behavioural trends as the tension reinforcement area
varies. A good agreement can also be seen between the trends predicted by the theoretical
models and the FEA. Shifting of the failure mode from intermediate span debond to end
debond is observed as the steel reinforcement amount increases.
Failure mode
predicted by IS debond Mixed debond End debond
FEA
100
Ultimate shear load (kN)
80
60
Exp
40 FE
Theory - IS debond
20 Theory - End debond
Theory - No debond
Theory - No FRP
0
0 50 100 150 200 250 300 350 400
2
A s (mm )
Figure 9.16 Comparison of the predicted trends with different tensile steel amounts
The behavioural trends as the CFRP bond length varies are depicted in Figure
9.17. The trend predicted by the theoretical model matches well with that predicted by the
FEA.
344
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
Critical unbonded
section near CFRP
end
Failure mode
predicted by End debond IS debond
FEA
120
Ultimate shear load (kN)
100
80
60
Exp.
FE
40 Theory - IS debond
Theory - End debond
20 Theory - No debond
Theory - No FRP
0
0.50 0.60 0.70 0.80 0.90 1.00
Lf / a
Figure 9.17 Comparison of the predicted trends with different CFRP bond lengths
As illustrated in Figure 9.18, the theoretical models also indicate that concrete
cover has an insignificant influence on the debond capacity of retrofitted beams.
Failure mode
predicted by End debond
FEA
Ultimate shear load (kN)
120
100
80
60
Exp.
40 FE
Theory - IS debond
Theory - End debond
20 Theory - No debond
Theory - No FRP
0
10 20 30 40 50
c (mm)
Figure 9.18 Comparison of the predicted trends with different concrete covers
345
CHAPTER 9
variables affecting the strength of structures. The development of these techniques has
been described in Ellingwood et al. (1980).
There are several reasons to conduct further reliability study on retrofitted beams.
Firstly, previous studies on retrofitted beams did not consider the model error or assumed
it to be the same as for normal RC beams. This was due to unavailability of verification
data for retrofitted beams. Secondly, more tests and analysis have become available
recently allowing more reliable estimation of material variability, especially for
composites. Thirdly, no study on the reliability of CFRP retrofitted RC beams has
considered debonding failures. The main reason for that was the lack of a reliable
theoretical model for these modes.
346
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
The following steps were carried out in this study as guided by Ellingwood et al.
(1980):
• Calculate the bias and variability of the computational procedures.
• Choose a series of representative beam configurations, defined by nominal material
strengths and nominal dimensions. Calculate the nominal resistance, Rn, using the
developed computational procedures.
• Generate a set of material strengths and dimensions randomly from the statistical
distributions of each variable. Calculate the theoretical capacities, R, using the set of
variables plus randomly generated values of the model error. Calculate the mean and
coefficient of variation of R.
• Use the statistical properties of R and select a target reliability index, β, to
recommend the resistance factor for the models.
Several beam configurations were analysed. The beams were assumed to fail in
only one of three following modes:
• Flexural failure: This mode was possible for all configurations. Other failure
mechanisms were assumed to be avoided through the use of end and intermediate span
anchorage.
• Intermediate span debond: This mode was likely when the sectional capacity is
moderately higher than the debond capacity. End debond, if it was more critical, was
assumed to be avoided by end anchorage.
• End debond: This mode was less likely compared to other two modes for large beams
bonded with a small amount of CFRP reinforcement. Intermediate span debond, if it was
more critical, was assumed to be avoided by intermediate span anchorage.
The procedures were followed separately for each failure mode, namely flexural
failure, intermediate span debond and end debond. One reason to adopt this procedure
was the complexity of the computational procedures. If the resistance could be calculated
347
CHAPTER 9
using ‘simple’ formulae, the beam variables would be able to be included in the limit-
state functions for each failure modes to compute their reliability index and a
stochastically most relevant mode would be found. Another reason was that this method
allowed determination of separate resistance factors for each case and therefore different
model errors could be incorporated.
