Basagiannis PDF

SEISMIC DESIGN AND EVALUATION
OF MOMENT RESISTING FRAMES

USING ELASTOMERIC DAMPERS
CHRISTOS A. BASAGIANNIS
A thesis submitted to the degree of
Doctor of Philosophy
University of Oxford
2018
Abstract
This study evaluates the characteristics of elastomeric dampers, and assesses their
effectiveness in mitigating the effects of dynamic loading. As part of this evalua-
tion a series of characterization tests were carried out in order to extract the main
mechanical properties of the elastomer material in a range of strain amplitudes, load-
ing frequencies, and ambient temperatures, since it has been proved that elastomeric
materials depend on these parameters. It was found that strain amplitude was the
most dominant factor not only regarding the values of their mechanical properties
(shear storage modulus, and loss factor) but also with regard to the shape of the
hysteresis loops. At the same time frequency was shown to have only a minor effect
on their behaviour.
Based on the Generalised Maxwell Model, a new hysteretic model was developed
and proposed in this thesis which is able to capture the behaviour of the material. As
part of this model an equation was derived which describes the force-displacement
relationship of the Generalised Maxwell Model for N Maxwell elements in time
domain, and formed the basis for the proposed model.
The seismic performance of a 10 storey steel moment resisting frame was evalu-
ated, and the effect of the dampers on the frame’s performance was examined. The
dampers were designed to provide an additional 10% damping to the Eurocode 8
compliant building. The structure was tested with and without dampers for two
different levels of earthquake intensity. It was shown that the elastomeric dampers
led to significant decrease of displacements, accelerations, base shear forces, and per-
manent damage of the structure, even after ground motions scaled to the Maximum
Considered Earthquake level.

Lastly, Real Time Substructure tests were performed in order to further val-
idate the proposed hysteretic model. Both SDOF and MDOF systems equipped
with elastomeric dampers were tested under scaled version of the ElCentro ground
motion and the corresponding force-displacement relationship of the dampers was
compared with the analytical model. The results showed that the proposed model is
able to capture the dynamic performance of the dampers under realistic earthquake
conditions.
ii
Acknowledgements
This study was conducted at the Engineering Department of Oxford University.
I deeply appreciate the guidance and support of my supervisor Professor Martin
Williams during all these years. I have to also thank Robin for his help regarding
not only the actual implementation of the experiments but the background theory
behind them, and Clive for his help at the laboratory. Special thanks to TARRC for
providing us with the elastomeric specimens. Furthermore, I would like to deeply
thank Mark for trusting me and believing in me, Antonis, Eleni, Maria, Marios,
and especially Onelia for their support during all these years, and to everyone that
helped me spend these years in Oxford as creative and happily as possible.
This dissertation is dedicated to the memory of my mother for her love, support,
and patience during my first 23 years of my life. I would also like to thank my father
for his innumerable patience and support, and for his guidance in a lot of aspects of
my life. Lastly, this dissertation is also dedicated to my sister Katerina, and to the
memory of my beloved uncle, Nasos.

Contents
1 Introduction 1
1.1 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Outline of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Literature Review 6
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Passive Energy Dissipation Systems . . . . . . . . . . . . . . . . . . 7
2.2.1 Dynamic Equations of Structures with Dampers . . . . . . . 10
2.2.2 Metallic Dampers . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.3 Friction Dampers . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.4 Viscous Fluid Dampers . . . . . . . . . . . . . . . . . . . . . 19
2.2.5 Tuned Mass Dampers . . . . . . . . . . . . . . . . . . . . . . 22
2.2.6 Tuned Liquid Dampers . . . . . . . . . . . . . . . . . . . . . 24
2.2.7 ViscoElastic Dampers . . . . . . . . . . . . . . . . . . . . . . 26
2.2.8 Elastomeric Dampers . . . . . . . . . . . . . . . . . . . . . . 36
2.2.9 Past applications of VE/Elastomeric Dampers . . . . . . . . 46
2.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3 Characterization Tests 51
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2 Experimental rig . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.3 Mechanical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
iv
4 Hysteretic Constitutive Model Elastomeric Dampers 74
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.2 Generalised Maxwell Model . . . . . . . . . . . . . . . . . . . . . . . 74
4.2.1 General Background . . . . . . . . . . . . . . . . . . . . . . . 74
4.2.2 GMM force-time relationship for N elements - Derivation . . 76
4.3 Hysteretic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.4 Parameter Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.5 Effect of Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5 Seismic Response of Steel Moment Resisting Frames using Elas-
tomeric Dampers 98
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.2 Prototype Building . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.3 Non Linear Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.3.1 Performance Criteria . . . . . . . . . . . . . . . . . . . . . . . 105
5.3.2 Conventional Moment Resisting Frame Model . . . . . . . . . 107
5.3.3 Ground Motions . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.3.4 Retrofitting Procedure . . . . . . . . . . . . . . . . . . . . . . 117
5.3.5 Earthquake Response of steel Conventional Moment Resisting
Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.3.6 Earthquake Response of steel Moment Resisting Frame with
Elastomeric Dampers . . . . . . . . . . . . . . . . . . . . . . 140
5.4 Effect of temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
6 Real Time Substructure Test 164
6.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
6.1.1 Shaking Tables . . . . . . . . . . . . . . . . . . . . . . . . . . 164
6.1.2 Pseudo Dynamic Tests . . . . . . . . . . . . . . . . . . . . . . 165
6.1.3 Effective Force Testing . . . . . . . . . . . . . . . . . . . . . . 167
6.1.4 Pseudo Dynamic Substructure Tests . . . . . . . . . . . . . . 168
v
6.1.5 Real Time Substructure Tests . . . . . . . . . . . . . . . . . . 170
6.2 RTS testing in Oxford . . . . . . . . . . . . . . . . . . . . . . . . . . 171
6.2.1 Time Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
6.2.2 Adaptive Time Series . . . . . . . . . . . . . . . . . . . . . . 178
6.2.3 Modified ATS . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
6.2.4 Implementation of ATS . . . . . . . . . . . . . . . . . . . . . 185
6.2.5 RTS test in SDOF structure . . . . . . . . . . . . . . . . . . . 195
6.2.6 RTS test in MDOF structure . . . . . . . . . . . . . . . . . . 203
6.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
7 Conclusions 209
7.1 Overview of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
7.2 Recommendations for future work . . . . . . . . . . . . . . . . . . . 211
A Mechanical Properties of the Elasomteric Dampers 213
B MGMM validated against sweep amplitude tests in different tem-
peratures 218
vi
List of Figures
2.1 Passive Control Systems[31] . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Active Control Systems[31] . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 SDOF structure with added damper . . . . . . . . . . . . . . . . . . 11
2.4 a: X-shaped plate damper, b: Triangular plate damper . . . . . . . . 16
2.5 Main components of BRB[14] . . . . . . . . . . . . . . . . . . . . . . 17
2.6 a: X-shaped ADAS damper - cyclic response, b: X-shaped ADAS
damper - hysteretic response [14] . . . . . . . . . . . . . . . . . . . . 17
2.7 Friction Dampers incorporated in the form of braced frames[101] . . 19
2.8 Hysteretic behaviour of limited bolted joints using different surface
treatments [102] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.9 Viscous Fluid Damper [11] . . . . . . . . . . . . . . . . . . . . . . . . 21
2.10 Orificed Viscous Fluid Damper [32] . . . . . . . . . . . . . . . . . . . 21
2.11 Force Displacement results for cyclic loading [82] . . . . . . . . . . . 22
2.12 Undamped Absorber and main mass subject to harmonic excitation
[45] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.13 Application of TMD to SDOF structure [31] . . . . . . . . . . . . . . 24
2.14 Configuration for: a) General TMD, b) TLD . . . . . . . . . . . . . 25
2.15 TLD Geometric definition[74] . . . . . . . . . . . . . . . . . . . . . . 26
2.16 Typical VE damper configuration . . . . . . . . . . . . . . . . . . . . 27
2.17 VE stress-strain behaviour under sinusoidal loading [145] . . . . . . . 27
2.18 VE stress-strain behaviour under sinusoidal loading . . . . . . . . . 29
2.19 Generalized Maxwell Model (GMM) presentation . . . . . . . . . . . 33
vii
2.20 Comparison of analytical and experimental values of shear storage
modulus for GMM, Fractional-Derivative models [42] . . . . . . . . . 34
2.21 Comparison of hysteretic behaviour of the 3 models( GMM, and
Fractional-Derivative models) [42] . . . . . . . . . . . . . . . . . . . . 35
2.22 Comparison of analytical and experimental values of roof displace-
ment’ time history of a test structure with diagonal-braced VE
dampers [42] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.23 Different regimes of polymers behaviour when temperature increases
[111] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.24 Typical hysteretic behaviour of elastomer [69] . . . . . . . . . . . . . 38
2.25 Comparison between analytical model proposed by Sause and exper-
imental data [107] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.26 Comparison between analytical model proposed by Lee and experi-
mental data [72] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.27 Comparison between experimental data and Bouc-Wen model [33] . 42
2.28 Comparison between experimental data and analytical model pro-
posed by Karavasilis [62] . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.29 Comparison between experimental data and analytical model pro-
posed by Kasai [64] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.30 Damper Installation at the World Trade Center [31] . . . . . . . . . 47
2.31 Santa Clara County building San Jose - Seismic application [31] . . . 47
2.32 Experimental rig for VE dampers replacing coupling beams [38] . . . 48
2.33 Elastomeric dampers used for seismic retrofit in the Gentile-Fermi
School in Fabriano [8] . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1 Elastomeric Damper provided by TARRC . . . . . . . . . . . . . . . 53
3.2 Elastomeric Damper Dimensions . . . . . . . . . . . . . . . . . . . . 54
3.3 Plan of the final rig: elastomeric damper connected with actuator . . 54
3.4 Final Rig: elastomeric damper connected with actuator . . . . . . . 55
3.5 Typical Displacements’ command - (30% shear strain, 2 Hz) . . . . . 55
3.6 Temperature chamber . . . . . . . . . . . . . . . . . . . . . . . . . . 56
viii
3.7 Typical force-displacement relationship for ED under frequency of
3Hz, strain amplitude of 40%, 20o C temperature . . . . . . . . . . . 57
3.8 ED’s hysteretic loops for 0.25 Hz and 20o C . . . . . . . . . . . . . . 60
3.9 ED’s hysteretic loops for 0.5 Hz and 20o C . . . . . . . . . . . . . . . 61
3.14 ED’s hysteretic loops for 2.0 Hz, 40% shear strain, under 20-35o C . . 66
3.15 Shear storage modulus, G0 , under various temperatures, and frequen-
cies for constant 40% shear strain . . . . . . . . . . . . . . . . . . . . 67
3.16 Loss factor, n, under various temperatures, and frequencies for con-
stant 40% shear strain . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.17 Shear storage modulus, G0 , at 20o C . . . . . . . . . . . . . . . . . . . 68
3.18 Loss factor, n, at 20o C . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.19 Shear storage modulus, G0 , under 25o C . . . . . . . . . . . . . . . . . 69
3.20 Loss factor, n, under 25o C . . . . . . . . . . . . . . . . . . . . . . . . 69
3.25 Comparison of EDs hysteretic Loops for degradation effects due to time 72
4.1 GMM representation in terms of Force-displacement . . . . . . . . . 76
4.2 Sweep amplitude test for frequency of 3 Hz . . . . . . . . . . . . . . 89
4.3 Comparison of force between experiment and MGMM model for
sweep amplitude test - 0.25 Hz, 20o C . . . . . . . . . . . . . . . . . . 91
sweep amplitude test - 1 Hz, 20o C . . . . . . . . . . . . . . . . . . . 91
ix
4.9 Model proposed by Lee [107] compared with 2 Hz sweep amplitude
test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.10 GMM compared with 2 Hz sweep amplitude test . . . . . . . . . . . 93
4.11 Alteration of temperature parameter γT with different temperatures 95
5.1 First 3 modes of the prototype structure . . . . . . . . . . . . . . . . 100
5.2 Elevation of the Prototype Building . . . . . . . . . . . . . . . . . . 102
5.3 Plan of the Prototype Building . . . . . . . . . . . . . . . . . . . . . 102
5.4 Design Spectrum for X axis according to Eurocode 8 . . . . . . . . . 103
5.5 Design Spectrum for Y axis according to Eurocode 8 . . . . . . . . . 103
5.6 Scaling Procedure of ground motions . . . . . . . . . . . . . . . . . . 110
5.7 Comparison of EC8 Spectrum with each of the 20 original ground
motions response spectrum . . . . . . . . . . . . . . . . . . . . . . . 113
5.8 Comparison of EC8 Spectrum with each of the 20 ground motions
scaled to DBE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.9 Comparison of EC8 Spectrum with the mean Spectrum of 20 ground
motions for DBE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.10 Ground motionL: IMPVALLI-I-ELC180 . . . . . . . . . . . . . . . . 115
5.11 Importance factor in terms of probability of exceedance in 50 years . 116
x
5.12 Comparison of EC8 Spectrum with the mean Spectrum of 20 ground
motions for MCE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.13 Dampers Design Procedure . . . . . . . . . . . . . . . . . . . . . . . 120
5.14 Location of Dampers . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.15 Simplified Frame-Braces-Damper representation . . . . . . . . . . . . 123
5.16 Dampers-Bracing geometric configuration [48] . . . . . . . . . . . . . 123
5.17 Storey maximum Displacements for SMRF under the DBE . . . . . 128
5.18 Storey Drift Ratios for SMRF under DBE . . . . . . . . . . . . . . . 128
5.19 Storey residual displacements for SMRF under the DBE . . . . . . . 129
5.20 Storey residual drifts for SMRF under the DBE . . . . . . . . . . . . 129
5.21 Storey maximum accelerations for SMRF under the DBE . . . . . . 130
5.22 Mean Shear forces for SMRF under the DBE . . . . . . . . . . . . . 130
5.23 Ground Motion: T abas − Iran − F ER − L1 scaled to DBE . . . . . 131
5.24 Roof displacement time history under the T abas − Iran − F ER − L1
grounnd motion scaled to DBE . . . . . . . . . . . . . . . . . . . . . 132
5.25 Roof acceleration time history under the T abas − Iran − F ER − L1
ground motion scaled to DBE . . . . . . . . . . . . . . . . . . . . . . 132
5.26 Moment-Rotation relationship for 1st story column of SMRF under
the T abas − Iran − F ER − L1 ground motion scaled to DBE . . . . 133
5.27 Moment-Rotation relationship for 1st story beam of SMRF under the
T abas − Iran − F ER − L1 ground motion scaled to DBE . . . . . . 133
5.28 Storey maximum Displacements for SMRF under the MCE . . . . . 135
5.29 Storey residual displacements for SMRF under the MCE . . . . . . . 135
5.30 Storey Drift Ratios for SMRF under MCE . . . . . . . . . . . . . . . 136
5.31 Storey residual drifts for SMRF under the MCE . . . . . . . . . . . 136
5.32 Storey maximum accelerations for SMRF under the MCE . . . . . . 137
5.33 Mean Storey shear forces for SMRF under the MCE . . . . . . . . . 137
5.34 Roof displacement time history under the T abas − Iran − F ER − L1
grounnd motion scaled to MCE . . . . . . . . . . . . . . . . . . . . . 138
5.35 Roof acceleration time history under the T abas − Iran − F ER − L1
ground motion scaled to MCE . . . . . . . . . . . . . . . . . . . . . . 138
xi
5.36 Moment-Rotation relationship for 1st story column of SMRF under
the T abas − Iran − F ER − L1 ground motion scaled to MCE . . . 139
5.37 Moment-Rotation relationship for 1st story beam of SMRF under the
T abas − Iran − F ER − L1 ground motion scaled to MCE . . . . . . 139
5.38 Storey maximum Displacements for DMRF under the DBE . . . . . 143
5.39 Storey Drift Ratios for DMRF under DBE . . . . . . . . . . . . . . . 143
5.40 Storey residual displacements for DMRF under the DBE . . . . . . . 144
5.41 Storey residual drifts for DMRF under the DBE . . . . . . . . . . . 144
5.42 Storey maximum accelerations for DMRF under the DBE . . . . . . 145
5.43 Mean Storey Shear Forces for DMRF and SMRF under the DBE . . 145
5.44 Top Storey Displacement comparison for the SMRF and DMRF under
DBE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.45 Top Storey Acceleration comparison for the SMRF and DMRF under
DBE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.46 Moment Rotation relationship for base column under theT abasI ran−
F ER − L1 ground motion scaled to DBE . . . . . . . . . . . . . . . 147
5.47 Moment Rotation relationship for base beam under theT abasI ran −
F ER − L1 ground motion scaled to DBE . . . . . . . . . . . . . . . 147
5.48 ED hysteretic response for each floor under DBE . . . . . . . . . . . 148
5.49 Storey maximum Displacements for DMRF under the MCE . . . . . 151
5.50 Storey residual displacements for DMRF under the MCE . . . . . . 151
5.51 Storey Drift Ratios for DMRF under MCE . . . . . . . . . . . . . . 152
5.52 Storey residual drifts for DMRF under the MCE . . . . . . . . . . . 152
5.53 Storey maximum accelerations for DMRF under the MCE . . . . . . 153
5.54 Mean Storey Shear Forces for DMRF and SMRF under the MCE . . 153
5.55 Top Storey Displacement comparison for the SMRF and DMRF under
MCE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
5.56 Top Storey Acceleration comparison for the SMRF and DMRF under
MCE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
5.57 Moment Rotation relationship for base column under theT abasI ran−
F ER − L1 ground motion scaled to MCE . . . . . . . . . . . . . . . 155
xii
5.58 Moment Rotation relationship for base beam under theT abasI ran −
F ER − L1 ground motion scaled to MCE . . . . . . . . . . . . . . . 155
5.59 ED hysteretic response for each floor under MCE . . . . . . . . . . . 156
5.60 Top storey displacement (mm) - 35o C . . . . . . . . . . . . . . . . . 159
5.61 Top storey acceleration (m/sec2 ) - 35o C . . . . . . . . . . . . . . . . 159
5.62 Storey shear forces - 35o C (kN) . . . . . . . . . . . . . . . . . . . . . 160
5.63 Top storey acceleration (m/sec2 ) . . . . . . . . . . . . . . . . . . . . 160
5.64 Top storey acceleration (m/sec2 ) . . . . . . . . . . . . . . . . . . . . 161
5.65 First storey ED’s force-displacement relationship . . . . . . . . . . . 161
6.1 Main concept of PsD substructure’ technique . . . . . . . . . . . . . 170
6.2 Basic steps of PsD substructure’ technique . . . . . . . . . . . . . . . 170
6.3 100 kN actuator used for tests . . . . . . . . . . . . . . . . . . . . . 172
6.4 Schematic overview of OUSDL equipment [15] . . . . . . . . . . . . . 173
6.5 Simplified Overview of Hybrid Procedure . . . . . . . . . . . . . . . 175
6.6 Comparison between input and output displacements of a servo-
hydraulic system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
6.7 a: Actuator delay compensator without feedback, b: Actuator delay
compensator with feedback . . . . . . . . . . . . . . . . . . . . . . . 179
6.8 Comparison of weighting of samples between ATS and modified ATS
compensators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
6.9 Schematic overview of the modified ATS delay compensator . . . . . 185
6.10 User defined time history a: Comparison of Instron controller dis-
placement with encoder’s displacement, b:Corresponding rig deflection 187
6.11 Comparison of target displacement displacement with encoder’s dis-
placement without modified ATS compensator . . . . . . . . . . . . 191
placement with modified ATS compensator . . . . . . . . . . . . . . 192
placement - Sychronization plots . . . . . . . . . . . . . . . . . . . . 193
6.14 Time variation of αi coefficients . . . . . . . . . . . . . . . . . . . . . 194
xiii
6.15 Physical substructure for RTS testing . . . . . . . . . . . . . . . . . 195
6.16 Physical and numerical part of the RTS test . . . . . . . . . . . . . . 195
6.17 Schematic overview of the RTS process . . . . . . . . . . . . . . . . . 196
6.18 Implementation of equation of motion for SDOF structure . . . . . . 197
6.19 ElCentro ground acceleration . . . . . . . . . . . . . . . . . . . . . . 199
6.20 Comparison of target and measured displacements during RTS test
for SDOF frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
6.21 Comparison of target and measured displacement during RTS test for
SDOF frame - Sychronization plot . . . . . . . . . . . . . . . . . . . 200
6.22 Time variation of the αi coefficients of the modified ATS compensator
during the scaled ElCentro Earthquake for SDOF frame . . . . . . . 201
6.23 Comparison of MGMM and RTS force under scaled ElCentro Earth-
quake for SDOF frame . . . . . . . . . . . . . . . . . . . . . . . . . . 202
6.24 Comparison of target and measured displacements during RTS test
for MDOF frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
6.25 Comparison of target and measured displacement during RTS test for
MDOF frame - Sychronization plot . . . . . . . . . . . . . . . . . . . 205
6.26 Time variation of the αi coefficients of the modified ATS compensator
during the scaled ElCentro Earthquake for MDOF frame . . . . . . . 206
6.27 Comparison of MGMM and RTS force under scaled ElCentro Earth-
quake for MDOF frame . . . . . . . . . . . . . . . . . . . . . . . . . 207
B.1 Comparison of force between experiment and MGMM model for
xiv

Chapter 1
Introduction
The basic concept of conventional earthquake-resistant design aims to provide an
acceptable structural performance level during the earthquakes. This level is usually
dictated by the importance of the building and by the local authorities. The most
common design philosophy is a non-collapse requirement under strong and rare
earthquakes, while structures are also expected to experience no damage under more
frequent and weaker earthquakes. However, the socio-economic reality, especially
in developed countries, has pushed the barrier considerable higher [28], leading to
design of damage-free or minimal-damage structures. Important structures such as
hospitals, schools, and museums are therefore desired to exhibit a more advanced
level of earthquake protection, in order to be functional even after a strong seismic
event.
A variety of approaches of earthquake protection has been proposed, including
base isolation, passive dissipative devices, and active or semi-active control devices.
Dissipative devices aim to absorb a portion of the seismic energy input to a structure,
which would otherwise be dissipated through mechanisms corresponding directly to
structural damage. The most promising and feasible approach is the use of passive
dissipative devices, which are likely to be most reliable in an extreme event, being
activated by the movements of the main structural system without the need of any
external power source.
The focus of this research is on passive dissipative devices, and more specifically
on elastomeric dampers. Among the various forms of passive dissipative devices,
1
Introduction
there has been much focus on fluid viscous, and metallic dampers, but elastomers
offer a promising alternative and have been under-investigated. Their main advan-
tages compared to other types of dissipative devices include their contribution of
both stiffness and damping to the structure regardless the type of external loading,
their low dependence on loading frequency compared to similar type of visco-elastic
dampers, and their full recovery after the removal of the source which created the
deformation.
The most common scenario is to apply the dampers between the connection of
steel diagonal braces and the middle length of the upper floor beam. This method
can be either designed in a new enhanced building, or as a retrofitting procedure
of an existing one. Due to the elastomer’s dependence on strain amplitude, loading
frequency, and ambient temperature, it is very difficult to accurately model it and
capture its dynamic characteristics during dynamic loading. Hence, the conven-
tional way of modelling this type of damper is to assign an equivalent Kelvin-Voigt
model with stiffness and damping values corresponding to the natural frequency of
the structure and an accepted level of lateral drift. A new hysteretic model has
been proposed as part of this research, based on sweep amplitude and frequency
tests which has shown that it can accurately model the dampers behaviour for a
specific range of strain amplitudes, loading frequencies, and ambient frequencies. It
was found that the dominant factor which mainly controls not only their mechanical
properties but also the shape of the hysteresis loop was the strain amplitude. How-
ever, both loading frequency and ambient temperature also affected the dampers’
capacity to dissipate energy.
1.1 Research Objectives
Elastomeric dampers provide a promising means of earthquake protection. However,
the application of this type of dampers as a mean of providing relatively uniform
distribution of stiffness and damping throughout the height of buildings has not been
widely used. In fact most of the elastomer applications in the earthquake community
until now have been focused on dissipating energy through base isolation systems.
2
Seismic design and evaluation of steel moment resisting frames using elastomeric dampers
Moreover, a lot of models which target to capture the dynamic characteristics of
the elastomeric dampers have been proposed and presented here 2.2.8. However,
lack of clear understanding of elastomer properties, especially how they vary with
frequency, amplitude, and time, and limited experimental validation of their per-
formance still exists. Most of the proposed models are either based on linearised
properties of the material or parameter optimization based on minimization of ana-
lytical formulas of shear storage modulus and the corresponding experimental values.
Even though these models can effectively capture the elastomer’s behaviour under
specific conditions (specific values of temperature, frequency, and limited range of
strain amplitudes), discrepancies may exist when the dampers function under a wider
range of these parameters. Furthermore, even though parameters of the analytical
models are optimised based on shear storage minimization techniques (analytical
expressions of shear storage modulus are extracted in frequency domain for specific
values of sinusoidal input loading and do not take into account any strain amplitude
alterations), this does not necessarily mean that can capture the actual hysteresis
behaviour of the dampers. The main innovative part of this research was to experi-
mentally evaluate elastomer’s characteristics, and define a new model which would
be able to describe the behaviour of the dampers due to dynamic loading. Therefore,
the main objectives of this research are:
• Evaluate the mechanical properties of the material
• Develop an analytical model for the dampers which would be adequate for non
linear analyses of structures under earthquake loading
• Demonstrate through non-linear structural analysis the structural seismic per-
formance enhancement that can be achieved through retrofitting with elas-
tomeric dampers
• Validation of the proposed Adaptive Time Series method as a new time delay
minimization technique applying it to elastomeric dampers
• Further validation of the proposed hysteretic model of the elastomeric dampers
under realistic conditions using Real Time Substructure tests
3
Introduction
1.2 Outline of Thesis
The remaining chapters of the thesis are organised as follows:
• Chapter 2 introduces the reader to the main concept of passive energy dissi-
pation and earthquake protection systems, the different types of dampers used
in the past, while at the same time reviewing the previous work carried out by
other researchers. At the end of this chapter the most widely used approaches
of modelling elastomeric dampers are described.
• Chapter 3 describes the experiments which were carried out in order to extract
the main mechanical characteristics of the material. Since the dampers depend
on strain amplitude, loading frequency, and ambient temperature, a series
of cyclic tests was carried out in a range of values for the aforementioned
parameters.
• Chapter 4 presents the proposed hysteretic model which is based on sweep
amplitude tests under different loading frequencies and ambient temperatures.
The first part of this chapter focuses on developing a force-displacement model
in time domain for the well known Generalised Maxwell Model, for N Maxwell
elements. This constitutes the basis of the proposed model described later in
the same chapter.
• Chapter 5 describes a design process of a 10 storey steel moment resisting frame
with elastomeric dampers, which targets an enhanced performance level. The
seismic response of this building with and without the dampers is evaluated
based on non linear time history analyses. The elastomeric dampers were
analytically modelled based on the findings of Chapter 4.
• Chapter 6 boosts our confidence with regard to the validation and the use of
the hysteretic model of Chapter 4, by carrying out Real Time Substructure
tests for both SDOF and MDOF simple steel frames equipped with elastomeric
dampers. In order to minimise the time delay which has been proved to cause
instability to the system, a new method proposed by Malloy and Blakeborough
4
(unpublished until now) was used.
• Chapter 7 summarises the findings and conclusions of this research and sug-
gests directions for future work.
5
Chapter 2
Literature Review
2.1 Introduction
Conventional seismic design is based on linear analysis and dictates that the perfor-
mance of structures during earthquakes depends on the lateral force resisting system
being able to absorb and dissipate energy for a number of cycles[31]. This design
concept therefore requires a combination of strength and ductility in order to provide
resistance to lateral loads [31, 114]. For major earthquakes, the structural design
engineer relies upon the inherent ductility of conscientiously detailed buildings to
prevent catastrophic results [114]. Energy dissipation is designed to occur in spe-
cially designed ductile plastic hinge regions of beams and columns bases. This design
approach has been considered acceptable because of economic considerations pro-
vided, of course, that structural collapse is prevented and life safety is ensured [31].
However, under large earthquakes the damage may be irrepairable, with potential
catastrophic and uneconomical results. Moreover, this conventional seismic philos-
ophy cannot easily be applied to all types of structures. New structures are often
designed to more demanding performance criteria, which may require the structure
to remain linear or to have minimal damage. At the same time there are numerous
existing structures with insufficient strength or ductility, which need strengthening
in order to meet current seismic demands.
A lot of different approaches have been developed in order to improve the earth-
quake response performance and damage control of buildings, bridges, and other
6
structures, and can be divided into four groups: base isolation, a technique which
is based on a separation of the main structure from its foundation incorporating
flexible bearings or pads; passive systems, such as supplemental energy dissipation
devices; active systems, which require the active participation of mechanical devices
whose characteristics are made to change during the building response on the basis
of current response measurements; and hybrid systems, which combine the passive
and active systems in a manner so that safety of the building is not compromised
even if the active system fails [48].
With regard to passive dissipative devices, much focus has been given to viscous
and metallic dampers; however elastomeric dampers, which could form an alter-
native solution regarding the seismic protection of buildings, have received much
less attention. Their characteristics alter in time when different strain amplitudes,
loading frequencies, and ambient temperatures are applied, making the proposal of
a hysteretic model a very difficult and challenging task. Furthermore, apart from
lack of clear understanding on how elastomers’ properties vary with time, limited
experimental validation of their performance is another limiting factor. Finally, com-
bining these limitations with the fact that the majority of elastomeric applications
and proposed models has been focused on base isolation methods creates the need
of further analytical evaluation and experimental validation of elastomeric dampers
as passive dissipative devices being incorporated throughout the height of a building
as uniformly as possible.
This section presents general background information and fundamental principles
of the aforementioned earthquake protection systems. A more detailed insight into
passive energy dissipation devices is given, and much more attention is paid to the
behaviour of the viscoelastic and elastomeric materials which are used as dampers.
2.2 Passive Energy Dissipation Systems
The four groups mentioned earlier, are summarised in table 2.1. Each one of these
approaches is different and may be applied based on the needs of the structure, either
if the main target is to achieve enhanced seismic performance or to strengthen and
7
Literature Review
retrofit an already existing structure to cope under modern anti-seismic codes.
Base isolation involves placing isolation devices at the foundation of a structure,
with the aim of protecting it from the damaging effects of earthquakes by uncoupling
it from the ground [65]. The main philosophy is to provide a layer of low lateral
stiffness between the foundation and the superstructure. This layer causes the nat-
ural period to highly increase, resulting in reduced input seismic loads, along with a
large displacement. However, this displacement is concentrated at the base isolation
system, while the structure’s internal deformations remain relatively low. Further-
more, modern seismic isolation systems incorporate energy dissipation mechanisms.
Examples are high damping elastomeric bearings, lead plugs in elastomeric bear-
ings, mild steel dampers, fluid viscous dampers, and friction in sliding bearings[31].
Finally, it is worth mentioning that the inherent damping of the base isolation mech-
anism and the corresponding hysteretic energy are of minor importance regarding
the reduced earthquake structural response. Base isolation is mainly so effective due
to the lengthening of the natural period [27].
Table 2.1: Structural Protective Systems

Seismic Isolation Passive Energy Dissipation Semi-Active and Active Control
Elastomeric Bearings Metallic Dampers Active Bracing Systems
Friction Dampers Active Mass Dampers
Viscoelastic Dampers Variable Stiffness
Lead Rubber Bearings Elastomeric Dampers or Damping Systems
Viscous Fluid Dampers Smart Materials
Tuned Mass Dampers
Sliding Friction Pendulum Tuned Liquid Dampers
Figure 2.1: Passive Control Systems[31]
8
Figure 2.2: Active Control Systems[31]
The distinction between active and passive control systems is shown in figs. 2.1
and 2.2. Passive control systems (energy dissipation devices, dynamic vibration ab-
sorbers) and seismic isolation systems develop motion control forces at the points
of attachment of the system. The power needed to generate these forces stems
from the motion of the points of attachment during the loading history. Contrary
to semi-active and active control systems, there is no need for external supply of
power[114]. Moreover, an essential advantage over base isolation systems is their
proven efficiency against both wind and earthquake forces[114]. In case of active
control systems, the motion control forces are determined by a controller based on
feedback information from the measured response, which is carried out by sensors, of
the structure and/or feedforward information from the external excitation[115]. The
recorded measurements from the response and/or excitation are monitored by a con-
troller (a computer) which, based on a pre-determined control algorithm, determines
the appropriate control signal for operation of the actuators [122]. Hybrid or semi-
active control systems on the other hand, unlike fully active systems, cannot add
energy to the system. They originate from passive control systems which have been
subsequently modified to allow for the adjustment of mechanical properties[122], re-
quiring only nominal amounts of external energy. For more information and actual
implementation for base isolation systems see Kelly [65], and for semi-active and
active control systems at Symans and Constantinou [122, 31], Soong[115].
The focus of this research is on passive energy dissipation devices, and more
specifically on elastomeric dampers. Their relative simplicity with regard to incor-
poration into buildings and their modelling into Finite Element softwares, alongside
9
Literature Review
the fact that passive devices in contrast with active/semi active systems do not
require any external power supply (which may become a critical factor during a
strong earthquake), and their 40 years of history of research and structural applica-
tions make them more relatively reliable and preferable. Hence, a brief introduction
about the main types of dampers, and a more comprehensive and detailed descrip-
tion of ViscoElastic and Elastomeric dampers will be presented in this section.
2.2.1 Dynamic Equations of Structures with Dampers
Consider the lateral motion of a Single Degree Of Freedom (SDOF) system, consist-
ing of mass, m, stiffness, k, and linear damper with viscous damping coefficient, c
(see Figure 2.3):
mẍ + cẋ + kx = −mẍg (2.1)
where x and xg represent the time-dependent structure’s displacement of the mass
relative to the ground, and ground displacement respectively. It can also be written
in the following form:
ẍ + 2ζω ẋ + ω 2 x = −ẍg (2.2)
Consider now the addition of a generic damper device (fig. 2.3), which may provide
not only supplemental damping, but also stiffness, and perhaps mass. Hence, the
symbol Γ corresponds for the whole contribution of the dampers to the SDOF system
[114]:
mẍ + cẋ + kx + Γ = −(m + m̄)ẍg (2.3)
where m represents the mass of the added element and Γ = k̄x + c̄ẋ + m̄ẍ. It can
be seen that adding a damper to the original structure leads to a new more stiff,
more damped, and heavier structure. This tends to reduce the seismic response of
the structure, even though this is not always the case, since it should be treated like
a new structure with different characteristics, instead of an improved version of the
10
prototype. However, if Γ has only damping contributions, then the SDOF response
will be always reduced, since it will be equivalent to the initial system with increased
damping. For all structures, careful analysis must be performed in order to ensure
the beneficial effects of the added elements.
Figure 2.3: SDOF structure with added damper
In the case of Multi Degree of Freedom (MDOF) systems, Eq. 2.3 is adjusted
to:
M ẍ + C ẋ + Kx + Γ = −(M + M̄ )ẍg (2.4)
where M , C, and K are the mass, damping, and stiffness matrices of the MDOF
structure, and Γ symbolizes a matrix of operators, which represent the added damper
to the MDOF system. Since, this section covers only the basic principles, when
added dampers are being used, the assumption that the damper’s behaviour is a
linear combination of mass, damping, and stiffness is made:
Γ = K̄x + C̄ ẋ + M̄ ẍ (2.5)
Then, Eq. 2.4 is rewritten as:
M̂ ẍ + Ĉ ẋ + K̂x = −M̂ ẍg (2.6)
where:
M̂ = M + M̄ (2.7a)
11
Literature Review
Ĉ = C + C̄ (2.7b)
K̂ = K + K̄ (2.7c)
Now, Eq. 2.6 is in the form of classical matrix dynamic analysis problem. Assuming
that all the matrix coefficients are constant, Eq. 2.6 represents a set of N linear sec-
ond order ordinary differential equations with constant coefficients. Its solution can
be computationally demanding, unless it can be uncoupled. This can be achieved, if
M̂ , Ĉ, and K̂ are diagonalized. This can not be done for arbitrary matrices M̂ , Ĉ,
and K̂. However, under certain restrictions on the damping matrix, Ĉ, the transfor-
mation to modal coordinates can be achieved as shown below[114, 31]. Considering
the generalized eigenvalue problem of the MDOF system:
(K̂ − ω02 M̂ )φ = 0 (2.8)
where ω0 represents the undamped natural frequency of the MDOF structure, in-
cluding the added dampers, and φ is the associated mode shape vector. The MDOF
system consists of N natural frequencies, and mode shapes, ω0i , and φi respectively.
Also, the mode shapes satisfy the following orthogonality conditions:


