tesis en ingles
TRANSCRIPT
Robust Control for Offshore Steel JacketPlatforms under Wave-Induced Forces
Submitted by
Dongsheng Han, Bachelor of Science
A Thesis Submitted for the DegreeMaster of Informatics
(Masters By Research)
School of Computing Sciences
Faculty of Business and Informatics
Central Queensland University
July 2007 (Firstly Submitted)
January 2008 (Finally Submitted)
Approved for Submission by Principal Supervisor
Professor Qing-Long Han
Declaration
The work contained in this thesis has not been previously submitted either in whole
or in part for a degree at Central Queensland University or any other tertiary in-
stitution. To the best of my knowledge and belief, the material presented in this
thesis is original except where due reference is made in text.
Dongsheng Han
20 July, 2007
Abstract
This thesis is concerned with robust control of an offshore steel jacket platform
subject to nonlinear wave-induced forces. Since time delay and uncertainty are
inevitably encountered for an offshore structure and their existence may induce in-
stability, oscillation and poor performance, it is very significant to study on how the
delay and uncertainty affect the offshore structure. In this thesis, a memory robust
control strategy is, for the first time, proposed to reduce the internal oscillations
of the offshore structure under wave-induced forces, so as to ensure the safety and
comfort of the offshore structure.
Firstly, when the system’s states are adopted as feedback, memory state feedback
controllers are introduced for the offshore structure. By using Lyapunov-Krasovskii
stability theory, some delay-dependent stability criteria have been established, based
on which, and by combining with some linearization techniques, memory state feed-
back controllers are designed to control the offshore structure. The simulation re-
sults show that such controllers can effectively reduce the internal oscillations of the
offshore structure subject to nonlinear wave-induced forces and uncertainties. On
the other hand, a new Lyapunov-Krasovskii functional is introduced to derive a less
conservative delay-dependent stability criterion. When this criterion is applied to
the offshore structure, an improved memory state feedback controller with a small
gain is obtained to control the system more effectively, which is sufficiently shown
by the simulation.
Secondly, when the system’s outputs are adopted as feedback, memory dynamic
output feedback controllers are considered for the offshore structure. By employing
a projection theorem and a cone complementary linearization approach, memory
dynamic output feedback controllers are derived by solving some nonlinear mini-
mization problem subject to some linear matrix inequalities. The simulation results
show that the internal oscillations of the offshore structure subject to nonlinear
wave-induced forces are well attenuated.
Finally, robust H∞ control is fully investigated for the offshore structure. By em-
ploying Lyapunov-Krasovskii stability theory, some delay-dependent bounded real
lemmas have been obtained, under which, via a memory state feedback controller or
a dynamic output feedback controller, the resulting closed-loop system is not only
asymptotically stable but also with a prescribed disturbance attenuation level. The
simulation results illustrate the validity of the proposed method.
Acknowledgements
I am deeply indebted to my Principal Supervisor, Professor Qing-Long Han, from
the Central Queensland University (CQU) whose help, sincere suggestions and en-
couragement helped me in all the time of research for and writing of this thesis. His
insight and enthusiasm for research have enabled me to accomplish this work and
are truly appreciated.
It is a pleasure to thank my Associate Supervisor, Dr. Xiefu Jiang, who made
many efforts on my research at CQU. A special thank goes to my another Associate
Supervisor, Dr. Xian-Ming Zhang, whose patient and meticulous guidance and
invaluable suggestions are indispensable to the completion of this thesis.
I would like to take this opportunity to thank the Faculty of Business and Infor-
matics for giving me permission to commence this thesis in the first instance, to do
the necessary research work and to use faculty’s source.
Last, but not least, I thank my parents, who have been a constant source of
support and love.
Table of Contents
Chapter 1: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Offshore Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Time-Delay Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.2 Robust Control . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Significance of This Thesis . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4 Outline and Contribution of the Thesis . . . . . . . . . . . . . . . . . 11
Chapter 2: Modeling of an Offshore structure . . . . . . . . . . . . . . 15
Chapter 3: State Feedback Control . . . . . . . . . . . . . . . . . . . . . 19
3.1 Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Nominal Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2.1 Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2.2 Controller Design . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2.3 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.4 An Improved Delay-Dependent Stabilization Criterion . . . . . 38
3.3 Uncertain Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.3.1 A Norm-Bounded Uncertainty . . . . . . . . . . . . . . . . . . 44
3.3.2 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . 46
3.3.3 A Polytopic Uncertainty . . . . . . . . . . . . . . . . . . . . . 57
3.3.4 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . 61
3.4 State Feedback H∞ Control . . . . . . . . . . . . . . . . . . . . . . . 69
3.4.1 An H∞ Control for Nominal Systems . . . . . . . . . . . . . . 70
3.4.2 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . 80
3.4.3 An Improved Controller Design Scheme . . . . . . . . . . . . . 81
3.4.4 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . 82
3.4.5 An H∞ Control for Uncertain Systems . . . . . . . . . . . . . 86
3.4.6 A Norm-Bounded Uncertainty . . . . . . . . . . . . . . . . . . 86
3.4.7 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . 88
3.4.8 A Polytopic Uncertainty . . . . . . . . . . . . . . . . . . . . . 99
3.4.9 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . 100
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
Chapter 4: Dynamic Output Feedback Control . . . . . . . . . . . . . .111
4.1 System Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.2 Nominal Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.2.1 Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.2.2 Controller Design . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.2.3 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . 120
4.3 Uncertain Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.3.1 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . 126
4.4 Output Feedback H∞ Control . . . . . . . . . . . . . . . . . . . . . . 131
4.4.1 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . 135
4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
Chapter 5: Conclusion and Future Work . . . . . . . . . . . . . . . . . .143
Appendix A: Lemmas Referred . . . . . . . . . . . . . . . . . . . . . . . .147
Appendix B: Useful Theories . . . . . . . . . . . . . . . . . . . . . . . . .149
B.1 Stability of Time-Delay Systems . . . . . . . . . . . . . . . . . . . . 149
B.1.1 Stability Concept: . . . . . . . . . . . . . . . . . . . . . . . . . 149
B.1.2 Lyapunov-Krasovskii Stability Theorem: . . . . . . . . . . . . 150
B.2 Linear Matrix Inequalities . . . . . . . . . . . . . . . . . . . . . . . . 150
B.3 The LMI Toolbox of Matlab . . . . . . . . . . . . . . . . . . . . . . 151
B.4 An H∞ Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
Appendix C: Simulation Diagrams . . . . . . . . . . . . . . . . . . . . .155
C.1 The Simulation Diagram of the Controller Design from Proposition 3 155
C.2 The Simulation Diagram of the Controller Design from Proposition 7 157
C.3 The Simulation Diagram of the Controller Design from Proposition 10 158
C.4 The Simulation Diagram of the Controller Design from Proposition 12 159
C.5 The Simulation Diagram of the Controller Design from Proposition 15 160
C.6 The Simulation Diagram of the Controller Design from Proposition 17 161
C.7 The Simulation Diagram of the Output Feedback Control . . . . . . . 162
C.8 The Simulation Diagram of the Robust Output Feedback Control . . 162
C.9 The Simulation Diagram of the Output Feedback H∞ Control . . . . 164
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .167
List of Tables
3.1 Comparison results about minimum of γ for various h. . . . . . . . . 86
3.2 The achieved minimum values of γ for various h . . . . . . . . . . . . 99
List of Figures
1.1 Offshore drill platforms . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Offshore steel jacket platform . . . . . . . . . . . . . . . . . . . . . . 3
2.1 Steel jacket structure with an AMD . . . . . . . . . . . . . . . . . . . 15
3.1 The displacement of the first floor with no control and ω = 1.8 . . . . 28
3.2 The displacement of the second floor with no control and ω = 1.8 . . 28
3.3 The displacement of the third floor with no control and ω = 1.8 . . . 29
3.4 The displacement of the first floor with no control and ω = 0.5773 . . 29
3.5 The displacement of the second floor with no control and ω = 0.5773 30
3.6 The displacement of the third floor with no control and ω = 0.5773 . 30
3.7 The displacement of the first floor via the controller (3.3) with (3.20)
for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.8 The displacement of the second floor via the controller (3.3) with
(3.20) for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.9 The displacement of the third floor via the controller (3.3) with (3.20)
for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.10 The displacement of the first floor via the controller (3.3) with (3.20)
for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.11 The displacement of the second floor via the controller (3.3) with
(3.20) for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.12 The displacement of the third floor via the controller (3.3) with (3.20)
for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.13 The displacement of the first floor via the controller (3.3) with (3.21)
for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.14 The displacement of the second floor via the controller (3.3) with
(3.21) for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.15 The displacement of the third floor via the controller (3.3) with (3.21)