348
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
1800
Cross section T1:
Tension reinforcement: 6Y22
180
692
Compression reinforcement: 2Y22
Span: 4.88 m
346
1800
Cross section T2:
Tension reinforcement: 4Y28 and 4Y26
Compression reinforcement: 2Y22
180
743
Span: 7.92 m
371
1800
Cross section T3:
Tension reinforcement: 10Y32
180
Span: 11.89 m
435
The strength models developed for retrofitted RC beams were dependant mainly
on the sectional dimensions and bending moment in the section. For the debonding
models, the shear span was the only additional input for the model. For simplicity, the
beams were loaded in four-point bending only even though, in practice, bridge girders are
under more complex loading pattern mainly from dead load and traffic load.
For each beams, two loading shear spans and two applied CFRP thicknesses were
considered. This led to 12 designed retrofitted cases. For all beams, the CFRP terminated
200 mm from the support. All beams were assumed to have a sufficient shear capacity.
The beam configurations are summarised in Table 9.1.
349
CHAPTER 9
350
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
The mean was taken as the average value of the ratio of the test strength divided
by calculated strength Rtest / Rcalc. The variability was given by
Vm = VT2 / C − Vtest
2
− Vspec
2
(9.6)
where VT/C is the coefficient of variation obtained directly from the comparison of
measured and calculated strength; Vtest represents the uncertainties in the measured loads
due to such things as the accuracies of the gages and errors in readings; and Vspec
represents the errors introduced by variations in the specimen material strengths and
dimensions. The typical values for Vtest are about 2 to 4 % and the values for Vspec are
about 4 % (Ellingwood et al., 1980). In the calculation for VT/C, it was assumed that the
verification results as described in Section 9.3.2 for mainly small to medium rectangular
beam sections were also valid for relatively large T beam sections.
The final results are shown in Table 9.2. It was clear that the variations of the
models proposed for three dominant failure modes were much higher than those of
normal RC beams. The variations are 14.8 %, 12.7 % and 18.9 % for the models to
predict the full composite action, intermediate span debond and end debond, respectively.
351
CHAPTER 9
Table 9.3 Statistical parameters for some random variables from CONTECVET
(2002)
CONTECVET report
Variable
Distribution COV (%)
Concrete compressive strength Log-normal 10-30
Steel yield strength Log-normal 5-10
Geometry Normal 5
Sanjayan and Candy (2004) tested six core specimens in their reliability study of a
bridge in Australia (Baandee Lakes Bridge No. 1049). They found that the variation of
concrete compressive strength was 25 %. The normal quoted range for the strength is
from 15 to 21 % (Ellingwood et al., 1980; Holicky and Markova, 2000). In this study, the
variation was chosen to be 20 %, which is the average value of the limits given in
CONTECVET report. It was assumed to have a log-normal distribution.
Some tests on the variation of steel yield strength are available in Sanjayan and
Candy (2004) and Sonnenberg and Boully (2004). In the first study, nine tensile tests
were performed. The variation found was 7.3 %. In the second study, steel reinforcement
samples were taken from seven old pre 1960 bridges in Victoria, Australia. They found
that the variations ranged from 2.2 % to 10.6 %. In this study, a value of 7.5 % was used,
which is the average value of the limits given in CONTECVET report. The steel yield
strength was assumed to have a log-normal distribution.
352
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
value of 7.1 mm was recommended later in Holicky and Markova (2000). In this study,
the standard deviation was taken to be 10 mm.
9.4.3.4 CFRP
In this study, only the variability in CFRP tensile strength is considered. The
composite stiffness and geometrical dimensions are assumed to be deterministic.
µ≈σ (9.7)
Okeil et al. (2002) used a value of 2.2 % for the COV and 1.10 for the bias in their
work. This based on the results from tensile testing of CFRP bars reported in Bakht et al.
(2000). Ayers and Van Erp (2002) reported an experimental program to investigate the
strength and stiffness variations of E-glass laminates formed from unidirectional fibres.