 1 for i = j

φTi M̂ φj = (2.9a)
 0 for i 6= j


 ω 2 for i = j

0i
φTi K̂φj = (2.9b)
0 for i 6= j


Eq. 2.9 provides the basis for transformation which can be applied to the whole
MDOF system. The corresponding mode shape vectors are the columns of the
N xN matrix, Φ. Hence, the relative displacements vector, x, can be rewritten:
PN
x= r=1 φr qr (t) = Φq(t) = Φq (2.10)
12
where q is the vector of modal coordinates. Combining, Eqs. 2.6, 2.9, and 2.10 the
equation of motion for the MDOF system can be written as:
ΦT M̂ Φq̈ + ΦT ĈΦq̇ + ΦT K̂Φq = −ΦT M̂ eẍg

(2.11)
M̂q̈ + Ĉq̇ + K̂q = −Ẍg
where e is a column vector full of ones. Since, M̂, and K̂ are diagonal matrices of
the generalized modal masses and stiffness respectively, Eq. 2.11 can be uncoupled
only if ΦT ĈΦ is also diagonal. Assuming proportional (Rayleigh) damping:
Ĉ = α0 M̂ + α1 K̂ (2.12)
The damping ratio for a system like this is[27]:
α0 1 α1 (2.13)
ζi = 2 ωi + 2 ωi
Similar rules apply when Caughey series are used [27], instead of the Rayleigh damp-
ing. Now the equations are uncoupled, and can be written in the following form for
each mode, i:
2q =− φTi M̂ e
q̈i + 2ζi ω0i q̇ + ω0i i ẍg (2.14)
φT
i M̂ φi
It can be seen that is similar equation to that of the SDOF system. The generalized
displacement, and therefore the relative displacement, x, can be determined for each
time step.
In reality, damping is not necessarily proportional; however for the low levels of
damping encountered in most structures, the proportional damping approximation is
accurate enough while considerably simplifying analysis. When damping is relatively
high, especially if added dissipative devices are used, or for structures exhibiting non
linearities, the errors incurred by this assumption become significant, and modal
superposition analysis based on Eq. 2.14 is potentially unreliable. The direct time
integration approach does not rely on any assumption about the form of the damping
matrix, and therefore is a better way to go. Using the implicit Newmark algorithm
13
Literature Review
[13, 27] a brief presentation of the time domain analysis follows. According to this
algorithm:
∆t
ẋi+1 = ẋi + (ẍi+1 + ẍi ) (2.15a)
2
∆t2 ∆t
xi+1 = xi + ẋi ∆t + (ẍi+1 + ẍi ) = xi + (ẋi + ẋi+1 ) (2.15b)
4 2
For every time step, Eq. 2.6 takes the following form:
M̂ ẍi + Ĉi ẋi + K̂i xi = −M̂ ẍg i (2.16)
Assuming, that the response of the structure at time step, i, is already known, the
aim is to obtain the MDOF system response for the i + 1 time step:
M̂ ẍi+1 + Ĉi+1 ẋi+1 + K̂i+1 xi+1 = −M̂ ẍg i+1 (2.17)
Eq. 2.15 can be rewritten:
∆t 2
ẍi+1 = (ẋi+1 − ẋi − ẍi ) (2.18a)
2 ∆t
∆t 2
ẋi+1 = (xi+1 − xi − ẋi ) (2.18b)
2 ∆t
Substituting Eq. 2.18 into 2.17:
tot x
Ki+1 tot (2.19)
i+1 = Fi+1
where
tot = K̂ 2 4 (2.20)
Ki+1 i+1 + ∆t Ĉi+1 + ∆t2
M̂
and
4M̂ 2Ĉi+1
tot = −M̂ ẍ
Fi+1 g,i+1 + [ ∆t2 + ∆t ]xi + [ 4∆t
M̂
+ Ĉi+1 ]ẋi + M̂ ẍi (2.21)
14
Finally, using iteration algorithms such as the modified Newton-Raphson method
[27] the displacement at time step i+1 can be determined, provided that the criteria
defined by the algorithm are satisfied. Usually , these criteria are based on the
convergence between xi+1

n and xi+1
n+1 , where n represents the number of iteration.
We can see that Eq. 2.21 can be simplified a lot, if the matrices K, and C are
constant during the whole time history analysis, as in the case of linear systems
with Rayleigh damping. In this case, K tot from Eq. 2.20 can be calculated only
once, at the beginning of the analysis.
2.2.2 Metallic Dampers
The concept of using metallic yielding dampers within a structure to absorb seismic
energy was first proposed by Kelly [66], and Skinner [113], considering torsional
beams, flexural beams, and other devices. Later research included the effect of more
efficient devices, such as the X-shaped and triangular-plate energy dissipators [130],
as shown in Figure 2.4. Other devices have the ability to dissipate energy either using
different materials such as lead and shape memory alloys [10], or even a complete
different configuration [131, 130, 139]. A different approach to the above type of
dampers is followed in case of the Buckling Restrained Braces (BRBs)[14], where
a yielding steel core is encased in a concrete-filled steel tube, avoiding buckling
phenomena, and targeting to a tension/compression steel brace (Figure 2.5). An
additional coating is provided at the steel core in order to minimize the friction with
the concrete.
Due to the detailed past and ongoing research a number of existing structures in
New Zealand, Mexico, Japan, Italy, and United States now include metallic dampers
[114]. They are usually added within a bay between chevron braces and the overlying
beam. The main concept behind metallic dampers is that they resist horizontal
forces via flexural deformation of all the steel plates combined. After a certain level
of force, the plates start to yield resulting in hysteretic behaviour. Even though most
metallic dampers are modelled as elastic perfectly-plastic materials or Ramsberg-
Osgood [104] idealizations, more rigorous models have been suggested, such as that
suggested by Ozdemir [46]. It is worth mentioning that even though that addition
15
Literature Review
of metallic dampers may be effective regarding the reduction of inter-storey drifts,
they are less effective at reducing accelerations and base shear forces [86].
Since the primary contribution of metallic dampers is additional stiffness to the
main structure, this may lead to large dynamic characteristics’ alterations of the
building (change of natural frequencies, mode shapes, and mass). Thus, this does
not guarantee an enhanced performance due to seismic loading. In some instances,
the structure may experience larger deformations, accelerations, and shear forces,
based on the frequency content of the imposed earthquake, compared to the original
unretrofitted building.
Whittaker et al. [14] carried out experimental evaluation of X-shaped steel
dampers (ADAS) under both cyclic and time history displacements, which were
determined from a scaled version of the ElCentro ground motion. The experimen-
tal rig consisted of a 3-storey, 1-bay moment resisting frame and the corresponding
results regarding the cyclic and the hysteretic response of the dampers are shown
at Figure 2.6. This Figure represents a typical hysteretic behaviour of metallic
dampers.
(a) X-shaped plate damper[48] (b) Triangular plate damper[48]
Figure 2.4: a: X-shaped plate damper, b: Triangular plate damper
16
Figure 2.5: Main components of BRB[14]
(a) Cyclic response (b) Hysteretic response
Figure 2.6: a: X-shaped ADAS damper - cyclic response, b: X-shaped ADAS damper
- hysteretic response [14]
2.2.3 Friction Dampers
Friction dampers have the ability to dissipate energy through friction developed
between two solid bodies sliding relative to each other. Several different types of
friction dampers have been developed [101, 102, 10, 98, 44]through the years.
Generally, friction dampers generate rectangular hysteretic loops, similar to the
characteristics of Coulomb friction, based on the following assumptions:
• The total frictional force that can be developed is independent of the apparent
surface area of contact
17
Literature Review
• The total frictional force that can be developed is proportional to the normal
force acting across the interface
• For the case of sliding with low relative velocities, the total frictional force is
independent of that velocity
The relationship between the frictional and the normal force is given as:
Ft = µFn (2.22)
where F t , and F n represent the friction and the normal force, respectively, and
µ the friction coefficient. Even though frictional forces are simple to measure or
calculate based on Eq. 2.22, the actual phenomena that friction action is based on
are much more complex. These main phenomena are discussed by Christopoulos and
Filiatrault [28], concluding that the main factors which affect the frictional forces
are the shape and contour of surfaces, the way asperities on the surfaces deform
when normal pressure is applied, how they adhere, the role of surface films, and how
energy is lost when the surfaces are deformed during sliding.
One of the earliest experimental efforts was carried out by A.S. Pall [102], who
conducted both static and dynamic testing on sliding elements, with different con-
figurations, incorporated in the form of braced frames (Figure 2.7). The tests were
based on the normal force applied by pre-tensioning 12.7 mm diameter high strength
bolts. Best results were determined when brake lining pads were used, in order to
limit the available slip deformation. A nearly elasto-plastic behaviour with negli-
gible degradation was observed for any sliding surface used (Figure 2.8), which is
a common friction damper behaviour. The friction dampers are designed not to
slip under normal service loads and moderate earthquakes. During strong ground
motions, however, the device slips at a predetermined load. It should be noted
that visco-plastic models have been proposed to model friction phenomena as well,
similarly to modelling of metallic dampers.
However, even though friction dampers can effectively dissipate hysteretic energy,
they are not very widely used. It has been shown [98] that addition of friction
dampers does not reduce structural response in the low frequency range. Moreover,
18
the very sharp non-linearities which friction dampers exhibit, the need for regular
maintenance, and the need for a high clamping force across the device are prohibiting
factors of friction dampers use.
Figure 2.7: Friction Dampers incorporated in the form of braced frames[101]
Figure 2.8: Hysteretic behaviour of limited bolted joints using different surface treat-
ments [102]
2.2.4 Viscous Fluid Dampers
Fluid dampers can also be very effective under seismic loads [54, 87, 12, 28]. An
effective concept, which was one of the first applications designed for seismic protec-
19
Literature Review
tion by Sumitomo Construction Company [11], consists of a viscous damping wall
(Figure 2.9), where a piston (steel plate) is constrained to move within a rectangular
steel container filled with viscous fluid. In order for this concept to be applied to a
real structure, the piston is attached at the overlying floor, while the container is at
the floor below. The energy dissipation is accomplished through relative inter storey
drifts. Adding a sufficient number of viscous damping walls within the structure,
will lead to a significant increase in damping.
In order to increase the energy dissipation capacity of the viscous fluid dampers,
another more efficient approach can be followed, using orificed fluid dampers incor-
porated within the bracing system of the structure(Figure 2.10) [32]. However, in
this case a high level of sophisticated analytical and detailed experimental work has
to be carried out before implementation to real structures. Makris and Constanti-
nou et al. [83, 82] carried out experimental tests on viscous walls, and found that
the following fractional derivative Maxwell force-displacement model was in a very
good agreement with the data they obtained:
v
P (t) + λ d dtPv(t) = C0 dx(t)
dt
(2.23)
where P is the applied force to the piston, x is the corresponding displacement. C0 ,
λ, and v are parameters that are either determined based on the experimental data.
Typical hysteretic loops of viscous dampers are presented at Figure 2.11. Another
modelling approach taking into account any non linearities is based on the following
equation:
P (t) = Cx(t)α (2.24)
Even though in reality viscous dampers exhibit non-linearities, they are frequently
treated as linear. For example Chang et. al. [26] carried out experimental tests on
viscous dampers. Modelling them using Eq. 2.24, he determined that a nonlinear
coefficient in the range of 0.48-0.55 was more suitable to capture the behaviour of
the viscous dampers. Furthermore, orificed viscous dampers tested by Constantinou
and Symans[32] were found to be less dependent on loading history and the proposed
20
linear analytical model was simplified to:
P (t) + λ dPdt(t) = C0 dx(t)

dt
(2.25)
Figure 2.9: Viscous Fluid Damper [11]
Figure 2.10: Orificed Viscous Fluid Damper [32]
21
Literature Review
Figure 2.11: Force Displacement results for cyclic loading [82]
2.2.5 Tuned Mass Dampers
The concept of the implementation of Tuned Mass Dampers (TMDs) is based on
the early studies carried out by Fram [45], and Den Hartog [34]. Figure 2.12 shows
Frahm’s absorber, which consists of the main mass connected with an additional
smaller mass through a spring with stiffness k. The main target was to select
appropriate values of k, and m, in order for the natural frequency of the added
absorber to be close to the excitation frequency. The result was that the main mass,
M , can stay completely stationary under a harmonic load. Despite its effectiveness,
this method was mostly developed for mechanical engineering systems. In the case of
buildings, where loads consist of many components of frequency, Figure 2.13 shows
the main concepts in simple form. The equations of motion describing the behaviour
of the SDOF structure along with the added TMD are:
mz̈(t) + cż(t) + kz(t) = −mÿ1 (t) + g(t) (2.26)
M ÿ1 (t) + C ẏ1 (t) + Ky1 (t) = cż(t) + kz(t) + f (t) (2.27)
where y1 (t) is the displacement of the SDOF system relative to its base, z(t) is
the relative displacement of the TMD with respect to the SDOF system, f (t) is
22
the external force of the main system, and g(t) the associated force applied on
m
TMD, which is zero in case of wind load, and M f (t) in case of earthquake loading.
Summation of Eqs. 2.26, and 2.27 leads to the final equation of motion for the total
system:
(M + m)ÿ1 (t) + C ẏ1 (t) + Ky1 (t) = −mz̈(t) + f (t) + g(t) (2.28)
Apart from the obvious decrease in natural frequency, and a slight increase in the
force in case of earthquake loading, the added inertia force of the TMD, −mz̈(t),
is the reason behind the TMD’s effectiveness. Maximum effectiveness of the added
TMD comes when the secondary mass lags the main SDOF mass by a phase angle
of 900 . This is as if every time the main mass trying to move towards one direction
another force pushes it back to its initial position. Optimizing techniques for both
the secondary mass and damping ratio have been developed by many researchers
[34, 105, 135, 71], which lead to the final design of the TMD. Recent versions of
TMDs are rested on top of elastomeric rubber bearings, which function as shear
springs, and rubber compound elements, which provide viscoelastic damping capa-
bility. The effectiveness of the TMDs is optimised based on the natural frequency
of the building. Hence, it is worth mentioning that inelastic deformations of the
structure may lead to a decrease of this effectiveness, since the TMD loses its tuning
effect. This fact makes TMDs use inefficient for seismic applications.
Figure 2.12: Undamped Absorber and main mass subject to harmonic excitation
[45]
23
Literature Review
Figure 2.13: Application of TMD to SDOF structure [31]
2.2.6 Tuned Liquid Dampers
The concept of adding Tuned Liquid Dampers (TLDs) in order to increase the
dynamic performance of a structure is similar to the TMD case, discussed earlier.
In the case of the TLD, a mass of water replaces the solid mass (Figure 2.14).
However, unlike TMDs, which often behave linearly, the behaviour of the TLDs is
governed by non linearities, either liquid sloshing or the presence of orifices[114].
Hence, most of the work, which is based on characterizing the dynamic response of
the TLDs, is based on experiments[31]. Assuming moderate amplitude sloshing, no
wave breaking shallow water gravity waves theory can be applied [90]. Lepelletier
and Raichlen [74] studied the response of a fluid layer within a rectangular tank,
due to horizontal excitation(Figure 2.15).
The motion of the fluid is governed by the wave elevation, n(x, t), and the depth-
averaged horizontal velocity, u(x, t):
ṅ + [(h + n)u],x = 0 (2.29)
24
2 1
u, t + u,x +gn,x − h3 u̇,x x + 1 vω
h 2
2 (1 + 2h
b + C)u + ẍb = 0 (2.30)
where the subscripts following commas denote differentiation.
Parameters v, ω, g, ẍb represent the kinematic viscosity of the fluid, the character-
istic frequency of the fluid motion, the gravitational acceleration, and the horizontal
acceleration of the tank respectively. Parameter, C governs the additional dissi-
pation due to free surface contamination. Assuming linearised response without
dissipation, the resulting natural frequency of the kt h mode of the associated waves
created by the fluid motion, can be written as:
1
ωk = (2k − 1)π (gh)
2 1 2 2 h 2
L [1 − 6 (2k − 1) π ( L ) ]
(2.31)
All the above parameters show the complexity of the TLDs, especially when ap-
plication in earthquake protection is involved, where time history analysis is often
necessary. Hence, another more recent approach, is modelling the TLD with a
TMD analogy [120, 143, 99]. Equivalent values for mass and damping coefficient
were extracted[120]:
2
(1−Ω )+(2ζα Ω) 2
mv = mα (1−Ω 2 )2 +(2ζ Ω)2
(2.32)
α
4
cv = cα (1−Ω2 )2Ω+(2ζα Ω)2 (2.33)
ω
where mα , cα , and Ω = ωα can be calibrated based on experimental data
Figure 2.14: Configuration for: a) General TMD, b) TLD
25
Literature Review
Figure 2.15: TLD Geometric definition[74]
Even though the mathematical theory behind TLDs is complicated, their appli-
cation to real structures has become quite popular due to their easy implementation,
and maintenance, along with the fact that they are not very sensitive to the actual
frequency ratio between the structure and the damper, in contrast with the TMDs.
2.2.7 ViscoElastic Dampers
Since ViscoElastic (VE) dampers show similar characteristics with the elastomeric
ones, they will be given much more attention and detailed description compared to
the types of damper presented earlier. In contrast with the viscous dampers, VE
dampers have the ability to provide both velocity-dependent force, which provides
supplemental viscous damping to the system (similar with viscous dampers), and
displacement-dependent elastic restoring force [28], adding stiffness to the system
(similar to metallic dampers). They typically consist of layers of polymers bonded
between steel plates (Figure 2.16). Visco-elasticity is a property which defines the
ability of a material to exhibit both elastic and viscous characteristics; typical ex-
amples are polymers which exhibit viscoelastic behaviour at all temperatures - they
are never simple elastic solids [89]. A more detail description of viscoelastic dampers
and their comparison to elastomeric ones is given in 2.2.8.
VE dampers are incorporated in the structure between concentrical braces and
the overlying beam. Hence, their ability to dissipate energy is based on their shear
deformation due to the drift between two adjacent floors. It appears that their
application began to gain attention after their first implementation at the World
Trade Center in New York (Figure 2.30), in order to improve the structural behaviour
due to wind loads. However, seismic applications usually require higher damping
26
contributions compared to wind.
In the case of low rise buildings, where moment resisting frames are mainly used
(as in the current research project), a rational assumption is that the supplemen-
tal damping devices aim to minimize the horizontal deformations, and therefore
the above implementation method with the concentric diagonal braces seems ratio-
nal. However, tall buildings usually consist of the combination of moment resisting
frames along with shear walls. The deflected shape of the shear walls is similar to a
cantilever, which actually highly affects the global deflected shape of the structure.
High axial deformations are created, and in order to take this into consideration,
the damping philosophy should be adapted as well, and designed in such a way so
they will resist these axial deformations through shear. Based on current research,
and on actual implementations of supplemental damping devices on skyscrapers,
VE dampers can actually replace the reinforced concrete coupling beams, resulting
in reducing the vertical deformations created at the ends of the beams due to high
shear forces. This solution has been already suggested and implemented into real
projects in order to mitigate both seismic and wind actions [91, 29]. It has been also
suggested to add both VE dampers along with diagonal braces for optimal results
[88].
Figure 2.16: Typical VE damper configuration
Figure 2.17: VE stress-strain behaviour under sinusoidal loading [145]
27
Literature Review
Even though the basic characteristics of VE materials are creep, stress relaxation,
and dynamic behaviour, the last one is the main focus here, due to the interest in
the dynamic loads applied at the structure. Considering the case of applying a
sinusoidal shear strain, γ, with frequency, ω, the corresponding viscoelastic stress
will oscillate at the same frequency but out of phase (Figure 2.17):
γ(t) = γ0 sinωt (2.34)
τ (t) = τ0 sin(ωt + δ) (2.35)
where γ0 , and τ0 represent the maximum strain and stress respectively, and δ is the
phase lag. Rewriting Eq. 2.35 in a different form:
τ (t) = γ0 [G0 (ω)sinωt + G00 (ω)cosωt] (2.36)
where
τ0 τ0
G0 (ω) = 00
γ0 cosδ, G (ω) = γ0 sinδ
(2.37)
p
Replacing γ0 sinωt with γ(t), and using the identity cosωt = 1 − sin2 ωt, Eq.2.36
can be rewritten as:
τ (t) = G0 (ω)γ(t) ± G00 (ω) γ02 − γ 2 (t)

p
(2.38)
Eq. 2.38 defines the characteristic VE relationship of the material in the form of the
ellipse shown at Figure 2.18. The area of this ellipse defines the energy dissipated
in each cycle of oscillation:
R 2π/ω
EH = 0 τ (t)γ̇(t)dt
R 2π/ω
= 0 γ02 ωcosωt[G0 (ω)sinωt + G00 (ω)cosωt]dt (2.39)
= πγ02 G00 (ω)
28
Using Eq. 2.34, and its derivative with respect to time, then Eq. 2.35 can take the
following form:
G00 (ω)
τ (t) = G0 (ω)γ(t) + ω γ̇(t)
(2.40)
where it can be noticed that the term G0 (ω) represents the elastic modulus, called
the shear storage modulus, and the term G00 (ω) represents the energy dissipation
component, and is called the shear loss modulus. Both G0 (ω), and G00 (ω) are func-
tions of frequency, and are considered as basic characteristics of every VE material
giving a measure of the energy stored and recovered, and energy dissipated per cycle,
respectively.
Figure 2.18: VE stress-strain behaviour under sinusoidal loading
Since, the quantity G00 (ω)/ω is the damping coefficient, the corresponding damp-
ing ratio is determined as:
G00 (ω) ω G00 (ω (2.41)

ζ= ω ( 2G0 (ω) ) = 2G0 (ω)
And the loss factor, n as:
G00 (ω) (2.42)

n= G0 (ω) = tanδ
29
Literature Review
Comparing, Eqs, 2.41, and 2.42:
n (2.43)
ζ= 2
In order to determine the shear complex modulus of the material (which forms
one of the basic parameters used to characterise a VE/elastomeric material) another
way of determining the stress-strain relationship is followed, using complex notation
and adjusting Eqs. 2.34, 2.35:
γ(t) = γ0 eiωt (2.44)
τ (t) = τ0 ei(ωt+δ) (2.45)
Eqs. 2.44, and 2.45 lead to:
τ (t) = G∗ (ω)γ(t) (2.46)
where G∗ (ω) is the complex modulus of the VE material:
τ (t) τ0 iδ
G∗ (ω) = γ(t) = γ0 e = G0 (ω) + iG00 (ω) = G0 (ω)(1 + in) (2.47)
and
τ0
|G∗ (ω)| = Geq =
p
γ0 = (G0 (ω))2 + (G00 (ω))2 (2.48)
Therefore, it is observed that the parameters, G0 and G00 , or G0 and n define
the dynamic behaviour of a VE material when it is subjected to sinusoidal load-
ing. As will be seen later, these parameters do not only depend on the frequency,
but on temperature and shear strain as well. Hence, a proper constitutive equa-
tion describing the stress-strain or the force-displacement relationship should take
these into account. VE mechanical response is usually simulated by different spring-
viscous damper combinations [141].
30
All the above equations are based on the stress-strain relationship of a VE mate-
rial. When this relationship has to be applied in terms of the force-displacement
relationship for a damper with shear area, A, and total thickness, h, then Eq. 2.38
is adjusted to:
F (t) = k̄(ω)u(t) + c̄(ω)u̇(t) (2.49)
where
AG0 (ω) AG00 (ω) (2.50)

k̄(ω) = h , c̄(ω) = ωh
Eq. 2.49 represents pure VE behaviour captured by a Kelvin-Voigt model (spring
and dashpot connected in parallel). Following a slightly different approach, the
equivalent stiffness can be also determined in this case, and its connection with the
equivalent shear modulus can be shown, assuming a sinusoidal imposed displace-
ment:
u(t) = u0 sinωt (2.51)
which gives the corresponding derivatives with respect to time:
u̇(t) = ωu0 cosωt (2.52)
ü(t) = −ω 2 u0 sinωt (2.53)
Replacing u(t), and u̇(t) in Eq. 2.49:
p
F (t) = k̄(ω)u(t) + c̄(ω)ω u20 − u2 (2.54)
which is the same as Eq. 2.38, but in force-displacement terms, and is representative
of pure VE behaviour. Since, the equivalent stiffness is defined as F0 /u0 , the maxi-
mum force has to be determined. Hence, assuming that the maximum force occurs
31
Literature Review
when its derivative with respect to time is zero, the displacement corresponding to
the maximum force (in elastic systems the maximum displacement corresponds to
the maximum force) can be calculated as:
dF (t) (2.55)
dt = k̄(ω)u̇(t) + c̄(ω)ü(t) = 0
Replacing Eqs. 2.52, and 2.53 to Eq. 2.55 the displacement corresponding to the
maximum force is extracted:
k̄(ω)
uF0 = √ 2
u0 (2.56)
k̄(ω) +c̄(ω)2 ω 2
Replacing Eq. 2.56 in Eq. 2.54, the maximum force is determined:
q
2
F0 = u0 k̄(ω) + c̄(ω)2 ω 2 (2.57)
Hence, the equivalent stiffness is:
q
2
keq = k̄(ω) + c̄(ω)2 ω 2 (2.58)
which is the force-displacement version of Eq. 2.48. Moreover, the force will cor-
respond with a time delay, δ, comparing with the imposed displacement, similarly
with Eq. 2.35:
F0 k̄(ω) (2.59)
k̄(ω) = u0 cosδ => F0 = cosδ u0
Combining Eqs. 2.57, and 2.59:
c̄(ω)ω (2.60)
tanδ = k
where again the association with the stress-strain terms is obvious. If k̄(ω), and c̄(ω)
are replaced using Eq. 2.50 then the loss factor obtained from Eq. 2.60 equals the
one from Eq. 2.42).
A full description of the VE behaviour was presented here. These parameters, even
32
though are very simple, form the basic formulas in order to capture the main charac-
teristics of VE/elastomeric material and will be used in the folllowing chapter. How-
ever, more sophisticated models have been proposed as well. Apart from the Kelvin-
Voigt (spring and dashpot connected in parallel) and Zener model (spring connected
in parallel with a spring-dashpot series connection), the Generalized Maxwell Model
(GMM) was found to be more adequate to capture the mechanical VE characteristics
[89, 141, 42, 75]. The GMM (Figure 2.19) consists of a linear spring, with stiffness,
k0 , connected in parallel with Maxwell elements (linear springs with stiffness, ki ,
connected in series with dashpots, ci ).
Figure 2.19: Generalized Maxwell Model (GMM) presentation
It can be expressed either in stress-strain or force-displacement form. When
dynamic displacement, in the form of Eq. 2.34 is applied, then the corresponding
shear storage modulus, and the shear loss modulus are determined as[89]:
N
X Gi ω 2 τi 2
G0 = G0 + (2.61)
1 + ω 2 τi 2
i=1
N
X Gi ωτi 2
G00 = (2.62)
1 + ω 2 τi 2
i=1
The above equations are valid only under sinusoidal loading with a specific frequency
ω. Minimizing the error between the experimental and analytical values of the shear
storage modulus and the shear loss modulus. Fan [42] carried out a comparative
analysis between Kelvin-Voigt, GMM, and two different Fractional–Derivatives
33
Literature Review
models (one proposed by Kasai [64] and one from Fan). The Fractional-Derivative
models are explained in the next subsection. A relatively good agreement was
achieved for the shear storage modulus, especially for the GMM (Figure 2.20).
However, the only available hysteretic loops which were found were comparing the
aforementioned models between them (Figure 2.21), and no comparison with exper-
imental values was possible. The only figure which was found to be comparing any
of these models with experimental data in time domain was the roof displacement
time history (Figure 2.22) of a 2/5-scale five-story 1-bay steel frame test structure
with diagonal-braced VE dampers [108]. However, careful comparison of these
results show the actual deviation of the GMM model.
Figure 2.20: Comparison of analytical and experimental values of shear storage

modulus for GMM, Fractional-Derivative models [42]
34
Figure 2.21: Comparison of hysteretic behaviour of the 3 models( GMM, and

Fractional-Derivative models) [42]
Figure 2.22: Comparison of analytical and experimental values of roof displacement’

time history of a test structure with diagonal-braced VE dampers [42]
35
Literature Review
2.2.8 Elastomeric Dampers
Elastomer Material Background
The ASTM standard D 1566 (ASTM: American Society for Testing and Materials)
defines elastomers as high-polymeric, organic networks that are capable of absorbing
large deformations in a reversible manner. Elastomers are rubber-like materials,
they are made up of long chains of atoms, mainly carbon, and hydrogen, which
have a degree of cross-linking with their neighbouring chains. It is these cross-
linking bonds that pull the elastomer back into shape when the deforming force
is removed. Polymers are characterized by their glass transition temperature, Tg .
Below the glass transition temperature, the proper motion of the molecules, which
is also known as Brownian motion – freezes. The material is in a rigid, glassy state
[6]. When the glass transition is exceeded, the molecules start becoming mobile
once again and the polymer undergoes a “transition” phase. When the temperature
increases to levels more than the “transition” phase, the polymer exhibits a more
rubbery behaviour, where stiffness and damping are generally lower but vary much
less with the temperature [77]. It is in this region where the elastomers are usually
used. If the temperature continues to rise, first rubbery and then viscous flow occurs
followed by decomposition of the polymer. During the rubbery flow phase and the
glass transition phase the polymer behaves as a visco-elastic material [111]. The
above description of how polymers are behaving with different temperatures can
be seen in Figure 2.23, where a typical specimen is subjected to a stress which is

σ
held for a time, t, at a temperature T , measuring the modulus: E(T, t) = (T,t) ,
where σ is the applied stress, and the corresponding displacement. It should be
mentioned that even though not all polymers are purely viscoelastic, and even fewer
are linearly viscoelastic [106], a variety of dampers has been modelled as pure linear
viscoelastic materials which is actually a very good approximation, especially for
low strain amplitudes or for polymers used with relatively high glass Tg .
Another interesting characteristic of the elastomeric materials, known as the
Mullin’s effect, is the softening of the material under the first few cycles of defor-
mation, after which the stress-strain behaviour approaches a steady state [50]. The
36
main properties of elastomers (at least in a macro-scale which is the main interest
here) can be summarised as:
• exhibit linear VE behaviour, especially under relatively small strains
• extremely high extensibility
• complete recovery after removal of the imposed stress
• strain hardening at relative large strains
Some of these define their main advantages comparing with other types of materials
used in passive dissipation devices, since even after strong earthquakes, the elas-
tomeric dampers will not exhibit any damage or permanent deformation, resulting
in more economical solutions as well.
Since elastomers, like many other polymers, show VE properties, they highly
depend on the strain amplitude, frequency, and ambient temperature. Moreover, a
distinct feature of the hysteresis loops of elastomers is that softening at a small strain
amplitude and hardening at a relatively large strain amplitude may occur in the
amplitude range of practical interest [107]. Similar softening-hardening hysteretic
behavior was observed for elastomeric dampers [107], and for high damping rubber
base isolators [127, 9]. Typical elastomeric hysteretic behaviour can be seen at
Figure 2.24.
It should be noted that most of the elastomer applications, with regard to seismic
protection of structures have been implemented for seismic isolated structures [68,
65, 144, 103] in the form of elastomeric bearings.
To capture the dynamic behaviour of elastomeric-rubber dampers several models
have been proposed. Five different models are presented here, which have been used
by either researchers or by design codes to model elastomeric and rubber dissipative
devices:
• Linearised model [5]
• Rate dependent model [107]
• Fractional derivatives [72, 68, 55]
37
Literature Review
• Bouc-Wen model [84]
• Modified Bouc-Wen model [62]
Figure 2.23: Different regimes of polymers behaviour when temperature increases

[111]
Figure 2.24: Typical hysteretic behaviour of elastomer [69]
Linearised Model
The most simplified, but at the same time practical way of presenting the basic
characteristic of either an elastomeric or VE damper/bearing, at least for design
purposes, is a linearised model with an equivalent stiffness, kef f , and equivalent
damping ratio, ζef f , proposed by FEMA [5]. These parameters are determined as:
|F + |+|F − | (2.63)
Kef f = |u+ |+|u− |
38
P
Wi (2.64)
ζef f = ζ + 4πWk
where:
• K 0 is the storage stiffness
• u+ , u− are the maximum, and minimum displacements
• F + , F − are the forces corresponding to u+ , u−
• ζ is the inherent damping ratio of the structure
• Wi is the work dissipated by the device i per cycle of oscillation
• Wk is the maximum strain energy
Rate Dependent Model
Sause [107] on the other hand proposed a rate-dependent hysteretic model, which
is far more sophisticated than the linearised model proposed by FEMA, based on
asymptote functions. Simulating softening and hardening behaviour, which the elas-
tomer experiences, with second and third order polynomials the final model is:
k[n−1][τ̄ (γ)−τ̄ (γ0 )] (−1)/(n−1) (2.65)