for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.16 The displacement of the first floor via the controller (3.3) with (3.21)
for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.17 The displacement of the second floor via the controller (3.3) with
(3.21) for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.18 The displacement of the third floor via the controller (3.3) with (3.21)
for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.19 The displacement of the first floor via the controller (3.3) with (3.28)
for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.20 The displacement of the second floor via the controller (3.3) with
(3.28) for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.21 The displacement of the third floor via the controller (3.3) with (3.28)
for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.22 The displacement of the first floor via the controller (3.3) with (3.28)
for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.23 The displacement of the second floor via the controller (3.3) with
(3.28) for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.24 The displacement of the third floor via the controller (3.3) with (3.28)
for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.25 The displacement of the first floor when no control is used to the
system with norm-bounded uncertainties and ω = 1.8 . . . . . . . . . 46
3.26 The displacement of the second floor when no control is used to the
system with norm-bounded uncertainties and ω = 1.8 . . . . . . . . . 47
3.27 The displacement of the third floor when no control is used to the
system with norm-bounded uncertainties and ω = 1.8 . . . . . . . . . 47
3.28 The displacement of the first floor when no control is used to the
system with norm-bounded uncertainties and ω = 0.5773 . . . . . . . 48
3.29 The displacement of the second floor when no control is used to the
system with norm-bounded uncertainties and ω = 0.5773 . . . . . . . 48
3.30 The displacement of the third floor when no control is used to the
system with norm-bounded uncertainties and ω = 0.5773 . . . . . . . 49
3.31 The displacement of the first floor via the controller (3.3) with (3.36)
for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.32 The displacement of the second floor via the controller (3.3) with
(3.36) for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.33 The displacement of the third floor via the controller (3.3) with (3.36)
for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.34 The displacement of the first floor via the controller (3.3) with (3.36)
for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.35 The displacement of the second floor via the controller (3.3) with
(3.36) for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.36 The displacement of the third floor via the controller (3.3) with (3.36)
for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.37 The displacement of the first floor via the controller (3.3) with (3.37)
for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.38 The displacement of the second floor via the controller (3.3) with
(3.37) for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.39 The displacement of the third floor via the controller (3.3) with (3.37)
for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.40 The displacement of the first floor via the controller (3.3) with (3.37)
for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.41 The displacement of the second floor via the controller (3.3) with
(3.37) for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.42 The displacement of the third floor via the controller (3.3) with (3.37)
for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.43 The displacement of the first floor when no control is used to the
system with polytopic uncertainties and ω = 1.8 . . . . . . . . . . . . 61
3.44 The displacement of the second floor when no control is used to the
system with polytopic uncertainties and ω = 1.8 . . . . . . . . . . . . 62
3.45 The displacement of the third floor when no control is used to the
system with polytopic uncertainties and ω = 1.8 . . . . . . . . . . . . 62
3.46 The displacement of the first floor when no control is used to the
system with polytopic uncertainties and ω = 0.5773 . . . . . . . . . . 63
3.47 The displacement of the second floor when no control is used to the
system with polytopic uncertainties and ω = 0.5773 . . . . . . . . . . 63
3.48 The displacement of the third floor when no control is used to the
system with polytopic uncertainties and ω = 0.5773 . . . . . . . . . . 64
3.49 The displacement of the first floor via the controller (3.3) with (3.52)
for ω = 1.8 (polytopic uncertainties) . . . . . . . . . . . . . . . . . . . 64
3.50 The displacement of the second floor via the controller (3.3) with
(3.52) for ω = 1.8 (polytopic uncertainties) . . . . . . . . . . . . . . . 65
3.51 The displacement of the third floor via the controller (3.3) with (3.52)
for ω = 1.8 (polytopic uncertainties) . . . . . . . . . . . . . . . . . . . 65
3.52 The displacement of the first floor via the controller (3.3) with (3.52)
for ω = 0.5773 (polytopic uncertainties) . . . . . . . . . . . . . . . . . 66
3.53 The displacement of the second floor via the controller (3.3) with
(3.52) for ω = 0.5773 (polytopic uncertainties) . . . . . . . . . . . . . 66
3.54 The displacement of the third floor via the controller (3.3) with (3.52)
for ω = 0.5773 (polytopic uncertainties) . . . . . . . . . . . . . . . . . 67
3.55 The displacement of the first floor when no control is used to the
system under disturbance and ω = 1.8 . . . . . . . . . . . . . . . . . 74
3.56 The displacement of the second floor when no control is used to the
system under disturbance and ω = 1.8 . . . . . . . . . . . . . . . . . 74
3.57 The displacement of the third floor when no control is used to the
system under disturbance and ω = 1.8 . . . . . . . . . . . . . . . . . 75
3.58 The displacement of the first floor when no control is used to the
system under disturbance and ω = 0.5773 . . . . . . . . . . . . . . . . 75
3.59 The displacement of the second floor when no control is used to the
system under disturbance and ω = 0.5773 . . . . . . . . . . . . . . . . 76
3.60 The displacement of the third floor when no control is used to the
system under disturbance and ω = 0.5773 . . . . . . . . . . . . . . . . 76
3.61 The displacement of the first floor via the controller (3.55) with (3.65)
for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.62 The displacement of the second floor via the controller (3.55) with
(3.65) for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.63 The displacement of the third floor via the controller (3.55) with
(3.65) for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.64 The displacement of the first floor via the controller (3.55) with (3.65)
for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.65 The displacement of the second floor via the controller (3.55) with
(3.65) for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.66 The displacement of the third floor via the controller (3.55) with
(3.65) for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.67 The displacement of the first floor via the controller (3.55) with (3.68)
for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.68 The displacement of the second floor via the controller (3.55) with
(3.68) for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.69 The displacement of the third floor via the controller (3.55) with
(3.68) for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.70 The displacement of the first floor via the controller (3.55) with (3.68)
for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.71 The displacement of the second floor via the controller (3.55) with
(3.68) for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.72 The displacement of the third floor via the controller (3.55) with
(3.68) for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.73 The displacement of the first floor of the uncontrolled system with
norm-bounded uncertainties under disturbance and ω = 1.8 . . . . . . 88
3.74 The displacement of the second floor of the uncontrolled system with
norm-bounded uncertainties under disturbance and ω = 1.8 . . . . . . 89
3.75 The displacement of the third floor of the uncontrolled system with
norm-bounded uncertainties under disturbance and ω = 1.8 . . . . . . 89
3.76 The displacement of the first floor of the uncontrolled system with
norm-bounded uncertainties under disturbance and ω = 0.5773 . . . . 90
3.77 The displacement of the second floor of the uncontrolled system with
norm-bounded uncertainties under disturbance and ω = 0.5773 . . . . 90
3.78 The displacement of the third floor of the uncontrolled system with
norm-bounded uncertainties under disturbance and ω = 0.5773 . . . . 91
3.79 The displacement of the first floor via the controller (3.55) with (3.74)
for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.80 The displacement of the second floor via the controller (3.55) with
(3.74) for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.81 The displacement of the third floor via the controller (3.55) with
(3.74) for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.82 The displacement of the first floor via the controller (3.55) with (3.74)
for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.83 The displacement of the second floor via the controller (3.55) with
(3.74) for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.84 The displacement of the third floor via the controller (3.55) with
(3.74) for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.85 The displacement of the first floor via the controller (3.55) with (3.75)
for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.86 The displacement of the second floor via the controller (3.55) with
(3.75) for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.87 The displacement of the third floor via the controller (3.55) with
(3.75) for ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.88 The displacement of the first floor via the controller (3.55) with (3.75)
for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.89 The displacement of the second floor via the controller (3.55) with
(3.75) for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.90 The displacement of the third floor via the controller (3.55) with
(3.75) for ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.91 The displacement of the first floor of the uncontrolled system with
polytopic type uncertainties under disturbance and ω = 1.8 . . . . . . 101