The strength reported was ‘Normalised Unit Strength’ or ultimate force per unit weight
per mass of reinforcement. They found that the variability of lamina strengths ranged
from 2.27 % to 15.34 %. A recent study on the variation of CFRP retrofitting materials
formed in the field by a wet lay-up method was by Atadero et al. (2004). The results from
the five date sets are listed in Table 9.4. The variations found were more consistent for
the last four samples with the values lying around 12 %. In this study, the COV for the
CFRP strength variability was taken to be 12 %.
353
CHAPTER 9
Table 9.4 Statistical parameters for FRP ultimate strength from Atadero et al.
Data set No. of samples Sample COV (%)
A 177 23.0
B1 49 12.1
B2 50 12.2
B3 20 13.6
C 260 12.7
354
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
9.4.5 Reliability
9.4.5.1 Reliability index
The performance of a structure can be represented by a limit-state function g,
which can be simplified as:
g = R − S = φR n (X 1 , X 2 , ..., X n ) − (γ D D + γ L L ) (9.9)
where R is the random resistance of the member and S is the random load effect acting on
the member. X1, X2…are the system random variables. D and L are dead and live loads,
respectively. The reliability index β can be defined as a measure of the probability that g
is less than zero, as follows
Pf = P(g ≤ 0) = Φ (− β ) (9.10)
where Φ is the cumulative distribution function of the standardised Normal distribution.
If g is normally distributed, β is taken as:
355
CHAPTER 9
µg
β= (9.11)
σg
For RC members, the target reliability index is about 3 (Ellingwood et al., 1980).
Allen (1992) suggested that the target reliability index should be increased by 0.25 for
components that fail suddenly with little warning but maintain their post failure capacity.
Since FRP retrofitted members maintain a capacity equal to that of the unstrengthened
member, β was taken as 3.25 in the present analysis.
The design value can be defined as the point on the failure surface closest to the
average point in the space of normalised variables as diagrammatically indicated in
Figure 9.21.
Figure 9.21 Design point and reliability index according to the first order reliability
method for normally distributed uncorrelated variables (Eurocode 2, 2002)
356
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
The design value of load effects, Sd, and resistance, Rd, should be defined such that the
probability of having a more unfavourable value is as follows
P(S > Sd) = Φ (+αSβ) (9.12)
P(R < Rd) = Φ (-αRβ) (9.13)
where αS and αR are the values of the FORM sensitivity factors. α is negative for
unfavourable load effects, and positive for resistances. Eurocode 2 allows independent
assessment of load and resistance if
0.16 < σS / σR < 7.6 (9.14)
where σS and σR are the standard deviations of the load effect and resistance,
respectively. Then, αR can be taken as 0.8 and Equation 9.14 becomes
P(R < Rd) = Φ (-0.8β) (9.15)
This condition was assumed to be satisfactory in the present study. With this
assumption and by reference to Figure 9.21, the capacity reduction factor was calculated
as:
R d µ R − α R βσ R ⎛ σ ⎞ µR
φ= = = ⎜⎜1 − α R β R ⎟⎟ (9.16)
Rn R ⎝ µR ⎠ Rn
µR n
µR
where σR, µR and Rn are the standard deviation, mean and nominal value of resistance,
respectively.
Table 9.7 lists the calculated reduction factors corresponding to the target
reliability index of 3.25. It can be seen from the table that the capacity reduction factor of
0.6 should be used for full composite action strength (no debond) and intermediate span
debond strength. A factor of 0.5 should be used for end debond strength. These values are
close to the lowest factor of 0.6 recommended by AS5100 (Table 9.8)
357
CHAPTER 9
358
DEBOND STRENGTH MODELS AND RELIABLITY ANALYSIS FOR RETROFITTED BEAMS
numerical results. The following steps are recommended to check the capacity of a
retrofitted beam:
1) Calculate the sectional capacity by following these steps
a. Calculate the sectional capacity using the beam theory.