τ (γ) = τ̄ (γ) − e0 [1 + e0 ]
when n > 1
τ (γ) = τ̄ (γ) − e0 exp[− k[τ̄ (γ)−τ̄

e0
(γ0 )]
] (2.66)
when n = 1 where:
• γ is the applied shear strain
• τ (γ) is the shear stress
• τ̄ (γ) is the asymptote function for loading and unloading direction
• e0 = τ̄ (γ0 ) − τ̄0 is the stress deviation at the most recent strain reversal, γ0 ,
and corresponding stress τ0
39
Literature Review
• k, and n are parameters which control the shape of the hysteretic loop
A linear dashpot was also added to the proposed model in order to take into ac-
count rate dependent effects. According to Sause[107], the slope of the stress-strain
path, τ (γ) asymptotically approaches the function τ̄ (γ). The parameters of this
model were determined using data from cyclic tests in a frequency range 0.5-3.0Hz.
Good correlation was achieved comparing the analytical model and experimental
results (see Figure 2.25). However, this model is based on a very similar hysteretic
model to previous studies [72], where even though a very good agreement between
analytical and experimental results had been achieved for selected strains, and fre-
quencies2, large discrepancies were observed for high frequencies, and strains below
100% (see Figure 2.26). The dynamic analysis results presented are limited, and
more validation methods and experimental data need to demonstrate the computa-
tional stability and general applicability of the current proposed model[107].
Figure 2.25: Comparison between analytical model proposed by Sause and experi-
mental data [107]
40
Figure 2.26: Comparison between analytical model proposed by Lee and experimen-
tal data [72]
Bouc-Wen Model
The Bouc-Wen model,which was originally proposed by Bouc [19] and later gen-
eralized by Wen [137], has been widely used to model the hysteretic behaviour of
structures experiencing non linearities, or dynamic behaviour of individual devices,
such as dampers, and base isolation systems [33, 144, 92, 118] . Based on the model
proposed by Bouc-Wen:
F (2.67)
F (x, ẋ) = α Yy x + (1 − α)Fy z
in which Z is a dimensionless hysteretic component satisfying the following nonlinear
differential equation:
Y ż = −γ|ẋ|z|z|n−1 − β ẋ|z|n + Aẋ (2.68)
Y , and Fy represent the the yield displacement and force of the hysteretic damper
respectively; x, and ẋ the displacement and velocity, respectively; and β, γ, A dimen-
sionless parameters. The parameter n is an integer which controls the smoothness
of the transition from elastic to plastic response, and α is the post to pre-yielding
stiffness ratio. The Bouc-Wen model is very efficient, however force and deformation
errors have been observed [128, 142] under certain conditions. Moreover, Ni et al [97]
concluded that the Bouc-Wen model can not capture the characteristics of rubbery
materials.
41
Literature Review
However, Constaninou [33] carried out a comparison between different dampers
and the Bouc-Wen model. One of the dampers was a lead-rubber damper, used for
base isolation. Figure 2.27 shows the comparison between the hysteretic loops of
the analytical model and experimental hysteresis. The parameters were optimised
based on minimizing the error between between the corresponding forces under 60
mm amplitude, 0.9 loading frequency, and a vertical load of 450 kN. Even though
the results are found to be in a very good agreement it should be noted that this
model was valid under this specific test under the aforementioned conditions. The
data were obtained from Skinner [112].
Figure 2.27: Comparison between experimental data and Bouc-Wen model [33]
Modified Bouc-Wen Model
Karavasilis et al [62] proposed a hysteretic model for a compressed elastomeric
damper. This model was based on a modified Bouc-Wen model [138], connected
in parallel with a non linear dashpot. The dampers’ behaviour consists of two dif-
ferent phases: before and after slip of the elastomer, which is allowed due to the
unbonding interface between the elastomer and the steel material. The force of the
dashpot is:
Fd = C|v|α sgn(v) (2.69)
42
where C is the damping coefficient, v is the deformation rate, andα is the velocity
exponent. The force of the dashpot was combined with the force from a modified
Bouc-Wen model, where the stiffness term FY /Y of Eq. 2.67 was modified to:
− uuref
max
(2.70)
k = k1 e + k2
where k1 , k2 ,uref are constants, and umax is the average of the maximum absolute
deformation amplitudes in both positive (umax,p ), and negative (umax,n ) direction:
|umax,n |+umax,p (2.71)

umax = 2
The aforementioned values were determined from characterization tests. The pro-
posed model was found to capture satisfactorily the dampers’ behaviour for both
pre-slip and post-slip displacements (Figure 2.28), and was used for non linear anal-
yses to evaluate the seismic performance of steel moment resisting frames equipped
with elastomeric dampers.
43
Literature Review
Figure 2.28: Comparison between experimental data and analytical model proposed
by Karavasilis [62]
Fractional Derivatives Model
Finally, Fractional Derivatives is also an approach of modelling the hysteretic
behaviour of both viscous[82], viscoelastic[42, 70], and elastomeric or rubber
materials[55, 68, 129] as passive dissipation devices. This model is based on the
standard linear model [95]:
σ + α dσ d
dt = E + βE dt
(2.72)
This model is a more complex stress-strain relationship between the simple Hooke’s
law (σ = E) or the dashpot-spring combination (σ = E + βEd/dt). Adding more
derivatives of σ, and in Eq. 2.72:
P∞ dn σ P∞ dn (2.73)
σ+ n=1 αn dtn = E + E n=1 βn dtn
44
Proper choices of βn , and αn lead to fit the experimental data with the proposed
model. Replacing the integral derivatives with fractional derivatives leads to:
P∞ αn σ(t)
P∞ βn (t)
σ+ n=1 αn D = E(t) + E n=1 βn D
(2.74)
where the generalized derivatives Dαn , and Dβn are defined as:
1 d
Rt x(τ )
Dαn [x(t)] = Γ(1−αn ) dt 0 (t−τ )αn dt
(2.75)
with 0 < αn < 1, and Γ is the gamma function. Hysteretic plots of derivative models
used for modelling VE materials can be seen at Figure 2.21. Moreover, Kasai et al
[64] used the fractional derivative model to model VE damper’s behaviour under
cyclic and ramp tests. Even though relatively good correlation was achieved for
both the shear storage modulus and the loss factor for 24, and 32 o C, results were
disappointing for the 15 o C case (see Figure 2.29).
Figure 2.29: Comparison between experimental data and analytical model proposed
by Kasai [64]
Another scenario is that a different form of Eq. 2.74 is used, called the 0 fractional
Kelvin Model0 , omitting any additional derivatives of the stress:
σ = E0 (1 + cDβ )(t) (2.76)
The most common approach of determining the parameters of the aforemen-
tioned models is to obtain analytical expressions of the shear storage modulus, G0 ,
45
Literature Review
and shear loss modulus, G00 under cyclic tests of specific strain amplitudes and fre-
quencies, and then optimising the models’ parameters fit the experimental data to
these expressions. However, as was shown, matching the experimental values of G0 ,
G00 does not necessarily mean that the proposed model can capture the dynamic
characteristics of the damper. The same approach is followed when minimizing the
experimental and analytical force is preferred.
2.2.9 Past applications of VE/Elastomeric Dampers
It should be mentioned that most of the elastomeric materials’ applications were
used in base isolation techniques, where they were usually incorporated between steel
plates, rather than actual implementation on frames along with diagonal braces. On
the other hand, the majority of the VE dampers were used in retrofitting structures
connecting the damper with the braces. The main target of the first implementations
of the VE dampers was to reduce wind induced vibrations. Therefore, one of the first
applications of VE dampers was in the World Trade Center (1969) in order to reduce
the acceleration levels due to wind [81]. About 10000 VE dampers were installed in
each tower from the 10th to the 110th floor. Their design was carried out in such a
way to assist the steel frame mitigating the wind induced movement (Figure 2.30).
The were located between the lower chords of the horizontal trusses and the the outer
columns of the structure. The total achieved damping was calculated and found to
be in the range of 2.5%-3% of critical. Seismic applications started a lot later, and
more specifically in 1993, when the 13 story Santa County building in San Jose, CA,
was retrofitted, since it was found that the viscous damping in the fundamental mode
was less than 1% of critical. VE dampers were chosen as the final solution, since
they could provide increased damping for both frequent low-level ground shaking
and strong ground motions. Two VE dampers were added to each building face per
floor (Figure 2.31), increasing the fundamental damping to 17% of the critical. More
recent applications of VE dampers connected with steel diagonal braces include the
Beijing 7 Star Morgan Plaza Hotel C (China, Beijing) in 2007 where 108 viscous
and viscoelastic dampers were added for both wind and earthquake protection, and
the Hotel Stockton (USA, Stockton, CA), where a combination of both viscous and
46
viscoelastic dampers were added to reduce the structure’s responce due to seismic
loads [4].
Figure 2.30: Damper Installation at the World Trade Center [31]
Figure 2.31: Santa Clara County building San Jose - Seismic application [31]
Another different approach was followed by Rant [38], where a 110-story, 630 m
mega-tall building was retrofitted with VE dampers, which replaced the coupling
beams connecting the shear walls. The main concept behind this decision is that
since the deflected shape of very tall buildings due to horizontal loads is similar
to a cantilever, rather than the shear deflection of moment resisting frames which
47
Literature Review
consists the main resisting system of low-rise buildings. Hence, the whole dampers
philosophy was adjusted as well (Figure 2.32). The dampers replaced about 60%
of the diagonally reinforced concrete coupling beams of the original structure. The
results showed that the peak inter-story drift ratios were reduced by up to 15%
under the Maximum Considered Earthquake, and the peak floor accelerations by
approximately 24%.
One of the rare applications of rubber-like dampers connected with the sur-
rounding structure with diagonal braces that could be found in the literature, was
carried out by Teramoto [125] on an 11 storey, 44 m tall building which was un-
der construction when the research paper was published. The main target of the
added dampers was to reduce the vibrations due to both seismic loads and traffic
motion since the building was close to subway lines. The results improved the build-
ing’s performance by reducing both the maximum accelerations and storey drifts by
approximately 20%. Another seismic retrofit was carried out in the Gentile-Fermi
School in Fabriano, after the building suffered sufficient damage during an earth-
quake in the Umbria-Marche region (September 1997). A total of 33 elastomeric
dampers were added at the second and the third floor of the building, absorbing
50% of the input energy. However, as already mentioned, the vast majority of the
applications of elastomeric dampers to structures were carried out as part of base
isolation methods.
Figure 2.32: Experimental rig for VE dampers replacing coupling beams [38]
48
Figure 2.33: Elastomeric dampers used for seismic retrofit in the Gentile-Fermi
School in Fabriano [8]
2.3 Conclusions
This section has presented the main principles of structures with supplemental dissi-
pation devices, and their main characteristics, along with alterations in the equation
of motion during earthquakes, for the case of both SDOF and MDOF structures.
Furthermore, a brief introduction into each of the passive energy devices was pre-
sented. Passive energy devices are generally preferred for their simplicity, and relia-
bility even during strong earthquakes. A more detailed insight was given in case of
viscoelastic and elastomeric dampers, with the latter case to be the research focus
of this study. Their main advantage is their ability to provide both stiffness and
damping in contrast with other types of dampers, while they can be also effective
under low amplitude forces not needing a specific value of either force or deformation
to trigger their mechanism. Finally, material models which have been used by previ-
ous researches to capture the dynamic characteristics of elastomeric materials were
briefly described, and actual applications of VE/elastomeric dampers to buildings
were presented.
Elastomers have been mainly used in base isolation methods when they have been
actually applied in the form of dampers distributed in buildings they were usually
modelled assuming visco-elastic behaviour, which has been proved to be valid under
only low strain amplitudes. Hence, a need for a more general hysteretic model which
49
Literature Review
would be able to capture the elastomers’ characteristics beyond the deformation for
which visco-elasticity can be accurate is covered by this research.
The fact that not only elastomers behaviour changes with frequency, strain am-
plitude, and temperature, but also elastomers hysteresis shape makes extremely
difficult the modelling process with regard to seismic interest, because the main
earthquake scope is to effectively capture the dynamic characteristics of the dampers
under random loading for a few seconds. Even though a lot of models have been
proposed, most of them are based on linearised properties of the elastomers or on
models parameters optimisation based on minimization of the error of analytical
expressions of shear storage modulus and corresponding experimental values un-
der imposed sinusoidal loading with specific values of frequencies. These analytical
formulas do not usually take into account any temperature and strain amplitude
alterations, and depend only on the imposed loading frequency. However, these
modelling techniques do not guarantee the actual representation of the elastomers
hysteresis loops. Furthermore, a lack of models predicting the dynamic behaviour of
the elastomeric materials under different temperatures was observed. Moreover, the
majority of research of elastomeric dampers is based on base isolation techniques
and not when the dampers are incorporated throughout the height of the structure
with additional steel diagonals braces. These techniques are highly different and
target in different expected structural performances. The uniform distribution of
the dampers throughout the height of buildings offers more damping and stiffness
to each floor level, as opposed to base isolation techniques where most of the seis-
mic energy is dissipated at the structure’s base. A new model is presented in this
thesis, which is based on a new time domain Generalised Maxwell Model equation,
and is able to predict the dynamic behaviour of elastomeric dampers for a range of
strain amplitudes, loading frequencies, and ambient temperatures. Non linearities
and strain-stress time history are taken into account, while the model was further
validated under earthquake loading using real time substructure tests (see Chapter
6).
50
Chapter 3
Characterization Tests
3.1 Introduction
This chapter presents the experimental procedure which was followed in order to
determine the main characteristics of two elastomeric dampers, provided by the
Tun Abdul Razak Research Centre (TARRC). It is well known that elastomeric
materials depend on strain amplitude, loading frequency, and ambient temperature.
Therefore, sinusoidal displacements under a specific range of strain, frequency, and
temperature were applied to the assemblage of the two dampers, and the corre-
sponding force was measured. Based on the force-displacement hysteretic loops the
main characteristics of the elastomer were determined.
3.2 Experimental rig
Figure 3.1 shows one of the two individual elastomeric dampers, which were pro-
vided by TARRC. It consists of an elastomeric material bonded between two steel
plates. Each of these dampers has overall dimensions of 260x260 mm2 , of which only
180x230 mm2 corresponds to the elastomer’s dimensions(fig. 3.2). The 230 mm di-
mension coincides with the loading direction. The four A-holes, and the two B-holes
allowed each of the dampers to be connected with adjacent steel plates (fig. 3.3):
an additional steel plate, A, was connected with both the outer plate of the damper
and with a fixed and stable reacting wall, which in turn was considered fixed at its
51
base; the interior plate of the damper was connected to a plate (B), which in turn
enables connection to the actuator via a central plate , C. The final configuration
of the rig can be seen in Figures 3.3 and 3.4.
It should be noted that the focus of this research is on the effect of the elastomeric
dampers on structures due to earthquake loads. Therefore, the elastomeric dampers
were tested for the earthquake scenario only.
In order to investigate the strain amplitude, frequency, and temperature depen-
dence, a series of tests was carried out based on sinusoidal displacement histories at
the Elastomeric Dampers (EDs). In order to avoid large initial displacement and
velocity, ramping cycles were implemented, along with 18 full sinusoidal cycles (see
Figure 3.5). This process, which is similar with the one proposed by FEMA356
[5], was repeated for frequencies of 0.25, 0.5, 1.0, 2.0, 3.0, and 4.0 Hz, for strain
amplitudes of 10, 20, 30, 40, and 50%, and for ambient temperatures of 20, 25, 30,
and 35o C. The maximum shear strain amplitude was limited to 50%, to avoid any
potential permanent damage, severe cracking, or debonding of the elastomer. The
first series of tests was conducted at room temperature (20o C), while the same pro-
cess was repeated for 25, 30, and 35o C, with the use of a heater, and a temperature
control chamber (fig. 3.6), so that the target temperature was kept stable during the
test. Two thermocouples were attached at the elastomer, one at the top and one at
the bottom surface of the elastomer, for measuring the ambient temperature along
with any increase in temperature during the loading history. For every test which
was carried out at higher than room temperature, the elastomer was heated until
convergence was achieved between the temperature shown at the two thermocouples,
typically between 8-12 hours.
A 100 kN Instron servo-hydraulic actuator was used to carry out the experi-
mental procedure. The actuator has a stroke of ±125 mm, receiving its power via
pressurised oil through Moog servo valves, which control the piston movements by
directing the oil to one side of it and connecting the other side to the return line
[18]. The pressurised oil is delivered through the laboratory’s substations, which are
installed directly on the main hydraulic high pressure line, driven by three 60 l/min
oil pumps, resulting in a peak flow of 180 l/min. Hydraulic accumulators which
52
are capable of storing highly pressurised oil, provide enhanced flow rates at times
of peak consumption when the instantaneous supply from the pumps is not enough
[18]. The substations are also connected to the return line to the oil reservoir. Dur-
ing low consumption, individual pumps can be switched off. In order to mitigate
any floor vibration effects all the hydraulic equipment is installed on a large con-
crete block of 9.1x4.2 m in surface area, 1.6m deep and weighing 166 tonnes. The
testing area of the floor is 7.5x3.5m. The communication between the actuators
and the substation (closed loop control) is controlled by an Instron Labtronic 8800
controller, which provides four channels of position or force control, and is based
on a proportional-integral-derivative-lag (PIDL) controller, with its parameters set
either by an auto-tuning process or by hand. The controller parameters are accessi-
ble to the user through the Instron program RS+. More information regarding the
laboratory equipment is given in Chapter 6, where real time substructure tests are
conducted.
A linear encoder is also attached to the damper in order to capture the actual
deformation, and not the one achieved by the actuator, since some flexibility of the
connecting elements is anticipated. It was connected at the mid-height of the central
plate of the rig assemblage. A thermocouple was attached to the top and bottom
surface of the elastomer in order to measure the ambient temperature, but also to
capture any increase in temperature during the tests. Hence, the experimental tests
were focused on verifying non linear dependence on strain amplitude, frequency,
and ambient temperature. Among these parameters, it is has been shown that
strain dependence is the dominant factor [72, 103].
Figure 3.1: Elastomeric Damper provided by TARRC
53
Figure 3.2: Elastomeric Damper Dimensions
Figure 3.3: Plan of the final rig: elastomeric damper connected with actuator
54
Figure 3.4: Final Rig: elastomeric damper connected with actuator
Figure 3.5: Typical Displacements’ command - (30% shear strain, 2 Hz)
55
Figure 3.6: Temperature chamber
3.3 Mechanical Properties
Figure 3.7 shows a typical hysteretic loop of the elastomeric dampers corresponding
to 40% maximum shear strain amplitude, 3 Hz frequency, and 20o C temperature.
It can be seen that after the first cycles, the hysteretic loops become more and
more steady and approximately repeatable, corresponding to smaller and smaller
stiffness degradation. Hence, in order to determine the mechanical properties of
the dampers the average value of the hysteretic loops of cycles 5-15 was taken. As
already mentioned in Chapter 2, the shear storage modulus, G0 , and either the shear
loss modus, G00 , or the loss factor, n, govern the main characteristics of the dynamic
behaviour of viscoelastic and elastomeric materials. Instead of G0 , the equivalent
shear modulus, Geq can be determined in a similar way that equivalent stiffness was
defined in Chapter 2 (see section 2.2.7). All the above parameters can be extracted
from each of the hysteretic loops as:
τ (γ=γmax )−τ (γ=γmin )

G0 = γmax −γmin
(3.1)
τmax −τmin (3.2)

Geq = γmax −γmin
56
1 ED (3.3)
n= 2π ES
where τ (γ = γmax ) is the shear stress corresponding to the maximum strain γmax ,
τ (γ = γmin ) is the shear stress corresponding to the minimum strain γmin , τmax is
the maximum shear stress, τmin is the minimum shear stress, ED and ES are the
dissipated energy per cycle of oscillation, and the maximum energy stored respec-
tively. ED can be calculated based on the area of the hysteretic loop. ES can be
calculated as:
ES = 21 G0 γmax
2 (3.4)
Figure 3.7: Typical force-displacement relationship for ED under frequency of 3Hz,

strain amplitude of 40%, 20o C temperature
3.4 Results
All the aforementioned parameters are summarised in Appendix 1 (tables A.1 to A.4)
for all the strain amplitudes, frequencies, and temperatures. Comparing the range
of the values of both the loss factor and the shear storage modulus shows that they
are close to the rubbery materials used by Lee[72], and other researchers [125, 114].
It should be also mentioned that any temperature variations during testing were
found to be almost negligible.
57
The results showed that strain amplitude had a larger impact on the EDs’ be-
haviour, in contrast with frequency, and temperature. This can be seen by two
different approaches. The first one is to compare the amplitude dependence at a
range of frequencies at fixed temperature, in the form of Figures 3.17 and 3.18.
These figures prove the above statement that the dominant factor which affects the
elastomers’ dynamic behaviour is the amplitude and not the frequency, especially
in the case of the loss factor, which practically remains the same regardless of any
change in frequency. It is also noticeable that when the strain amplitude increases
above 30% (which is the key point, as be seen in the second approach, for the
transition between the viscoelastic phase to elastomeric phase) the rates of change
of both the shear storage modulus and the loss factor reduce significantly. Below
30%, a large decrease, especially in the loss factor, can be observed. Furthermore,
it seems that amplitude and frequency have exactly opposite effects, since increas-
ing strain amplitude leads to decreasing the mechanical properties of the elastomer,
while increasing the loading frequency leads to increasing them.
The second approach of demonstrating that the strain amplitude has the largest
effect on the elastomer’s behaviour is by plotting the hysteretic loops of the dampers
keeping the other two parameters (frequency, and temperature) constant. Hence, ob-
serving Figures 3.8 to 3.13, it can be seen that the elastomer exhibits approximately
viscoelastic behaviour for strain amplitudes of 10%-30%, proving that elastomers
exhibit viscoelastic behaviour under relatively low amplitudes, which is one of the
main material characteristics addressed in Chapter 2. The more the strain am-
plitude increases, the more the hysteretic behaviour tends to form a more typical
elastomeric shape (compare for example Figures 3.9a and 3.9e), which as explained
in Chapter 2 has slightly different characteristics from the conventional viscoelastic
shape. On the other hand, this is not the case when the frequency is altered (com-
pare for example Figures 3.8a and 3.13a or Figures 3.8e and 3.13e ). In that case,
even though the mechanical properties change, the overall shape of the hysteresis
loops remain the same (Figures 3.8 to 3.13) . In case of elastomer dependence on
strain amplitude, not only the mechanical characteristics change, but also the whole
behaviour of the material changes as well. This fact makes extremely difficult the
58
modelling procedure of the material. But, at the same time it shows that when
the material has larger strain amplitudes, and hence behaves more like a typical
elastomer, it tends to be more and more independent of the frequency, in contrast
with its viscoelastic phase when the strain amplitudes are relatively low.
With regard to the effect of ambient temperature, both the shear storage mod-
ulus, and the loss factor decrease with the increase of temperature. This can be
observed by plotting the hysteretic curves of the dampers for different temperatures,
while keeping the frequency, and the strain amplitude constant. Figure 3.14 shows
how the ability of the dampers to dissipate energy becomes smaller as the ambient
temperature increases. However, it can also be seen that the shape of the hystere-
sis remains unchanged, leading to the conclusion that only strain amplitude leads
to different hysteresis shapes, in contrast with frequency and temperature, which
affect, more or less, the dampers’ characteristics but not their hysteresis shape.
The effect of the temperature on the elastomers’ properties can be more clearly
seen in Figures 3.15 and 3.16, where the values of loss factor and shear storage
modulus are plotted against frequency for different temperatures, while keeping the
strain amplitude constant at 40%. Two different conclusions can be made through
these figures: a) increase of temperature leads to decrease of dampers’ efficiency.
However, the rate of this decrease slows as the temperature gets higher, b) the
dampers’ mechanical properties remain almost constant, especially in the case of
the loss factor, with any frequency alteration. Moreover, similar observations with
the 20o C case, can be determined from Figures 3.19 to 3.24. These figures validate
the observations determined from ambient temperature, and lead to the estimation
that the elastomer behaves similarly at different temperatures. Its mechanical char-
acteristics are reduced. However, at the same time it seems that any alteration in
strain amplitude and frequency results in a quite similar change in the loss factor,
and the shear storage modulus. Finally, it should be noted that the tests at the
ambient temperature of 20o C were repeated after approximately two years and no
noticeable ageing effects were observed. Figure 3.25 shows this resemblance com-
paring the hysteretic response of the EDs for a strain amplitude of 40%, frequency
2Hz and ambient temperature of 20o C. The figures are almost identical.
59
(a) 10% maximum strain (b) 20% maximum strain
(d) 40% maximum strain

(c) 30% maximum strain
(e) 50% maximum strain
Figure 3.8: ED’s hysteretic loops for 0.25 Hz and 20o C
60
(a) 10% strain (b) 20% strain
(c) 30% strain (d) 40% strain
(e) 50% strain
61
(e) 50% strain
62
(e) 50% strain
63
(e) 50% strain
64
(b) 20% strain

(a) 10% strain
(e) 50% strain
65
(a) 20o C (b) 25o C
(c) 30o C (d) 35o C
Figure 3.14: ED’s hysteretic loops for 2.0 Hz, 40% shear strain, under 20-35o C
66
Figure 3.15: Shear storage modulus, G0 , under various temperatures, and frequencies
for constant 40% shear strain
Figure 3.16: Loss factor, n, under various temperatures, and frequencies for constant
40% shear strain
67
Figure 3.17: Shear storage modulus, G0 , at 20o C
Figure 3.18: Loss factor, n, at 20o C
68
Figure 3.19: Shear storage modulus, G0 , under 25o C
Figure 3.20: Loss factor, n, under 25o C
69
70
71
Figure 3.25: Comparison of EDs hysteretic Loops for degradation effects due to time
3.5 Conclusions
This chapter describes the experimental procedure which was followed in order to
obtain the main mechanical characteristics of the elastomeric dampers, which were
provided by TARRC. Similar to viscoelastic materials, elastomers’ behaviour de-
pends on strain amplitude, frequency, and ambient temperature. Hence, the tests
were based on sinusoidal strain time histories in a range of different maximum strain
amplitudes, loading frequencies, and temperatures. The results showed that the
dominant factor which highly affected elastomers’ performance was the strain am-
plitude, since it not only affected the values of loss factor, and shear storage modulus,
but also was responsible for alteration of the hysteresis shape from viscoelastic to
pure elastomeric. On the other hand, the elastomers’ parameters, especially the loss
factor, were found to be insensitive to any frequency alteration. This was observed
under every different ambient temperature that the dampers were tested. With
regard to temperature effect, it seems that any increase in ambient temperature
resulted in decreasing the dissipative capacity of the dampers. However, the dete-
rioration becomes smaller and smaller as the temperatures gets higher. One other
72
important observation is that elastomer’s mechanical properties tended to have only
a low variability when the strain amplitude was more than 30%, which, as explained
earlier, was the transition point from the viscoelastic to the elastomeric phase. This
leads to the conclusion that, comparing with conventional viscoelastic materials,
elastomers can be much less dependent not only on frequency but on strain ampli-
tude as well, providing that the strain amplitude has relatively large values.
73
Chapter 4
Hysteretic Constitutive Model

Elastomeric Dampers
4.1 Introduction
The description of the characterization tests was presented in Chapter 3, in order
to evaluate the elastomer’s dependence on strain amplitude, loading frequency, and
ambient temperature. It was observed that the elastomer exhibited highly non lin-
ear behaviour, which was dependent on these factors, especially strain amplitude.
In order to represent the damper’s dynamic behaviour a modified version of the
Generalised Maxwell Model (GMM) is proposed in this chapter, with an effort to
take into account any non linearities. Moreover, the GMM force-displacement re-
lationship for N Maxwell elements, in the time domain, is also proposed. This
equation was then modified, assuming N = 1 Maxwell elements, and to form the
final proposed hysteretic constitutive model.
4.2 Generalised Maxwell Model
4.2.1 General Background
The proposed model for the hysteretic behaviour of the EDs is based on the well
known Generalized Maxwell Model (GMM). The GMM has been proposed as a
numerical representation of VE materials [60, 89] and can be graphically represented
74
as shown in Figure 4.1. It was briefly presented in Chapter 2, as a modelling option
of VE materials, and will be presented in more detail here, since it forms the basis
of the proposed hysteretic model. As already mentioned, it consists of a linear
spring, with stiffness k0 (or shear modulus G0 in case of a stress-strain relationship),
connected in parallel with N Maxwell elements. Each Maxwell element is a linear
spring, with stiffness ki (N/m), connected in series with a linear dashpot, with
damping parameter ci (N sec/m) (or shear modulus Gi (N/m2 ) and shear damping
coefficient ni (N sec/m2 ) in case of a stress-strain relationship).
The most common procedure of determining the parameters of a VE or elas-
tomeric material, when the GMM is used, is to minimize the error between the
analytical values of the shear storage and the shear loss modulus, and the corre-
sponding values which are based on characterization experiments. With regard to
how these values are extracted based on the experimental data, Chapters 2 and 3
provide relevant information. With regard to analytical expressions which predict
the shear storage modulus, and the shear loss modulus, or loss factor, these can be
determined from Eqs 4.1, and 4.2. These equations are based on the assumption
that the applied strain has sinusoidal form, with a frequency ω. Therefore, the corre-
sponding shear storage modulus, G0 and the shear loss modulus, G00 are determined
as[89]:
N
X Gi ω 2 τi 2
G0 = G0 + (4.1)
1 + ω 2 τi 2
i=1
N
X Gi ωτi 2
G00 = (4.2)
1 + ω 2 τi 2
i=1
where
ci (4.3)
τi = ki
is known as the relaxation time.
75
Figure 4.1: GMM representation in terms of Force-displacement
As already mentioned, the above procedure is based on matching the results of
shear storage modulus and shear loss factor for discrete values of the loading fre-
quencies, and is valid only for sinusoidal loading. However, in order to be generalised
and validated under random loading, as in the case of seismic loading, these values
are normally determined based on a frequency range close to the frequency content
of earthquakes (usually 0.25-4 Hz), which is also similar with the content of the
natural frequencies of many structures.
A different approach was used here, where the main focus of the GMM is on
extracting the force-displacement relationship in the time domain for any N number
of Maxwell elements, and therefore the model can be validated under any random
loading. Although, it is very easy to extract this relationship when only one Maxwell
element is used, this becomes much more complicated when more elements are being
used. Therefore, a force-displacement relation for N Maxwell elements has been
also developed here based on Laplace transformation, and presented in the following
section.
4.2.2 GMM force-time relationship for N elements - Derivation
For every Maxwell element of the GMM the force displacement relationship in time
domain is given as:
dFi dui
ki Fi + ci = ki ci (4.4)
dt dt
76
where k i and ci are the stiffness and the dashpot coefficients of the ith Maxwell
element, ui is the imposed displacement and F i the corresponding force for every i
Maxwell element. In the case of parallel connection of Maxwell elements ui = u for
every Maxwell element. Assuming zero initial conditions for both the displacement
and the force, and applying Laplace Transformation to Eq.4.4:
(ki + ci s)Fi,L = ki ci suL (4.5)
where F i,L is the Laplace transform of the force corresponding to the ith Maxwell
element, and uL is the Laplace transform of the displacement. Equation 4.5 can be
rewritten in the following form:
ki ci s
Fi,L = uL (4.6)
ki + ci s
And for N Maxwell elements the Laplace transform of the total force can be deter-
mined as:
N N
X X ki ci s
FL = Fi,L + F0,L = uL + F0,L (4.7)
ki + ci s
i=1 i=1
where F0,L represents the Laplace transform of the additional force corresponding
to the linear spring (in this case), which is connected in parallel with the N Maxwell
elements. Multiplying throughout by the minimum common multiple of the denom-
inators of the operators of the right side of Equation 4.7 and rearranging:
N
hY i N
hX N
Y i N
hY i
(kh + ch s) FL = (ki ci s) (kj + cj s) uL + (kg + cg s) F0,L (4.8)
h=1 i=1 j=1 g=1
where Σ represents summation, Π represents a series multiplication, and j 6= i.
77
Focusing only on the left hand side of Equation 4.8:
N
hY i
(kh + ch s) FL =
h=1
h i
(k1 + c1 s)(k2 + c2 s)(k3 + c3 s)...(kN −1 + cN −1 s)(kN + cN s) FL =
h
(k1 k2 k3 ...kN −2 kN −1 kN )+
(k1 k2 k3 ...kN −2 kN −1 cN + k1 k2 k3 ...kN −2 kN cN −1 + ...+
k1 k3 ...kN −2 kN −1 kN c2 + k2 k3 ...kN −2 kN −1 kN c1 )s+
(k1 k2 k3 ...kN −2 cN −1 cN + k1 k2 k3 ...kN −4 kN −3 kN −1 cN −2 cN + ...+

(4.9)
k1 k4 k5 ...kN −2 kN −1 kN c2 c3 + k3 k4 k5 ...kN −2 kN −1 kN c1 c2 )s2 +
(k1 k2 k3 ...kN −4 kN −3 cN −2 cN −1 cN + k1 k2 k3 ...kN −4 kN −2 cN −3 cN −1 cN + ...+
k1 k5 k6 ...kN −1 kN c2 c3 c4 + k4 k5 k6 ...kN −1 kN c1 c2 c3 )s3
+...+
(k1 c2 c3 ...cN −2 cN −1 cN + k2 c1 c3 ...cN −2 cN −1 cN + ...+
kN −1 c1 c2 ...cN −3 cN −2 cN + kN c1 c2 ...cN −2 cN −1 )sN −1 +

i
(c1 c2 c3 ...cN −2 cN −1 cN )sN FL
N
Y
Dividing 4.9 by (kd ):
d=1
h
cN −1
1 + ( kcNN + kN −1 + ... + c2
k2 + c1
k1 )s+
cN −1 cN cN −2 cN cN −3
( kcNN kN −1 + kN kN −2 + kN kN −3 + ... + c2 c3
k2 k3 + c1 c2 2
k1 k2 )s +
cN −1 cN −2 cN cN −1 cN −3
( kcNN kN −1 kN −2 + kN kN −1 kN −3 + ... + c2 c3 c4
k2 k3 k4 + c1 c2 c3 3
k1 k2 k3 )s
+...+
c cN cN −1 cN cN −3 cN −2 cN cN −2 cN −1 N −1
( kc22 kc33 ... kNN−1
−1 kN
+ c1 c3
k1 k3 ... kN −1 kN + ... + c1 c2
k1 k2 ... kN −3 kN −2 kN + c1 c2
k1 k2 ... kN −2 kN −1 )s +
i
c cN
( kc11 kc22 ... kNN−1
−1 kN
)s N F
L
(4.10)
Equation 4.10 can be rewritten in the following format:
N h
X X Y cr i
FL + sa FL { { ( )}} (4.11)
kr
a=1 mi ∈Ai r∈B
78
where:
• i∈{1,2,...,α}
• Ai ∈{i,i+1,i+2,...,N -α+i}
• mi >mi-1 , for i>1
• B∈{m1 ,m2 ,...,mα-1 ,mα }
. Focusing now on the first part of the right hand side of Equation 4.8:
N
hX N
Y i
(ki ci s) (kj + cj s) uL =
i=1 j=1
h
k1 c1 s(k2 + c2 s)(k3 + c3 s)...(kN −1 + cN −1 s)(kN + cN s)+
k2 c2 s(k1 + c1 s)(k3 + c3 s)...(kN −1 + cN −1 s)(kN + cN s)+ (4.12)