3.92 The displacement of the second floor of the uncontrolled system with
polytopic type uncertainties under disturbance and ω = 1.8 . . . . . . 101
3.93 The displacement of the third floor of the uncontrolled system with
polytopic type uncertainties under disturbance and ω = 1.8 . . . . . . 102
3.94 The displacement of the first floor of the uncontrolled system with
polytopic type uncertainties under disturbance and ω = 0.5773 . . . . 102
3.95 The displacement of the second floor of the uncontrolled system with
polytopic type uncertainties under disturbance and ω = 0.5773 . . . . 103
3.96 The displacement of the third floor of the uncontrolled system with
polytopic type uncertainties under disturbance and ω = 0.5773 . . . . 103
3.97 The displacement of the first floor via the controller (3.55) with (3.80)
for ω = 1.8 (polytopic uncertainties) . . . . . . . . . . . . . . . . . . . 104
3.98 The displacement of the second floor via the controller (3.55) with
(3.80) for ω = 1.8 (polytopic uncertainties) . . . . . . . . . . . . . . . 104
3.99 The displacement of the third floor via the controller (3.55) with
(3.80) for ω = 1.8 (polytopic uncertainties) . . . . . . . . . . . . . . . 105
3.100The displacement of the first floor via the controller (3.55) with (3.80)
for ω = 0.5773 (polytopic uncertainties) . . . . . . . . . . . . . . . . . 105
3.101The displacement of the second floor via the controller (3.55) with
(3.80) for ω = 0.5773 (polytopic uncertainties) . . . . . . . . . . . . . 106
3.102The displacement of the third floor via the controller (3.55) with
(3.80) for ω = 0.5773 (polytopic uncertainties) . . . . . . . . . . . . . 106
4.1 The displacement of the first floor when the output feedback con-
troller is used and ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . 120
4.2 The displacement of the second floor when the output feedback con-
troller is used and ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . 121
4.3 The displacement of the third floor when the output feedback con-
troller is used and ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . . . 121
4.4 The displacement of the first floor when the output feedback con-
troller is used and ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . 122
4.5 The displacement of the second floor when the output feedback con-
troller is used and ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . 122
4.6 The displacement of the third floor when the output feedback con-
troller is used and ω = 0.5773 . . . . . . . . . . . . . . . . . . . . . . 123
4.7 The displacement of the first floor when the robust output feedback
controller is used and ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . 127
4.8 The displacement of the second floor when the robust output feedback
controller is used and ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . 127
4.9 The displacement of the third floor when the robust output feedback
controller is used and ω = 1.8 . . . . . . . . . . . . . . . . . . . . . . 128
4.10 The displacement of the first floor when the robust output feedback
controller is used and ω = 0.5773 . . . . . . . . . . . . . . . . . . . . 128
4.11 The displacement of the second floor when the robust output feedback
controller is used and ω = 0.5773 . . . . . . . . . . . . . . . . . . . . 129
4.12 The displacement of the third floor when the robust output feedback
controller is used and ω = 0.5773 . . . . . . . . . . . . . . . . . . . . 129
4.13 The displacement of the first floor when the robust H∞ output feed-
back controller is used and ω = 1.8 . . . . . . . . . . . . . . . . . . . 136
4.14 The displacement of the second floor when the robust H∞ output
feedback controller is used and ω = 1.8 . . . . . . . . . . . . . . . . . 136
4.15 The displacement of the third floor when the robust H∞ output feed-
back controller is used and ω = 1.8 . . . . . . . . . . . . . . . . . . . 137
4.16 The displacement of the first floor when the robust H∞ output feed-
back controller is used and ω = 0.5773 . . . . . . . . . . . . . . . . . 137
4.17 The displacement of the second floor when the robust H∞ output
feedback controller is used and ω = 0.5773 . . . . . . . . . . . . . . . 138
4.18 The displacement of the third floor when the robust H∞ output feed-
back controller is used and ω = 0.5773 . . . . . . . . . . . . . . . . . 138
C.1 The simulation diagram of the Controller Design from Proposition 3 . 155
C.2 The simulation diagram of the Controller Design from Proposition 7 . 157
C.3 The simulation diagram of the Controller Design from Proposition 12 159
C.4 The simulation diagram of the Controller Design from Proposition 15 160
C.5 The simulation diagram of the output feedback control . . . . . . . . 163
C.6 The simulation diagram of the robust output feedback control . . . . 163
C.7 The simulation diagram of the output feedback H∞ control . . . . . . 164
List of Symbols
Rn n-dimension real space
Rm×n set of all real m by n matrices
L2[0, +∞) space of square integrable functions on [0, +∞)
AT (resp. xT ) transpose of matrix A (resp. vector x)
‖x‖ Euclidean norm of vector x
‖A‖ norm of matrix A
A ≤ B B − A is a symmetric positive semi-definite matrix
P > 0 P is a symmetric positive definite matrix
P < 0 P is a symmetric negative definite matrix
I identity matrix of appropriate dimensions
∗ denotes a symmetric term in a symmetric, i.e.[P W∗ Q
]=
[P W
W T Q
]
diag... denotes a block-diagonal matrix
Chapter 1
Introduction
1.1 Offshore Structures
The history of structural designs can be roughly divided into three eras: classical,
modern, and post-modern. Classical civil structural designs deal only with static
loads. The modern designs add specifications on the dynamic response. Today, in
the post-modern era, major civil infrastructures must be designed to satisfy both
static and dynamic requirements in the presence of a specified class of environmental
loads. In practice, many civil structures, for example, vibratory platforms, bridges,
tall buildings, etc., are subjected to different types of environmental loading such as
strong winds, earthquake motions, etc., which induce severe vibration. Moreover,
excessive displacement and velocity of the structures endanger the safety. Since the
theory of structural vibration control was built, much attention has been paid on
this topic in the past decades. As a well-known application, the offshore structure
has been widely investigated recently, see, for example, [32, 34, 35, 37, 54, 56, 61].
Offshore structures are a class of typical vibrating structures under external
loads, which have been widely applied as operating station for the offshore exploita-
tion, floating breakwater, offshore fish-farming platform and combination of the
entertainment facilities. They have evolved from very stiff and relatively shallow-
water structures in 1940s to very flexible deep-water ones in recent years. Shortly
after the turn of this century, more than 6,500 offshore oil and gas installations
2 CHAPTER 1. INTRODUCTION
Figure 1.1: Offshore drill platforms
have been set up around the world, in some 53 countries [72]. There are a number
of different types of permanent offshore platforms used for a particular range, see
Figure 1.1 [74]. The most common type is the steel tubular jacket type structure
which is one kind of fixed platform. The steel structures are built on pile ground-
work composed of welded steel pipes. These pipes with 1 to 2 meters diameter are
fixed into the seabed and the depth up to 100 meters [73]. A drawing of steel jacket
structure is shown in Figure 1.2 [71].
Located in hostile marine environments, offshore structures are commonly equipp-
ed with a helicopter pad, drilling derrick, cranes, offices and accommodations. They
are typically subjected to severe loads due to not only gravity and operating loads
but also winds, waves and also currents. And all these loads may induce the exces-
sive vibration of the structures. Therefore, the risk of failure in these structures is
not only higher than other structures, but also the possibility of local or major dam-
age and considerable human discomfort due to vibrations are more likely. To ensure
safety, the horizontal displacement of the structures needs to be limited. Moreover,
for the comfort of people who work on the platforms, acceleration also needs to be
1.1. OFFSHORE STRUCTURES 3
Figure 1.2: Offshore steel jacket platform
restricted.
Vibration control of offshore structures can be achieved by increasing the cross-
sectional area of individual elements and/or adding bracing members to the frames,
so as to shift the natural frequencies away from the resonating of frequencies. How-
ever, it will multiply cost for meeting excessive construction materials. An alter-
native approach is to implement a passive or active damping method to regulate
the motion of structures. The passive approach requires the detailed understanding
of structural dynamics and materials properties. In [53], Patil and Jangid (2005)
studied the response of offshore steel jacket platforms installed with energy dissipa-
tion devices such as viscoelastic, viscous and friction dampers. In [44], Lee et al.
(2006) presented a typical tension-leg type of floating platform incorporated with
the tuned liquid column damper (TLCD) device. Although passive control devices
can mitigate the vibration of offshore structures without requiring external energy,
the performance may be limited by the environment and choices of materials. On
the other hand, an active control mechanism can be effective over a wide frequency
range with desired reduction in the dynamic response. It requires sensors and ac-
4 CHAPTER 1. INTRODUCTION
tuators connected through a feedback control. In [71], Zribi et al. (2004) presented
nonlinear and robust control schemes for offshore steel jacket platforms by using
Lyapunov theory and an optimal control approach. In [45], Li et al. (2003) devel-
oped an H2 control algorithm for controlling the lateral vibration of a jacket-type
offshore platform by using an active mass damper (AMD). In [48], Ma et al. (2006)
studied the structural vibration control of the same platform as the one in [45], and
presented a feedback and feedforward optimal control (FFOC) law which was shown
more efficient than a classical state feedback optimal control (SFOC) in reducing the
displacement of the same platform. Based on the stochastic dynamic programming
principle and stochastic averaging method for quasi-Hamiltonian systems, Luo and
Zhu (2006) [47] proposed a nonlinear stochastic optimal control (NSOC) strategy
for offshore platforms under wave loading.