b. Calculate the corresponding shear capacity and compare with applied
shear force using a reduction factor of 0.6, as follows
V* < 0.6 Vu,no debond (9.17)
2) Calculate the intermediate span debond capacity by following these
steps
a. Calculate the maximum force P that the composite can take using
P = βf βL f c b f L e (9.18)
359
CONCLUSIONS
CHAPTER 10 - CONCLUSIONS
A literature review was conducted on the bond behaviour of FRP on concrete. The
review showed that a great deal of experimental work has been done in this area and
different approaches to calculate the bond strength have been proposed. The main failure
modes were identified in the review and the existing prediction models were assessed
based on a large database of test results. It was shown that two failure modes of shear-lap
specimens with properly bonded composites are interfacial debond and shear-tension
failure. It was found that most theoretical models to predict the shear-lap bond strength
are based on an assumed bond-slip relationship. From the assessment study, it is clear that
despite the models being based on different relationships, they generally produce
reasonably accurate predictions. However, the review also highlighted that the validity of
this assumption needs further investigation especially when transverse cracks also exist.
The literature review also revealed that failure of retrofitted RC beams is more
complex with a larger number of failure modes. The two main debonding modes
identified are intermediate span debond and end debond. It was established that the
models proposed to predict debond loads are based on very different approaches, and
because of the complexity of the problem, many assumptions and simplifications are
required in the analyses. As part of the present study, the models were assessed with a
large test database. It was shown that that while some models produce inaccurate and
scattered results, some are not based on the actual mechanisms and therefore lacks
grounding.
361
CHAPTER 10
362
CONCLUSIONS
Finite element modelling was carried out for the shear-lap specimens using
smeared cracks. A fine mesh was generated near the bond line to simulate the bond
behaviour. The finite element model made use of the plane stress elements and a rotating
crack model. A reasonable correlation was achieved between the outcome of the finite
element analyses and the experimental results. The finite element analyses also showed
that the CFRP strain varies significantly along its thickness near the loaded edge, the end
and the transverse cracks. The analyses confirmed that even though the bond-slip curves
can vary along the bond due to the presence of transverse cracks, they appear to have
similar shapes and the bond behaviour can be approximated using a bond-slip relationship
following Popovics’ equation.
A different modelling approach was also used to simulate the behaviour of the
shear-lap specimens and the retrofitted beams under testing. In these analyses, a
combination of discrete and smeared cracks was utilised. The analyses made use of an
interface crack model developed recently, in which concrete degradation in both tension
and shear was modelled. The simulations yielded similar results to those of the smeared
crack models. The simulations also confirmed that the two main debonding modes of
363
CHAPTER 10
retrofitted beams are the result of shearing of concrete along the tensile rebar level in the
shear span and of concrete along the bond surface near the tip of a flexure-shear crack.
A design method was developed based on the findings from the experimental
programs and numerical simulations. The method can be used to predict the capacity of a
retrofitted RC beam failing by one of three modes: flexural failure, intermediate span
debond and end debond. The method is relatively simple. It is based on the well-known
beam theory, which has been verified to be able to predict the FRP strain level at ultimate
quite accurately. The end debond model uses the average shear stress concept, which can
be checked against the average concrete shear strength at the tensile rebar level. The
intermediate span debond model is based on the maximum force that the composite could
take, which can be calculated using an equation modified from that for a shear-lap test. A
simple equation relating the shear strength to the concrete compressive strength was also
proposed. From the verification results of the design method against both the data from
the experimental programs carried out in this research and the large database of tests
available in the literature, it was found that the method is reliable and can be used for
design.
364
CONCLUSIONS
365
REFERENCES
REFERENCES
ACI Committee 363 (1992), State of the art report on high-strength concrete. Detroit,
American Concrete Institute.
ACI Committee 440F (1999), Guidelines for the selection, design and installation of fiber
reinforced polymer (FRP) systems for external strengthening of concrete structures.
Detroit, American Concrete Institute.
ACI Committee 440F (2002), Guide for the design and construction of externally bonded
FRP systems for strengthening concrete structures.
Ahmed, O. and Van Gemert, D. (1999), Effect of longitudinal carbon fiber reinforced
plastic laminates on shear capacity of reinforced concrete beams. Proceedings of the
Fourth International Symposium on Fiber Reinforced Polymer Reinforcement for
Reinforced Concrete Structures, Maryland, USA, pp. 933-943.