+...+
kN −1 cN −1 s(k1 + c1 s)(k2 + c2 s)...(kN −2 + cN −2 s)(kN + cN s)+

i
kN cN s(k1 + c1 s)(k2 + c2 s)...(kN −2 + cN −2 s)(kN −1 + cN −1 s) uL
79
Rearranging Equation 4.12:

"
(c1 k1 k2 ...kN −1 kN + c2 k1 k2 ...kN −1 kN + ...+
cN −1 k1 k2 ...kN −1 kN + cN k1 k2 ...kN −1 kN )s+

c1 (c2 k1 k3 k4 ...kN −1 kN + c3 k1 k2 k4 ...kN −1 kN + ...+
cN −1 k1 k2 ...kN −2 kN + cN k1 k2 ...kN −2 kN −1 )+
c2 (c1 k2 k3 ...kN −1 kN + c3 k2 k1 k4 ...kN −1 kN + ...+
cN −1 k2 k1 k3 ...kN −2 kN + cN k2 k1 k3 ...kN −2 kN −1 )+
c3 (c1 k3 k2 k4 ...kN −1 kN + c2 k3 k1 k4 ...kN −1 kN + ...+
cN −1 k3 k1 k2 ...kN −2 kN + cN k3 k1 k2 ...kN −2 kN −1 ) + ...+
cN −1 (c1 kN −1 k2 k3 ...kN −2 kN + c2 kN −1 k1 k3 ...kN −2 kN + ...+
cN −2 kN −1 k1 k2 ...kN −3 kN + cN kN −1 k1 k2 ...kN −3 kN −2 )+
cN (c1 kN k2 k3 ...kN −2 kN −1 + c2 kN k1 k3 ...kN −2 kN −1 + ...+
cN −2 kN k1 k2 ...kN −3 kN −1 + cN −1 kN k1 k2 ...kN −3 kN −2 ) s2 +

c1 (c2 c3 k1 k4 ...kN + c2 c4 k1 k3 k5 ...kN + ... + cN −1 cN k1 k2 ...kN −2 )+
c2 (c1 c3 k2 k4 ...kN + c1 c4 k2 k3 k5 k6 ...kN + ... + cN −1 cN k2 k1 k3 ...kN −2 )+
c3 (c1 c2 k3 k4 ...kN + c1 c4 k3 k2 k5 ...kN + ... + cN −1 cN k3 k1 k2 k4 ...kN −2 ) + ...+
cN −1 (c1 c2 kN −1 k3 ...kN −2 kN + c1 c3 kN −1 k2 k4 ...kN −2 kN + ... + cN −2 cN kN −1 k1 ...kN −3 )+
cN (c1 c2 kN k3 ...kN −1 + c1 c3 kN k2 k4 ...kN −1 + ... + cN −2 cN −1 kN k1 ...kN −3 ) s3

+...+

c1 (c2 ...cN −1 k1 kN + c2 ...cN −2 cN k1 kN −1 + ... + c3 ...cN k1 k2 )+
c2 (c1 c3 ...cN −1 k2 kN + c1 c4 ...cN k2 k3 + ... + c3 ...cN k2 k1 ) + ...+
cN −1 (c1 ....cN −2 kN −1 kN + ... + c2 ...cN −2 cN kN −1 k1 )+
cN (c1 ...cN −2 kN kN −1 + ... + c2 ...cN −1 kN k1 ) sN −1 +

#
N
k1 c1 ...cN + k2 c2 c1 c3 ...cN + ... + kN cN c1 ...cN −1 s uL
(4.13)
80
Equation 4.13 can be rewritten as:

"

(c1 + c2 + ... + cN −1 + cN )k1 k2 ...kN −1 kN s+

c1 c2 (k1 + k2 )k3 ...kN + c1 c3 (k1 + k3 )k2 k4 ...kN + ...+
cN −1 cN (kN −1 + kN )k1 k2 ...kN −2 s2 +

c1 c2 c3 (k1 + k2 + k3 )k4 ...kN + c1 c2 c4 (k1 + k2 + k4 )k3 k5 ...kN + ...+
cN −2 cN −1 cN (kN −2 + kN −1 + kN )k1 ...kN −3 s3

(4.14)
+...+

c1 c2 ...cN −2 cN −1 (k1 + k2 + ... + kN −2 + kN −1 )kN + ...+
c2 c3 ...cN −1 cN (k2 + k3 ... + kN −1 + kN )k1 sN −1 +

#
N
c1 c2 ...cN −1 cN (k1 + k2 + ... + kN −1 + kN ) s uL
N
Y
And now dividing Equation 4.14 by (kd ), as was done with Equation 4.9:
d=1
"

c1 + c2 + ... + cN −1 + cN s+
c1 c2 c1 c3
k1 k2 (k1 + k2 ) + k1 k3 (k1 + k3 ) + ...+
cN −2 cN −1 cN −1 cN
+ kN ) s2 +

kN −2 kN −1 (kN −2 + kN −1 ) + kN −1 kN (kN −1
c
1 c2 c3 c1 c2 c4
k1 k2 k3 (k1 + k2 + k3 ) + k1 k2 k4 (k1
+ k2 + k4 ) + ...+
cN −2 cN −1 cN 3
kN −2 kN −1 kN (kN −2 + kN −1 + kN ) s
(4.15)
+...+
c
1 c2 cN −2 cN −1
k1 k2 ... kN −2 kN −1 (k1 + k2 + ... + kN −2 + kN −1 )+
c1 c3 cN −1 cN
k1 k3 ... kN −1 kN (k1
+ k3 + ... + kN −1 + kN ) + ...+
c2 c3 cN −1 cN N −1
k2 k3 ... kN −1 kN (k2 + k3 + ... + kN −1 + kN ) s # +
c c cN −1 cN N
1 2
k1 k2 ... kN −1 kN (k1 + k2 + ... + kN −1 + kN ) s uL
Which leads to:
N h
X X Y cz X i
sβ uL { { ( ) kz }} (4.16)
kz
β=1 xc ∈Qc z∈W z∈W
where:
81
• c∈{1,2,...,β}
• Qc ∈{c,c+1,c+2,...,N -β+c}
• xc >xc-1 for c>1
• W∈{x1 ,x2 ,...,xβ-1 ,xβ }
The last part of the right hand side of the Equation 4.8 can be rearranged
similarly to its left part and can be presented in its final format:
N
X X Y cf
d
F0,L + s F0,L { { ( )}} (4.17)
kf
d=1 pv ∈Hv f ∈L
where:
• v∈{1,2,...,d}
• H v ∈{v,v+1,v+2,...,N -d+v}
• pv >pv-1 for v>1
• L∈{p1 ,p2 ,...,pd-1 ,pd }
Combining Equations 4.11,4.16, and 4.17, then 4.8 can be rewritten as:
XN h X Y cr i
sa FL {
FL + { ( )}} =
kr
a=1 mi ∈Ai r∈B
N h X Y cz X i h N X Y cf
X X i
sβ uL {
d
{ ( ) kz }} + F0,L + s F0,L { { ( )}}
kz kf
β=1 xc ∈Qc z∈W z∈W d=1 pv ∈Hv f ∈L
(4.18)
Applying inverse Laplace transformation to 4.18:
N h a
X ∂ F X Y cr i
F+ { { ( )}} =
∂ta kr
a=1 mi ∈Ai r∈B
N h β N
X ∂ u X Y cz X i h X ∂ d F0 X Y cf i
{ { ( ) kz }} + F0 + { { ( )}}
∂tβ kz ∂td kf
β=1 xc ∈Qc z∈W z∈W d=1 pv ∈Hv f ∈L
(4.19)
82
where:
• i∈{1,2,...,α}, c∈{1,2,...,β}, v∈{1,2,...,d}
• Ai ∈{i,i+1,i+2,...,N -α+i}, Qc ∈{c,c+1,c+2,...,N -β+c}, H v ∈{v,v+1,v+2,...,N -
d+v}
• B∈{m1 ,m2 ,...,mα-1 ,mα }, W∈{x1 ,x2 ,...,xβ-1 ,xβ }, and L∈{p1 ,p2 ,...,pd-1 ,pd }
• mi >mi-1 , xc >xc-1 , pv >pv-1 for i, c, v>1
For illustrative purposes, the cases of N =1, N =2, and N =3 will be presented.
If N =1 Eq. 4.19 becomes:
dF c1 dF0 c1 du (4.20)
F+ dt [ k1 ] = F0 + dt k1 + dt c1
In the case of N =2 Eq. 4.19 is determined as:
dF c1 c2 d2 F c1 c2 dF0 c1 c2 d2 F0 c1 c2
F+ dt [ k1 + k2 ] + [
dt2 k1 k2
] = F0 + dt [ k1 + k2 ] + [
dt2 k1 k2
]+
(4.21)
du d2 u c1 c2
dt [c1 + c2 ] + [
dt2 k1 k2 1
(k + k2 )]
In the case of N =3 Eq. 4.19 is determined as:
dF c1 c2 c3 d 2 F c1 c2 c1 c3 c2 c3 d3 F c1 c2 c3
F+ dt [ k1 + k2 + k3 ] + [
dt2 k1 k2
+ k1 k3 + k2 k3 ] + [
dt3 k1 k2 k3
] =
dF0 c1 c2 c3 d2 F0 c1 c2 c1 c3 c2 c3 d3 F0 c1 c2 c3
F0 + dt [ k1 + k2 + k3 ] + [
dt2 k1 k2
+ k1 k3 + k2 k3 ] + [
dt3 k1 k2 k3
]+
(4.22)
du d2 u c1 c2 c1 c3 c2 c3
dt [c1 + c2 + c3 ] + [
dt2 k1 k2 1
(k + k2 ) + k1 k3 (k1 + k3 ) + k2 k3 (k2 + k3 )]+
d3 u c1 c2 c3
[
dt3 k1 k2 k3 1
(k + k2 + k3 )]
It is obvious, from the above equations that the number of derivatives of both the
force and the displacement is the same as the number N of Maxwell elements. The
advantage of Eq. 4.19, apart from the fact that it may be useful in the mathematical
and physical fields, is that it can be used for random loading and not only sinusoidal
loading, and the force-displacement relationship can now be extracted in the time
domain for any number N of Maxwell elements. Furthermore, comparing Eq. 4.19
with Eq.2.73, which is based on generalised derivatives (see Chapter 2), it can be
concluded that they are very similar. However, the parameters αn , βn of Eq.2.73 are
83
now determined, and are functions of the stiffness ki , and the damping coefficients
ci of the GMM. Even though fractional derivative models are usually based on
fewer parameters, it was found that additional parameters were necessary to capture
the dynamic characteristics of the EDs tested. However, for preliminary analysis
of structures, a simplified Kelvin-Voigt model (spring and dashpot connected in
parallel) can be used, the parameters of which can be determined based on the
natural frequency of the structure and a design dampers’ displacement.
84
4.3 Hysteretic Model
In trying to fit the GMM to the experimental data, it was found that the GMM
was only adequate to describe VE behaviour, regardless of the number of Maxwell
elements used, and could not adequately represent the frequency or amplitude de-
pendence. Hence, a Modified Generalised Maxwell Model (MGMM) is proposed
here. In order to take into account the non linear part of the elastomer’s behaviour,
a non linear damper was added to the main GMM equation:
FN L = cN L |u̇|α sgn(u̇) (4.23)
It was also found that the ED’s behaviour depended on both the maximum
displacement and the maximum velocity obtained during the loading history. Kar-
avasilis et al [62] proposed a modified stiffness based on the maximum displacement
to take into account the softening of the elastomer:
− uumax
kmod = ka e ref + kb (4.24)
where k a , k b and uref are constants, and umax is the average of the maximum absolute
deformation amplitudes in the negative (umax,n ) and positive (umax,p ) directions, and
can be determined as follows:
umax,p + |umax,n |
umax = (4.25)
2
The values of umax,p and umax,n are updated for every time step and are fed into the
constitutive equation to calculate the overall response. Similar modelling approach
was followed by Summers [119]. Following the same philosophy in terms of velocity
instead of displacement, a modified damping coefficient is introduced here, which
was found to significantly contribute to the elastomer’s dynamic behaviour:
− vvmax
cmod = ca e ref + cb (4.26)
85
The parameters ca , cb , v ref , and v max are defined in a similar way to the terms in
k mod , of Eq. 4.24.
Hence, the final form of the Modified Generalized Maxwell Model (MGMM)
consists of the following parts:
• The linear GMM obtained from Equation 4.19
• The final version of F 0 , which consists of a linear spring:
F0 = k0 u (4.27)
• The non linear damper (see Eq. 4.23)
• The modified stiffness from Equation 4.24
• The modified damping coefficient from Equation 4.26
Since a generalised equation for N Maxwell elements is now available, efforts
were made in order to use more than one element. More specifically, analytical
models were determined with N = 2, 3 and 4 Maxwell elements respectively. The
Normalised Root Mean Square (NRMS) error, defined by Eq. 4.28, was determined
for every combination (Table 4.1) and it was found that using only one Maxwell
element was adequate for the specific material without any significant compromise
in the accuracy.
N
X
(FA − FE )2
i=1
N RM S = N
(4.28)
X
2
(FA )
i=1
where FA , and FE represent the force extracted from the analytical model and the
experiments respectively.
Hence, the final equation describing the MGMM is:
c1 c1
F+ Ḟ = (k0 + kmod )u + ( k0 + c1 + cmod )u̇ + cN L |u̇|α sgn(u̇) (4.29)
k1 k1
86
Table 4.1: NRMS error based on different values of Maxwell Elements
Frequency (Hz) 1 Maxwell Element 2 Maxwell Elements 3 Maxwell Elements 4 Maxwell Elements
0.25 1.60 1.59 1.59 1.59
0.5 1.45 1.44 1.44 1.44
1 1.72 1.70 1.70 1.70
2 1.73 1.73 1.73 1.73
3 1.41 1.43 1.42 1.42
4 1.57 1.60 1.60 1.60
Average NRMS Error 1.580 1.582 1.580 1.580
It should be noted that this forms an equation which can mechanically capture
the dynamic behaviour of the elastomeric dampers under specific conditions on a
macroscopic level. Unlike the model proposed by Lee [107], the MGMM takes into
account the loading history. In order for Lee to take loading history into consid-
eration and avoid potential overshooting under variable cyclic loading, additional
modifications were made in order to adjust the stress-strain path, which were not
taken into account on the proposed MGMM, which may cause discrepancies if sud-
den strain reversals occur. Examples of more realistic loading histories are presented
in Chapter 6 along with the Real-Time Substructure tests. For comparison reason
both the GMM (using one Maxwell element), and the model proposed by Lee [107]
were compared with experimental data. More specifically, these models were com-
pared with the 2 Hz sweep amplitude test (which is analytically described later)
and presented at Figures 4.9 and 4.10. It can be concluded that both of these mod-
els in a good overall agreement with corresponding NRMS error of 2.1 %, and 2.5
% for Lee’s model and GMM respectively. However, when frequency was altered
the results were far from satisfying. Specifically, GMM model could only capture
linear VE behaviour under a target frequency, while Lee’s model presented some
discrepancies when different frequencies and strain amplitudes compared with the
ones which were used to calibrate the model were used.
4.4 Parameter Estimation
Least squares minimization is one of the most widely used methods [42, 72, 60] of
fitting the numerical parameters of a proposed model into the experimental data.
However, as mentioned earlier, in case of GMM this minimization technique has
been based on the frequency domain, trying to minimize the shear storage modulus
87
G00
(from Eq. 4.1), and the loss factor, n = G0 (combining Eq. 4.1, and Eq. 4.2) for
every frequency and temperature that the damper was experimentally tested, as:
N h
nX i2 o
0 0
min Ganal − Gexp (4.30)
j=1
N h
nX i2 o
min nanal − nexp (4.31)
j=1
where N denotes the number of experimental data available. With regard to fitting
the numerical model to the experimental data, researchers have used various tech-
niques. The most common one is to assume constant one of the three main param-
eters [42] that the material is dependent upon (strain amplitude, loading frequency,
ambient temperature). On the other hand, minimization of the corresponding error
for every cyclic experiment that was carried out [62] has been used as well, taking
the average value of these parameters in order to generate the model through the
whole range of shear strain amplitudes, frequencies, and temperatures. Finally, an-
other widely used method is to include all the available data into one least squares
equation [72, 42].
Furthermore, in the case of the models which were minimised based on their shear
storage modulus, and loss factor for specific frequencies, even though the numerical
values of G0 , and n seem to fit well with the experimental data, the corresponding
model will not necessarily be valid in the time domain for random loading (and more
specifically for random displacements), since these parameters will only be adequate
to predict the shear storage modulus, and the loss factor of the damper under cyclic
testing with constant frequency, and constant maximum displacement, and not the
actual elastomer’s dynamic behaviour.
Another approach was therefore used here, which takes into account the alter-
ation of displacements (since it was found to be the most dominant factor in the
elastomer’s dependence). Sweep amplitude sinusoidal tests (the shear strain ampli-
tude was ranging from 10%-50%), which were carried out for the same frequency
range used in the characterization tests of Chapter 3, formed the basis for the afore-
88
mentioned minimization. The sweep amplitude tests were based on proposals from
Dorka and Garcia[39]. Keeping the frequency and the temperature constant within
each test, and setting the maximum strain amplitude to be 50%, the sweep am-
plitude tests consisted of the following steps: 1.5 cycles of umax /2, 1.5 cycles of
umax , 1.5 cycles of umax /3, 1.5 cycles of umax /6, 1.5 cycles of umax /2, 1.5 cycles of
2umax /3, 1.5 cycles of 5umax /6, 3 cycles of umax , 1.5 cycles of 5umax /6, 1.5 cycles of
2umax /3, 1.5 cycles of umax /2, 1.5 cycles of umax /3, 1.5 cycles of umax /6. Figure 4.2
shows a typical example of both the sweep sinusoidal command displacement and
the ED’s response for a test at a frequency of 3.0 Hz.
Figure 4.2: Sweep amplitude test for frequency of 3 Hz
The sweep amplitude test is believed to be a more representative test, at least for
fitting the parameters because of the strong amplitude dependence of the elastomer.
In contrast, previous characterization tests were based on a fixed strain amplitude
and frequency. The MGMM (Eq. 4.29) contains 11 parameters which need to be
determined: k 1 , c1 , k 0 , k a , k b , uref , ca , cb , v ref , cNL , α. In order to optimize these
parameters a least squares method was used:
nXXh i2 o
min Fanal(i,k) − Fexp(i,k) (4.32)
i k
where i is the index for frequency content (e.g., i = 2 is equivalent with frequency of
0.5 Hz), and k stands for the force-displacement data set at a specific frequency and
displacement. Stress relaxation and very low frequency tests were used as starting
89
points of determining initial assumptions of the aforementioned parameters.
Table 4.2 provides the parameters of the hysteretic model, while Figures 4.3
to 4.8 show the hysteretic response of the damper for both the analytical and ex-
perimental case. It can be seen that the results are in a very good agreement. It
is also worth mentioning that the proposed model is valid under any random earth-
quake loading, and its parameters do not depend on the imposed displacement or
frequency. Chapter 6 provides further validation of the proposed model, where real
time hybrid tests with alterations of both displacement and frequency are carried
out. Furthermore, Table 4.3 shows the NRMS error calculated based on comparison
of the proposed model and the fixed amplitude characterization tests described in
Chapter 3. The error determined for these tests was slightly larger compared to the
sweep amplitude tests; however the agreement observed was very good as well.
Table 4.2: Material Parameters
k1 (kN/mm) c1 (kNsec/mm) k0 (N/mm) c NL(kN(sec/mm)0.374 ) α

0.515 0.0151 0.107 1.585 0.374
k a (kN/mm) k b (kN/mm) u ref (mm) c a (kNsec/mm) c b (kNsec/mm) v ref (mm/sec)
3.741 3.613 7.966 0.902 0.190 13.468
Table 4.3: NRMS error based on comparison of the proposed model and the char-
acterization tests described in Chapter 3
Frequency
Shear Strain %
0.25Hz 0.5Hz 1Hz 2Hz 3Hz 4Hz
10 1.65 1.48 1.78 1.75 1.46 1.65
20 1.61 1.49 1.81 1.79 1.5 1.64
30 1.6 1.49 1.76 1.81 1.49 1.64
40 1.59 1.52 1.79 1.8 1.51 1.71
50 1.63 1.53 1.84 1.81 1.55 1.68
90
Figure 4.3: Comparison of force between experiment and MGMM model for sweep
amplitude test - 0.25 Hz, 20o C
amplitude test - 1 Hz, 20o C
91
92
Figure 4.9: Model proposed by Lee [107] compared with 2 Hz sweep amplitude test
Figure 4.10: GMM compared with 2 Hz sweep amplitude test
4.5 Effect of Temperature
One of the most widely used procedure in order to take into account temperature
effects of polymeric materials tested under loading is the one proposed by Ferry [43].
According to this method a shifting function, αT with two parameters p1 , p2 was
proposed:
−p1 (T −Tref )
αT (T, Tref ) = e p2 +(T −Tref ) (4.33)
G0 increases as the temperature decreases from a reference temperature Tref to a
temperature T at a reference frequency ωref . This increase of G0 is equivalent to
the increase of G0 as the frequency changes from ωref to αT (T, Tref ωref ) at Tref .
Incorporating the shifting function αT into the analytical values of G0 , and G00 cap-
93
tures any alterations of the material’s characteristics due to change of temperature.
A much simpler method was followed in this research. In order to capture the be-
haviour of the EDs under different ambient temperatures, both the characterization
tests and the sweep amplitude tests were repeated at different temperatures: 25, 30,
and 35o C. The same model (MGMM), which was developed and validated for the
experiments which were carried out at room temperature (20o C), will be used here
as well, by introducing a new parameter, γT . However, a simplifying assumption
that all parameters change at the same rate with temperature was made. This led
to very good agreement between experimental and numerical results. The original
proposed hysteretic model in this case has been adjusted:
c1
FT + k1 ḞT =
(4.34)
γT (k0 + kmod,T )u + ( kc11 k0 γT + γT c1 + cmodT )u̇ + γT kN L (u̇)αγT
where:
− γvmax
cmod,T = γT ca e v ef
T r + γT cb (4.35)
and
− γumax
kmod,T = γT ka e u ef
T r + γT kb (4.36)
where T denotes the ambient temperature, and γT the coefficient which controls the
parameters of the model in different temperatures. Again, the same minimization
techniques introduced earlier were used to determine the γT parameter for different
temperatures. These values are summarised in Table 4.4.
Table 4.4: Values of γT for different ambient temperatures
γ20 γ25 γ30 γ35

1 0.935 0.875 0.820
As expected from the results of the characterization tests carried out at different
temperatures (Chapter 3), the values of γT reduce with the increase of temperature.
This reduction can be schematically seen at Figure 4.11 and can be approximately
94
determined as 0.116/o C. Figures 4.12 to 4.14 show the comparison between the
experimental data and the proposed model for the 4Hz sweep amplitude tests. It
can be seen that the MGMM can capture adequately the dynamic behaviour of the
EDs under every temperature. More figures regarding the full comparison of the
proposed model against the dynamic behaviour of the EDs in different temperatures
are available in Appendix 2.
Figure 4.11: Alteration of temperature parameter γT with different temperatures
4.6 Conclusions
A new model has been developed in this chapter, the Modified Generalised Maxwell
Model (MGMM), based on the well known Generalised Model (GMM), in order to
capture the dynamic behaviour of elastomeric dampers. In order to extract this
model, the force-displacement relationship of the GMM for N Maxwell elements
in the time domain was determined (Eq. 4.19). However, it was observed that the
GMM was able to capture only visco-elastic behaviour, and not follow the elastomer’s
performance, regardless the number N used. Hence, the MGMM, based on N =1
Maxwell element, was proposed which was found to be able to predict the dynamic
performance of the elastomeric dampers under a range of strain amplitudes and
frequencies. In order to take into consideration the effect of the ambient temperature,
a new parameter was introduced into the proposed model, and the analytical results
were found to be in very good agreement with the experimental data for every strain
amplitude, frequency, and temperature, with the maximum NRMS error observed
during these tests to reach 1.8%.
95
96
The proposed model was also evaluated under the characterization tests carried
out in Chapter 3; a very good agreement was achieved as well even though the
NRMS error was slightly larger, reaching 1.84%. The proposed hysteretic model
was later implemented in an incremental way as a new material into OpenSees, so
that an in depth analysis of the effect of the EDs could be carried out in Chapter 5.
97
Chapter 5
Seismic Response of Steel

Moment Resisting Frames using
Elastomeric Dampers
5.1 Introduction
In Chapter 3, the characterization tests carried out at the Dynamics Laboratory of
the Oxford University Engineering Department were analytically described. Based
on these tests, a hysteretic model was proposed at Chapter 4, and incorporated into
the Finite Element software OpenSees[3]. The current chapter describes the analysis
of a 10-storey steel moment-resisting frame, designed according to Eurocode 3 and
Eurocode 8, and its dynamic response under non-linear time histories. In order to
improve the building’s performance, elastomeric dampers were added at a second
stage. A uniform damper distribution throughout the height of the structure was
used, and elastomers’ deformation limit was constrained to be in the shear strain
target range. With regard to the dampers’ selection, a design process [114] was
followed, so the target performance level was achieved. Both the initial and the
retrofitted steel frame seismic performance were evaluated under seismic actions
with 2% and 10% probability of exceedance in 50 years.
98
5.2 Prototype Building
Figures 5.2 and 5.3 show the 10-storey, 3-bay by 4-bay prototype building, which
was used for this study. The length of the beams was assumed 7.2m, making the
plan of the building 21.6m x 28.8m, while the height of each storey was assumed
3.5m, leading to a total height of 35m. The building, which was assumed to be
located on a dense sand (type C ground), was designed according to Eurocode 3 [1],
and Eurocode 8 [2]. The structure was designed based only on q factor and not on
interstorey drift ratio (IDR) in order to obtain more flexible structure. Concentric
braces provide the main resistance to horizontal forces in the global Y direction, and
Moment Resisting Frames (MRFs) in the global X direction. The strong axis of the
columns was aligned with the global X axis [56].
The design dead load used at each floor was assumed 4.5 kN/m2 , and the live
load 2 kN/m2 , made up as shown in table 5.1. An additional dead load of 1.2 kN/m
was assigned to perimeter beams, due to glazing.
The structure was designed for Ductility Class High (DCH), and behaviour fac-
tor of 4 along the global Y axis, where the diagonal braces are mainly responsible
for resisting the seismic forces, and 6.5 in the global X direction in which moment
resisting frames is the main mechanism for the horizontal actions. DCH was imple-
mented in this research in order to obtain larger q factor, and therefore more flexible
structure. This will potentially result in larger lateral displacements, and will even
further highlight the efficiency of the dampers.
The following assumptions were made in structural modelling of the building:
• Beam column connections were assumed as fully rigid for the moment resisting
frames, and simple connections for the direction where diagonal braces were
used
• The yield stress of the steel members is assumed to be 275 MPa
• The columns were assumed fixed at their bases
• Rigid Diaphragms were assigned to each floor
99
Seismic response of steel moment resisting frames using elastomeric dampers
The FE model was first validated under simple static and cyclic loading before pro-
ceeding to more complex seismic loading. The first 3 natural modes of the building
are shown in Figure 5.1.
(a) Mode 1 (b) Mode 2 (c) Mode 3
Figure 5.1: First 3 modes of the prototype structure
With regard to the representation of seismic action, the EC8 elastic response
spectrum type 1 was used, and the design of the structure was based on modal
response spectrum analysis according to Eurocode 8 [2]. The design seismic action
is expressed in terms of two parameters: 1) the probability of exceedance in a time
period of 50 years, and 2) the importance factor, γ 1 . Both of these factors are ex-
plained later in detail (see 5.3.3). In case of Eurocode 8, structures shall be designed
for a seismic action having 10% probability of exceedance in 50 years. The two or-
thogonal components of the horizontal seismic action are assumed independent and
represented by the same response spectrum. The ability of the structure to exhibit
non linearities during seismic actions permits to design for smaller seismic forces
than those corresponding to a linear elastic response. Hence, to avoid complicated
and explicit inelastic analyses, the capacity of the structure to dissipate energy in
a ductile way, can be taken into consideration by performing elastic analysis based
on the elastic response spectrum reduced by the behaviour factor, q. Briefly, the
horizontal components of the seismic action of the design spectrum are:
0 ≤T ≤T B : S d (T ) = ag S[ 23 + T 2.5
TB ( q − 23 )] (5.1)
T B ≤T ≤T C : S d (T ) = ag S 2.5
q
(5.2)
100

 ag S 2.5 [ T c ]

q T
T C ≤ T ≤ T D : S d (T ) = (5.3)
≥ βag



 ag S 2.5 [ T C T2 D ]

q T
T D ≤ T : S d (T ) = (5.4)
≥ βag


where:
S D (T ) is the design spectrum
T is the vibration period of a linear single-degree-of-freedom system
ag = γ 1 agR is the design ground acceleration
agR is the reference peak ground acceleration on type A ground (=0.36g based on
the National Annex of Greece for zone 3 areas)
γ 1 is the importance factor
T B , T C , and T D are given values by Eurocode 8, depending on the corresponding
elastic response spectrum
q is the behaviour factor
β is the lower bound factor for the horizontal design spectrum
Figures 5.4, and 5.5 show the corresponding design spectrum according to EC8.
101
Figure 5.2: Elevation of the Prototype Building
Figure 5.3: Plan of the Prototype Building
102
Figure 5.4: Design Spectrum for X axis according to Eurocode 8
Figure 5.5: Design Spectrum for Y axis according to Eurocode 8
The masses were determined from the load combination of dead + 0.3 live load,
the design of the structure was based on the following load combinations:
1.35G + 1.5Q
G+Q
G + 0.3Q + E X + 0.3E Y
G + 0.3Q + E X − 0.3E Y
G + 0.3Q − E X + 0.3E Y
G + 0.3Q − E X − 0.3E Y
G + 0.3Q + E Y + 0.3E X
G + 0.3Q + E Y − 0.3E X
103
G + 0.3Q − E Y + 0.3E X
G + 0.3Q − E Y − 0.3E X
where
G is the dead load
Q is the live load
E is the seismic action The final sections of the structural members are summarized
in table Table 5.2, while the dynamic characteristics of the structure are given in
table Table 5.3. It can be noticed that the difference between the periods corre-
sponding to the first 2 modes is slightly large. This is due to the design philosophy
followed for this structure: design as flexible as possible the direction where the
dampers are going to be added, whereas provide stiff diagonal braces at the other
direction.
Table 5.1: Permanent and Live loads of the Prototype Building
Item Equivalent Uniform Load

Slab with Metal Deck 3.2 kN/m2
Ceiling and Services 0.5 kN/m2
Partitions 0.8 kN/m2
Glazing 1.2 kN/m
Live Load 2 kN/m2
Table 5.2: Sections of Steel Structure
Floors Columns Beams (X axis) Beams (Y axis) Braces (Y axis)

1 HEB450 IPE450 IPE360 SHS200x200x10
2 HEB450 IPE450 IPE360 SHS200x200x10
3 HEB450 IPE450 IPE360 SHS200x200x10
4 HEB450 IPE450 IPE360 SHS200x200x10
5 HEB360 IPE400 IPE360 SHS180x180x10
6 HEB360 IPE400 IPE360 SHS180x180x10
7 HEB360 IPE400 IPE360 SHS180x180x10
8 HEB320 IPE360 IPE360 SHS160x160x10
9 HEB320 IPE360 IPE360 SHS160x160x10
10 HEB320 IPE360 IPE360 SHS160x160x10
104
Table 5.3: Dynamic Characteristics of the Prototype Building
Number of Natural Natural Mass Participation Ratios

Mode Shape Period (sec) Frequency (Hz) UX UY RZ
1 2.651286 0.37718 0.74817 6.617E-11 0.19531
2 0.973412 1.0273 5.796E-09 0.79867 0.36919
3 0.923146 1.0833 0.12323 3.262E-08 0.02966
4 0.747592 1.3376 0.000005825 0.000001858 0.22303
5 0.51529 1.9407 0.04826 3.038E-12 0.01216
6 0.32702 3.0579 0.02625 1.52E-10 0.00682
7 0.320467 3.1204 7.637E-12 0.13171 0.06094
8 0.248166 4.0296 0.000004159 0.000005787 0.03576
9 0.232642 4.2985 0.01844 1.355E-09 0.00518
Sum 0.96435999 0.930387679 0.93805
5.3 Non Linear Analysis
5.3.1 Performance Criteria
FEMA356[5] defines three discrete levels of structural damage, referred to as Per-
formance Levels (PLs):
• Immediate Occupancy (IO): Post-earthquake damage state that remains safe
to occupy, essentially retains the pre-earthquake design strength and stiffness
of the structure
• Life Safety (LS): Post-earthquake damage state that includes damage to struc-
tural components but retains a margin against onset of partial or total collapse
• Collapse Prevention (CP): Post-earthquake damage state that includes damage
to structural components such that the structure continues to support gravity
loads but retains no margin against collapse
Every structure can be designed according to a target performance level under
a specific earthquake hazard level. According to Table 5.4[5], rehabilitation objec-
tives h in combination with l (blue color) are the basic objectives under which the
structures are usually designed. Hence, the main target of conventional structures
is the design of a structure for a seismic action of 10% probability of exceedance
in 50 years for Life Safety performance level (Design Basis Earthquake: DBE) or
2% probability of exceedance in 50 years (Maximum Considered Earthquake: MCE)
105
for Collapse Prevention performance level. However, as already mentioned earlier,
under different circumstances the design of the structure may be determined based
on different combinations of earthquake hazard levels and target performance levels.
The rehabilitation objectives coloured with yellow are considered as enhanced design
objectives (a,d,g,j,k), while those without colour (objectives: b,c,e,f,i) are considered
as reduced rehabilitation objectives. Table 5.5 summarizes the expected behaviour
of the primary elements of steel frames, and the expected drift for every performance
level. These drifts have been used as the main evaluation tool for seismic response
of structures in previous studies[72, 62].
The steel frame in this study was analysed under both DBE, and MCE hazard
levels. In order to prevent or to minimize damage, and potential economic loss which
is associated with the repair and limited occupancy after a strong earthquake [72],
elastomeric dampers were later added and the Damped Moment Resisting Frame
(DMRF) was evaluated under the same hazard levels of the seismic action. The
target performance levels used in this study, in case of the retrofitted frame, are
different for the hazard level of 10% probability of exceedance in 50 years and the
hazard level of 2% probability of exceedance in 50 years compared to the ones
described in Table 5.4, 5.5, and are summarised as follows:
• All columns remain elastic under the DBE
• All columns remain either elastic or are allowed to exhibit minor plastic de-
formations under the MCE
• All beams remain either elastic or exhibit minor plastic deformations under
DBE
• Plastic hinges are allowed to be created at beams under the MCE, if plastic
deformations are minor in most of the columns
• 0.7%-1.3% interstorey drift under the DBE
• 0.2%-0.5% residual interstorey drift under the DBE
• 2.3%-2.7% drift under the MCE
106
• 0.5%-1.0% residual interstorey drift under the MCE
It should be mentioned, that the main target the retrofit design which will be de-
scribed in detail in Section 5.3.4 is based on the DBE, and not the MCE. Hence, the
main focus should be on fulfilling the acceptance criteria set for DBE. However, im-
proved behaviour is anticipated for the MCE as well, and therefore some additional
criteria have been set.