In a word, the offshore steel jacket platforms subject to wave forces have been
extensively investigated in recent years, and some nice fruits have been obtained on
this issue. However, all the studies aforementioned are under such an assumption
that the system model is exactly known, also, the delay effects have never been
considered in feedback channels. In fact, the delay and uncertainty are inevitably
encountered for a practical system in control process and signal processing fields
due to actuator speed limit of mechanical systems, transportation time and so on.
Their existences usually result in instability and degrade the performances of the
system under consideration. Therefore, it is of much more practical significance to
investigate how the delay and uncertainty affect the offshore steel jacket platforms,
which motivates this thesis.
1.2. TIME-DELAY SYSTEMS 5
1.2 Time-Delay Systems
1.2.1 Stability Analysis
Consider the following system
x(t) = A0x(t) + Bx(t− h)
x(t) = φ(t), t ∈ [−h, 0](1.1)
where x(t) ∈ Rn denotes the system state, h > 0 is a constant time delay and φ(t) is
an initial condition, which guarantees the existence and uniqueness of the solution
of (1.1).
As is well known, during the past decades, the study on stability for system (1.1)
has been attracted much attention, see, for example, [8, 10, 11, 12, 21, 22, 23, 25,
26, 27, 31, 40, 46, 59, 60, 62, 69]. Two types of stability conditions have been formu-
lated. One is delay-independent, while the other is delay-dependent. As the names
imply, the former doesn’t contain any delay information, while the later depends
on information of a delay. As expected, delay-dependent stability criteria are less
conservative than delay-independent ones in applications. The achieved maximum
admissible upper bound (MAUB) of a delay then becomes a main performance index
to measure the conservatism of a delay-dependent stability condition. The larger
MAUB, the less conservatism.
In the existing literature, there are two kinds of Lyapunov-Krasovskii function-
als, i.e. complete Lyapunov-Krasovskii functionals and simple Lyapunov-Krasovskii
functionals, for estimating the maximum time-delay bound the system can tolerate
and still retain stability. Using the complete Lyapunov-Krasovskii functionals as
those in [17, 18, 19, 25, 26], one can obtain the maximum time-delay bound which
is very close to the analytical limit. Employing the simple Lyapunov-Krasovskii
functionals usually yields conservative results. However, the results based on the
simple Lyapunov-Krasovskii functionals can be easily applied to controller synthe-
sis and filter design. Hence, it is still an attractive topic for finding some simple
6 CHAPTER 1. INTRODUCTION
Lyapunov-Krasovskii functionals, by which one can have less conservative results.
For comparison, we use a numerical example for system (1.1) with
A =
( −2 00 −0.9
), B =
( −1 0−1 −1
).
The analytical limit for stability for the numerical example is calculated to be
hanalytical = 6.17258. In the existing literature, in order to derive a delay-dependent
stability criterion, one transforms system (1.1) into a system with a distributed
delay, i.e.
x(t) = (A + B)x(t)−B
∫ t
t−h
[Ax(ξ) + Bx(ξ − h)]dξ. (1.2)
Choose a Lyapunov-Krasovskii function
V (t, xt) = xT (t)Px(t), P = P T > 0, (1.3)
and apply Razumikhin Theorem to obtain hmax = 0.9041. As pointed out by Gu
et al. (2003) [19] (Example 5.3), for this example, the stability of system (1.1) is
equivalent to that of system (1.2). The conservatism of the result is due to the
application of the Razumikhin Theorem. For some systems (1.1) with different
system’s matrices, the model transformation (1.2) may induce additional dynamics.
To reduce the conservatism, instead of transforming system (1.1) into (1.2), one
transforms it into
x(t) = (A + B)x(t)−B
∫ t
t−h
x(ξ)dξ. (1.4)
Then choosing a Lyapunov-Krasovskii functional
V (t, xt) = xT (t)Px(t) +
∫ t
t−h
xT (ξ)Qx(ξ)dξ
+
∫ 0
−h
∫ t
t+θ
xT (ξ)BT RBx(ξ)dξdθ, (1.5)
where P = P T > 0, Q = QT > 0, R = RT > 0, and using the bounding technique
for some cross term yield hmax = 4.3588 [52]. This result was also derived by
1.2. TIME-DELAY SYSTEMS 7
decomposing delayed term matrix as B = B1 + B2 in [22]. In [8], the author
introduced a descriptor transformation
x(t) =y(t), (1.6)
y(t) =(A + B)x(t)−B
∫ t
t−h
y(ξ)dξ. (1.7)
A Lyapunov-Krasovskii functional was chosen as
V (t, xt)=
(x(t)y(t)
)T (I 00 0
)(P1 0P2 P3
)(x(t)y(t)
)
+
∫ t
t−r
xT (ξ)Qx(ξ)dξ +
∫ 0
−r
∫ t
t+θ
yT (ξ)ATd RAdy(ξ)dξdθ, (1.8)
where P1 = P T1 > 0, Q = QT > 0 and R = RT > 0. Use this model transformation
and bounding technique for cross terms [52] to obtain hmax = 4.4721 in [9]. In [28],
the authors introduce some slack variables (free-weighting matrices) to derive the
same result.
In [27], the author avoided using model transformation on system (1.1) and
proposed the following Lyapunov-Krasovskii functional
V (t, xt) = xT (t)Px(t) +
∫ t
t−h
xT (ξ)Qx(ξ)dξ
+
∫ t
t−h
(h− t + ξ)xT (ξ)(hR)x(ξ)dξ, (1.9)
which is equivalent to
V (t, xt)=xT (t)Px(t) +
∫ t
t−r
xT (ξ)Qx(ξ)dξ +
∫ 0
−r
∫ t
t+θ
xT (ξ)(rR)x(ξ)dξdθ (1.10)
where P = P T > 0, Q = QT > 0 and R = RT > 0. Instead of using the bounding
technique for some cross term, the author used the following bounding
−∫ t
t−h
xT (ξ)(hR)x(ξ)dξ ≤(
x(t)x(t− h)
)T ( −R RR −R
)(x(t)
x(t− h)
)(1.11)
to derive the maximum allowed delay bound as hmax = 4.4721. Compared with
the above mentioned results, the most advantage of the result in [27] is that the
8 CHAPTER 1. INTRODUCTION
stability condition which was formulated in an LMI form, was very simple and
easily applied to controller design, and did only include the Lyapunov-Krasovskii
functional matrices variables P,Q and R, which means that no additional matrix
variable was involved. From the computation point of view, it is clear to see that
testing the result in [27] is less time-consuming than some existing results in the
literature. However, the result hmax = 4.4721 is not close enough to the analytical
limit hanalytical = 6.17258 and work needs to be done to arrive at a value much closer
to the analytical limit. Therefore, the natural question is: How can one improve the
result by using simple Lyapunov-Krasovskii functionals? Answer to this question
will significantly enhance the stability analysis and controller synthesis of time-delay
systems. It seems that using the existing simple Lyapunov-Krasovskii functionals
can not realize the outcome even if one introduces more additional matrices variables
apart from Lyapunov-Krasovskii functional matrices variables. One way to solve the
problem is to choose a new Lyapunov-Krasovskii functional. For this purpose, we
propose the following new simple Lyapunov-Krasovskii functional
V (t, xt) = xT (t)Px(t) +
∫ 0
−h2
∫ t
t+s
xT (θ)Rx(θ)dθds
+
∫ t
t−h2
[x(s)
x(s− h2)
]T [Q1 Q2
QT2 Q3
] [x(s)
x(s− h2)
]ds (1.12)
where xt is defined as xt = x(t + θ),∀θ ∈ [−h, 0] and P = P T > 0, R = RT > 0,[Q1 Q2
QT2 Q3
]=
[Q1 Q2
QT2 Q3
]T
> 0. Based on (1.12), one can have a delay-dependent
stability criterion. Applying this new criterion, one obtains the maximum allowed
delay bound as hmax = 5.7175, which significantly improves the result hmax = 4.4721
in the above mentioned references. One can clearly see that we have made a very
significant step towards the analytical limit for stability of the system.
1.2.2 Robust Control
The robust control of a linear system with time delay is also an important problem
both in theory and in practice, much effort has been done on this topic, see, for
1.2. TIME-DELAY SYSTEMS 9
example, [5, 7, 13, 16, 24, 29, 36, 38, 39, 41, 42, 55, 57, 58, 64, 66, 67, 68] etc. For
simplicity, we consider the following system
x(t) = A0(t)x(t) + A1(t)x(t− h) + B(t)u(t)
x(t) = φ(t), t ∈ [−h, 0](1.13)
where u ∈ Rm is the control input; the coefficient matrices A0(t), A1(t), B(t) are
uncertain, which possibly belong to one of the following uncertainties.