Al-Sulaimani, G. J., Sharif, A., Basunbul, I. A., Baluch, M. H. and Ghaleb, B. N. (1994),
“Shear repair for reinforced concrete by fiberglass plate bonding”, ACI Structural
Journal, Vol. 91, No. 4, pp. 458-464.
An, W., Saadatmanesh, H. and Ehsani, M. R. (1991), “RC beams strengthened with FRP
plates. II. Analysis and parametric study”, Journal of Structural Engineering ASCE, Vol.
117, No. 11, pp. 3434-3455.
Arduini, M., Di Tommaso, A. and Nanni, A. (1997), “Brittle failure in FRP plate and
sheet bonded beams”, ACI Structural Journal, Vol. 94, No. 4, pp. 363-370.
Arya, C. and Farmer, N. (2001), Design guidelines for flexural strengthening of concrete
members using FRP composites. FRPRCS-5 - Fibre-reinforced plastics for reinforced
concrete structures, UK, pp. 167-176.
367
REFERENCES
Atadero, R. A., Lee, L. S. and Karbhari, V. M. (2004), Materials variability and reliability
of FRP rehabilitation of concrete. 4th International Conference on Advanced Composite
Materials in Bridges and Structures, Calgary, Canada.
Ayers, S. R. and Van Erp, G. M. (2002), An Australian structural design standard for
fibre composites. ACUN - 4: Composite Systems -Macrocomposites, Microcomposites,
Nanocomposites, pp. 160-170.
Bakht, B., Al-Bazi, G., Banthia, N., Cheung, M., Erki, M.-A., Faoro, M., Machida, A.,
Mufti, A. A., Neale, K. W. and Tadros, G. (2000), “Canadian Bridge Design Code
provisions for fiber-reinforced structures”, Journal of Composites for Construction, Vol.
4, No. 1, pp. 3-15.
Bakis, C. E., Bank, L. C., Brown, V. L., Cosenza, E., Davalos, J. F., Lesko, J. J.,
Machida, A., Rizkalla, S. H. and Triantafillou, T. C. (2002), “Fiber-reinforced polymer
composites for construction - state-of-the-art review”, Journal of Composites for
Construction, Vol. 6, No. 2, pp. 73-87.
Bazant, Z. P. and Oh, B. H. (1983), “Crack band theory for fracture of concrete”, Vol. 16,
No. 93, pp. 155-177.
Bizindavyi, L. and Neale, K. W. (1999), “Transfer lengths and bond strength for
composites bonded to concrete”, Journal of Composites for Construction, Vol. 3, No. 4,
pp. 153-160.
Blaschko, M., Niedermeier, R. and Zilch, K. (1998), Bond Failure Modes of Flexural
Members Strengthened with FRP. Second International Conference on Composites in
Infrastructures, Tucson, Arizona, pp. 450-461.
368
REFERENCES
Brosens, K. and Van Gemert, D. (1997), Anchoring stresses between concrete and carbon
fibre reinforced laminates. Non-Metallic (FRP) Reinforcement for Concrete Structures,
Proc., 3rd Int. Symp, Sapporo.
Camata, G., Spacone, E., Al-Mahaidi, R. and Saouma, V. (2004), “Analysis of test
specimens for cohesive near-bond failure of fiber-reinforced polymer-plated concrete”,
Journal of Composites for Construction, Vol. 8, No. 6, pp. 528-538.
Cha, J. Y., Balaguru, P. and Chung, L. (1999), Experimental and analytical investigation
of partially prestressed concrete beams strengthened with carbon reinforcement.
Proceedings of the Fourth International Symposium on Fiber Reinforced Polymer
Reinforcement for Reinforced Concrete Structures, Maryland, USA, pp. 625-633.
Chajes, M. J., Finch, W. W., Jr., Januszka, T. F. and Thomson, T. A. J. (1996), “Bond and
force transfer of composite material plates bonded to concrete”, ACI Structural Journal,
Vol. 93, No. 2, pp. 208-217.