Table 5.4: Rehabilitation Objectives (blue colour: main rehabilitation objectives,
yellow colour: enhanced rehabilitation objectives)
Target Building Performance

Earthquake Hazard Levels
Immediate Occupancy Life Safety Collapse Prevention
50% in 50 years a b c
20% in 50 years d e f
10% in 50 years g h i
2% in 50 years j k l
Table 5.5: Expected drifts
Type Immediate Occupancy Life Safety Collapse Prevention

Transient Drift 0.70% 2.50% 5%
Permanent Drift 0% 1% 5%
5.3.2 Conventional Moment Resisting Frame Model
All the non linear analyses and the evaluation of the retrofit of the steel frame,
are carried out using the finite element software OpenSees, focusing on an interior
steel frame, in the global X direction of the prototype building (see fig Figure 5.3).
The main target of this chapter is to demonstrate the enhancement of the struc-
tural performance level of the prototype Conventional Moment Resisting Frame
(MRF), adding Elastomeric Dampers (EDs), and evaluate their effectiveness under
strong ground motions. In order to achieve this, the 10-storey MRF was modelled
in OpenSees and tested under ground motions, which were scaled in order to be
compatible with the design hazard level, according to EC8. Elastomeric dampers
were later added to the MRF, following a design method proposed by Lee [114]. The
hysteretic model described in chapter 4 formed the basis of modelling the EDs used
107
in this analysis. These frames were compared with regard to displacements, drift
ratios, moments, and shear forces in order to fully evaluate the EDs’ efficiency.
The MRF is illustrated in Figure 5.2, and is modelled based on the principles of
the prototype building. Steel01 was selected from the OpenSees[3] Material class,
with 275MPa yield strength, and zero strain hardening ratio, corresponding to an
elastic-perfectly plastic material representing the non linear behaviour of the build-
ing, following the principles of Eurocode 3 [1]. P-Delta effects are taken into account
through the P-Delta geometric transformation command in OpenSees. Distributed
plasticity force-based beam column elements [96, 126] were used to capture the non
linear behaviour of the beams and columns elements, based on fiber sections. More
specifically, after a preliminary sensitivity analysis in a 1-storey, 1-bay steel frame
subjected to ground motions, 16 fibers were assigned along the width of the flange
and the height of the web, while 4, and 2 fibers were assigned along the thickness of
the flange and the thickness of the web respectively. The beams were loaded with
a dead load of 23 kN/m and live load of 9.6 kN/m and the corresponding masses
were assigned at the nodes of the columns as: 29.5 tons at the interior columns and
19.35 tons at the corner columns. A Rayleigh damping matrix was assigned to the
model, representing the 2% inherent critical damping of the structure at the first
two modes of vibration. In order to take into account any changes on the Rayleigh
damping due to yielding, the stiffness matrix was updated at each time step, instead
of using the initial matrix.
5.3.3 Ground Motions
A total of 20 ground motions were used to evaluate the seismic performance of both
the SMRF and DMRF. In EC8[2], it is clearly stated that in order to use ground
motions for nonlinear analysis the following criteria should be met:
• A minimum of 3 accelerograms should be used
• The mean of the zero period spectrum response acceleration values (calculated
from the individual time histories) should not be smaller than the value of ag S
for the site in question, where ag is the design ground acceleration, and S is
108
the soil factor.
• In the range of periods between 0.2T 1 and 2T 1 , where T 1 is the fundamental
period of the structure in the direction where the accelerogram will be applied;
no value of the mean 5% damping elastic spectrum, calculated from all time
histories, should be less than 90% of the corresponding value of the 5% damping
elastic response spectrum.
In order for the above conditions to be satisfied, the spectrum matching method [41,
133] was selected. The basis of this method is for the time history response spectrum
to gradually match the response spectrum of EC8, by increasing some components
of the frequency spectrum, and reducing some others. In essence, this method does
not create new ground motions, but it modifies selected records in order to satisfy
the above criteria. The whole procedure is shown in Fig. 5.6 and can be summarized
as follows: After calculating the EC8 target spectrum, the response spectrum of the
ground motion is calculated [27], and is compared with the target spectrum from
EC8, by dividing the values of the two spectra for each time increment. Then,
the outcome ratio is transmitted from the time domain to the frequency domain
(Figures 5.6d and 5.6e). The Fourier transform [100] of the accelerogram is then
being determined (Figure 5.6f). Then the components of the Fourier transform from
Figure 5.6f are multiplied by the ratio components of Figure 5.6e, (only for the
frequency range of interest), giving a new Fourier transform (Figure 5.6g). The
inverse Fourier transform is now used in order to obtain new ground motion, which
is shown at Figure 5.6h. If the response spectrum of the scaled record is in good
agreement with the one proposed from EC8, and the above criteria are satisfied,
then the whole procedure stops. Otherwise, it has to be repeated until a satisfactory
convergence has been achieved. This method could be very useful in regions that
strong ground motions recordings are unavailable[49].
109
(a) Target Spectrum
(b) Ground motion (c) Response Spectrum of ground motion
(d) Ratio of Target Spectrum to Response (e) Ratio of Target Spectrum to Response
Spectrum (Time domain) Spectrum (Frequency domain)
(f) Fourier Transform of original ground mo-

(g) Fourier Transform of scaled ground motion
tion
(h) Scaled ground motion
Figure 5.6: Scaling Procedure of ground motions
110
Response-history analysis was performed using both near-field and far-field
ground motions. The magnitude of the ground motions ranges between 6.5 and
7.36 and the site to source distance ranges between 0.07 and 222.42 km. These
records have been selected from PEER Ground Motion Database[67]. The original
ground motions used in this study are summarized in table 5.6. However, from this
point on the scaled versions of these ground motions are going to be used, and there-
fore, any reference to any of the names of these ground motions will be referred to
their scaled version, unless specified otherwise. The response spectra of each origi-
nal unscaled ground motion, and the resulting spectra from the scaled earthquake
histories are compared with the EC8 spectrum in Figure 5.7, and 5.8. It can be
clearly seen that there is a very good agreement between the response spectrums
of the scaled ground motions with the EC8 spectrum. Figure 5.9 shows the mean
value of the response spectra of both the scaled and the unscaled time histories in
comparison with the EC8 spectrum. All these ground motions were scaled in order
to comply under EC8 principles. Figure 5.10 presents a typical ground motion (Im-
perial Vallley-02) compared with its scaled to DBE version, with the latter being
the one which was used for the non-linear analyses of the steel frame.
In order to evaluate the steel frames under stronger ground motions, representing
a seismic action of 2% probability of exceedance in 50 years, EC8 adopts the following
philosophy: The main goal is the alteration of the design seismic action, which is
represented by a hazard level of 10% probability of exceedance in 50 years. According
to EC8, this can be achieved not by upgrading the performance level for a given
earthquake level, but by modifying the reference seismic action[76]. This can be
done, by altering the importance factor, γ 1 depending on the target performance
level[20].
111
Table 5.6: Ground Motions used for non linear analyses
Earthquake Name Year Station Name Magnitude Distance (km) Component

Imperial Valley-02 1940 El Centro Array #9 6.95 6.09 I-ELC180
Northwest Calif-02 1941 Ferndale City Hall 6.6 91.15 C-FRN045
Borrego 1942 El Centro Array #9 6.5 56.88 B-ELC000
Kern County 1952 LA - Hollywood Stor FF 7.36 114.62 PEL090
Northern Calif-03 1954 Ferndale City Hall 6.5 26.72 H-FRN044
Imperial Valley-06 1979 El Centro - Meloland Geot. Array 6.53 0.07 H-EMO000
Imperial Valley-06 1979 El Centro Array #7 6.53 0.56 H-E07140
Superstition Hills-02 1987 Parachute Test Site 6.54 0.95 B-PTS225
Imperial Valley-06 1979 El Centro Array #8 6.53 3.86 H-E08140
Imperial Valley-06 1979 Chihuahua 6.53 7.29 H-CHI012
Imperial Valley-06 1979 EC County Center FF 6.53 7.31 H-ECC002
Imperial Valley-06 1979 Brawley Airport 6.53 8.54 H-BRA225
El Alamo 1956 El Centro Array #9 6.8 121 ELC180
Borrego Mtn 1968 El Centro Array #9 6.63 45.12 A-ELC180
Borrego Mtn 1968 LA - Hollywood Stor FF 6.63 222.42 A-PEL090
Borrego Mtn 1968 LB - Terminal Island 6.63 199.84 A-TLI249
San Fernando 1971 2516 Via Tejon PV 6.61 55.2 PVE065
Tabas Iran 1978 Ferdows 7.35 89.76 FER-L1
Tabas Iran 1978 Kashmar 7.35 193.91 KSH-L1
Tabas Iran 1978 Sedeh 7.35 150.33 SED-L1
EC8 associates the importance factor with the target performance level by:
PL
γ1 = ( P )−1/k (5.5)
LR
where: P LR is the reference probability of exceedance in 50 years
P L is the probability of exceedance in 50 years, of the target performance level
k is recommended to be 3
Figure 5.11 graphically shows how importance factor and different probabilities
of exceedance in 50 years are related. Hence, in the case of the the hazard level
of 2% probability of exceedance in 50 years: P LR =0.1, P L =0.02, and therefore the
reference seismic action has to be multiplied by γ 1 =1.7. This will be particularly
important in the following sections, since the frames were tested for both the hazard
levels (2% and 10% probability of exceedance). What is worth mentioning is that
the same spectral shape is meant to be used for the seismic action for both of the
aforementioned performance levels, with a universal multiplicative factor reflecting
the difference in hazard level. The corresponding response spectrum, given by EC8,
under which the original ground motions have to be scaled for a seismic action of
112
2% probability of exceedance in 50 years, can be seen at Figure 5.12.
Figure 5.7: Comparison of EC8 Spectrum with each of the 20 original ground mo-
tions response spectrum
Figure 5.8: Comparison of EC8 Spectrum with each of the 20 ground motions scaled
to DBE
113
Figure 5.9: Comparison of EC8 Spectrum with the mean Spectrum of 20 ground
motions for DBE
114
(a) Original
(b) Scaled to DBE
Figure 5.10: Ground motionL: IMPVALLI-I-ELC180
115
Figure 5.11: Importance factor in terms of probability of exceedance in 50 years
Figure 5.12: Comparison of EC8 Spectrum with the mean Spectrum of 20 ground
motions for MCE
116
5.3.4 Retrofitting Procedure
Dampers Design Recommendations
Although several few retrofitting procedures have been proposed in the same aca-
demic literature [23, 31, 25, 24, 48, 73, 85, 110], while others have been applied to
real structures [59, 136, 40, 47], only the most widely used will be presented here.
The first method was initially proposed by Fan [42] for implementation of Vis-
coElastic dampers, and later extended to elastomeric dampers by Lee [73]. It is
a Simplified Design Procedure (SDP), which is used for performance based design
of MRFs, and uses an elastic-static analysis to determine the frame’s response and
compare it with the initially assumed earthquake performance criteria, assuming
the dampers’ behaviour is purely viscoelastic. More specifically it consists of the
following steps:
1. Establish seismic performance criteria:
A target performance level, as explained in 5.3.1, is selected
2. Idealize the non-linear damping material as a linear VE material:
Since elastomeric materials exhibit non linear behaviour which is dependant on
strain amplitude, frequency, and temperature, an equivalent linear VE model
has to be assumed for the static analysis. This model would have an equivalent
shear stiffness, and loss factor, in the same way as they were determined and
presented in Chapter 3, based on the natural frequency of the building.
3. Choose the design temperature:
Since, the characteristics of elastomers depend on temperature, a design tem-
perature must be selected.
4. Select an appropriate value for α and β:
α is the ratio of the brace stiffness per storey in the global direction to the
storey stiffness without dampers and braces, and β is the ratio of the damper
stiffness per storey in the global direction to the storey stiffness without
dampers and braces. Moreover, the locations where the dampers are going
117
to be applied must be determined as well. Kasai [63], Chang [22], Hanson and
Soong [48], Karavasilis [62], and Constantinou [31] have used these ratios for
preliminary design analyses. A value between 10 and 30 is recommended for
α, in order for the braces to be sufficiently stiff, and between 0.5 and 5.0 for
β.
5. Perform elastic-static analysis based on equivalent lateral forces
6. Compare the response from the static analysis with the target earthquake
performance level
7. Select minimum β that satisfies design criteria and provides target seismic
performance
8. Determine the structural response at the low-end of the design temperature
range and compare structural response with design criteria. Since a temper-
ature decrease leads to an increase of stiffness, the structural response under
the minimum allowed temperature is investigated, under which the structure
will not be so flexible.
9. Determine damper area and thickness
10. Verify the procedure through non linear time history analysis (optional step)
The second method [114, 48, 24] is an iterative procedure, based on the principle
that the stiffness added due to the dampers should be proportional to the storey
stiffness of the structure. The design process consists of the following steps:
1. Determine structural properties of the building and perform structural analy-
sis. An analysis of the structure without dampers is a necessary step in order
to capture the dynamic behaviour of the building, and how this can be further
improved.
2. Select target performance criteria. Either define new performance criteria (as
was done in this study), or adapt already defined criteria from design codes
(see section 5.3.1). This will be the target structural behaviour under the
design loads when dampers are added
118
3. Determine target desired damping ratio. Either through time history analyses
or through modal response analysis, the approximate required damping ratio
can be extracted.
4. Select desirable and available damper locations. Even though in this study,
this does not seem to create any important issues, in the case of irregular 3D
buildings, the location in which the dampers will be allocated could highly
contribute to the seismic response of the structure.
5. Select damper stiffness and loss factor. In order to achieve the target damping
ratio, the required dampers’ characteristics have to be determined (see 5.3.4).
These characteristics have to be extracted based on dampers’ characterization
tests which have to be carried out at frequencies and amplitudes, which the
structure, and therefore also the dampers, are expected to experience during
a seismic event
6. Estimate the equivalent damping ratio using the modal strain energy method,
at section 5.3.4 [132, 116, 57, 108]
7. Perform structural analysis using the desired damping ratio. Time history
analyses can now be carried out, in order to evaluate the performance of the
structure with added dampers
When steps 6 and 7 satisfy the desired damping ratio and the target performance
criteria then the analysis is complete. Otherwise, a new analysis has to be carried out
with either new structural properties or damper locations, dimensions or properties
(see Fig. 5.13). Based on Lee’s [73] principle to assume that the elastomer behaves
in a pure viscoelastic way, the second design process will be performed here.
Modal Strain Energy Method
Before proceeding to the damper design for the MRF, a brief presentation of
the modal strain energy method, mentioned earlier in the second dampers design
proposal, will be given. The ıth modal damping ratio can be determined as [57]:
119
Figure 5.13: Dampers Design Procedure
n(ω i ) φT
i (K D )φi
ζ D,i = T
2 φi (K S +K D )φi
(5.6)
where n(ω i ) is the loss factor of the material, extracted from characterization tests
(see Chapter 3) at the frequency ω i of the original structure, φi is the ıth mode shape
vector corresponding to the frequency ω i , K S is the stiffness matrix of the original
structure, and K D is the stiffness matrix corresponding only to the dampers. The
frequency ωi can be written as:
φT
i (K S +K D )φi 1/2
ωi = [ ] (5.7)
φT
i (M )φi
where M is the mass matrix of the structure. Assuming that the addition of the
dampers will not affect the mode shapes, and combining Eqs. 5.6, 5.7, the ith modal
120
damping ratio can be rewritten as:
n(ω i ) φT
i K S φi n(ω i ) ωi2
ζ D,i = 2 [1 − T ]= 2 (1 − 2 )
ωD,i
(5.8)
φi (K S +K D )φi
Elastomeric Dampers Design
One of the first steps of the retrofitting procedure described earlier, is the determi-
nation of the target performance criteria, which are going to be repeated here for
convenience:
• All columns remain elastic, while specifically the base columns can exhibit
minor plastic deformations under the DBE
• All columns remain either elastic or are allowed to exhibit minor plastic de-
formations under the MCE
• All beams remain either elastic or exhibit minor plastic deformations under
DBE
• Plastic hinges are allowed to be created at beams under the MCE, if plastic
deformations are minor in most of the columns
• 0.7%-1.3% drift under the DBE
• 2.3%-2.7% drift under the MCE
The design process is based on fulfilling the target criteria set for the DBE, and not
the MCE. Nevertheless, additional criteria have been set in case the structure expe-
riences a seismic action equivalent to the MCE. Now that the performance criteria
have been set, a target damping ratio of 12% will be assumed, corresponding to an
additional 10% of critical damping. This number was chosen based on a trial and
error procedure, carrying out time history analyses. Based on theses analyses, it was
found that an additional 10% damping should be adequate in order for the structure
to comply with the performance criteria. Regarding the location of the dampers,
121
bays 1 and 4 have been selected (Fig. 5.14 shows the locations of the dampers). The
selection of the loss factor and the dampers stiffness can be determined based on a
modified version of Eq. 5.8[114]:
2ζ (5.9)
k D,j = n−2ζ k S,j
where ζ is the target damping ratio, k S,j is k D,j is the story stiffness and the
damper stiffness at the th story. It should be noted here, that k D,j is the total con-
tribution of the dampers along with the diagonal braces, and not only the dampers.
This ensemble can be modelled assuming that the brace stiffness is connected in se-
ries with the damper, and then both the damper and the brace connected in parallel
with the structure (Figure 5.15) [48].
Figure 5.14: Location of Dampers
122
Figure 5.15: Simplified Frame-Braces-Damper representation
In this figure, k S , and cS are the story stiffness of the structure, and the damping
coefficient between the story j and j + 1. k b is the braces stiffness, while k d , and
cd are the damper stiffness and damping respectively. With regard to αb , and
αd , these values correspond to attachment coefficients, and depend on the dampers-
braces geometric configuration (see Figure 5.16). In this specific case, where chevron
diagonal braces are used, and the damper main direction is horizontal and coincides
with the loading direction of the frames, αb = 2cos2 θ (taking into account the
stiffness of both braces), and αD = 1 [48].
(a)
(b)
Figure 5.16: Dampers-Bracing geometric configuration [48]
123
The equivalent storey stiffness and damping are derived as:
αd αb kd kb (αd kd +αb kb )+ω 2 αb α2d kb c2d

k = kS + (5.10)
(αd kd +αb kb )2 +ω 2 α2d c2d
α2b αd kb2 cd
c = cS + (5.11)
(αd kd +αb kb )2 +ω 2 α2d c2d
In the case where the braces are assumed rigid, which is approximately the case here
since they were modelled to be 7 times stiffer than the storey stiffness, the above
equations can be simplified to:
k = k S + αd k d (5.12)
c = cS + αd k d (5.13)
Taking into consideration the above parameters, it can be concluded that k D,j
in Eq. 5.9 equals with the dampers stiffness. Now, the design of the EDs can
be determined. One of the first assumptions will be that the maximum allowed
elastomer strain would be 50%, which coincides with the maximum strain amplitude
under which the EDs were tested (see Chapter 3 for more information). For design
purposes, the characteristics of the dampers which are going to be used are based on
the characterization tests under 40% of the maximum allowed strain (corresponds to
20% strain amplitude), and under frequency range 0.25-0.5 Hz (the natural frequency
of the building lies between these values). The shear storage modulus, and the loss
factor of the elastomer can now be determined as 1.06 MPa, and 0.3535 respectively.
Therefore, Eq. 5.9 can now be used. For example, for the first storey the stiffness
is 51,776 kN/m, and based on Eq. 5.9, the total stiffness that needs to be provided
by the dampers is 67,309 kN/m.
It is recalled here that the target performance criterion is the achievement of
1% lateral drift. Furthermore, as already mentioned, all the characterization tests
124
have been carried out with a 50% maximum shear strain, which results in a damper
thickness of h = 35/0.5 = 70 mm. Assuming that each damper consists of two layers,
and a total of two dampers are added per story, the area of each elastomeric layer
is 1.11 m2 and is determined from Eq 5.14.
AD = 41 k D Gh (5.14)
S
A similar process is repeated for each floor, and the final stiffnesses of the dampers
along with the storey stiffnesses of the structure are summarized in Table 5.7. Now
using the modal strain energy and Eq. 5.8 the estimated damping is 9.86%, which
validates our initial assumption of a 10% damping ratio.
Table 5.7: Story and Damper Stiffness
Story Story Stiffness (kN/m) Total Damper Stiffness (kN/m) Damper Area/layer (m2) AD/AP
1 51776 67309 1.11 26.84
2 26820 34867 0.58 13.90
3 22744 29567 0.49 11.79
4 20863 27121 0.45 10.82
5 16934 22014 0.36 8.78
6 15144 19687 0.33 7.85
7 14515 18870 0.31 7.52
8 12776 16608 0.27 6.62
9 11591 15069 0.25 6.01
10 9592 12469 0.21 4.97
However, all of these properties are based on the designed dampers which are dif-
ferent from the prototype dampers which were tested in the Laboratory, and formed
the basis for the hysteretic model proposed in Chapter 4. In order to adjust this
model for the designed dampers, some modifications needed to be made. Assuming
that the shear storage modulus, the loss factor, and the shear strain of the design
and the prototype damper are the same:
125
GS,D = GS,P (5.15)
nD = nP = n (5.16)
γD = γP (5.17)
where P and D denote for Prototype and Design dampers. Equations 5.15, and 5.17
can be rewritten as:
AD hP
k S,D = AP hD k S,P
(5.18)
hD
uD = hP uP
(5.19)
where k S denotes for the storage stiffness, associated with the shear storage modulus.
Taking into account Eq. 5.18, the equivalent stiffness of the design damper can be
determined as:
p
k eq,D = k S,D 2 + k D,D 2 =
p
k S,D 2 + (nk S,D )2 =
√
k S,D n2 + 1 = (5.20)
AD hP √ 2
A h n + 1k P =
P D
AD hP
AP hD k eq,P
where k D,D is the stiffness loss modulus, corresponding to the shear loss modulus.
Combining Eq. 5.19, and 5.20 the force of the design damper in terms of the proto-
type damper force is given:
AD
FD = AP F P
(5.21)
126
The above procedure was based on transforming the force displacement relationship
for the tested dampers to the equivalent shear strain relationship, and then back to
the force displacement relationship for the design dampers, based on the design di-
mensions. All of these parameters are summarized in Table 5.7. Non linear analyses
with added dampers can now be performed. Regarding the braces stiffness, it was
assumed that were 7 times stiffer than the stiffness of the corresponding storey. The
aim was to add braces which were adequate to produce elastomer rather than brace
deformation, and at the same time capable of resisting horizontal loads without
buckling.
5.3.5 Earthquake Response of steel Conventional Moment Resist-

ing Frame
This section describes the earthquake behaviour of the undamped SMRF under
the ground motions scaled to DBE and MCE (see section 5.3.3). Special attention
should be given to the columns’ behaviour, since they might experience larger axial
forces than those for which they were initially designed. An ambient temperature of
20o C was assumed. Results include the seismic response of the MRF with respect
to shear forces, interstorey drifts, residual displacements, roof displacements and
accelerations. Since 20 different ground motions were used, the response of the
structure is presented in terms of mean and maximum values.
With regard to the performance of the MRF under the DBE, the maximum
roof displacement, according to Figure 5.17, is 78cm, which results in 2.2% drift
ratio, while the mean value is approximately 42cm and 1.2% lateral drift. The DBE
level also led to permanent deformations (Figure 5.19) of 38 cm and 12 cm being
the maximum and the mean value. Figure 5.18, shows the inter storey drift ratios
determined for each ground motion, and their corresponding mean value which is
shown with a clear bold blue line. A maximum drift ratio of 3.75% was observed
under the BorregoM tn − A − ELC180 ground motion, while the maximum drift
ratio of the mean values was found to be 2.1%. This is very close to the Life Safety
performance level (see 5.3.1). The associated mean and maximum residual drifts
reached 0.6% and 1.8% respectively (Figure 5.20).
127
Figure 5.17: Storey maximum Displacements for SMRF under the DBE
Figure 5.18: Storey Drift Ratios for SMRF under DBE
128
Figure 5.19: Storey residual displacements for SMRF under the DBE
Figure 5.20: Storey residual drifts for SMRF under the DBE
129
Figure 5.21: Storey maximum accelerations for SMRF under the DBE
Figure 5.22: Mean Shear forces for SMRF under the DBE
Regarding the roof accelerations in Figure 5.21, 6.8 m/sec2 and 5.5 m/sec2 were
130
observed during the non linear time history analyses, as the maximum and mean
values. The mean storey shear forces are also presented in Figure 5.22. The detailed
time history response of the SMRF under the T abas − Iran − F ER − L1 ground
motion scaled to DBE (Figure 5.23), with regard to roof displacement, and acceler-
ation are shown in Figures 5.24 and 5.25. The beam and the left corner column of
the 2nd bay of the 1st floor were found to be the most critical. Hence, the moment-
rotations relationships are determined and shown at Figures 5.26 and 5.27, where
plastic hinges have been created in both of these elements. Running all these 20
ground motions, it was generally observed that plastic hinges were created at almost
all the beams, and at the base joints of the 1st floor columns, which is practically
the wanted and anticipated form of behaviour, since the steel frame was designed
according to EC8 philosophy which has adopted the design capacity rule.
Figure 5.23: Ground Motion: T abas − Iran − F ER − L1 scaled to DBE
131
Figure 5.24: Roof displacement time history under the T abas − Iran − F ER − L1
grounnd motion scaled to DBE
Figure 5.25: Roof acceleration time history under the T abas − Iran − F ER − L1
ground motion scaled to DBE
132
Figure 5.26: Moment-Rotation relationship for 1st story column of SMRF under the
T abas − Iran − F ER − L1 ground motion scaled to DBE
Figure 5.27: Moment-Rotation relationship for 1st story beam of SMRF under the
T abas − Iran − F ER − L1 ground motion scaled to DBE
Regarding the response of the SMRF under MCE, the concept of section 5.3.3
133
was applied. For comparison, the same parameters used to evaluate the response
of the MRF under DBE will be used here as well. Hence, Figure 5.28 shows the
maximum and mean storey displacements, corresponding to 127 cm and 62 cm re-
spectively. Figure 5.29 shows that the MCE led to severe damage, resulting in a
77 cm and 35 cm maximum and mean values of residual displacements. The corre-
sponding storey drifts can be seen at Figure 5.30, where a value of 5.1% has been
observed during T abas − Iran − F ERl1 ground motion, while the mean value in-
creased from 2.1% under DBE to 3.1% when the frame was subjected to the MCE.
The corresponding residual drifts are presented at Figure 5.31, where a maximum
drift of 2.8%, and a mean drift of 1.3% was observed. With respect to the floors’
accelerations, Figure 5.32 shows that they have also been increased, compared to
the DBE level, reaching a maximum and mean value of 13.8 m/sec2 and 9.1 m/sec2
respectively. The shear forces (Figure 5.33) were also increased to 1449kN (max-
imum). And finally, the time history response under the T abas − Iran − F ERl1
ground motion scaled to MCE level is presented at (Figures 5.34 and 5.35). The cor-
responding moment-rotation relationship of the same beam and column used for the
DBE is shown at Figures 5.36 and 5.37. It can be clearly seen that these elements
have experienced much larger plastic rotations, in comparison with DBE. The MRF
being tested under the MCE hazard level has exhibited large permanent displace-
ments, while plastic hinges were created at all the beams and most of the columns.
The behaviour of the frame is far from achieving a Life Safety performance level,
even if we take into account only the mean values, and not the maximum values of
the frame’s response.
134
Figure 5.28: Storey maximum Displacements for SMRF under the MCE
Figure 5.29: Storey residual displacements for SMRF under the MCE
135
Figure 5.30: Storey Drift Ratios for SMRF under MCE
Figure 5.31: Storey residual drifts for SMRF under the MCE
136
Figure 5.32: Storey maximum accelerations for SMRF under the MCE
Figure 5.33: Mean Storey shear forces for SMRF under the MCE
137
Figure 5.34: Roof displacement time history under the T abas − Iran − F ER − L1
grounnd motion scaled to MCE
Figure 5.35: Roof acceleration time history under the T abas − Iran − F ER − L1
ground motion scaled to MCE
138
Figure 5.36: Moment-Rotation relationship for 1st story column of SMRF under the
T abas − Iran − F ER − L1 ground motion scaled to MCE
Figure 5.37: Moment-Rotation relationship for 1st story beam of SMRF under the
T abas − Iran − F ER − L1 ground motion scaled to MCE
139
5.3.6 Earthquake Response of steel Moment Resisting Frame with