Norm-bounded uncertainty
[A0(t) A1(t) B(t)] = [A0 A1 B] + DF (t)[E0 E1 Eb] (1.14)
where A0, A1, B,E0, E1, Eb are known real matrices with appropriate dimen-
sions and F (t) is a unknown time-varying matrix satisfying F T (t)F (t) ≤ I.
Polytopic uncertainty
[A0(t) A1(t) B(t)] ∈
q∑i=1
αi[Ai0 Ai
1 Bi];
q∑i=1
αi = 1, αi ≥ 0
(1.15)
where Ai0, A
i1, B
i, (i = 1, 2, · · · , q) are constant real matrices.
The basic robust control problem is to seek a suitable controller such that the
resulting closed-loop system of (1.13) is asymptotically stable for all uncertainties
satisfying (1.14) or (1.15). To this aim, two control schemes are well used. One is
to design a state feedback controller as
u(t) = K1x(t) + K2x(t− h) (1.16)
where K1, K2 are called controller gains, which are to be determined. The other one
is to design a dynamic output feedback controller to stabilize the original system
(1.13). Assume the output of the system (1.13) is
y(t) = C0x(t) + C1x(t− h) (1.17)
10 CHAPTER 1. INTRODUCTION
where C0 and C1 are constant real matrices with appropriate dimensions. The
dynamic output feedback controller is of form
xc(t) = Acxc(t) + Bcy(t)
u(t) = Ccxc(t) + Dcy(t)(1.18)
where Ac, Bc, Cc, Dc are controller parameters to be determined.
In the recent years, the robust control for system (1.13) has been widely inves-
tigated based on linear matrix inequality technique and a lot of controller design
methods have been proposed. To mention it, the parameter tuning approach is ad-
dressed to design a state feedback controller in [9, 67]. Since the parameter tuning
approach needs to restrict some matrix variables to some special structures, this
method usually leads to much conservative results. To achieve much better results,
Moon et al proposed an iterative algorithm to work out the controller parameters in
[49], which, recently, has been gained much attention in control synthesis, one can
see [16, 66] etc. On the other hand, dynamic output feedback control has also been
extensively studied [2, 29, 30, 37]. A projection approach has been a main tool to
obtain an output feedback controller, see, for example, [2, 30] and reference therein.
1.3 Significance of This Thesis
• Since humankind stepped into the 21st century, the need of energy has been
augmenting sharply. Therefore, accompanied by increasingly exhausted ter-
restrial energy, more and more attention has been devoted into offshore energy
drilling. Meanwhile, offshore structures provide the necessary possibility for
such an effort. Nevertheless, comparing with terrestrial exploring platforms,
complex ocean environment brings much larger challenge to the safety of the
offshore structures. Specifically, dynamic loading, such as earthquake, wind,
wave, deep water current and so on, can induce large motion of offshore plat-
forms, further potentially endanger the safety and comfort of the platforms.
Hence, it is a meaningful research topic on how to control an offshore structure.
1.4. OUTLINE AND CONTRIBUTION OF THE THESIS 11
• On the other hand, time delays are unavoidable for practical systems, and
their existences usually lead to instability or performance degradation of a
system. It is of great significance to study the effect of time delays on an
offshore structure both in practice and in theory.
1.4 Outline and Contribution of the Thesis
In this thesis, the robust stability and control problem is investigated for offshore
steel jacket platforms subject to nonlinear wave-induced forces. In order to reduce
the internal oscillations of the platforms subject to irregular wave forces, two types of
control schemes are proposed, i.e. state feedback control and dynamic output feed-
back control. Based on Lyapunov-Krasovskii stability theory, some delay-dependent
stabilization criteria are obtained in terms of linear matrix inequalities. By employ-
ing an iterative algorithm, the controllers are then easily obtained thanks to the
Matlab LMI Toolbox.
The outline of the thesis is given as follows:
In Chapter 2, a typical offshore steel jacket platform is described. After analyzing
its structure and parameter characteristics, a mathematical model is built, based
on which, two control strategies are introduced to reduce the internal oscillations in
the next sections.
In Chapter 3, a memory state feedback control is considered for the offshore steel
jacket platform with uncertainty. Firstly, a delay-dependent stability condition is
obtained for a nominal closed-loop system, based on which, the controller is then
designed by employing linearization approaches. The simulation results show that
internal oscillations are effectively reduced when this controller is applied to the
offshore steel jacket platform. Secondly, the above derivatives are successfully ex-
tended to the system with either norm-bounded or polytopic uncertainties. Finally,
an H∞ control stabilization is also developed for this offshore platform, and some
12 CHAPTER 1. INTRODUCTION
delay-dependent criteria are obtained in terms of LMIs, which guarantee the con-
sidered system is asymptotically stable with a prescribed disturbance attenuation
level γ via a state feedback controller.
In Chapter 4, a memory dynamic output feedback control is studied for the
offshore steel jacket platform with uncertainties. A sufficient condition for the ex-
istence of such a controller is first established for a nominal system. In virtue of a
projection theorem and a cone complementary linearization algorithm, a dynamic
output feedback controller is then designed, which can be easily solved by using the
Matlab LMI Toolbox. Simulation results illustrate that the dynamic output feed-
back controller can more effectively reduce the internal oscillations of the offshore
platform subject to wave forces when the system outputs are adopted as feedback.
Then, these results are extended to suit for the uncertain system. Similar to the pre-
vious chapter, the robust H∞ control is also discussed for the system via a dynamic
output feedback controller.
In Chapter 5, this thesis is summarized and some future research topics on
offshore steel jacket platforms are also proposed.
The contributions of the thesis can be summarized in the following
(1) In practice, we study, for the first time, the effect of time delays on the offshore
structure system. Two types of controllers with time delays are proposed to
reduce the internal oscillations of the offshore structure platforms. When the
system states are adopted as feedback, a memory state feedback controller
is designed; while, a memory dynamic output feedback controller is achieved
when the system outputs are considered as feedback;
(2) Two types of uncertainties, namely norm-bounded uncertainty and polytopic
uncertainty, are considered for offshore structure platforms and some sufficient
conditions for robust control are obtained;
1.4. OUTLINE AND CONTRIBUTION OF THE THESIS 13
(3) An H∞ control problem is investigated for the offshore structure platforms,
such that not only the internal oscillations can be reduced via above two types
of controller, but also the system is of a prescribed disturbance attenuation
level; and
(4) In theory, in order to reduce the conservatism of delay-dependent stability
conditions in the literature, a new Lyapunov-Krasovskii functional is first in-
troduced.
Chapter 2
Modeling of an Offshore structure
Because nonlinear self-excited hydrodynamic force is the most dominant external
persistence, we consider an offshore steel jacket platform under wave-induced forces,
see Figure 2.1 [1]. An active tuned mass damper (AMD) mechanism mounted on
top is connected to a hydraulic servo mechanism. The operation of the hydraulic
servo is driven by active control forces regulated by a designed controller.
Figure 2.1: Steel jacket structure with an AMD
The natural frequencies and mode shapes of the undamped free vibration can
be calculated by one of the available structural dynamic softwares. For simplicity
of presentation, we will only consider the first two modes of vibration because these
two modes are the most dominant and hence the most important for the design of
16 CHAPTER 2. MODELING OF AN OFFSHORE STRUCTURE
the control. The equation of motion of the first two modes of vibrations with the
coupled AMD are
z1 = −2ξ1ω1z1 − ω21z1 − ΦT
1 (Fa − FAMD) + ft1 + ft2,
z2 = −2ξ2ω2z2 − ω22z2 − ΦT
2 (Fa − FAMD) + ft3 + ft4,
y = −2ξT ωT (y − U8)− ω2T (y − U8) + 1
mTu,
(2.1)
where z1 and z2 are the generalized coordinates of vibration modes 1 and 2, re-
spectively; y is the horizontal displacement of the AMD; ω1 and ω2 are the natural
frequencies of the first two modes of vibrations; ξ1 and ξ2 are the damping ratio in
the first two modes of vibrations; Φ1 and Φ2 are the first and second mode shapes
vectors, respectively; FAMD is the passive control force vector due to the AMD; ξT
is the damping ratio of the AMD; CT , mT and KT are the damping, the mass and
the stiffness of the AMD, respectively; ωT =√
KT /mT is the natural frequency of
the AMD; U8 is the horizontal displacement of joint 8 in the offshore structure; u is
the control action of the system; and ft1, ft2, ft3, ft4 are the nonlinear self-excited
hydrodynamic force terms [71].