Chen, J. F. and Teng, J. G. (2001), “Anchorage Strength Models for FRP and Steel Plates
Bonded to Concrete”, Journal of Structural Engineering ASCE, Vol. 127, No. 7, pp. 784-
791.
Chen, J. F., Yang, Z. J., Pan, X. M. and Holt, G. D. (2001), Effect of test methods on
plate-to-concrete bond strength. FRPRCS-5 - Fibre-reinforced plastics for reinforced
concrete structures, UK, pp. 429-438.
Chen, W.-F. and Duan, L. (2000), Bridge Engineering Handbook. Boca Raton, CRC
Press.
369
REFERENCES
Comite Euro-International du Beton (1991), CEB-FIB Model Code 1990. London, Great
Britain, Thomas Telford.
CONTECVET (2002), A validated users manual for assessing the residual service life of
concrete structures - A manual for assessing corrosion-affected concrete structures, EC
Innovation Programme, IN30902I, GEOCISA.
Dai, J. G. and Ueda, T. (2003), Local bond stress slip relations for FRP sheets-concrete
interfaces. FRPRCS-6 - Fibre-reinforced polymer reinforcement for concrete structures,
Singapore, pp. 143-152.
David, E., Djelal, C., Ragneau, E. and Bodin, F. B. (1999), Use of FRP to strengthen and
repair RC beams: experimental study and numerical simulations. Proceedings of the
Eighth International Conference on Advanced Composites for Concrete Repair, London,
UK.
De Lorenzis, L., Miller, B. and Nanni, A. (2001), “Bond of FRP laminates to concrete”,
ACI Materials Journal, Vol. 98, No. 3, pp. 256-264.
Dorr, K. (1980), Ein Beitrag zur Berechnung von Stahlbetonscheiben unter besonderer
Berücksichtigung des Verbundverhaltens, PhD thesis, University of Darmstadt (in
German).
Fanning, P. J. and Kelly, O. (2000), “Smeared crack models of RC beams with externally
bonded CFRP plates”, Journal of Computational Mechanics, Vol. 26, No. 4, pp. 325-
332.
fib Bulletin 14 (2001), Design and use of externally bonded FRP reinforcement for RC
structures.
370
REFERENCES
Fu-quan, X., Jian-guang, G. and Yu, C. (2001), Bond strength between CFRP sheets and
concrete. FRP composites in civil engineering, Hong Kong, Elsevier, pp. 357-363.
Gao, B., Leung, W.-H., Cheung, C.-M., Kim, J.-K. and Leung, C. K. Y. (2001), Effects of
adhesive properties on strengthening of concrete beams with composite strips. FRP
composites in civil engineering, Hong Kong, Elsevier, pp. 423-432.
Gendron, G., Picard, A. and Guerin, M.-C. (1999), “A Theoretical Study on Shear
Strenthening of Reinforced Concrete Beams Using Composite Plates”, Composite
Structures, Vol. 45, pp. 303-309.
Hillerborg, A., Modeer, M. and Petersson, P.-E. (1976), “Analysis of crack formation and
crack growth in concrete by means of fracture mechanics and finite element.s”, Cement &
Concrete Research, Vol. 6, pp. 773-782.
371
REFERENCES
Hussain, M., Sharif, A., Basunbul, I. A., Baluch, M. H. and Al-Sulaimani, G. J. (1995),
“Flexural behaviour of precracked reinforced concrete beams strengthened externally by
steel plates”, ACI Structural Journal, Vol. 92, No. 1, pp. 14-22.
ISIS Canada (2001), Design Manual No. 4 - Strengthening reinforced concrete structures
with externally-bonded FRP.
Izumo, K., Asamizu, T., Saeki, N. and Shimura, K. (1997), “Shear strengthening of PRC
members by fiber sheets”, Transactions of the Japan Concrete Institute, Vol. 19, pp. 105-
112.
Izumo, K., Asamizu, T., Saeki, N. and Shimura, K. (1998), “Bond Behaviour of Aramid
and Carbon Fiber Sheeting”, Transactions of the Japan Concrete Institute, Vol. 20, pp.
269-279.