Elastomeric Dampers
While energy dissipation systems have the ability to lead to lateral drift reduction,
this is not always the case for the columns axial forces [10] and the base shear of
the structure. If the dissipation devices are attached to moment frames in chevron
brace form, then this changes the building’s configuration from moment resisting
frame to braced frame configuration, resulting in changes at the load paths as well.
Therefore, even though the retrofitted structure aims not to highly alter the dynamic
characteristics of the prototype building, it should be evaluated not as an enhanced
and better version of the prototype structure but as a new and different structure.
Of course, if this “new” structure fails to achieve the target performance level,
then a different type of retrofitting procedure is required. Applying the retrofitting
procedure of section 5.3.4, and testing the DMRF under the same ground motions
under which the MRF was tested, evaluation of the effectiveness of the EDs can be
determined.
Figures 5.38 and 5.40 show the effect of the EDs on the maximum displacements
and on the corresponding residual displacements. It can be observed that the maxi-
mum displacements were decreased dramatically with a maximum and a mean value
of 35 and 29 cm respectively. EDs were able to minimize residual displacements,
resulting in a mean value of 2.5 cm, and a maximum value of 13.5 cm. Further-
more, Figure 5.39 shows the large storey drift reduction from 3.7% to 1.48% for the
maximum values, and from 2% to 1.1% for the mean values. Therefore, the DMRF
complies with the target performance criteria, with respect to the lateral drift for
the majority of the ground motions (the target performance criteria was to achieve
interstorey drifts from 0.7-1.3%). EDs had also huge effect on the reduction of the
residual displacements (Figure 5.41).
Accelerations (Figure 5.42) did not follow the same rate of decrease, as storey
displacements did, but a noticeable mitigation was observed. A comparison of the
maximum storey shear forces between the MRF and DMRF under the DBE is also
presented in Figure 5.43, where a slight reduction is observed in every floor. In
140
order for evaluation of the influence of the EDs on the time history response of the
DMRF to become more clear, Figures 5.44 and 5.45 show both the time history
response of the MRF and DMRF under the T abas − Iran − F ER − L1 ground
motion scaled to DBE level, where it can be clearly seen that the addition of the
EDs led to large deformation reductions and more moderate accelerations reductions
at the top storey of the building. Regarding the moment rotation relationship for
the same beam and column evaluated in MRF case, Figures 5.46 and 5.47 show the
EDs effect on both of them. A plastic hinge was not created in the column, which
remained completely linear in the DMRF, in contrast with the MRF case. In the
case of the beam, a plastic hinge was not avoided, but the the plastic rotations, along
with the total energy absorbed by the element were greatly reduced. Regarding the
dampers hysteretic response, Figure 5.48 shows that the initial design assumption
of a maximum elastomer deformation of 35 mm has been satisfied as well.
It can be concluded that EDs enhanced the seismic behaviour of the prototype
frame, decreasing both the displacements, accelerations, shear forces, and structural
elements seismic demands leading the structure to comply with all the target per-
formance criteria which had been set at the beginning of the analysis, and perform
very close to the Immediate Occupancy performance level. Plastic hinges were not
created to any of the columns, including those of the 1st floor, while only a few
beams, especially on the first 2-3 floors exhibited non linearities.
However, the shear forces exhibited the smallest reductions comparing with the
all the other aforementioned parameters. Similar problems or even worse have been
noted in case of metallic and friction dampers as well [114]. This is mainly due to
two reasons: a) the nature of the dissipation mechanism of elastomeric dampers is
based on the forces being extracted from the dampers and transferred to the rest of
the structure through the diagonal braces and eventually creating additional shear
forces to each storey, and b) the non-linear nature of the materials used in the
analyses sometimes leads to large decreases of displacements, which can be reduced
to a level close to the yield point of the structural elements; however the forces are
still going to be near the yield level.
It can be also noted that some ground motions which caused high responses in the
141
DMRF, caused only moderate structural response in the SMRF. Even though the
initial target was to alter the dynamic properties of the original frame as less as
possible, the addition of the EDs led to a slight reduction of the natural period.
This caused some ground motions which have very high frequency peaks on these
specific periods to amplify the response of the DMRF.
142
Figure 5.38: Storey maximum Displacements for DMRF under the DBE
Figure 5.39: Storey Drift Ratios for DMRF under DBE
143
Figure 5.40: Storey residual displacements for DMRF under the DBE
Figure 5.41: Storey residual drifts for DMRF under the DBE
144
Figure 5.42: Storey maximum accelerations for DMRF under the DBE
Figure 5.43: Mean Storey Shear Forces for DMRF and SMRF under the DBE
145
Figure 5.44: Top Storey Displacement comparison for the SMRF and DMRF under
DBE
Figure 5.45: Top Storey Acceleration comparison for the SMRF and DMRF under
DBE
146
Figure 5.46: Moment Rotation relationship for base column under theT abasI ran −
F ER − L1 ground motion scaled to DBE
Figure 5.47: Moment Rotation relationship for base beam under theT abasI ran −
F ER − L1 ground motion scaled to DBE
147
(a) 1st floor-DBE (b) 2nd-DBE
(c) 3rd floor-DBE (d) 4th floor-DBE
(e) 5th floor-DBE (f) 6th floor-DBE
(g) 7th floor-DBE (h) 8th floor-DBE
(i) 9th floor-DBE (j) 10th floor-DBE
Figure 5.48: ED hysteretic response for each floor under DBE
148
Regarding, the effect of the EDs under the MCE, Figure 5.49 presents the cor-
responding displacements, which were highly decreased, dropping from 127 cm to
56 cm and from 62 cm to 41 cm for the maximum and mean values respectively.
The residual displacements (Figure 5.50) were highly reduced as well; a fact which
validates the EDs effectiveness, since a total reduction of 60.7% and 83.1% for the
maximum and mean value was achieved, leading the structure to behave slightly
above its linear limit. Figures 5.51 and 5.52 shows the effect of the EDs on the
inter storey drift ratios, where a maximum value of 2.4% and a mean value of 1.6%
was achieved, leading the frame to achieve lateral drifts much smaller than the ones
proposed for the Life Safety performance level. Similar reductions were observed
in terms of residual displacements, where exceptional structural behaviour was ob-
served.
Regarding the accelerations (Figure 5.53), EDs seem to be more efficient, when
the frame is tested under MCE rather than DBE, since a reduction of 36.2% and
30.8% for maximum and mean values was achieved. The maximum shear forces
obtained from the analyses were again compared here, under the MCE, (Figure 5.54).
However, they were found to be approximately the same (1% decrease at the 1st
floor). In addition, time histories of roof displacements and accelerations under
the T abas − Iran − F ER − L1 ground motion scaled to the MCE are presented
in Figures 5.55 and 5.56, according to which EDs led the total displacements not
only to highly decrease but also the residual displacements to be almost eliminated.
Regarding the accelerations, again they did not follow the same reduction rate as the
displacements. The comparison of the relationship of the same beam and column
with the corresponding rotation for the MRF and DMRF under the MCE is shown
at Figures 5.57 and 5.58. It can be seen that even though plastic hinges were not
avoided for either of these two elements, plastic rotations were greatly reduced.
Finally, the dampers’ hysteretic response under MCE is shown at Figure 5.59.
The DMRF behaved satisfactorily under the MCE hazard level. Actually, it
behaved better even comparing with the MRF seismic performance under the DBE.
Another important parameter, as already mentioned, is the evaluation of the
axial forces acting at the columns, which were anticipated to be increased. Indeed,
149
table 5.9 shows that the axial forces of the left corner base column of the second bay
(which was found to be the most critical) increased by 28% and 40% under DBE
and MCE respectively. However, as Figures 5.46 and 5.57 show, even though the
axial forces were increased, at the same time the total behaviour of the element was
highly improved, due to decreased rotations and displacements. Hence, it can be
concluded that all the target performance criteria can be achieved under the DBE
hazard level even by using lighter and weaker column sections.
The main target in the case of frame tested under MCE, as in the conventional
design of buildings, is the creation of plastic hinges at the bottom joints of the base
columns, and at the ends of the beams, leading to a ductile seismic performance, and
not a brittle collapse. This target has been achieved for this frame, and therefore
all the target performance criteria have been achieved. It can also be observed
that EDs’ effectiveness increases when the intensity of the ground motion increases.
Finally, the total response of MRF and DMRF under both DBE and MCE can be
summarized at Tables 5.8 and 5.9 with respect to maximum and mean values of the
parameters examined earlier.
150
Figure 5.49: Storey maximum Displacements for DMRF under the MCE
Figure 5.50: Storey residual displacements for DMRF under the MCE
151
Figure 5.51: Storey Drift Ratios for DMRF under MCE
Figure 5.52: Storey residual drifts for DMRF under the MCE
152
Figure 5.53: Storey maximum accelerations for DMRF under the MCE
Figure 5.54: Mean Storey Shear Forces for DMRF and SMRF under the MCE
153
Figure 5.55: Top Storey Displacement comparison for the SMRF and DMRF under
MCE
Figure 5.56: Top Storey Acceleration comparison for the SMRF and DMRF under
MCE
154
Figure 5.57: Moment Rotation relationship for base column under theT abasI ran −
F ER − L1 ground motion scaled to MCE
Figure 5.58: Moment Rotation relationship for base beam under theT abasI ran −
F ER − L1 ground motion scaled to MCE
155
(a) 1st floor-MCE (b) 2nd floor - MCE
(c) 3rd floor-MCE (d) 4th floor-MCE
(e) 5th floor-MCE (f) 6th floor-MCE
(g) 7th floor-MCE (h) 8th floor-MCE
(i) 9th floor-MCE (j) 10th floor-MCE
Figure 5.59: ED hysteretic response for each floor under MCE
156
Table 5.8: Comparison of mean response of SMRF and DMRF under DBE and
MCE
Performance DBE Performance MCE
Evaluation Parameters
Criteria Without Eds With Eds Difference (%) Criteria Without Eds With Eds Difference (%)
Roof Displacement (cm) - 42 29 -30.95 - 62 41 -33.9
Residual Roof Displacement (cm) - 12 2.6 -78.33 - 34 4.8 -85.9
Roof Drift (%) - 1.2 0.67 -44.17 - 1.71 1.17 -31.6
InterStorey Drift (%) 0.7-1.3 2 1.1 -45.00 2.3-2.7 3.1 1.7 -45.2
Residual InterStorey Drift (%) 0.2-0.5 0.6 0.13 -78.33 0.5-1.0 1.3 0.22 -83.1
Base Shear (kN) - 1020 915 -10.3 - 1276 1170 -8.3
Roof Acceleration (m/sec2) - 5.5 4 -27.27 - 9.1 6.3 -30.8
Beam Plastic Rotation (Rad) - 0.0086 0.0015 -82.56 - 0.0104 0.0076 -26.9
Column Plastic Rotation (Rad) - 0.006 0.0002 -97.43 - 0.008 0.0023 -71.3
Column Axial Force (kN) - 2052 2631 28.22 - 2103 2951 40.3
Table 5.9: Comparison of maximum response of SMRF and DMRF under DBE and
MCE
Performance DBE Performance MCE
Evaluation Parameters
Criteria Without Eds With Eds Difference (%) Criteria Without Eds With Eds Difference (%)
Roof Displacement (cm) - 79 35 -55.70 - 127 56 -55.9
Residual Roof Displacement (cm) - 38 13.5 -64.47 - 77 23 -70.1
Roof Drift (%) - 2.25 1 -55.56 - 3.6 1.6 -55.6
InterStorey Drift (%) 0.7-1.3 3.7 1.48 -60.00 2.3-2.7 5.1 2.4 -52.9
Residual InterStorey Drift (%) 0.2-0.5 1.8 0.6 -66.67 0.5-1.0 2.8 1.1 -60.7
Base Shear (kN) - 1280 1180 -7.81 - 1449 1347 -7
Roof Acceleration (m/sec2) - 6.7 5.5 -17.91 - 13.8 8.8 -36.2
Beam Plastic Rotation (Rad) - 0.0124 0.0044 -64.52 - 0.0256 0.0137 -46.5
Column Plastic Rotation (Rad) - 0.0094 0.0006 -93.73 - 0.0157 0.0048 -69.4
Column Axial Force (kN) - 2150 2792 29.85 - 2171 3124 43.9
5.4 Effect of temperature
All the previous analyses were based on an ambient temperature of 20o C. In order to
evaluate the effect of temperature, the worst case scenario will be taken into account
in this section. Hence, the analysis of the DMRF will be shown here, assuming
an ambient temperature of 35o C, where the decrease of the dampers’ efficiency is
greatest, according to Chapters 3 and 4. Since, the overall target of this section is
only to simply illustrate the temperature’s effect, and not to carry out design of the
DMRF with different temperatures, only the results of the modified El Centro (I-
ELC180 component) will be presented here. In the case of a full design of a DMRF,
a parametric analysis should be carried out, and the worst case scenario should be
taken into account for the design of the dampers in order for the DMRF to cope
under the target performance criteria.
The adjusted proposed proposed model (see Chapter 4), with γ = 0.82 will be
considered in this section. The analysis carried out here was focused on the DBE
157
level, since this hazard level formed the basis for the design of the EDs. The results
were compared with the ones from the initial analysis carried out at 20o C. Fig-
ure 5.60 shows the top storey displacement under the modified El Centro (I-ELC180
component) ground motion for both 20o C, and 35o C. It can be observed a slightly
better behaviour, but not significant, since the displacements’ time histories almost
coincide for both temperatures. This is the case for both top storey accelerations
(Figure 5.61), and storey shear forces (Figure 5.62), where the difference is negli-
gible. However, this is not the case when it comes to the hysteretic behaviour of
individual members, where the beams’ plastic deformations were increased, as can
be seen from Figure 5.63. Moreover, the first storey columns’ hysteretic behaviour,
was also affected by the alteration of ambient temperature. Figure 5.64, shows that
the column exhibits minor plastic deformations at 35o C, as opposed to the cor-
responding moment-rotation relationship under 20o C, where the column remained
elastic. Therefore, even though the overall behaviour of the structure with regard to
inter storey displacements, accelerations, and shear forces was not highly affected,
it can be concluded that even this slight temperature alteration led the DMRF to
fail the target performance criteria (see section 5.3.1), even though the performance
of the structure was highly enhanced. The temperature alteration also affected the
EDs’ hysteresis loops. Figure 5.65 shows this effect on the dampers implemented
on the first floor, comparing the EDs’ behaviour under both temperatures. It can
be seen that the dampers seem to experience a stiffness degradation, as expected.
The above results and conclusions can only be extracted for the DMRF behaviour
under this specific ground motion, and can not be generalised as a design tool. It
should also be noticed at this point, that the alteration of the ambient temperature
from 20o C to 35o C led the dampers to become less stiff, and therefore the structure
more flexible. This has an overall effect on the applied seismic loads on the DMRF,
even though this change is not huge. Hence, even though the dampers are less
stiff (or probably less effective), this does not necessary means that the structure’s
behaviour will worsen as well. Depending on the frequency content of the applied
ground motion, the structure may even have improved dynamic performance. It can
be then concluded that a detailed design procedure, similar to section 5.3.4 should
158
be carried out for every temperature, in order to design the dampers.
Figure 5.60: Top storey displacement (mm) - 35o C
Figure 5.61: Top storey acceleration (m/sec2 ) - 35o C
159
Figure 5.62: Storey shear forces - 35o C (kN)
Figure 5.63: Top storey acceleration (m/sec2 )
160
Figure 5.64: Top storey acceleration (m/sec2 )
Figure 5.65: First storey ED’s force-displacement relationship
161
5.5 Conclusions
The current chapter described the analysis of a 10 storey steel SMRF under both
DBE and MCE hazard levels. 20 ground motions, scaled to these levels, were used
to evaluate its seismic performance. EDs were then added at specific locations of the
original frame, and the structure was tested again under the same seismic intensity, in
order for the EDs efficiency to be determined. EDs were specifically designed, based
on the characterization tests carried out at Dynamics Laboratory and the material
model described in previous chapters, in order to add 10% of critical damping to
the bare frame.
The results showed satisfying behaviour regarding the seismic response of the
DMRF under both DBE, and MCE, while the initial performance criteria which
had been set before the beginning of the analysis were all satisfied as well. The
DMRF remained linear under most of the ground motions under the DBE, while
the mean lateral drift was restricted to 1.1%, and the residual displacements were
almost eliminated. The DMRF showed even more impressive results under the
MCE, where the EDs efficiency increased since the displacements were larger. The
mean inter storey drifts reduced by 45%, and the residual drifts by 83%. On the
other hand, axial forces were found to be increased, which was anticipated, but at
the same time the total behaviour of the members under axial loads were found
to behave much more satisfactorily, where the plastic deformations were greatly
reduced. Even though in this case, the increased axial forces do not cause any issues
with regard with the capacity of the column, it is suggested that under certain
circumstances columns may have to be strengthened.
A simple study of the temperature effect was also shown. Even though a full
design process was not followed, but only illustration of results under a specific
ground motion, the ambient temperature, did not seem to highly alter the dynamic
behaviour of the DMRF. Individual structural members were found to be affected
more, while the target performance criteria with regard to the behaviour of the
individual members behaviour were slightly below the anticipated level. However, a
more robust parametric analysis is recommended for the whole temperature range
162
of interest, similar to section 5.3.4.
163
Chapter 6
Real Time Substructure Test
6.1 Background
In seismically active regions, buildings are usually designed to exhibit a level of
inelasticity under strong ground motions. Therefore, as mentioned earlier, by pro-
viding the structure the ability to dissipate energy, collapse should be avoided. In
order to fully capture the inelastic structural performance experimental tests are
often used. There are several experimental methods available for evaluating the
inelastic performance of structures, or even isolated parts of the structure, such as
dampers and bearings. Since many structures have their lowest natural frequency
within the loading range of earthquakes (0-8 Hz), their damping and inertia char-
acteristics may make a large contribution to their behaviour [18]. Some of the most
well established experimental techniques are described in this section.
6.1.1 Shaking Tables
The shaking table test is one of the most realistic and reliable experimental methods
for evaluating the inelastic seismic performance of structures. However, the size,
weight and stiffness properties of the test structure are significantly limited by the
capacity of available shaking tables [30]. Shaking table tests are carried out on a
stiff platform, which is shaken in order to apply the appropriate base motion [140].
This way the ground motion is applied at the base, generating realistic inertia forces
in every element of the structure. The motion of the shaking table is controlled
164
by servo-hydraulic actuators. The current trend is to provide active control in all
six degrees of freedom, even when in some cases the motion in several directions is
zero. The largest shaking table in the world is E-Defense, constructed by Japan’s
National Institute for Earth Science and Disaster Prevention (NIED), which is able
to simulate at full scale the response of all types of structures to the full force of a
magnitude-7 earthquake [123].
Regarding the control of the shaking table, accuracy requires knowledge of the
system dynamics, which is generally not known prior to the test. Hence, an it-
erative tuning procedure can be followed, according to which a demand signal is
adjusted, while the system remains in its linear phase, until the desired table mo-
tion is achieved. Therefore, the control of the table can be deternined. However, it
is based on properties of the system being in its linear phase. The shaking table test
is likely to generate non linearities in the test specimen during ground motions, and
therefore the parameters of the controller cease to be optimal, and consequently the
desired base motion will not be accurately reproduced. Adaptive controllers may be
used to overcome this issue. The main concept behind this, is that the parameters
of the controller are updated during the test, so they take into account any changes
of the system’s characteristics [117].
While the shaking table technique is the most reliable and able to produce the
actual seismic forces to the structure, it is also very expensive while requiring state
of the art testing facilities. In case of shaking tables where full scale models cannot
be tested, primarily due to either cost or specimen’ size, scaling extrapolation of
the results has to be carried out. However, Nakashima [93] showed that scaling
introduces discrepancies, mostly in the fidelity of ductility and yielding properties
of structures.
6.1.2 Pseudo Dynamic Tests
In the Pseudo-Dynamic (PsD) method the displacements are computed, using inte-
gration algorithms, such as the central difference method, and then are imposed in a
quasi-static way on the structure using actuators. The resulting resistance forces are
measured and fed back into the computational model, so that the next displacement
165
step can be computed. Because the rate of the imposed displacements is relatively
slow, it is possible to observe the behaviour of the specimen in detail during testing
[35].
Using the central difference integration method, the main philosophy of the
pseudo dynamic test will be presented. In an iterative procedure, the equation
of motion for the i + 1 time step of the tested structure can be described as:
Müi+1 + Cu̇i+1 + Ri+1 = Fi+1 (6.1)
where M is the mass matrix, C is the damping matrix, R is the restoring force
vector, and F is the applied force vector, and dots represent time differentiation.
According to the central difference integration method:
ui+1 = ui + dtu̇i + 12 dt2 üi (6.2)
u̇i+1 = u̇i + 12 dt(üi + üi+1 ) (6.3)
where dt is the time-step length. Since the restoring forces are measured experi-
mentally, there is no uncertainty over the non linear stiffness characteristics of the
structure [140]. Knowing the state of the structure at time step, i, the main steps
of this method for the i + 1 time step can be summarised as:
• Apply external force Fi+1
• Compute displacement ui+1 from Eq. 6.2
• Impose displacement ui+1 on test specimen
• Measure restoring force, Ri+1
• Compute velocity, u̇i+1 , and acceleration üi+1 based on Eqs 6.2, and 6.1
• Repeat process for i + 2 time step
166
The PsD method has been proved to be a very effective and reliable technique with
regard the simulation of the response of structural systems [94, 93, 79, 37]. However,
Shing and Mahin [109] showed that the PsD method is very sensitive to errors,
possibly resulting in an unstable response. The reason behind that is that the errors
from an individual time step are fed back to the system and so introduce errors
in computing the displacements for the new time step, and consequently wrong
restoring forces. Hence, errors have a cumulative nature. These effects become
more pronounce at higher frequencies, a fact which makes PsD method extremely
inefficient for stiff structures. In order to minimize the errors, several approaches
have been proposed such as increased viscous damping, numerical integrations to
be based on previous computed displacements rather than measure displacements
[109], PsD method to be based on an integrated form of the equation of motion
[121], and finally integration algorithms to damp the higher modes [51].
However, this method may not be suitable for certain types of structures. Struc-
tures with concentrated masses are most easily modelled and tested, and this method
is difficult to be applied to distributed mass structures [35]. Furthermore, since the
method is based on imposing the computed displacements in a quasi-static way,
structures constructed from materials which are sensitive to any loading rate al-
terations are not suitable to be tested using the Pseudo-Dynamic method. Lastly,
structures where damping is anticipated to have a significant effect on the overall
behaviour are not suitable for this methods as well, since both inertia and damping
characteristics are numerically modelled [35]. On the other hand, imposing loads
quasi-statically allows the test to be carried out with simpler static hydraulic ac-
tuators, which are able to produce higher forces comparing with dynamically rated
equipment, making it possible to avoid scaling [18].
6.1.3 Effective Force Testing
The main philosophy of the Effective Force Testing (EFT) method is very similar to
the PsD method. However, in this case the actuators are dynamically rated and force
controlled. The structure is loaded in real time by actuators with an earthquake force
equal to the ground acceleration multiplied with the lumped mass of the specimen.
167
The corresponding motion is equal to the relative motion of an equivalent system
subjected to a ground motion. It is a very appealing method, since the loading can
be determined in advance, while at the same time no numerical integration is needed
[36, 140]. A major drawback however, is that the whole structural mass needs to be
included in the specimen, which may be difficult to achieve in the majority of the
available testing facilities. Another disadvantage of the EFT method is the inability
of the hydraulic actuators to produce force close to the natural frequency of the
structure, especially for lightly damped structures.
6.1.4 Pseudo Dynamic Substructure Tests
The PsD technique can be also implemented with the substructure concept. The
main reason that the substructure technique is so appealing is its efficiency in terms
of both economic and geometric restrictions. Full scale models are expensive to
build, and require special large scale and high capacity test facilities. Therefore, it
is not usually practical to construct a full scale model. For example in case of eval-
uating the effect of elastomeric bearings of a bridge, constructing the whole bridge
in a laboratory is not a realistic project. Elastomeric bearings, however, could be
experimentally tested at the laboratory, while the remainder of the bridge is nu-
merically modelled. The first attempts at this approach were made by Takanashi
and Nakashima [58], and Mahin [80]. According to this method the total structural
system is divided into the analytical part, with predictable and well known be-
haviour, and the experimental substructure, which is tested in the laboratory. The
experimental part is the main item of interest. Properly specified boundary condi-
tions between the numerical and experimental substructure are necessary in order
to provide force equilibrium and displacement compatibility. It is preferable to have
a small number of experimental Degrees of Freedoms (DOFs), while care must be
taken in order for the DOFs to accurately represent the true dynamic behaviour of
the test specimen [35].
As an illustrative example Figures 6.1 and 6.2 show the PsD substructure method
concept, and the main PsD substructure technique steps. The total structure con-
sists of a 1-storey 1-bay frame connected with dampers. In this example the main
168
interest is on the behaviour of the damper. Hence, the damper along with the diago-
nal braces makes up the physical substructure, while the remainder of the structure
is numerically modelled. The test commences by determining the response of the nu-
merical substructure for the first ground motion acceleration step. The displacement
at the interface between the experimental and analytical substructures is calculated
based on a time integration algorithm, and then is applied at the test specimen using
hydraulic actuators. The restoring force, which is experimentally calculated is then
fed back to the numerical model along with the next ground motion increment. This
process is repeated until the test is complete. In the PsD substructure case Eq. 6.1
takes a different form:
Müi+1 + Cu̇i+1 + RE A
i+1 + +Ri+1 = Fi+1
(6.4)
where RA is the restoring force corresponding to the analytical part, and RE is
the restoring force corresponding to the physical specimen. In the case of a linear
elastic frame , RA is expressed as RA = k A u, where u represents the displacements.
With regard to the displacement calculations, a similar process to that described in
Section 6.1.2 can be followed, where the central difference method was used. The
effect of a range of numerical time integration algorithms on the dynamic behaviour
and stability of MDOF systems was the focus of the research of Bonnet [18].
Even though the PsD substructure method seems to be very appealing and
more efficient than the PsD method, with regard to the facilities and economical
restrictions, is still not suitable in cases when loading rate effects become important
parameters on the behaviour of the specimens tested.
169
Figure 6.1: Main concept of PsD substructure’ technique
Figure 6.2: Basic steps of PsD substructure’ technique
6.1.5 Real Time Substructure Tests
PsD methods are actually time-scaled expanded versions of the real seismic events,
normally restricted to materials assumed to behave in a rate-independent manner
[78]. In order to test structural components which exhibit velocity dependent char-
acteristics, the concept of Real Time Substructure (RTS) tests has been the focus of
several researchers [78, 93, 18, 61]. The concept remains almost the same as the PsD
substructure technique described earlier. The total structure is again divided into
the physical and the numerical substructure. Also, the test again proceeds according
to Figure 6.2. However, now the tests are performed in real time, where the actuator
needs to act continuously, eliminating the quasi-static phase described in 6.1.4. In
170
order to achieve this, each cycle described in Figure 6.2 has to be completed within
the time interval between the load increments, so that the loading and structural
response occur at the same rate in the test as in a real dynamic loading event in
a prototype structure [140], meaning that each cycle has to be completed within a
few milliseconds. It should be noted that the force feedback in RTS tests includes
both inertia and damping components, and these therefore do not need to be taken
into account in the numerical substructure, as opposed to the PsD method. One of
the first RTS tests was conducted by Nakashima [94], where the physical substruc-
ture was a viscous damper located at the base of a multi-storey building, which was
numerically modelled.
However, one of the main problems that RTS tests face is the time delay, mainly
due to the finite response of the hydraulic actuators. This delay is the time difference
between the command signal and the actual response of the actuator to the desired
position. Consequently, the corresponding forces are based on wrong displacements,
creating cumulative error leading to instability. More specifically, it has been proved
[124] that the effect of this error is equivalent to adding negative damping to the
system. Several procedures have been proposed [124, 52, 18, 17, 21] in order to
mitigate the effect of time delay. The focus of this chapter is to carry out RTS tests
using a new method for minimizing the delay, which is based on an extension of the
work of Chae et al [21].
6.2 RTS testing in Oxford
The main equipment in the Oxford University Structural Dynamics Laboratory
(OUSDL) consists of six servo-hydraulic actuators: two 10 kN actuators (piston

+ +
stroke: − 75 mm), two 100 kN actuators (piston stroke: − 125 mm), and two 250
+
kN actuators (piston stroke: − 125 mm). For the needs of the tests carried out as
part of this thesis, one of the 100 kN actuators was used (Figure 6.3). The actu-
ators can move via pressurised oil through servo-valves, which control the piston
movement. The oil is distributed to the valves by a hydraulic substation. In order
to achieve a large supply of pressurised oil, three 60 l/min pumps are connected in
171
parallel over a large oil reservoir at one end of the high pressure line with a set of
hydraulic accumulators, which are capable of storing 180 l at the other end. Hence,
during peak flow consumptions the pressurised oil flow rate can be doubled for a
short period of time, when the instantaneous supply of the pumps is not enough.
The hydraulic substation is connected with the pressure line as well. All this equip-
ment is installed on a 166 tonnes concrete block, with 9.1 m x 4.2m surface area,
and 1.6 m deep. An overview of the main equipment installation of the OUSDL can
be seen at Figure 6.4.
Figure 6.3: 100 kN actuator used for tests
172
Figure 6.4: Schematic overview of OUSDL equipment [15]
The closed loop control between the substation and the actuator is conducted
through an Instron Labtronic 8800 controller, which provides four channels of posi-
tion or force control. The controller parameters are accessible to the user through the
Instron program RS+. The Instron controller is based on a Proportional-Integral-
Derivative-Lag (PIDL) algorithm, the parameters of which can be determined either
manually or by an auto-tuning process based on minimizing the error of the com-
mand signal and the actual response of the actuator with regads to a square wave
displacement command. The output displacement u(t) of the controller is based
on the error between the desired displacement, udes , and the actual displacement
achieved by the actuator, uact . The displacement feedback, uact , is determined by
a linear variable differential transformer (LVDT) fitted inside the actuators’ body
and outputs the effective piston displacement relative to the actuators’ body [18].
Moreover, the actuator is also equipped with a load cell which connects the end of
the piston with the test specimen and is responsible to determine the force. The two
transducers send their signals to the Instron controller. Although both force and
displacement control of the actuators can be achieved, only displacement control
was used in this research. However, in each case, the signal of the corresponding
transducer is determined, and compared with the command signal. Here is where
173
the aforementioned PIDL controller is being used, producing a closed loop control
generating command signals every 0.2 ms (sampling frequency: 5 kHz).
Real time control of the substructure testing loop, however, is achieved via a
Digital Signal Processing (DSP) board, which is connected with a 32-channel Data
Acquisition (DAQ) system, creating an additional control loop around the system.
During RTS tests, monitoring the force and displacement of the actuator, along
with the encoder’s displacement is possible through the PC hosting the DSP. The
numerical substructure, which is modelled using Simulink [7], is downloaded to the
DSP for execution. The DSP’s host PC communicates with the Instron controller
to fully capture the interaction between the numerical and physical substructure. A
main and simplified overview of the RTS procedure is shown at Figure 6.5.
The ground acceleration is the input to the numerical part. The corresponding
displacements at the physical-numerical interface, xn , are calculated at the DSP’s
host PC, and sent to the delay compensator (outer loop control) producing uc . In
order to take into account any rig flexibilities, the difference between the actual
displacement extracted by the linear encoder and the actuator’s LVDT is added
to uc to generate the final command displacement which is sent to the Instron
controller and from there to the actuators. The resulting force extracted from the
actuator is sent back to the numerical part in order to produce the next desired
displacement increment. The corresponding measured displacements, xm are used
as input commands to the delay compensator, so the final command displacements,
uc , can be determined. All this procedure will be described in detail in the following
sections. It is worth mentioning that the integration algorithm, along with the
delay compensator, is solved in a different time step to the one used for the Instron
controller, called the main step, taken as 1 ms for the tests in current chapter. Then,
the controller generates the required displacements based on the PIDL algorithm
described earlier in finer time steps called sub-steps (0.2 ms), in order to obtain good
experimental control. However, actuators are known not to respond instantaneously.
The characterization of actuators in terms of time delay and amplitude error time
histories is a very practical way of approaching this non-instantaneous actuator
response, as will be seen in this chapter.
174
Figure 6.5: Simplified Overview of Hybrid Procedure
6.2.1 Time Delay
Among other researchers [124, 52, 17, 21], the effect of time delay has been thor-
oughly studied by Bonnet [18]. He carried out a large series of tests at the OUSDL,
and estimated that the minimum time delay of a 10 kN actuator was approximately
at 4.7 ms. This value was not constant, since most of Bonnet’s research was fo-
cused on the parameters which may affect the delay. He concluded, that the time
delay highly depended on loading frequency, specimen stiffness, proportional gain
parameter of the PID controller, and loading magnitude, where the maximum value
of time delay observed during tests Bonnet conducted was approximately 30 ms.
The amplitude error on the other hand, was estimated as of secondary importance
compared to the time delay. These delays can be generally ignored in an open loop
test (where there is no feedback from the actuator’s response). However, it has been
proved [124] that time delays in closed loop tests, as in the case of RTS tests, can
considerably change the behaviour of the system, make it unstable. Specifically, Ho-
riuchi [124] proved that for an SDOF system, with displacement x(t), a time delay
δt, leads to addition of negative damping to the system:
ceq = −kδt (6.5)
175
ζeq = −ωn δt/2 (6.6)
p
where wn = k/m is the natural frequency of the system. If the negative damping
is larger than the inherent damping of the system, the system becomes unstable.
As already mentioned a lot of procedures and algorithms have been considered in
previous research [124, 52, 18, 17, 21], in order to improve the stability of the system
in RTS testing. These actuation compensation schemes aim to force the actual
output to mimic the desired output [18]. Once the displacement, xt , of the numerical
integration algorithm is solved, it is fed into the delay compensation algorithm, and
a new displacement command, uc , is generated based on the delay scheme, and sent
to the actuator producing the actual displacement measured experimentally, xm .
The general aim of the delay compensation schemes is to correct the displacement
determined from the numerical part of the RTS. This is done in such a way, so
the error between the actual displacement which will be achieved by the actuator
after a time equal with the time delay, and xt is minimised, extrapolating xt by an
amount equal with an estimation of the time delay. The most widely used actuation
compensation schemes are briefly presented here:
• Exact Polynomial Fitting Extrapolation
Horiuchi [124] proposed a mathematical model which is based on fitting a
third order polynomial equation on the last data points. The polynomial is
then used to extrapolate, xt by an amount equal with an estimation of the
time delay. The number of data points used for the fitting of the polynomial
is equal with the chosen polynomial order, n, plus one. The estimation of the
time delay is assumed constant during the RTS test, and is based on previous
tests and highly depends on the specimens and loadings requirements.
• Least Square Polynomial Fitting
A similar approach has been followed by Wallace [134], and Zhu [146]. How-
ever, more points than n+1 are taken into account for the fit of the polynomial
in this case. This method is based on a least square approximation and not an
exact fit, however. In terms of noise this can be very effective, since simulation
176
results showed that the fewer data points used the more sensitive the system
becomes to noise. Nevertheless, this may cause some potential problems when
the polynomial changes from point to point, since each polynomial of n order
is a best fit between m data points (where m > n), no point actually lies on
it.
• Linear Predicted Acceleration
This method is not based on a polynomial fitting as the previous ones. Instead,
assuming that the acceleration at the interface varies linearly, the correspond-
ing displacement can be then determined through integration. The need for
the knowledge of the acceleration of the current step leads to the assumption of
extrapolating the acceleration in a time equal with the estimated delay, since
most of the explicit time integration algorithms do not provide this informa-
tion. Horiuchi [53] showed that this method is theoretically superior in terms
of stability compared to the exact polynomial fitting.
• Live Delay Estimation
All the aforementioned delay compensation schemes assume an estimation of
the time delay, which remains constant during the RTS tests, and is usually
based on open loop tests which have preceded the RTS tests. To eliminate
this uncertainty Darby [15] suggested a live estimate of the time delay which
is updated for every time step. The update of the time delay, δ, is based on
the error between the target (or desired) displacement and the measured dis-
placement, along with the time delay of the previous time step. Additional
empirical parameters along with a velocity estimate are used in Darby’s al-
gorithm. The time delays are determined based on the zero-crossing method.
According to this method, the time delay is calculated each time the target and
measured displacement become zero. However, this also creates a limitation
of the method since no updates are available between zero crossings.
177
6.2.2 Adaptive Time Series
A new method, based on an adaptive actuator compensation scheme was proposed
by Chae [21], called Adaptive Time Series (ATS). The ATS method continuously
updates its coefficients, taking into account any non linearities of both the actuator
and the test specimen, resulting in an accurate actuator control. Unlike most existing
adaptive methods, the ATS scheme does not involve user-defined adaptive gains.
Furthermore, most of the adaptive delay compensation techniques are based on
gains extracted from calibration tests prior to the RTS testing, or on a tuning and
optimizing procedures with moderate force or displacement amplitudes. However,
these gains are not valid for larger amplitudes where the behaviour of the testing
specimen may be different, whether due to its inherent material properties or due
to damage/fracture. The ATS method is based on the following basic relationship
between the input displacement command, u, and the output displacement of the
servo-hydraulic system, x:
1 (6.7)
u(t) = A x(t + τ)
where A is the amplitude error which represents the ratio between the output and the
input displacement, and τ is a constant time delay. Figure 6.6 shows the relationship
between u, and x. Using the Taylor series Eq. 6.7 can be approximated as:
h i
1 τ2 τ n dn x (6.8)
u(t) = A x(t) + τ ẋ(t) + 2 ẍ(t) + ... + n! dtn
According to Eq. 6.8 the input command of the actuator, can be determined
based on the constant time delay, amplitude error, and the output displacement
at time t. If we want to achieve a target displacement, xt , at a time-step k (the
measured displacement xm would ideally be equal to the target displacement xt ),
then the command signal uc is approximated as :
dn xt,k
uc,k = α0 xt,k + α1 ẋt,k + . . . + αn dtn
(6.9)
178
where
τj
αj = Aj , j = 0, 1, 2, .., n (6.10)
If in Eq. 6.9 the measured displacement is used instead of the target one, the
estimated command signal would have been:
dn xm,k
uest,k = α0 xm,k + α1 ẋm,k + . . . + αn dtn
(6.11)
Varying adaptively, the coefficients αj of Eq. 6.9, ATS can take into account any
alterations of time delay and amplitude error during the RTS tests, and not assum-
ing these parameters as constant. The main concept of the ATS can be shown at
Figure 6.7, where it is compared with delay compensators without feedback.
Figure 6.6: Comparison between input and output displacements of a servo-

hydraulic system
(a) (b)
Figure 6.7: a: Actuator delay compensator without feedback, b: Actuator delay

compensator with feedback
In order to minimize the time delay and amplitude error, the measured dis-
placement and the target displacement should be as close as possible. This can
179
be achieved by minimizing the error between uc , and uest , using the least squares
method:
q
X
Jk = (uc,k−1 − uest,k−1 )2 (6.12)
i=1
Where q is the number of time steps which are taken into account for the mini-
mization of the error, and represents the size of the time window over which the αj
coefficients are determined. Minimizing the cost function Jk the following relation-
ship can be obtained:
T X )−1 X T U
A = (Xm (6.13)
m m c
n
Where A = [α0,k α1,k α2,k . . . αn,k ]T , Xm = [xm , ẋm , . . . ddtxnm ], xm =
[xm,k−1 , xm,k−2 , xm,k−q ]T , Uc = [uc,k−1 , uc,k−2 , . . . , uc,k−q ]T Since the coefficients of
A are now known, they can be substituted into Eq. 6.9, and the new command
signal at time tk can be determined:
dn xt,k
uc,k = α0,k xt,k + α1,k ẋt,k + . . . + αn,k dtn
(6.14)
From Eq. 6.13 the amplitude error and the time delay for the specific time step, tk
can be determined based on the first two A coefficients:
1 α1,k
Ak = α0,k , τk = α0,k
(6.15)
It should be noted that when the amplitude error is not significant (Ak = 1), then
α1,k is approximately equal to the time delay, τk . Even though good displacement
control could be, theoretically, achieved using higher order derivatives in Eq. 6.14,
this might cause problems in RTS tests amplifying noise and leading to probable
instability of the system. Choosing n = 2, Eq 6.14 can be simplified to:
uc,k = α0,k xt,k + α1,k ẋt,k + +α2,k ẍt,k (6.16)
180
The velocity and the acceleration are calculated by using the finite difference method:
xtk −xtk −1 ẋtk −ẋtk−1

ẋtk = T , ẍtk = T
(6.17)
The measured displacement, however, will contain noise and must therefore be fil-
tered. A 6th order Butterworth filter was used by Chae [21] in order to achieve better
estimate of ẋm , and ẍm . The implementation of the filter unavoidably introduces a
time delay. Hence, the exact same filter should be used in the compensated actuator
command displacement, uc,k , so a set of synchronised data of uc,k−1 , and uest,k−1 is
obtained. Furthermore, user-defined limit values were used for the ATS coefficients
in order to avoid any erroneous values, and prevent any potential instability. With
regard to the value, q, of the number of time steps which are taken into account in
the minimization of the function Jk of Eq. 6.12, if it is too short, then the corre-
sponding coefficients lack accuracy. If, on the other hand, q takes a very large value
then the calculation time increases and the coefficients will not respond as quickly to
a change in delay. Based on the speed of the computer, the coefficients may not be
available by the end of the simulation time step (1/1024 in case of Chae tests [21]).
Hence, this data was obtained based on a 1 second time-window, corresponding to
a value of q equal to 1024 time steps. In order to minimize computational time,
data was retained every 16th time step decreasing the size of Xm at Eq. 6.13 from
1024x3 to 64x3.
6.2.3 Modified ATS
A modified version of the ATS compensator has been proposed by Malloy and Blake-
borough (unpublished). When these lines were written, no journal paper or written
work had been submitted with regard to this method. The main target of this
method was the further minimization of the computational time when the ATS co-
efficients were determined. This was achieved by changing the time window of 1
second, from a rectangular to an exponentially decaying window. Every sample has
the same weighting when the rectangular time window is used. However, when the
proposed decaying window is used a new parameter, ρ, is introduced where 0 < ρ ≤ 1
181
(see Eq. 6.18). Hence, the most recent samples are weighted more, in contrast with
the older samples. A schematic overview of the comparison of these two methods
can be seen at Figure 6.8 where the influence of different values of ρ are illustrated
and compared with the original ATS method. In comparison with the original ATS
where only 1 in every 16 data points were used, all the sample data are used in the
modified ATS compensator in order to determine the corresponding coefficients.
The minimization function of Eq. 6.12 has to be adjusted as well. The new
weighted function, Zk can be written as:
q
X
Zk = ρi−1 (uc,k−1 − uest,k−1 )2 (6.18)
i=1
If the parameter, ρ = 1, then Zk = Jk , and the delay compensator method is the
ATS method described earlier. In order to obtain the new parameters of the modified
ATS compensator, Zk has to be minimized taking the following form:
XTρ Uc = XTρ Xm Ak ⇒ Ak = (XTρ Xm )−1 XTρ Uc (6.19)
where:
 
2
xm,k−1 ρxm,k−2 ρ xm,k−3 ...
XTρ = 
 
2 (6.20)
ẋm,k−1 ρẋm,k−2 ρ ẋm,k−3 ...