The horizontal displacement of joint 8 in the offshore structure is
U8 = φ1z1 + φ2z2, (2.2)
where φ1 and φ2 are the contributions of the first two mode shapes. Noting that
ΦT1 Fa = φ1u , ΦT
2 Fa = φ2u, (2.3)
the control forces due to the AMD can explicitly be written as
ΦT
1 FAMD = φ1[KT (y − U8) + CT (y − U8)],
ΦT2 FAMD = φ2[KT (y − U8) + CT (y − U8)]
(2.4)
The nonlinear self-excited hydrodynamic force terms can be modeled by using
Morison equation.
Using Equations (2.1)-(2.4), the equations of motion of a steel jacket platform
subjected to nonlinear self-excited hydrodynamic forces can be written as
17
z1 =− 2ξ1ω1z1 − ω21z1 − φ1KT (φ1z1 + φ2z2) + φ1KT y
− φ1CT (φ1z1 + φ2z2) + φ1CT y − φ1u + ft1 + ft2,
z2 =− 2ξ2ω2z2 − ω22z2 − φ2KT (φ1z1 + φ2z2) + φ2KT y
− φ2CT (φ1z1 + φ2z2) + φ2CT y − φ2u + ft3 + ft4,
y =− 2ξT ωT y + 2ξT ωT (φ1z1 + φ2z2)− ω2T y + ω2
T (φ1z1 + φ2z2)
+1
mT
u.
(2.5)
Let x1(t) = z1(t), x2(t) = z1(t), x3(t) = z2(t), x4(t) = z2(t), x5(t) = y(t), and
x6(t) = y(t). The system (2.5) can be written as
x(t) = Ax(t) + Bu(t) + Fg(x, t) (2.6)
where
x(t) =[
x1(t) x2(t) x3(t) x4(t) x5(t) x6(t)]T
, (2.7)
B =[
0 −φ1 0 −φ2 0 1mT
]T, (2.8)
F =
[0 1 0 0 0 00 0 0 1 0 0
]T
, (2.9)
g(x, t) =
[ft1 + ft2
ft3 + ft4
], (2.10)
and
A =
0 1 0 0 0 0A21 A22 −KT φ1φ2 −CT φ1φ2 φ1KT φ1CT
0 0 0 1 0 0−KT φ1φ2 −CT φ1φ2 A43 A44 φ2KT φ2CT
0 0 0 0 0 1ω2
T φ1 2ξT ωT φ1 ω2T φ2 2ξT ωT φ2 −ω2
T −2ξT ωT
(2.11)
with A21 = −ω21 − KT φ2
1, A22 = −2ξ1ω1 − CT φ21, A43 = −ω2
2 − KT φ22, and A44 =
−2ξ2ω2 − CT φ22.
Remark 1. The nonlinear term g(x, t) in (2.6) is uniformly bounded and it can be
assumed that the nonlinear term g(x, t) satisfies the following cone-bounding con-
straint:
‖g(x, t)‖ ≤ µ ‖x(t)‖ , (2.12)
18 CHAPTER 2. MODELING OF AN OFFSHORE STRUCTURE
where µ is a positive scalar.
For the purpose of simulation, the same system’s parameters setting from Abdel-
Rohman [1] is used for simulation studies. The data for the waves are H = 12.19m,
h = 76.2m, λ = 182.88m. The structure consists of cylindrical steel tube members.
The density of steel is 7730.7kg/m3; the density of water, ρw = 1025.6kg/m3. The
weight of the concrete deck carried by the steel members is 6672.3 × 103N, and
Uow = 0.122m/ sec . The dimensions of the structural elements are given in Figure
1. The project areas, volumes, and masses of each member in the structure can
be found in [1]. Using these data, the wave force parameters at each joint can be
calculated [1].
The natural frequencies of the first two modes of vibrations are ω1 = 1.818rps
and ω2 = 10.8717rps. The structural damping in each mode is considered to be
0.005, hence ζ1 = ζ2 = 0.005. The contribution of the first two modes shapes are
φ1 = −0.003445 and φ2 = 0.003463.
The TMD parameters are chosen to tune with the first mode such that ωT =
1.8rps, ζT = 0.15 and KT = 1551.5, CT = 256.
Using the values of the parameters of the system, we find that
A =
0 1 0 0 0 0−3.3235 −0.0212 0.0184 0.0030 −5.3449 −0.8819
0 0 0 1 0 00.0184 0.0030 −118.1385 −0.1117 5.3468 0.8822
0 0 0 0 0 1−0.0114 −0.0019 0.0114 0.0019 −3.3051 −0.5454
;
B =[
0 0.003445 0 −0.00344628 0 0.00213]T
.
In the following chapters, two kinds of control approaches, i.e. state feedback
control and dynamic output feedback control, will be proposed for the above offshore
steel jacket platform.
Chapter 3
State Feedback Control
For practical systems, time delay and uncertainty are unavoidable due to modeling
errors and data transmission, which usually result in instability and degrade per-
formance of the corresponding systems. In this chapter, for an offshore steel jacket
platform, in order to effectively reduce the internal system oscillations, a memory
state feedback controller will be introduced. Based on Lyapunov-Krasovskii func-
tional stability theory and some linearization methods, a memory state feedback
controller will be designed, under which, via numerical simulation, the considered
system can be effectively controlled, that is, the amplitudes of the internal oscilla-
tions are greatly decreased.
In addition, robust and H∞ control via a state feedback control will also be
investigated for the system with uncertainties, either norm-bounded or polytopic.
3.1 Problem Formulation
Consider an offshore steel jacket platform of a general form
x(t) = A(t)x(t) + B(t)u(t) + Fg(x, t)
x(t) = φ(t), t ≤ 0(3.1)
where x(t) ∈ Rn is the state vector, u(t) ∈ Rm is the control, g(x, t) ∈ Rp is the
nonlinear self-excited hydrodynamic force vector, which is uniformly bounded and
20 CHAPTER 3. STATE FEEDBACK CONTROL
satisfies the following cone-bounding constraint:
‖g(x, t)‖ ≤ µ ‖x(t)‖ , (3.2)
where µ is a positive scalar; A(t) and B(t) are system matrices, which may be
time-varying, but are known to belong to a certain compact set Ω, that is,
(A(t), B(t)) ∈ Rn×(n+m) ⊂ Ω
and φ(t) denotes an initial condition. In this chapter, suppose that the states of
system (3.1) are adopted as feedback, in this situation, we introduce a state feedback
controller
u(t) = K1x(t) + K2x(t− h), (3.3)
where K1 and K2 are constant matrices, which are to be determined, delay h is
assumed to be constant satisfying h > 0.
Remark 2. Clearly, the controller (3.3) is dependent on both the current and the
past states, which is called a memory state feedback controller when K2 6= 0, while
a memoryless one when K2 = 0.
The main aim of this chapter is to design a controller of form (3.3) such that
the resulting closed-loop system by (3.1) and (3.3) is asymptotically stable. In the
sequel, we first consider the controller design for the nominal system of (3.1), and
then these results are extended to suit for system (3.1) with either norm-bounded
or polytopic uncertainty. Finally, a robust H∞ control problem of the system is
considered.
3.2 Nominal Systems
In this section, we focus on the stabilization of the nominal system of (3.1). In this
case, we suppose A(t) ≡ A and B(t) ≡ B for all t ≥ 0, where A and B are constant
3.2. NOMINAL SYSTEMS 21
real matrices. The resulting closed-loop system of (3.1) with (3.3) is given as
x(t) = (A + BK1)x(t) + BK2x(t− h) + Fg(x, t),
x(t) = φ(t), t ∈ [−h, 0],(3.4)
3.2.1 Stability Analysis
Based on an integral inequality established recently and the Lyapunov-Krasovskii
stability theory, a delay-dependent stability criterion is first derived, which is stated
as follows.
Proposition 1. For some given scalars µ > 0 and h > 0, system (3.4) is asymp-
totically stable if there exist matrices P > 0, Q > 0, R > 0 and M1, M2, Z1, Z2, Z3
of appropriate dimensions such that
Ξ :=
Ξ11 Ξ12 PF h(A + BK1)T
∗ Ξ22 0 hKT2 BT
∗ ∗ −I hF T
∗ ∗ ∗ −hR−1
< 0, (3.5)
Ψ :=
R M1 M2
∗ Z1 Z2
∗ ∗ Z3
≥ 0, (3.6)
where
Ξ11 := P (A + BK1) + (A + BK1)T P
+ Q + µ2I + MT1 + M1 + hZ1,
Ξ12 := PBK2 −MT1 + M2 + hZ2,
Ξ22 := −Q−MT2 −M2 + hZ3.