Jones, R., Swamy, R. N. and Charif, A. (1988), “Plate Separation and Anchorage of
Reinforced Concrete Beams Strengthened By Epoxy-Bonded Steel Plates”, Structural
Engineer, Vol. 66, No. 12, pp. 187-188.
Kanakubo, T., Furuta, T. and Fukuyama, H. (2003), Bond strength between fiber-
reinforced polymer laminates and concrete. FRPRCS-6 - Fibre-reinforced polymer
reinforcement for concrete structures, Singapore, pp. 133-142.
Karbhari, V. M., Engineer, M. and Eckel II, D. A. (1997), “On the durability of
composite rehabilitation schemes for concrete: use of a peel test”, Journal of Materials
Science, Vol. 32, pp. 147-156.
372
REFERENCES
Kishi, N., Mikami, H., Matsuoka, K. G. and KUrihashi, Y. (2001), Failure behavior of
flexural strengthened RC beams with AFRP sheet. FRPRCS-5 - Fibre-reinforced plastics
for reinforced concrete structures, UK, pp. 85-95.
Lee, K. (2003), Shear strength of concrete T-beams repaired using carbon fibre reinforced
polymer, PhD thesis, Department of Civil Engineering, Monash University.
Lee, K., Al-Mahaidi, R., Taplin, G. and Millar, D. (2002), Behaviour of Reinforced
Concrete T-beams Strengthened with Prefabricated L-shaped CFRP Plates. IABSE
symposium, Melbourne.
Lees, J. M., Winistörfer, A. and Meier, U. (2002), “External prestressed carbon fiber-
reinforced polymer straps for shear enhancement of concrete”, Journal of Composites for
Construction, Vol. 6, No. 4, pp. 249-256.
Maeda, T., Asano, Y., Sato, Y., Ueda, T. and Kakuta, Y. (1997), A Study on Bond
Mechanism of Carbon Fiber Sheet. Proc., 3rd Symp. - Non-Metallic (FRP)
Reinforcement for Concrete Struct., Japan, pp. 279-286.
Menetrey, P. and Willam, K. J. (1995), “Triaxial failure criterion for concrete and its
generalization”, ACI Structural Journal, Vol. 92, No. 3, pp. 311-318.
373
REFERENCES
Naaman, A. E., Park, S. Y., Lopez, M. M. and Till, R. D. (2001), Parameters influencing
the flexural response of RC beams strengthened using CFRP sheets. FRPRCS-5 - Fibre-
reinforced plastics for reinforced concrete structures, UK, pp. 117-125.
Nakaba, K., Kanakubo, T., Furuta, T. and Yoshizawa, H. (2001), “Bond Behavior
between Fiber-Reinforced Polymer Laminates and Concrete”, ACI Structural Journal,
Vol. 98, No. 3, pp. 359-367.
Neubauer, U. and Rostasy, F. S. (1999), Bond failure of CFRP plates at inclined cracks -
experiments and fracture mechanics model. Fourth International Symposium - Fiber
Reinforced Polymer Reinforcement for Reinforced Concrete, American Concrete
Institute, pp. 369-381.
Nguyen, D. M., Chan, T. K. and Cheong, H. K. (2001), “Brittle Failure and Bond
Development Length of CFRP-Concrete Beams”, Journal of Composites for
Construction, Vol. 5, No. 1, pp. 12-17.
Niu, H. D. and Wu, Z. (2001a), Peeling-off criterion for FRP-strengthened R/C flexural
members. FRP composites in civil engineering, Hong Kong, Elsevier, pp. 571-578.
Oehlers, D. J. (1992), “Reinforced concrete beams with plates glued to their soffits”,
Journal of Structural Engineering ASCE, Vol. 118, No. 8, pp. 2023-2038.
374
REFERENCES
Raoof, M. and Zhang, S. (1997), An insight into the structural behaviour of reinforced
concrete beams with externally bonded plates. Proceedings of the Institition of Civil
Engineers: Structures and Buildings, pp. 477-492.
Rots, J. G. (1989b), “Crack models for concrete: discrete or smeared? fixed, multi-
directional or rotating”, Heron, Vol. 34, No. 1.