 
ẍm,k−1 ρẍm,k−2 ρ2 ẍm,k−3 ...
 
uc,k−1
 
 
uc,k−2 
Uc =  (6.21)
 

u
 
 c,k−3 
 
...
182
 
x ẋm,k−1 ẍm,k−1
 m,k−1 
 
xm,k−2 ẋm,k−2 ẍm,k−2 
Xm = (6.22)
 

xm,k−3 ẋm,k−3 ẍm,k−3 
 
 
... ... ...
 
α0,k 
 
Ak = α 
 1,k  (6.23)
 
α2,k
Figure 6.8: Comparison of weighting of samples between ATS and modified ATS
compensators
The matrices Xρ and Xm have dimensions qx3, the vector Uc has dimensions
qx1, and the vector Ak 3x1. Defining Mk = (XTρ Xm ) and Wk = XTρ Uc , Eq. 6.19
becomes:
Ak = M−1
k Wk
(6.24)
183
It can be seen that Mk is a 3x3 matrix, and Wk a 3x1 vector, regardless of the value
of q. It can be seen that Mk can be updated at each time step according to:
Mk = ρMk−1 + xk xTk (6.25)
Where
 
x
 m,k 
 
xk = 
ẋm,k 
 (6.26)
 
ẍm,k
Similarly, Wk can be also updated to:
Wk = ρWk−1 + uc,k xk (6.27)
From the above equations it can be concluded that the modified ATS method is no
longer based on Eq. 6.13, where a 64x3 matrix Xm is used every time the coefficients
are updated. The modified ATS is much more efficient since only current values of
xm , ẋm , and ẍm are used as feedback to Eq. 6.24 in order to obtain the ATS
coefficients, instead of vectors of previous measurements.
Figure 6.9 shows how the modified ATS method was implemented using Simulink.
The whole process described earlier can be overviewed in this Figure, where the
Pseudo-Inverse block takes measured displacements and their derivatives, along with
the command displacements as inputs, and the ATS coefficients are extracted and
updated for every time step, based on Eq. 6.24. In this block the matrices Wk , and
Mk are calculated in order to produce the vector Ak and the corresponding coeffi-
cients: α0 , α1 , and α2 . These coefficients are used as an input to the Compensator
block where the final command signal is determined based on Eq. 6.16 and sent to
the actuator.
184
Figure 6.9: Schematic overview of the modified ATS delay compensator
6.2.4 Implementation of ATS
A preliminary set of experiments was carried out in order to safely assign the param-
eters for the RTS test. Based on both the characterization tests described in Chapter
3, along with the tests of this section, the following values of the PID controller were
used, based on the auto-tuning process: P = 34.98dB, I = 0, D = 0, L = 0.0008.
An additional parameter had to be taken into consideration with regard to the
final implementation of the modified ATS compensator. Rig flexibility is something
that occurs in all experimental rigs to some extent. As already mentioned in Chap-
ter 3, where the characterization tests were described, a linear encoder was attached
at the mid-height of the central steel plate of the EDs in order to capture the actual
elastomer’s deformation, since some rig flexibility is expected. The displacements
extracted from the encoder are used to determine the ATS coefficients, whereas
the PID controller uses data from the LVDT mounted on the actuator. The dif-
ference between the actuator’s displacement xm , and the encoder’s displacement,
xenc , which is determined based on the actual movement of the damper gives the
rig deflection, xrig . This displacement is added to the already determined command
displacement, extracted from the modified ATS compensator, in order to give the
corrected displacement, which is also the displacement which is sent to the actu-
185
ator. A predefined displacement time history (Figure 6.10a), which was extracted
by analysis of an SDOF steel frame equipped with elastomeric dampers subjected
to scaled El Centro ground motion and corresponds to the dampers displacements,
was used in order to evaluate the importance of rig deflection. This time history is
based on a user-defined command which included cycling tests in a range of +

− 5 mm
amplitudes, and 0.25 − 5Hz frequencies. Despite the fact that the rig was designed
to be as stiff as possible Figure 6.10 shows that this factor is not negligible.
Two experiments were carried out in order to provide an initial estimate of the mod-
ified ATS coefficients, with no real time substructure feedback loop. Both of them
were based on pre-defined displacement time histories. The first one is based on
a white noise signal with a maximum displacement 5.5 mm (corresponding to 45%
shear strain of the damper) and a 5 Hz bandwidth in order to cover the whole range
of displacements and frequencies where the main interest is. The second test was
the user-defined time history with the range of +

− 5 mm amplitudes, and 0.25 − 5Hz
frequencies, which was used in order to present the effect of the rig deflection. Fig-
ure 6.11 shows the displacement time histories and the corresponding measured
displacements, xm , without the implementation of the modified ATS method.
186
(a)
(b)
Figure 6.10: User defined time history a: Comparison of Instron controller displace-
ment with encoder’s displacement, b:Corresponding rig deflection
Amplitude error was measured based on the peaks of these signals, while the
time delay was measured based on the zero-crossings method. The aforementioned
tests had an average delay of 10.2 ms, and an amplitude error of 0.6%. Based on
minimization of the difference of uc , and xm , and Eq.6.13 the mean values of the αi
coefficients are determined for the duration of the tests: α0 = 1.0, α1 = 0.01s, and
187
α2 = 0.00006s2 . These values will be used as initial estimates of the αi coefficients
when modified ATS compensator is going to be used in these initial tests. Moreover,
similar values for the time delay and the amplitude error can be determined from
Eq. 6.10, using the ATS coefficients as functions of time delay and amplitude error.
The aforementioned values were also in a very good agreement with the cyclic tests
which were carried out as part of the thesis, and are described in detail in Chapter 3.
Based on these values, limits were assigned to the coefficients. These limits prevent
system instability by providing minimum and maximum values for α0 , α1 , and α2 ,
along with the rate of change of these coefficients. Furthermore, the modified ATS
method holds these coefficients to their initial values until the actuator has moved
through enough displacements, so that Eq. 6.24 is able to provide accurate estimates.
When the measured actuator’s displacements are relatively small, then the matrix
XρT Xm may cause instability to the system producing unrealistic coefficients. Hence,
user-defined limit values have been assigned.
Eq. 6.15 shows the relationship between the coefficient α0,k and the amplitude
error, Ak . Since the actuator response is not anticipated to achieve more than +
− 30%
of the actuator’s command during an RTS test, the permitted range of values for
the α0,k would be 0.7 < α0,k < 1.3. A maximum rate of change for α0,k of 2 per
second is assumed, implying that the change in the amplitude response cannot be
more than 2 within a second. The lower permitted values for α1,k , and α2,k are taken
as 0, and correspond to zero time delay. Assuming that the maximum time delay
cannot surpass 40 ms, which is a reasonable assumption based on the tests described
max = 0.04, and αmax = 0.008, under the
in Chapter 3, then based on Eq. 6.10: α1,k 2,k
assumption that A = 1. The maximum rate of change of α1,k is assumed not to
exceed 0.05. Based on Eq. 6.15 this corresponds to a change in the time delay of 50
ms. Larger values were considered as irrational. The maximum limit value for α2,k is
max = (τ max )2 /2 = 0.0008s2 . Differentiating
calculated based on Eq. 6.10. Hence α2,k k
dα2 d τ2 dα1
α2 with respect to time in 6.10: dt = dt ( 2 ) = τ dτ
dt = α1 dt . Therefore, the
maximum rate of change of α2 , assuming A = 1.0, equals with [α1max ( dα

dt )max ]/2.
1
The product α1max ( dα

dt )max was divided by 2, because is not expected that both α1 ,
1
dα1
and dt will reach their maximum values. These limits are set for the given system,
188
and may be updated based on the actuator and specimen characteristics and the
anticipated time delay and amplitude error.
Figure 6.11 shows the displacement time history when no compensator is used,
and Figure 6.12 shows the same time history displacements when the modified ATS
compensator has been used, comparing the target and the encoder displacements.
A much better performance has clearly been achieved. A trial and error procedure
was followed in order to determine the decay factor, which corresponded to using
ρ = 0.996. Table 6.1 shows the effectiveness of the modified ATS method in terms
of Normalised Root Mean Square (NRMS) error, time delay, and amplitude error.
NRMS is calculated as:
N
X
(xt,k − xm,k )2
i=1
N RM S = N
(6.28)
X
2
(xt,k )
i=1
It can be noticed that the compensator is highly effective for both of these tests.
The NRMS error was decreased from 10.1% to 0.9% for the Multi-Frequency and
from 15.9% to 1.7% for the White-Noise displacement time history. The time delay
was substantially decreased, producing approximately 1 ms time delay for both of
the tests. With regard to the amplitude response, the associated error was almost
eliminated. The modified ATS compensator therefore seems to be highly effective
regardless of the displacement and frequency alterations, since its coefficients are
updated in every time step, and therefore it is able to take into account any changes
of the time delay and amplitude error during the RTS test. Figure 6.12 shows the
comparison between the target and encoder’s displacement, where the influence of
the compensator can be clearly seen, comparing with Figure 6.11. However, in order
to have a better insight into these comparisons, Figure 6.13 compares the target and
encoder displacements and their synchronization for both of the aforementioned
tests, with and without the compensator. The effectiveness of the proposed delay
compensator is clearly shown.
With regard to the values of the αi coefficients, and their behaviour under the
time histories, Figure 6.14 shows that at the start of the experiments these coeffi-
189
cients are the same as the initial values. Once, xm exceeds the limit value of 0.1
mm then the modified ATS compensator starts to calculate the αi coefficients. The
mean values for both of these tests are summarised in Table 6.2. These mean values
will be used as better estimates for the RTS tests which are going to be described
in the following sections. Observing Table 6.2, and Figure 6.14 it can be noticed
that α0 coefficients are almost 1, implying that the amplitude error in the servo-
hydraulic system is almost negligible, comparing to the time delay. Furthermore, α1
coefficients are very close to the initial assumptions carried out before the beginning
of the tests, which correspond to a mean time delay of approximately 10 ms. The
time variation of the coefficients shows that they correspond very well within the
maximum and minimum values which were assigned to the αi coefficients and their
corresponding rate of change.
190
(a) White noise displacement time history
(b) Multi-Frequency displacement time history
Figure 6.11: Comparison of target displacement displacement with encoder’s dis-

placement without modified ATS compensator
191
(a) White noise displacement time history
(b) Multi-Frequency displacement time history

placement with modified ATS compensator
192
Table 6.1: Evaluation of modified ATS delay compensator
MultiFrequency Command White-Noise command

Parameters
without modified ATS with modified ATS without modified ATS with modified ATS
NRMS (%) 10.1 0.9 15.9 1.7
Time Delay (ms) 10.3 1.1 10 0.9
Amplitude error (%) 0.3 0.02 0.6 0.09
Table 6.2: Mean values of αi coefficients
Mean Values
Displacement Time History
α0 α1 (sec) α2 (sec2)
Multi-Frequency command 1.0045 0.0106 0.000108
White-Noise command 1.0058 0.0108 0.000081
(a) Multi-Frequency without modified ATS(b) Multi-Frequency with modified ATS com-
compensastor pensastor
(c) White-Noise without modified ATS com-(d) White-Noise with modified ATS compen-
pensastor sastor

placement - Sychronization plots
193
(a) α0 (White Noise) (b) α0 (Multi-Frequency)
(c) α1 (White Noise) (d) α1 (Multi-Frequency)
(e) α2 (White Noise) (f) α2 (Multi-Frequency)
Figure 6.14: Time variation of αi coefficients
194
6.2.5 RTS test in SDOF structure
The total emulated system which is going to be tested in real time, will be a 1-storey,
1-bay steel frame connected with diagonal braces equipped with EDs (Figure 6.16).
The behaviour of the EDs along with the proposed hysteretic model was described
in Chapters 3 and 4. The EDs described in these chapters will comprise the physical
substructure here (Figure 6.15).
Figure 6.15: Physical substructure for RTS testing
Figure 6.16: Physical and numerical part of the RTS test
The numerical part consists of the remaining part of the emulated structure:
the steel frame. As already mentioned in Section 5.3.4 (where the influence of the
bracing to the frame’s lateral stiffness and damping is being considered), assuming
a large value of the stiffness of the diagonal braces leads to the rational assumption
of neglecting the influence of the diagonals in terms of lateral resistance. There-
fore, only the EDs along with the steel frame were assumed to contribute at the
195
ground motions, which is consistent with the research carried out in Chapter 5. The
schematic presentation of the RTS process was shown at Figure 6.5, and is presented
here as well for convenience at Figure 6.17
Figure 6.17: Schematic overview of the RTS process
Movement has been assumed to be restrained to the horizontal, so only one DOF
will be tested. The equation of motion for the Single Degree of System (SDOF) is
expressed as:
mẍ(t) + cẋ(t) + kx(t) + FD (t) = −mẍg (t) (6.29)
where k, m, and c are the elastic story stiffness, damping coefficient, and mass
which were assigned to the analytical model, and FD is the force component which
is extracted by the EDs. This equation was modelled in Simulink, and is presented
in Figure 6.18. The equation of motion numerically determines the target position,
velocity, and acceleration, which are used as input to Eq. 6.16 in order to produce
the signal sent to the actuator. The integration of the differential equation of motion
was based on the Bogacki and Shampine method [16]. The numerical models used in
this chapter were very simple, and the integration algorithm selected is not critical
factor for the system’s stability.
196
Figure 6.18: Implementation of equation of motion for SDOF structure
The storey elastic stiffness was taken as 5.2 times larger than the effective stiffness
of the EDs (based on the results extracted from the characterization tests under
2Hz loading frequency, 30% shear strain, and 20o C), which corresponds to a story
stiffness of 22.6x106 N/m. A mass of 143.26 tons was numerically assigned to the
system, so that an undamped natural frequency of the SDOF structure of 1.99Hz
has been achieved. However, this frequency corresponds to the frame without the
EDs. Assuming the aforementioned EDs’ parameters, the natural frequency of the
emulated system increases to 2.18 Hz, which is in very good agreement with natural
frequencies obtained from free response tests of the total system. With regard to
the inherent structural damping, a value of ζ = 0.02 was used.
It should be noted that relatively small value of natural frequency of the frame
(which resulted from the large value of the mass) has been intentionally used in the
RTS tests. Based on Eq. 6.6 higher frequency corresponds to larger values of neg-
ative damping and higher probability of causing instability. The main aim of this
chapter was to carry out simple RTS tests as a validation tool of the new proposed
modified ATS method. The results of these tests were also used as further validation
techniques of the the hysteretic model proposed in Chapter 3 under more realistic
conditions and loading histories compared to simple cyclic or sweep amplitude tests.
Hence, not the full range of capabilities of this method will be captured here; there-
fore small values of natural frequencies were used for stability purposes. However,
Malloy and Blakeborough, who developed this method, are currently evaluating it
197
under more realistic, and demanding conditions.
For the initial estimates of the αi coefficients of the modified ATS delay com-
pensator, the mean values of the aforementioned tests were used: α0 = 1.00515,
α1 = 10.7s, and α2 = 9.45x10−6 s2 . The ground motion which was used in order
to excite the frame, was the 1940 El Centro earthquake (Figure 6.19), scaled to
0.0576 g (equivalent to 0.18 scaled factor) in order not to produce EDs’ displace-
ments larger than the ones corresponding to 50% shear strain. The relatively small
scaling factor used in the El Centro ground motion was due to the size of the EDs.
In reality, bigger dampers would have been used (as was also seen in Chapter 5) in a
realistic building. This could have been achieved by scaling the dampers’ forces and
displacement during a test applying stronger ground motions, or by directly scaling
down the ground motion keeping the size of the dampers and their corresponding
displacements the same. The latter approach was preferred here.
The results (Figures 6.20 and 6.21) showed a very good agreement between mea-
sured and target displacement, while the time delay was decreased from approxi-
mately 13.5 ms (this value was based on Figure 6.22) to 1.2 ms. The amplitude
error reached a maximum of 0.03%, while the NRMS error between target and mea-
sured displacement was 1.57%, demonstrating the effectiveness of the method. The
time variation of the αi coefficients shows that even though a relative large peak was
observed for the α0 , and α1 coefficients at the beginning of the RTS test, it can be
concluded that the initial estimates of these coefficients were capable of capturing the
characteristics of the servo-hydraulic system. The corresponding force-displacement
relationship of the EDs and its comparison with the hysteretic model proposed in
Chapter 4 is shown in Figure 6.23. A very good agreement (1.9% NRMS error) was
achieved between the proposed MGMM and the force data extracted from the RTS
test, giving confirmation of the model’s predictive ability under realistic dynamic
loading conditions.
198
Figure 6.19: ElCentro ground acceleration
Figure 6.20: Comparison of target and measured displacements during RTS test for
SDOF frame
Table 6.3: Evaluation of the modified ATS compensator for SDOF frame under
ElCentro
Evaluation Parameters ElCentro Earthquake

NRMS (%) 1.57
Time Delay (ms) 1.2
Amplitude error (%) 0.03
199
Figure 6.21: Comparison of target and measured displacement during RTS test for
SDOF frame - Sychronization plot
200
(a) α0 - SDOF frame
(b) α1 - SDOF frame
(c) α2 - SDOF frame
Figure 6.22: Time variation of the αi coefficients of the modified ATS compensator
during the scaled ElCentro Earthquake for SDOF frame
201
(a) Force-displacement domain
(b) Force-time domain
Figure 6.23: Comparison of MGMM and RTS force under scaled ElCentro Earth-
quake for SDOF frame
202
6.2.6 RTS test in MDOF structure
The final series of tests was focused on capturing the behaviour of the EDs under
the El Centro earthquake in a very simple MDOF frame. A 1-bay, 3 storey shear-
type structure was modelled in Simulink, in a similar way to the SDOF case. EDs
were added at the first story only. The storey stiffnesses assigned were 22.6x106 ,
22.6x106 , and 18.08x106 , N/m for the first, second, and third floor respectively. The
analytical frame also included masses of 143.26, 143.26, and 71.63 tons assigned at
the first, second, and third floor respectively. The corresponding natural frequencies
are 1.0826, 2.7746, and 3.67 Hz. The same limitations with the SDOF numerical
substructure which was tested in the previous section apply to the MDOF frame as
well, resulting in relatively low values of natural frequencies. The damping matrix
assigned to the numerical substructure was based on Rayleigh damping, and a ratio,
ζ, of 0.02 for the first and third mode. Since, the addition of the dampers alters
the dynamic characteristics of the MDOF frame, it was decided not to analyse the
RTS test based on modal time history analysis, but proceed with the original time
integration time history analysis:
[M ]{ẍ} + [C]{ẋ} + [K]{x} + [FD ] = −[M ]{e}{ẍg } (6.30)
where K is the elastic storey stiffness, C is the damping matrix, M is the mass
matrix, {e} is a 3x1 vector of ones, {ẍ}, {ẋ}, and {x} are 3x1 vectors representing
the acceleration, velocity, and displacement of each floor. [Fd ] is a 3x1 vector rep-
resenting the EDs contribution to the MDOF frame: [0 0 FED ]T , where FED is the
force feedback from the device during the RTS test. The same ground motion was
selected (1940 El Centro earthquake), but in order to produce slightly less than 50%
EDs’ shear strain, it had to be scaled down to 0.0288 g (equivalent to a 0.09 scale
factor).
RTS tests for the MDOF frame validated the effectiveness of the proposed delay
compensator as well. The NRMS error reached the value of 1.17%, while Figures 6.24
and 6.25 show that good agreement was achieved between the measured and the
target displacement. The maximum time delay observed during the RTS test was
203
1.6 ms, while the amplitude error was shown once again to play a minimal role to
the behaviour of the servo-hydraulic system since the corresponding error was not
more than 0.1%. Figure 6.26 shows the time variation of the αi coefficients. The
mean values which were used as initial assumptions were found to effectively capture
the mean values of these coefficients during the RTS test. The α0 coefficient was
almost 1 during the whole test, concluding that the amplitude error was minimal,
while the time variation of the α1 coefficient shows that the mean actuation time
delay is approximately 11 ms, a value which is very close to the initial assumption.
With regard to the hysteretic behaviour, Figure 6.27 shows again that the proposed
MGMM is capturing the dynamic characteristics of the EDs very well under realistic
conditions (1.95% NRMS error).
Figure 6.24: Comparison of target and measured displacements during RTS test for
MDOF frame
204
Figure 6.25: Comparison of target and measured displacement during RTS test for
MDOF frame - Sychronization plot
Table 6.4: Evaluation of the modified ATS compensator for MDOF frame under
ElCentro
Evaluation Parameters ElCentro Earthquake

NRMS (%) 1.17
Time Delay (ms) 1.6
Amplitude error (%) 0.09
205
(a) α0 - MDOF frame
(b) α1 - MDOF frame
(c) α2 - MDOF frame
Figure 6.26: Time variation of the αi coefficients of the modified ATS compensator
during the scaled ElCentro Earthquake for MDOF frame
206
(a) Force-displacement domain
(b) Force-time domain
Figure 6.27: Comparison of MGMM and RTS force under scaled ElCentro Earth-
quake for MDOF frame
207
6.3 Conclusions
The use of real time substructure testing to achieve additional validation of the
MGMM and the effectiveness of elastomeric dampers was the focus of this chapter.
A new method based on the Adaptive Time Series (ATS) method was proposed by
Malloy and Blakeborough (unpublished), in order to minimize the delay between
the target and measured displacement, since it has been proven that this parameter
highly affects the stability of servo-hydraulic systems. This method is based on the
original ATS method proposed by Chae [21]. However, in comparison with the orig-
inal ATS where only 1 every 16 data was used, all the sample data are being used in
the modified ATS compensator in order to determine the corresponding coefficients.
The results showed that this method reduces the time delay by approximately 90%
to a range of 0.5-1.6 ms. The amplitude error was shown to have minimal effect.
The corresponding NRMS errors were largely decreased as well.
Relatively small values of natural frequencies were selected for both SDOF and
MDOF numerical substructures. Since the primary aim of this chapter was to eval-
uate the performance of the EDs hysteretic model, which was proposed in Chapter
4, and its dynamic behaviour under relevantly realistic conditions, and not to exten-
sively evaluate and optimize every parameter of the proposed modified ATS method,
the numerical substructures were kept simple and stability of the actuation system
was prioritised. Due to the dependence of instability of the system of the natural
frequency of the system, it was preferred to use low frequency frames. Further re-
search is currently being carried out by Malloy and Blakeborough in order to further
validate this new modified ATS method on Buckling Restrained Braces with more
sophisticated numerical substructures, and parametric analyses.
Regarding the validation of the dampers hysteresis model, a very good agree-
ment was achieved for both SDOF and MDOF cases (1.9%, and 1.95% NRMS error
respectively). This further boosts our confidence that the specific model can be used
in the analysis of real structures, and not necessarily be useful only for academic
purposes.
208
Chapter 7
Conclusions
7.1 Overview of thesis
Passive dissipative devices have been proved to be effective under strong earthquakes
with regard to the seismic protection of buildings. Elastomeric dampers can be a
very efficient method of seismic enhancement as well. However, limited experimental
studies, lack of information regarding how elastomeric dampers behave in a macro-
scopic way through time, along with the fact that the majority of the elastomers
have been used in base isolation techniques creates the need for further analytical
and experimental validation regarding their effectiveness and modelling approach
for improvement of earthquake response of structures.
Two elastomeric dampers were provided by TARRC. In order to determine the
main characteristics of these dampers a series of tests was carried out in a range of
strain amplitudes, frequencies, and ambient temperatures in a form of sinusoidal dis-
placement time history commands. Based on the corresponding force-displacement
relationships the main characteristics of the dampers were extracted: shear storage
modulus, G0 , and loss factor, n. It was shown that the displacement was the dom-
inant factor which affected the dampers’ characteristics the most, in contrast with
the frequency. Especially the loss factor remained almost constant regardless of any
frequency alteration. Moreover, it was shown that for strain amplitudes below 30%
both G0 , and n exhibit large decrease. However, above 30% the rate of decrease
massively decreases. It also seems that loading frequency and strain amplitude have
209
Conclusions
the exact opposite effects. Another noticeable conclusion is that the dampers ex-
hibit approximately linear visco-elastic behaviour for strain below 30%. The value of
strain amplitude of 30% works as some type of trigger point of the material. Above
this point the dampers exhibit a more classical elastomeric behaviour, which shows
to be less and less dependent on the frequency. With regard to the temperature
effect, both G0 and n decrease with any increase of temperature. However, it was
shown that only strain amplitude effect could change the dampers hysteretic loop;
whereas the frequency and the temperature affect the mechanical properties of the
damper, but not its hysteretic shape.
As part of this thesis a new hysteretic model (MGMM) for capturing the be-
haviour of elastomeric dampers under dynamic loading has been developed and
experimentally validated. This model was based on a modified version of the well
known Generalised Maxwell Model. Part of the proposed model, is the development
of a force-displacement relationship for N Maxwell elements in time domain. How-
ever, it was seen that using N = 1 Maxwell element was able to capture well the
dynamic characteristics of the damper, where the maximum NRMS error observed
was 1.83%. Furthermore, based on a simplified assumption that all the parameters
of the MGMM alter with the same rate with temperature, a new parameter γT was
introduced. Hence, keeping all the parameters of the MGMM constant and change
only the new parameter γT was adequate and efficient enough to capture the be-
haviour of the dampers under different temperatures. Hence, the proposed model
was able to represent the dampers performance under a range of strain amplitudes,
loading frequencies, and ambient frequencies with accuracy.
MGMM was incorporated into OpenSees in order to evaluate the dampers’ con-
tribution to the seismic resistance of a 10 storey steel moment resisting frame. The
frame was analytically evaluated under 20 ground motions, which were scaled in
order to match the EC8 response spectrum, using non linear analyses. Great im-
provement was observed, since the addition of the dampers led the structure to
achieve an enhanced performance level compared to the one which was initially de-
signed. The structure was tested under both DBE and MCE levels for two different
cases: with and without elastomeric dampers. When dampers were added an addi-
210
tional damping of 10% was achieved which caused the building to exhibit only minor
damage even under ground motions scaled to MCE level. More specifically, DMRF
exhibited a mean lateral drift of 1.1% under DBE which was one of the main initial
target performance criteria, while residual displacements were almost eliminated.
The results were more impressive when the frame was tested under MCE level; the
mean interstorey drifts were decreased by 45%, and the residual drifts by 83%.
In order to further validate the proposed constitutive model RTS tests were
carried out for both simple SDOF and MDOF frames equipped with elastomeric
dampers. The analytical model was shown to be able to capture the dynamic char-
acteristics of the damper even under realistic conditions. In order to minimise the
unavoidable time delay and amplitude error an unpublished method (modified ATS
method) proposed by Malloy and Blakeborough was used, which was found to highly
decrease the time delay.
Overall, it can be concluded that essential progress was achieved in advancing
the usability of elastomeric dampers for seismic mitigation. Furthermore, a clear
understanding on how elastomeric dampers behave with regard to their dynamic
characteristics was achieved.
7.2 Recommendations for future work
The proposed model was validated under a series of tests in a range of strain am-
plitudes, loading frequencies and ambient temperatures. The model can be further
validated or optimised under wider range of these parameters, especially under lower
values of ambient temperature. Unfortunately due to lack of funds no approach could
be followed in this research to carry out series of tests while keeping the temperature
constant under lower values of the ambient temperature (lower than 20o C).
One other possible idea for future work would be the effect of the elastomeric
dampers in structures designed based on different materials, such as concrete or
timber. Furthermore, keeping the material the same but changing the dynamic
characteristics of the building could provide us with an optimization tool regarding
the effect of the dampers in different type of structures. Then a total comparison of
211
Conclusions
the effect of the dampers in any type of structure would be possible. Another useful
modelling approach of elastomeric dampers could be the combination of elastomers
with a type of either steel dampers/fuses which could be able to control the maximum
force extracted from the dampers during strong ground motions, probably resulting
in a more visco-plastic behaviour.
Lastly, a more advanced RTS tests could give a better understanding and con-
fidence to the proposed hysteretic model. Furthermore, more advanced numerical
substructures could be carried out to evaluate the effect of the dampers in more
realistic conditions.
212
Appendix A
Mechanical Properties of the

Elasomteric Dampers
213
Generalized Maxwell Model
Table A.1: Mechanical Properties at a range of strain amplitudes, frequencies at

20o C
Shear Strain Frequency (Hz) G0 (Mpa) Geq (Mpa) n

%
0.25 1.356 1.677 0.407
0.5 1.362 1.656 0.414
1 1.469 1.770 0.409
10 2 1.616 1.912 0.402
3 1.701 2.002 0.398
4 1.765 2.068 0.394
0.25 1.042 1.295 0.355
0.5 1.075 1.258 0.352
1 1.149 1.344 0.354
20 2 1.244 1.438 0.350
3 1.298 1.507 0.349
4 1.338 1.559 0.347
0.25 0.875 1.089 0.334
0.5 0.916 1.074 0.333
1 0.976 1.139 0.334
30 2 1.048 1.226 0.336
3 1.090 1.278 0.333
4 1.119 1.321 0.332
0.25 0.807 1.046 0.329
0.5 0.843 0.990 0.326
1 0.892 1.051 0.330
40 2 0.948 1.121 0.331
3 0.983 1.167 0.329
4 1.013 1.209 0.327
0.25 0.766 0.946 0.301
0.5 0.801 0.937 0.304
1 0.844 0.986 0.308
50 2 0.893 1.049 0.310
3 0.920 1.095 0.311
4 0.948 1.132 0.309
214

25o C

%
0.25 1.314 1.732 0.388
0.5 1.328 1.622 0.377
1 1.431 1.722 0.371
10 2 1.547 1.843 0.367
3 1.622 1.921 0.363
4 1.663 1.992 0.364
0.25 1.007 1.245 0.329
0.5 1.010 1.201 0.330
1 1.076 1.274 0.330
20 2 1.157 1.359 0.327
3 1.207 1.419 0.325
4 1.246 1.470 0.323
0.25 0.829 1.046 0.319
0.5 0.860 1.022 0.312
1 0.911 1.076 0.312
30 2 0.974 1.158 0.315
3 1.008 1.202 0.310
4 1.041 1.243 0.309
0.25 0.746 0.931 0.302
0.5 0.773 0.916 0.299
1 0.816 0.970 0.303
40 2 0.861 1.036 0.306
3 0.893 1.079 0.304
4 0.917 1.108 0.301
0.25 0.696 0.878 0.296
0.5 0.720 0.854 0.291
1 0.755 0.904 0.297
50 2 0.795 0.962 0.301
3 0.823 1.002 0.296
4 0.839 1.030 0.300
215
Generalized Maxwell Model