Proof. Choose a Lyapunov-Krasovskii functional candidate as
V (xt) = xT (t)Px(t) +
∫ t
t−h
xT (s)Qx(s)ds
+
∫ 0
−h
dθ
∫ t
t+θ
xT (s)Rx(s)ds, (3.7)
where xt = x(t + α), α ∈ [−h, 0] and P > 0, Q > 0, R > 0 are to be determined.
Taking the derivative of V (xt) with respect to time t along the trajectory of system
22 CHAPTER 3. STATE FEEDBACK CONTROL
(3.4) yields
V (xt) =2xT (t)Px(t) + xT (t)Qx(t)
− xT (t− h)Qx(t− h)
+ hxT (t)Rx(t)−∫ t
t−h
xT (s)Rx(s)ds (3.8)
According to Lemma 3 (see Appendix A), for any Mi, Zj (i = 1, 2, j = 1, 2, 3) with
appropriate dimensions, if (3.6) holds, then
−∫ t
t−h
xT (s)Rx(s)ds ≤[
x(t)x(t−h)
]T [ρ11 ρ12
∗ ρ22
] [x(t)
x(t−h)
](3.9)
where
ρ11 := MT1 + M1 + hZ1
ρ12 := −MT1 + M2 + hZ2
ρ22 := −MT2 −M2 + hZ3
In addition, noting from (3.2) that
0 ≤ µ2xT (t)x(t)− gT (x, t)g(x, t). (3.10)
Introducing a new vector
η(t) = [xT (t) xT (t− h) gT (x, t)]T
and substituting (3.9) and (3.10) into (3.8) gives
V (t, xt) ≤ ηT (t)Φ + hΓT RΓη(t)
where
Φ :=
Ξ11 Ξ12 PF∗ Ξ22 0∗ ∗ −I
,
Γ := [A + BK1 BK2 F ]
3.2. NOMINAL SYSTEMS 23
with Ξ11, Ξ12 and Ξ22 are defined in (3.5). Clearly, if the matrix inequalities (3.5)
and (3.6) hold, then using the Schur complement yields Φ+hΓT RΓ < 0. Therefore,
there exists a sufficient small number δ > 0 such that V (t, xt) ≤ −δxT (t)x(t) < 0
for x(t) 6= 0, which guarantees the asymptotical stability of the closed-loop system
(3.4) by Lyapunov-Krasovskii stability theorem. This completes the proof.
Proposition 1 provides a delay-dependent stability condition for system (3.4).
Clearly, this condition is nonlinear due to nonlinear terms, such as PBK1 etc. Below,
in order to get controller gains K1 and K2, we will propose two schemes to linearize
the matrix inequalities (3.5) and (3.6).
3.2.2 Controller Design
In this section, we are concerned with the controller design based on Proposition 1.
To this aim, an equivalent version of Proposition 1 is first presented in the following.
Proposition 2. For some given positive scalars µ > 0 and h > 0, system (3.4) is
asymptotically stable if there exist matrices X > 0, Q > 0, R > 0 and Y1, Y2, M1, M2,
Z1, Z2, Z3 of appropriate dimensions such that
Ξ :=
Ξ11 Ξ12 F hXAT + hY T1 BT µX
∗ Ξ22 0 hY T2 BT 0
∗ ∗ −I hF T 0∗ ∗ ∗ −hR 0∗ ∗ ∗ ∗ −I
< 0 (3.11)
Ψ :=
XR−1X M1 M2
∗ Z1 Z2
∗ ∗ Z3
≥ 0 (3.12)
where
Ξ11 := XAT + AX + BY1 + Y T1 BT + Q + MT
1 + M1 + hZ1,
Ξ12 := BY2 − MT1 + M2 + hZ2,
Ξ22 := −Q− MT2 − M2 + hZ3.
Moreover, the controller gains are given by K1 = Y1X−1 and K2 = Y2X
−1.
24 CHAPTER 3. STATE FEEDBACK CONTROL
Proof. Now, we will prove the matrix inequalities (3.5) and (3.6) are equivalent
to (3.11) and (3.12), respectively. In fact, noting that P > 0 in (3.5), then P is
invertible, define
T1 := diagP−1, P−1, I, I
T2 := diagP−1, P−1, P−1
and set X = P−1, Y1 = K1P−1, Y2 = K2P
−1, R = R−1 and
Q M1 M2
∗ Z1 Z2
∗ ∗ Z3
:=
X 0 0∗ X 0∗ ∗ X
Q M1 M2
∗ Z1 Z2
∗ ∗ Z3
X 0 0∗ X 0∗ ∗ X
then we have
T T1 ΞT1 =
Ξ11 + µ2X2 Ξ12 F hXAT + hY T1 BT
∗ Ξ22 0 hY T2 BT
∗ ∗ −I hF T
∗ ∗ ∗ −hR
T T2 ΨT2 = Ψ
where Ξ and Ψ are defined in (3.5) and (3.6), respectively. Thus, by the Schur
complement, one obtains
Ξ < 0 ⇐⇒ Ξ < 0
Ψ ≥ 0 ⇐⇒ Ψ ≥ 0
which completes the proof.
We are now in a position to design the desired controller. From Proposition 2,
matrix inequality (3.11) is linear on matrix variables, while (3.12) not. However,
only one nonlinear term, i.e. XR−1X, is included in (3.12). This means that only
if this term is linearized, then (3.12) is converted into an LMI. By employing the
Matlab LMI Toolbox, the controller can be easily solved out based on the LMIs. In
the following, two methods will be introduced to linearize this nonlinear term.
3.2. NOMINAL SYSTEMS 25
According to the fact that
XR−1X ≥ 2X − R, (3.13)
we have
XR−1X M1 M2
∗ Z1 Z2
∗ ∗ Z3
≥
2X − R M1 M2
∗ Z1 Z2
∗ ∗ Z3
This leads to a controller design approach based on the solutions of two LMIs, which
is stated in the following.
Proposition 3. For some given positive scalars µ > 0 and h > 0, system (3.4)
with K1 = Y1X−1 and K2 = Y2X
−1 is asymptotically stable if there exist matrices
X > 0, Q > 0, R > 0 and Y1, Y2, M1, M2, Z1, Z2, Z3 of appropriate dimensions such
that (3.11) and
2X − R M1 M2
∗ Z1 Z2
∗ ∗ Z3
≥ 0 (3.14)
Although Proposition 3 provides an approach to design a controller, the obtained
results are usually conservative since it requires 2X − R > 0, which is a restriction
on variables X and R. To remove this constraint, we can introduce a new matrix
variable S > 0 such that
XR−1X ≥ S (3.15)
which is equivalent to
[R−1 X−1
X−1 S−1
]≥ 0 (3.16)
Following the above line yields the second method to design the suitable controller
of form (3.3), which is shown in the following proposition.
Proposition 4. For some given positive scalars µ > 0 and h > 0, system (3.4)
with K1 = Y1X−1 and K2 = Y2X
−1 is asymptotically stable if there exist matrices
26 CHAPTER 3. STATE FEEDBACK CONTROL
X > 0, Q > 0, R > 0, S > 0,X > 0, R > 0,S > 0 and Y1, Y2, M1, M2, Z1, Z2, Z3 of
appropriate dimensions such that (3.11) and
S M1 M2
∗ Z1 Z2
∗ ∗ Z3
≥ 0 (3.17)
XX = I, RR = I, SS = I. (3.18)
Clearly, Proposition 4 is a non-convex feasibility problem due to the equality
constraints in (3.18). So it cannot be directly solved by using Matlab LMI Toolbox.
However, employing the cone complementary approach proposed in [4], we can con-
vert the non-convex feasibility problem into a nonlinear minimization one subject
to LMIs.
A Nonlinear Minimization Problem
Minimize Tr(XX + RR+ SS)
Subject to (3.11), (3.17) and[R XX S
]≥ 0,
[X II X
]≥ 0,
[R II R
]≥ 0,
[S II S
]≥ 0. (3.19)
Similar to [4], an iterative algorithm can be proposed to solve the above nonlinear
minimization problem, which is stated below.
Algorithm 2.1: Maximize h for a given scalar µ > 0.
Step 1 Choose a sufficiently small initial value hini such that (3.11),(3.17)
and (3.19) are feasible. Set hso = hini.
Step 2 Find a feasible set
(X0, Q0, R0, S0,X 0, R0,S0, Y 0i , M0
i , Z0j (i = 1, 2, j = 1, 2, 3)) satisfy-
ing (3.11),(3.17) and (3.19). Set l = 0.