Rots, J. G., Nauta, P., Kusters, G. M. A. and Blaauwendraad, J. (1985), “Smeared crack
approach and fracture localization in concrete”, Heron, Vol. 30, No. 1.
375
REFERENCES
Saafi, M. and Buyle-Bodin, F. (1997), Shear capacity and ductility of reinforced concrete
beams strengthened by plate bonding. Non-Metallic (FRP) Reinforcement for Concrete
Struct. Proc., 3rd Symp., Japan, pp. 351.
Sharif, A., Al-Sulaimani, G. J., Basunbul, I. A., Baluch, M. H. and Ghaleb, B. N. (1994),
“Strengthening of initially loaded reinforced concrete beams using FRP plates”, ACI
Structural Journal, Vol. 91, No. 2, pp. 160-168.
Standards Australia (1991), AS1391-1991: Methods for tensile testing of metals. Sydney,
Australia.
376
REFERENCES
Täljsten, B. (1997), “Defining anchor lengths of steel and CFRP plates bonded to
concrete”, International Journal of Adhesion & Adhesives, Vol. 17, No. 4, pp. 319-327.
Teng, J. G., Smith, G., Yao, J. and Chen, J. F. (2001), Strength model for intermediate
flexural crack induced debonding in RC beams and slabs. FRP composites in civil
engineering, Hong Kong, Elsevier, pp. 579-587.
Toutanji, H. and Ortiz, G. (2001), “The effect of surface preparation on the bond interface
between FRP sheets and concrete members”, Composite Structures, Vol. 53, pp. 457-462.
Tumialan, G., Serra, P., Nanni, A. and Belarbi, A. (1999), Concrete cover delamination in
reinforced concrete beams strengthened with carbon fiber reinforced polymer sheets.
Proceedings of the Fourth International Symposium on Fiber Reinforced Polymer
Reinforcement for Reinforced Concrete Structures, Maryland, USA, pp. 725-735.
Ulaga, T. and Vogel, T. (2003), Bilinear stress-slip bond model: theoretical background
and significance. FRPRCS-6 - Fibre-reinforced polymer reinforcement for concrete
structures, Singapore, pp. 153-162.
377
REFERENCES
Wittmann, F. H. (2002), “Crack formation and fracture energy of normal and high
strength concrete”, Sadhana, Vol. 27, No. 4, pp. 413-423.
Wittmann, F. H., Mihashi, H. and Nomura, N. (1990), “Size effect on fracture energy of
concrete”, Engineering Fracture Mechanics, Vol. 35, No. 1-3, pp. 107-115.
Wu, Z., Yuan, H. and Niu, H. D. (2002), “Stress transfer and fracture propagation on
different kinds of adhesive joints”, Journal of Engineering Mechanics, Vol. 128, No. 5,
pp. 562-573.
Yang, Z. J., Chen, J. F. and Proverb, D. (2002), “Finite element modelling of concrete
cover separation failure in FRP plated RC beams”, Construction and Building Materials,
Vol. 17, pp. 3-13.
378
REFERENCES
Zarnic, R. and Bosiljkov, V. (2001), Behaviour of Beams Strengthened with FRP and
Steel Plates. The 2001 Structural Congress and Exposition, Washington, D.C.
Zarnic, R., Gostic, S., Bosiljkov, V. and Bosiljkov, B. (1999), Improvement of Bending
Load-Bearing Capacity by Externally Bonded Plates. Proc. Creating with Concrete,
London, Thomas Telford, pp. 433-442.
Zhang, J.-P. (1997), “Diagonal cracking and shear strength of reinforced concrete
beams”, Magazine of Concrete Research, Vol. 49, No. 178, pp. 55-65.
Ziraba, Y. N., Baluch, M. H., Basunbul, I. A., Sharif, A. M., Azad, A. K. and Al-
Sulaimani, G. J. (1994), “Guidelines toward the design of reinforced concrete beams with
external plates”, ACI Structural Journal, Vol. 91, No. 6, pp. 639-646.
379