30o C

%
0.25 1.126 1.373 0.361
0.5 1.129 1.370 0.369
1 1.206 1.449 0.367
10 2 1.307 1.553 0.361
3 1.375 1.619 0.355
4 1.420 1.684 0.355
0.25 0.875 1.052 0.317
0.5 0.893 1.050 0.319
1 0.951 1.116 0.320
20 2 1.021 1.194 0.319
3 1.068 1.254 0.315
4 1.096 1.294 0.314
0.25 0.764 0.937 0.306
0.5 0.797 0.935 0.301
1 0.841 0.987 0.305
30 2 0.894 1.053 0.304
3 0.932 1.106 0.300
4 0.952 1.135 0.302
0.25 0.674 0.775 0.278
0.5 0.707 0.818 0.284
1 0.749 0.873 0.289
40 2 0.788 0.934 0.294
3 0.819 0.976 0.291
4 0.839 1.005 0.290
0.25 0.656 0.806 0.281
0.5 0.680 0.802 0.280
1 0.710 0.837 0.283
50 2 0.748 0.892 0.286
3 0.778 0.940 0.286
4 0.817 0.969 0.278
216

35o C

%
0.25 1.005 1.208 0.350
0.5 1.017 1.192 0.352
1 1.079 1.266 0.352
10 2 1.169 1.370 0.349
3 1.219 1.431 0.346
4 1.266 1.484 0.343
0.25 0.804 0.934 0.297
0.5 0.819 0.946 0.305
1 0.874 1.006 0.306
20 2 0.936 1.082 0.306
3 0.972 1.124 0.302
4 1.004 1.166 0.300
0.25 0.726 0.833 0.284
0.5 0.765 0.882 0.288
1 0.814 0.945 0.290
30 2 0.814 0.945 0.290
3 0.848 0.987 0.287
4 0.864 1.017 0.287
0.25 0.644 0.779 0.281
0.5 0.669 0.775 0.279
1 0.705 0.818 0.283
40 2 0.746 0.875 0.285
3 0.770 0.908 0.282
4 0.791 0.936 0.279
0.25 0.609 0.719 0.260
0.5 0.632 0.728 0.266
1 0.660 0.767 0.272
50 2 0.695 0.817 0.274
3 0.715 0.851 0.274
4 0.732 0.877 0.274
217
Appendix B
MGMM validated against sweep

amplitude tests in different
temperatures
This appendix presents how the MGMM is efficient to capture the dynamic response
of the EDs with regard to test under different temperatures:
Figure B.1: Comparison of force between experiment and MGMM model for sweep
218
219
MGMM validated against sweep amplitude tests in different temperatures
220
221
222
223
224
Bibliography
[1] Eurocode 3: Design of steel structures —part 1-1:general rules and rules for
buildings, cen, brussels.
[2] Eurocode 8: Design of structures for earthquake resistance—part 1: General
rules, seismic actions and rules for buildings, cen, brussels.
[3] Peer (pacific earthquake engineering research center). 2015. opensees v2.4.6.
berkeley: University of california berkeley.
[4] Structural applications of taylor fluid viscous dampers.
http://taylordevices.com/pdf/Yearly-updates/
YearlyUpdatedStructuralApplicationChart.pdf. Accessed: 2017-12-
11.
[5] FEMA. Prestandard and Commentary for the Seismic Rehabilitation of Build-
ings. Report No. FEMA356, Federal Emergency Management Agency: Wash-
ington, DC, 2001.
[6] VERLAG MODERNE INDUSTRIE Technical Elastomers, The basis of high-
tech sealing and vibration control technology solutions, 2006.
[7] MathWorks, Mathworks MATLAB Simulink, Available:
http://www.mathworks.com/products/simulink/, 2007.
[8] Rodolfo A., Francesco B., Gabriella C. M., Ahmadi H., Goodchild I, and Fuller
K. Viscoelastic dampers for seismic protection of buildings: An application to
an existing building. In Fifth World Congress on Joints, Bearings and Seismic
Systems for Concrete Structures, 2002.
225
[9] I Aiken. An analytical hysteresis model for elastomeric seismic isolation bear-
ings. Earthquake Engineering and Structural Dynamics, 26(2), 1997.
[10] I.D. Aiken and J.M. Kelly. Earthquake simulator testing and analytical stud-
ies of two energy-absorbing systems for multistory structures. University of
California, Berkeley, 1990.
[11] F. Arima, M. Miyazaki, H. Tanaka, and Y. Yamazaki. A study on buildings
with large damping using viscous damping walls. In Proceedings of the 9th
World Conference on Earthquake Engineering, volume 821, 1988.
[12] Michael J Astrella and Andrew Stuart Whittaker. The performance-based
design paradigm. Number 11. Multidisciplinary Center for Earthquake Engi-
neering Research, 2005.
[13] Klaus-Jürgen Bathe and Edward L Wilson. Numerical methods in finite ele-
ment analysis. Prentice-Hall Englewood Cliffs, NJ, 1976.
[14] C.J. Black, N. Makris, and I.D. Aiken. Component testing, seismic evalua-
tion and characterization of buckling-restrained braces. Journal of Structural
Engineering, 130(6):880–894, 2004.
[15] A Blakeborough, MS Williams, AP Darby, and DM Williams. The devel-
opment of real–time substructure testing. Philosophical Transactions of the
Royal Society of London A: Mathematical, Physical and Engineering Sciences,
359(1786):1869–1891, 2001.
[16] P. Bogacki and L.F. Shampine. A 3(2) pair of Runge - Kutta formulas. Applied
Mathematics Letters, 2(4):321 – 325, 1989.
[17] P. A. Bonnet, C. N. Lim, M. S. Williams, A. Blakeborough, S. A. Neild, D. P.
Stoten, and C. A. Taylor. Real-time hybrid experiments with newmark inte-
gration, m.c.smd outer-loop control and multi-tasking strategies. Earthquake
Engineering & Structural Dynamics, 36(1):119–141, 2007.
[18] P.A. Bonnet. The development of multi-axis real-time substructure testing.
PhD thesis, University of Oxford, 2006.
226
[19] R. Bouc. Forced vibration of mechanical system with hysteresis. In Proc., 4th
Conf. on Nonlinear Oscillation, Prague, Czechoslovakia., 1967.
[20] Eduardo C.C. Overview of Eurocode 8, 2011 (accessed Febru-
ary 3, 2016). http://eurocodes.jrc.ec.europa.eu/doc/WS_335/S1_
EC8-Lisbon_E%20CARVALHO.pdf.
[21] Y. Chae, K. Kazemibidokhti, and J.M. Ricles. Adaptive time series com-
pensator for delay compensation of servo-hydraulic actuator systems for real-
time hybrid simulation. Earthquake Engineering & Structural Dynamics,
42(11):1697–1715, 2013.
[22] K.C. Chang, ML Lai, Tsu T Soong, DS Hao, and YC Yeh. Seismic behav-
ior and design guidelines for steel frame structures with added viscoelastic
dampers, technical report nceer-93-0009. 1993.
[23] K.C. Chang, YY Lin, and ML Lai. Seismic analysis and design of structures
with viscoelastic dampers. ISET Journal of Earthquake Technology, 143-166,
1998.
[24] K.C. Chang, TT Soong, and ML Lai. Development of a design procedure
for structure with added viscoelastic dampers. Applied Technology Council,
ACT-17-1, 1993.
[25] K.C. Chang, TT Soong, ML Lai, and EJ Nielsen. Viscoelastic dampers as en-
ergy dissipation devices for seismic applications. Earthquake Spectra, 9(3):371–
387, 1993.
[26] Kuo-Chun Chang, Yu-Yuan Lin, and Chang-Yu Chen. Shaking table study
on displacement-based design for seismic retrofit of existing buildings using
nonlinear viscous dampers. Journal of structural engineering, 134(4):671–681,
2008.
[27] A.K. Chopra. Dynamics of structures: theory and applications to earthquake
engineering. Prentice-Hall, 2001.
227
[28] C. Christopoulos and A. Filiatrault. Principles of Passive Supplemental Damp-
ing and Seismic Isolation. IUSS Press, 2006.
[29] C. Christopoulos and M. Montgomery. Viscoelastic coupling dampers (vcds)
for enhanced wind and seismic performance of high-rise buildings. Earthquake
Engineering & Structural Dynamics, 42(15):2217–2233, 2013.
[30] W.J. Chung, C.B. Yun, N.S. Kim, and J.W. Seo. Shaking table and pseudo-
dynamic tests for the evaluation of the seismic performance of base-isolated
structures. Engineering Structures, 21(4):365 – 379, 1999.
[31] M.C. Constantinou, T.T. Soong, and G.F. Dargush. Passive energy dissipa-
tion systems for structural design and retrofit. Multidisciplinary Center for
Earthquake Engineering Research Buffalo, New York, M.C.EER Monograph
No. 1, 1998.
[32] M.C. Constantinou, MD Symans, P Tsopelas, and DP Taylor. Fluid viscous
dampers in applications of seismic energy dissipation and seismic isolation.
Proceedings ATC 17, 1:581–592, 1993.
[33] M.C. Constantinou and I.G. Tadjbakhsh. Hysteretic dampers in base isolation:
random approach. Journal of Structural Engineering, 111(4):705–721, 1985.
[34] JP Den Hartog and JP Den Hartog fourth edition Mechanical. Vibrations.
M.C.Graw-Hill Book Company, Inc., New York, 1956.
[35] S.N. Dermitzakis and S.A. Mahin. Development of substructure techniques
for online computer controlled seismic performance testing. Technical report,
College of Engineering, University of California, Berkeley, California, 1985.
[36] J. Dimig, C. Shield, C. French, F. Bailey, and A. Clark. Effective force testing
a method of seismic simulation for structural testing. Journal of Structural
Engineering, 125(9):1028–1037, 1999.
[37] J. Donea, G. Magonette, P. Negro, P. Pegon, A. Pinto, and G. Verzeletti.
Pseudodynamic capabilities of the ELSA laboratory for earthquake testing of
large structures. Earthquake Spectra, 12(1):163–180, 1996.
228
[38] Pant D.R., Montgomery M., Christopoulos C., and Poon D. Viscoelastic cou-
pling dampers for the enhanced seismic resilience of a megatall building. In
16th World Conference on Earthquake Engineering, 16WCEE 2017, volume
1318, 2017.
[39] U. Droka and J. Garcia. Seismic qualification of passive mitigation devices.
Report no.1,2005, Cooperative Advancements in Seismic and Dynamic Exper-
iments (CASCADE), 2005.
[40] RM Elhassan, A Arminak, GC Hart, G Brandow, and X Liu. The design
of analysis of a seismic retrofit of a tall concrete building using supplemental
viscous dampers. In 1996 Tall Buildings Annual Meeting. Los Angeles Tall
Buildings Structural Design Council, pages 96–7, 1996.
[41] Y Fahjan and Z Ozdemir. Scaling of earthquake accelerograms for non-
linear dynamic analysis to match the earthquake design spectra. In The 14th
World Conference on Earthquake Engineering, pages 12–17. Chinese Society
for Earthquake Engineering Beijing, China, 2008.
[42] C.P. Fan. Seismic Analysis, Behavior, and Retrofit of Non-Ductile Reinforced
Concrete Frame Buildings with Viscoelastic Dampers. Ph.D. Dissertation, De-
partment of Civil and Environmental Engineering, Lehigh University, 1998.
[43] John D Ferry. Viscoelastic properties of polymers. John Wiley & Sons, 1980.
[44] TF Fitzgerald, Thalia Anagnos, Mary Goodson, and Theodore Zsutty. Slotted
bolted connections in aseismic design for concentrically braced connections.
Earthquake Spectra, 5(2):383–391, 1989.
[45] H. Frahm. Device for damping vibrations of bodies., April 18 1911. US Patent
989,958.
[46] Ozdemir H. Nonlinear Transient Dynamic Analysis of Yielding Structures.
Ph.D. Dissertation, Department of Civil and Environmental Engineering,
Lehigh University, 1976.
229
[47] T Hale and R Pall. Seismic upgrade of the freeport water reservoir, sacramento,
california. In Twelfth World Conference on Earthquake Engineering, Auckland,
New Zealand, 2000.
[48] R.D. Hanson and T.T. Soong. Seismic design with supplemental energy dissi-
pation devices. Earthquake Engineering Research Institute, 2001.
[49] CB Haselton, AS Whittaker, A Hortacsu, JW Baker, J Bray, and DN Grant.
Selecting and scaling earthquake ground motions for performing response-
history analyses. In Proceedings of the 15th World Conference on Earthquake
Engineering, 2012.
[50] C. Hepburn and R.J. Reynolds. Elastomers: Criteria for Engineering Design.
Applied Science Publishers, London, 1979.
[51] H.M. Hilber, T.J. R. Hughes, and R.L. Taylor. Improved numerical dissipation
for time integration algorithms in structural dynamics. Earthquake Engineer-
ing and Structural Dynamics, 5(3):283–292, 1977.
[52] T. Horiuchi, M. Inoue, T. Konno, and Y. Namita. Real-time hybrid experimen-
tal system with actuator delay compensation and its application to a piping
system with energy absorber. Earthquake Engineering & Structural Dynamics,
28(10):1121–1141, 1999.
[53] T. Horiuchi and T. Konno. A new method for compensating actuator
delay in real–time hybrid experiments. Philosophical Transactions of the
359(1786):1893–1909, 2001.
[54] Chien-Yuan Hou, Deh-Shiu Hsu, and Hsing-Yuan Chen. Experimental verifica-
tion of the restoring-stiffness concept used in fluid dampers for seismic energy
dissipation. The Structural Design of Tall and Special Buildings, 14(1):1–13,
2005.
230
[55] J.S. Hwang and J.C. Wang. Seismic response prediction of HDR bearings using
fractional derivative Maxwell model. Engineering Structures, 20(9):849–856,
1998.
[56] Vayas I., Ermopoulos I., and G. Ioannidis. Steel Structures Design. Klidarith-
mos, Athens, 2005.
[57] Conor D Johnson and David A Kienholz. Finite element prediction of damping
in structures with constrained viscoelastic layers. AIAA Journal, 20(9):1284–
1290, 1982.
[58] M. Nakashima. K. Takanashi. Japanese activities on online testing. Journal
of Engineering Mechanics, 113:1014–1032, 1987.
[59] R. Kanitkar, M. Harms, P. Crosby, and M.L. Lai. Seismic retrofit of a steel mo-
ment frame structure using viscoelastic dampers. J. of Earthquake Technology,
354:207–219, 1998.
[60] T.L. Karavasilis, A. Blakeborough, and M.S. Williams. Development of non-
linear analytical model and seismic analyses of a steel frame with self-centering
devices and viscoelastic dampers. Computers & Structures, 89(11):1232–1240,
2011.
[61] T.L. Karavasilis, J.M. Ricles, R. Sause, and Cheng Chen. Experimental eval-
uation of the seismic performance of steel mrfs with compressed elastomer
dampers using large-scale real-time hybrid simulation. Engineering Structures,
33(6):1859 – 1869, 2011.
[62] T.L. Karavasilis, R. Sause, and J.M. Ricles. Seismic design and evaluation
of steel moment-resisting frames with compressed elastomer dampers. Earth-
quake Engineering & Structural Dynamics, 41(3):411–429, 2012.
[63] K. Kasai, Y. Fu, and A. Watanabe. Passive control systems for seismic damage
mitigation. Journal of Structural Engineering, 124(5):501–512, 1998.
231
[64] K. Kasai, J.A Munshi., M.L. Lai, and B.F. Maison. Viscoelastic Damper
Hysteretic Model: Theory, Experiment and Application. ATC-17-1, 521-532,
San Franscisco, CA, March, 1993.
[65] J.M. Kelly. Earthquake-resistant design with rubber. 1993.
[66] J.M. Kelly, R.I. Skinner, and A.J. Heine. Mechanisms of energy absorption
in special devices for use in earthquake resistant structures. Bulletin of NZ
Society for Earthquake Engineering, 5(3):63–88, 1972.
[67] Donald Knuth. PEER (Pacific Earthquake Engineering Research Center).
2013. PEER NGA-WEST 2 ground motion database., 2011 (accessed February
3, 2016). http://ngawest2.berkeley.edu/.
[68] C.G. Koh and J.M. Kelly. Application of fractional derivatives to seismic anal-
ysis of base-isolated models. Earthquake engineering & structural dynamics,
19(2):229–241, 1990.
[69] A. Kontopanos. Experimental investigation of a prototype elastomeric struc-
tural damper, MSc Thesis in Civil Engineering, Lehigh University, 2006.
[70] ML Lai, P Lu, DA Lunsford, K Kasai, and K.C. Chang. Viscoelastic damper:
a damper with linear or nonlinear material. In Proceedings of 11th World
Conference on Earthquake Engineering, 1996.
[71] J. Lee. Optimal weight absorber designs for vibrating structures exposed
to random excitations. Earthquake Engineering & Structural Dynamics,
19(8):1209–1218, 1990.
[72] K.S. Lee. Seismic Behavior of Structures with Dampers Made from Ultra
High Damping Natural Rubber. Ph.D. Dissertation, Department of Civil and
Environmental Engineering, Lehigh University, 2003.
[73] K.S. Lee, C.P. Fan, R. Sause, and J. Ricles. Simplified design procedure for
frame buildings with viscoelastic or elastomeric structural dampers. Earth-
quake engineering & structural dynamics, 34(10):1271–1284, 2005.
232
[74] TG Lepelletier and F Raichlen. Nonlinear oscillations in rectangular tanks.
Journal of Engineering Mechanics, 114(1):1–23, 1988.
[75] Lyan-Ywan Lu, Ging-Long Lin, and Ming-Hsiang Shih. An experimental study
on a generalized maxwell model for nonlinear viscoelastic dampers used in
seismic isolation. Engineering Structures, 34:111–123, 2012.
[76] Fardis M. EARTHQUAKE-RESISTANT DESIGN OF CONCRETE BUILD-
INGS ACCORDING TO EN1998-1 (EUROCODE 8), 2008 (accessed Octo-
ber 21, 2015). http://eurocodes.jrc.ec.europa.eu/doc/WS2008/Fardis_
2008.pdf.
[77] Paul Macioce. Viscoelastic damping 101.
[78] G. Magonette. Development and application of large–scale continuous
pseudo–dynamic testing techniques. Philosophical Transactions of the Royal
Society of London A: Mathematical, Physical and Engineering Sciences,
359(1786):1771–1799, 2001.
[79] S.A. Mahin and P.S.B. Shing. Pseudodynamic method for seismic testing.
Journal of Structural Engineering, 111(7):1482–1503, 1985.
[80] S.A. Mahin, P.S.B. Shing, C.R. Thewalt, and R.D. Hanson. Pseudodynamic
test method;current status and future directions. Journal of Structural Engi-
neering, 115(8):2113–2128, 1989.
[81] P Mahmoodi, LE Robertson, M Yontar, C Moy, and L Feld. Performance of
viscoelastic dampers in world trade center towers. In Dynamics of structures,
pages 632–644. ASCE, 1987.
[82] N. Makris and M.C. Constantinou. Fractional-derivative Maxwell model for
viscous dampers. Journal of Structural Engineering, 117(9):2708–2724, 1991.
[83] N. Makris, M.C. Constantinou, and GF Dargush. Analytical model of vis-
coelastic fluid dampers. Journal of Structural Engineering, 119(11):3310–3325,
1993.
233
[84] A. Manzoori and H. Toopchi-Nezhad. Application of an extended Bouc-Wen
model in seismic response prediction of unbonded fiber-reinforced isolators.
Journal of Earthquake Engineering, pages 1–18, 2016.
[85] M Martinez-Rodrigo and ML Romero. An optimum retrofit strategy for mo-
ment resisting frames with nonlinear viscous dampers for seismic applications.
Engineering Structures, 25(7):913–925, 2003.
[86] E. Martinez-Romero. Experiences on the use of supplementary energy dissi-
pators on building structures. Earthquake Spectra, 9(3):581–625, 1993.
[87] R.L. Mayes and I.N. Wassim. Comparative seismic performance of four struc-
tural systems and assessment of recent aisc brb test requirements. In SEAOC
74th Annual Convention, pages 251–264, 2005.
[88] Fabio Mazza and Alfonso Vulcano. Control of the earthquake and wind dy-
namic response of steel-framed buildings by using additional braces and/or vis-
coelastic dampers. Earthquake Engineering & Structural Dynamics, 40(2):155–
174, 2011.
[89] N.G. M.C.Crum, C.P. Buckley, and C.B. Bucknall. Principles of polymer
engineering. Oxford University Press, USA, 1997.
[90] C.C. Mei. The Applllied Dynamics of Ocean Waves. John Wiley, New York,
1983.
[91] M.S. Montgomery. Fork configuration damper (FCDs) for enhanced dynamic
performance of high-rise buildings, University of Toronto. PhD thesis, 2011.
[92] S. Nagarajaiah, A.M. Reinhorn, and M.C. Constantinou. Nonlinear dynamic
analysis of 3-d-base-isolated structures. Journal of Structural Engineering,
117(7):2035–2054, 1991.
[93] M. Nakashima. Development, potential, and limitations of real–time online
(pseudo–dynamic) testing. Philosophical Transactions of the Royal Society of
London A: Mathematical, Physical and Engineering Sciences, 359(1786):1851–
1867, 2001.
234
[94] M. Nakashima, H. Kato, and E. Takaoka. Development of real-time pseudo
dynamic testing. Earthquake Engineering and Structural Dynamics, 21(1):79–
92, 1992.
[95] A.D. Nashif, D.I.G. Jones, and J.P. Henderson. Vibration damping. John
Wiley & Sons, 1985.
[96] A. Neuenhofer and F.C. Filippou. Evaluation of nonlinear frame finite-element
models. Journal of structural engineering, 123(7):958–966, 1997.
[97] YQ Ni, JM Ko, and CW Wong. Nonparametric identification of nonlinear
hysteretic systems. Journal of Engineering Mechanics, 125(2):206–215, 1999.
[98] Douglas K Nims, Phillip J Richter, and Robert E Bachman. The use of the
energy dissipating restraint for seismic hazard mitigation. Earthquake Spectra,
9(3):467–489, 1993.
[99] D.E. Olson and D.A. Reed. A nonlinear numerical model for sloped-
bottom tuned liquid dampers. Earthquake engineering & structural dynamics,
30(5):731–743, 2001.
[100] A.V. Oppenheim, A.S. Willsky, and S.H. Nawab. Signals & Systems (2Nd
Ed.). Prentice-Hall, Inc., Upper Saddle River, NJ, USA, 1996.
[101] A.S. Pall and C. Marsh. Response of friction damped braced frames. Journal
of Structural Engineering, 108(9):1313–1323, 1982.
[102] A.S. Pall, C. Marsh, and P. Fazio. Friction joints for seismic control of large
panel structures. PCI JOURNAL, 25(6):38–61, 1980.
[103] K.D. Papoulia and J.M. Kelly. Material characterization of elastomers used
in earthquake base isolation. Earthquake Engineering Research Center, Uni-
versity of California, 1994.
[104] W. Ramberg and W.R. Osgood. Description of stress-strain curves by three
parameters. 1943.
235
[105] SE Randall, DM Halsted, and DL Taylor. Optimum vibration absorbers for
linear damped systems. Journal of Mechanical Design, 103(4):908–913, 1981.
[106] David Roylance. Lecture notes in engineering viscoelasticity, 2001. Depart-
ment of Materials Science and Engineering Massachusetts Institute of Tech-
nology Cambridge, MA 02139.
[107] R. Sause, K.S. Lee, and J. Ricles. Rate-independent and rate-dependent mod-
els for hysteretic behavior of elastomers. Journal of Engineering Mechanics,
133(11):1162–1170, 2007.
[108] KL Shen, TT Soong, K.C. Chang, and ML Lai. Seismic behaviour of rein-
forced concrete frame with added viscoelastic dampers. Engineering Struc-
tures, 17(5):372–380, 1995.
[109] P.S.B. Shing and S.A. Mahin. Experimental error propagation in pseudody-
namic testing. Technical report, College of Engineering, University of Califor-
nia, Berkeley, California, 1983.
[110] S. Silvestri, G. Gasparini, and T. Trombetti. A five-step procedure for the
dimensioning of viscous dampers to be inserted in building structures. Journal
of Earthquake Engineering, 14(3):417–447, 2010.
[111] Clive Siviour. Lecture notes in polymers, 2017. University of Oxford.
[112] R. I. Skinner, J. M. Kelly, and A. J. Heine. Hysteretic dampers for
earthquake-resistant structures. Earthquake Engineering and Structural Dy-
namics, 3(3):287–296, 1974.
[113] RI Skinner, JM Kelly, and AJ Heine. Hysteretic dampers for earthquake-
resistant structures. Earthquake Engineering & Structural Dynamics,
3(3):287–296, 1974.
[114] Tsu T Soong and Gary F Dargush. Passive energy dissipation systems in
structural engineering. Wiley, 1997.
[115] TT Soong. Active structural control: theory and practice. 1990. Essex, UK:
Longman Scientific and Technical.
236
[116] TT Soong and ML Lai. Correlation of experimental results with predictions
of viscoelastic damping for a model structure. In Proceedings of Damping,
volume 91. San Diego, California:[sn], 1991.
[117] D.P. Stoten and E.G. Gómez. Adaptive control of shaking tables using
the minimal control synthesis algorithm. Philosophical Transactions of the
359(1786):1697–1723, 2001.
[118] R.H. Sues, Y.K. Wen, and A.H.S. Ang. Stochastic evaluation of seismic struc-
tural performance. Journal of Structural Engineering, 111(6):1204–1218, 1985.
[119] P. Summers, P. Jacob, J. Marti, G. Bergamo, L. Dorfmann, G. Castellano,
A. Poggianti, D. Karabalis, H. Silbe, and S. Triantafillou. Development of
new base isolation devices for application at refineries and petrochemical facil-
ities. In 13th World Conference on Earthquake Engineering, Vancouver, BC,
Canada, pages 1–6, 2004.
[120] L.M. Sun, Y. Fujino, P. Chaiseri, and B.M. Pacheco. The properties of tuned
liquid dampers using a TMD analogy. Earthquake engineering & structural
dynamics, 24(7):967–976, 1995.
[121] Chang S.Y. A time integration pseudodynamic algorithm. In Proceedings of the
11th World Conference on Earthquake Engineering, Paris, volume Rotterdam:
A.A. Balkema.
[122] M.D. Symans and M.C. Constantinou. Semi-active control systems for seismic
protection of structures: a state-of-the-art review. Engineering structures,
21(6):469–487, 1999.
[123] Alastair T. Testing Buildings to Destruction, 2015 (accessed September 19,
2016). http://www.alatown.com/testing-buildings-destruction/.
[124] Horiuchi T., Nakagawa M., Sugano M., and Konno T. Development of a
real-time hybrid experimental system with actuator delay compensation. In
237
Eleventh World Conference on Erthquake Engineering, volume paper 660,
1996.
[125] Teramoto T., Kitamura H., and Ozaki H. Practical application of high-
damping rubber dampers to a slender building. In 11th World Conference
on Earthquake Engineering, 11WCEE 1996, volume 1801, 1996.
[126] Vesna T. Force based Element vs.Displacement-based Element, 2011 (ac-
cessed February 3, 2016). http://opensees.berkeley.edu/wiki/images/
c/c5/FBEvsDBE.pdf.
[127] W Taniwangsa and JM Kelly. Studies on seismic isolation for housing in
developing regions. In Proceedings of the 11th World Conference on Earthquake
Engineering, Acapulco, Mexico, 1996.
[128] R.S. Thyagarajan. Modeling and analysis of hysteretic structural behavior.
1989.
[129] P. J. Torvik and R. L. Bagley. On the appearance of the fractional derivative
in the behavior of real materials. Journal of Applied Mechanics, Transactions
ASME, 51(2):294–298, 6 1984.
[130] Keh-Chyuan Tsai, Huan-Wei Chen, Ching-Ping Hong, and Yung-Feng Su. De-
sign of steel triangular plate energy absorbers for seismic-resistant construc-
tion. Earthquake spectra, 9(3):505–528, 1993.
[131] RG Tyler. Tapered steel energy dissipators for earthquake resistant struc-
tures. Bulletin of the New Zealand National Society for Earthquake Engineer-
ing, 11(4):282–294, 1978.
[132] E.E. Ungar and E.M. Kerwin Jr. Loss factors of viscoelastic systems in terms of
energy concepts. Journal of the Acoustical Society of America, 34(7):954–957,
1962.
[133] E.H. Vanmarcke and D.A. Gasparini. Simulated earthquake ground motions.
In Structural mechanics in reactor technology. 1977.
238
[134] M. I. Wallace, J. Sieber, S. A. Neild, D. J. Wagg, and B. Krauskopf. Stability
analysis of real-time dynamic substructuring using delay differential equation
models. Earthquake Engineering & Structural Dynamics, 34(15):1817–1832,
2005.
[135] GB Warburton and EO Ayorinde. Optimum absorber parameters for simple
systems. Earthquake Engineering & Structural Dynamics, 8(3):197–217, 1980.
[136] D. Way, B. Taylor, T. Dooley, and G. Stice. Seismic retrofit of the trans
world bank building with the use of friction dampers for energy dissipation.
In Proceedings of the 5th US National Conference on Earthquake Engineering,
volume 1, pages 765–774, 1994.
[137] Y. K. Wen. Method for random vibration of hysteretic systems. Journal of
the Engineering Mechanics Division, 102(2):249–263, 1976.
[138] Yi-Kwei Wen. Method for random vibration of hysteretic systems. Journal of
the engineering mechanics division, 102(2):249–263, 1976.
[139] A.S. Whittaker, V.V. Bertero, C.L. Thompson, and L.J. Alonso. Seismic
testing of steel plate energy dissipation devices. Earthquake Spectra, 7(4):563–
604, 1991.
[140] M. S. Williams and A. Blakeborough. Laboratory testing of structures under
dynamic loads: an introductory review. Philosophical Transactions of the
359(1786):1651–1669, 2001.
[141] A.S. Wineman and K.R. Rajagopal. Mechanical response of polymers: an
introduction. Cambridge University Press, 2000.
[142] CW Wong, YQ Ni, and SL Lau. Steady-state oscillation of hysteretic dif-
ferential model. i: Response analysis. Journal of engineering mechanics,
120(11):2271–2298, 1994.
[143] Jin Kyu Yu. Nonlinear characteristics of tuned liquid dampers. PhD thesis,
1997.
239
[144] Y. Yu, Y. Li, J. Li, and X. Gu. A hysteresis model for dynamic behaviour of
magnetorheological elastomer base isolator. Smart Materials and Structures,
25(5):055029, 2016.
[145] R.H. Zhang, T.T. Soong, and P. Mahmoodi. Seismic response of steel frame
structures with added viscoelastic dampers. Earthquake engineering & struc-
tural dynamics, 18(3):389–396, 1989.
[146] F. Zhu, J. Wang, F. Jin, M. Zhou, and Y. Gui. Simulation of large-scale numer-
ical substructure in real-time dynamic hybrid testing. Earthquake Engineering
and Engineering Vibration, 13(4):599–609, 2014.
240

Basagiannis PDF

Uploaded by

Copyright:

Available Formats

Basagiannis PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Basagiannis PDF

Uploaded by

Copyright:

Available Formats

SEISMIC DESIGN AND EVALUATION

OF MOMENT RESISTING FRAMES

A thesis submitted to the degree of

effectiveness in mitigating the effects of dynamic loading. As part of this evalua-

mechanical properties of the elastomer material in a range of strain amplitudes, load-

relationship of the Generalised Maxwell Model for N Maxwell elements in time

domain, and formed the basis for the proposed model.

dampers were designed to provide an additional 10% damping to the Eurocode 8

Considered Earthquake level.

motion and the corresponding force-displacement relationship of the dampers was

This study was conducted at the Engineering Department of Oxford University.

I deeply appreciate the guidance and support of my supervisor Professor Martin

providing us with the elastomeric specimens. Furthermore, I would like to deeply

helped me spend these years in Oxford as creative and happily as possible.

memory of my beloved uncle, Nasos.

1.1 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Outline of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.2 Passive Energy Dissipation Systems . . . . . . . . . . . . . . . . . . 7

2.2.1 Dynamic Equations of Structures with Dampers . . . . . . . 10

2.2.2 Metallic Dampers . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.2.3 Friction Dampers . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.2.4 Viscous Fluid Dampers . . . . . . . . . . . . . . . . . . . . . 19

2.2.5 Tuned Mass Dampers . . . . . . . . . . . . . . . . . . . . . . 22

2.2.6 Tuned Liquid Dampers . . . . . . . . . . . . . . . . . . . . . 24

2.2.7 ViscoElastic Dampers . . . . . . . . . . . . . . . . . . . . . . 26

2.2.8 Elastomeric Dampers . . . . . . . . . . . . . . . . . . . . . . 36

2.2.9 Past applications of VE/Elastomeric Dampers . . . . . . . . 46

3.2 Experimental rig . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3.3 Mechanical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . 56

4.2 Generalised Maxwell Model . . . . . . . . . . . . . . . . . . . . . . . 74

4.2.1 General Background . . . . . . . . . . . . . . . . . . . . . . . 74

4.2.2 GMM force-time relationship for N elements - Derivation . . 76

4.3 Hysteretic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

4.4 Parameter Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . 87

4.5 Effect of Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . 93

5 Seismic Response of Steel Moment Resisting Frames using Elas-

5.2 Prototype Building . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

5.3 Non Linear Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

5.3.1 Performance Criteria . . . . . . . . . . . . . . . . . . . . . . . 105

5.3.2 Conventional Moment Resisting Frame Model . . . . . . . . . 107

5.3.3 Ground Motions . . . . . . . . . . . . . . . . . . . . . . . . . 108

5.3.4 Retrofitting Procedure . . . . . . . . . . . . . . . . . . . . . . 117

5.3.5 Earthquake Response of steel Conventional Moment Resisting

5.3.6 Earthquake Response of steel Moment Resisting Frame with

Elastomeric Dampers . . . . . . . . . . . . . . . . . . . . . . 140

5.4 Effect of temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

6 Real Time Substructure Test 164

6.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

6.1.1 Shaking Tables . . . . . . . . . . . . . . . . . . . . . . . . . . 164

6.1.2 Pseudo Dynamic Tests . . . . . . . . . . . . . . . . . . . . . . 165

6.1.3 Effective Force Testing . . . . . . . . . . . . . . . . . . . . . . 167

6.1.4 Pseudo Dynamic Substructure Tests . . . . . . . . . . . . . . 168

6.2 RTS testing in Oxford . . . . . . . . . . . . . . . . . . . . . . . . . . 171

6.2.1 Time Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175

6.2.2 Adaptive Time Series . . . . . . . . . . . . . . . . . . . . . . 178

6.2.3 Modified ATS . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

6.2.4 Implementation of ATS . . . . . . . . . . . . . . . . . . . . . 185

6.2.5 RTS test in SDOF structure . . . . . . . . . . . . . . . . . . . 195