Step 3 Solve the following LMI problem for the variables (X,X , R, R, S,S)
Minimize Tr(X lX + X lX + RlR+ RlR + SlS + S lS)Subject to (3.11), (3.17), (3.19)
Set X l+1 = X,X l+1 = X , Rl+1 = R, Rl+1 = R, Sl+1 = S,S l+1 = S.
3.2. NOMINAL SYSTEMS 27
Step 4 If matrix inequality (3.12) is satisfied, then set hso = hini and in-
crease γini to some extent and go back to Step 2. If the condition
(3.12) is not satisfied within a specified number of iterations, then
exit, otherwise, set l = l + 1 and go to Step 3.
Remark 3. The proposed algorithm provides a procedure to obtain a suboptimal
upper bound h of delay size. It is worth noting that the inequality (3.12) is used as
the stopping criterion in the above algorithm since it is numerically very difficult to
exactly obtain the minimum value, 36, of Tr(X lX +X lX + RlR+RlR+SlS+S lS).
Remark 4. Based on Proposition 1, two controller design schemes are proposed, i.e.
the Controller Design from Proposition 3 and the Controller Design from Proposition
4. In general, the former scheme can tolerate a larger upper bound h than the later
one. Moreover, when these two schemes are applied to the offshore structure in
(2.6), even for the same admissible bound h, the obtained controller gain by the
later is smaller than that by the former, which is shown in the next section.
3.2.3 Simulation Results
In the previous section, a memory state feedback controller is introduced to control
an offshore structure platform and two controller design schemes are proposed. Now,
we aim to show the effectiveness of the proposed methods via numerical simulations.
For simulation to the offshore structure platform (2.6), two cases of wave fre-
quency are taken into consideration, ω = 1.8 rps and ω = 0.5773 rps. We set
µ = 0.1. In the sequel, we compare the state curves under controller (3.3) with
h = 0.1 by two schemes with those by no control.
The responses of the offshore structure with no control are shown in Figures 3.1
- 3.6. The uncontrolled lateral displacements of the three floors of jacket for ω =
1.8 and ω = 0.5773 are shown in Figures 3.1-3.3 and Figures 3.3-3.6, respectively.
Clearly, for ω = 1.8, the oscillation amplitudes of the three floors range from -2.25
28 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−2.5
−2
−1.5
−1
−0.5
0
0.5
1
1.5
2
2.5First floor − No Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.1: The displacement of the first floor with no control and ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−2.5
−2
−1.5
−1
−0.5
0
0.5
1
1.5
2
2.5Second floor − No Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.2: The displacement of the second floor with no control and ω = 1.8
3.2. NOMINAL SYSTEMS 29
0 10 20 30 40 50 60 70 80 90 100−3
−2
−1
0
1
2
3Third floor − No Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.3: The displacement of the third floor with no control and ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8First floor − No Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.4: The displacement of the first floor with no control and ω = 0.5773
30 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1Second floor − No Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.5: The displacement of the second floor with no control and ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1Third floor − No Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.6: The displacement of the third floor with no control and ω = 0.5773
3.2. NOMINAL SYSTEMS 31
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4First floor − State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.7: The displacement of the first floor via the controller (3.3) with (3.20)for ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4Second floor − State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.8: The displacement of the second floor via the controller (3.3) with (3.20)for ω = 1.8
32 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5Third floor − State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.9: The displacement of the third floor via the controller (3.3) with (3.20)for ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5First floor − State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.10: The displacement of the first floor via the controller (3.3) with (3.20)for ω = 0.5773
3.2. NOMINAL SYSTEMS 33
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
Second floor − State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.11: The displacement of the second floor via the controller (3.3) with (3.20)for ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
Third floor − State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.12: The displacement of the third floor via the controller (3.3) with (3.20)for ω = 0.5773
34 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4First floor − State Feedback Control (Nonlinear Minimization)
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.13: The displacement of the first floor via the controller (3.3) with (3.21)for ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4Second floor − State Feedback Control (Nonlinear Minimization)
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.14: The displacement of the second floor via the controller (3.3) with (3.21)for ω = 1.8
3.2. NOMINAL SYSTEMS 35
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4Third floor − State Feedback Control (Nonlinear Minimization)
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.15: The displacement of the third floor via the controller (3.3) with (3.21)for ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5First floor − State Feedback Control (Nonlinear Minimization)
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.16: The displacement of the first floor via the controller (3.3) with (3.21)for ω = 0.5773
36 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
Second floor − State Feedback Control (Nonlinear Minimization)
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.17: The displacement of the second floor via the controller (3.3) with (3.21)for ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
Third floor − State Feedback Control (Nonlinear Minimization)
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.18: The displacement of the third floor via the controller (3.3) with (3.21)for ω = 0.5773
3.2. NOMINAL SYSTEMS 37
ft to 2.25 ft, from -2.5 ft to 2.5 ft, and from -2.6 ft to 2.6 ft, respectively. The
responses for ω = 0.5773 oscillate within the range of 1.5 ft, 1.6 ft and 1.7 ft peak
to peak, respectively (see Figures 3.2 - 3.4).
However, under the controller (3.3), the amplitudes of displacements for both
ω = 1.8 and ω = 0.5773 can be effectively controlled. In fact, for the Controller
Design from Proposition 3, the controller gains K1 and K2 in (3.3) are given by
Ks1 :
K1 = [−0.3168 −0.0612 4.8474 0.3767 −0.1314 −0.3290]× 104
K2 = [0.0359 0.0061 −1.3969 0.0348 0.1165 0.0597]× 104
(3.20)
The responses of the offshore structure under controller Ks1 are also given in Figures
3.7 - 3.12. Compared with Figures 3.1 - 3.6, clearly, the displacements of the three
floors are greatly decreased. When the system reaches a steady state, the amplitudes
are just 14%, 16% and 16% of those with no control for ω = 1.8, respectively. For
ω = 0.5773, the amplitudes of displacement are also decreased, which can be seen
from Figures 3.10 - 3.12. (See Appendix C.1 for the simulation diagram)
Next, we will show the effectiveness of the Controller Design from Proposition
4. The obtained controller gains by this scheme are
Ks2 :
K1 = [−0.9575 −0.6353 0.1208 1.7778 1.1862 −1.4151]× 103
K2 = [0.0012 −0.0022 0.0348 0.0008 0.0003 0.0021]
(3.21)
Obviously, compared with the Controller Design from Proposition 3, the gains of
Ks2 are much smaller than those of Ks1, which means that the Controller Design
from Proposition 4 requires less energy than the Controller Design from Proposition
3 to control the offshore structure. On the other hand, Figures 3.13 - 3.18 plot the
response curves of offshore structure by Ks2, from which it is not difficult to see
that these amplitudes are little smaller than those in Figures 3.7 - 3.12, respectively.
From the view of both energy consumption and the decrease of internal oscillation,
the Controller Design from Proposition 4 is more effective than the Controller De-
sign from Proposition 3. (The simulation diagram of the Controller Design from
38 CHAPTER 3. STATE FEEDBACK CONTROL
Proposition 4 is similar to that of the Controller Design from Proposition 3. It’s
omitted.)
3.2.4 An Improved Delay-Dependent Stabilization Crite-rion
In the previous section, Proposition 1 provides a delay-dependent stability condition
for system (3.4) by using a widely employed Lyapunov-Krasovskii functional (LKF)
(3.7). As pointed out in [46], it seems impossible to reduce the conservatism of the
obtained delay-dependent stability conditions by using this LKF. In this section,
we introduce a new LKF, by which a less conservative delay-dependent stability
criterion is obtained. Interestingly, based on this new stability condition, the result-
ing controller can control the offshore structure more effectively than the Controller
Design from Proposition 3.
The new LKF candidate is given as
V (xt) = xT (t)Px(t) +
∫ 0
−h2
ds
∫ t
t+s
xT (θ)Rx(θ)dθ
+
∫ t
t−h2
[x(s)
x(s− h2)
]T [Q1 Q2
QT2 Q3
] [x(s)
x(s− h2)
]ds (3.22)
where P, Q1, Q2, Q3, R ∈ Rn, P > 0, Q =
[Q1 Q2
QT2 Q3
]> 0, R > 0 to be deter-
mined.
Similar to Proposition 1, the following proposition can be easily obtained, the
proof is omitted.
Proposition 5. For some given positive scalars µ and h, system (3.4) is asymptot-
ically stable if there exist matrices P > 0, Q =
[Q1 Q2
QT2 Q3
]> 0, R > 0 and Mi, Zj
(i = 1, 2, j = 1, 2, 3) of appropriate dimensions such that
Ω11 Ω12 PBK2 PF h2(A + BK1)
T
∗ Ω22 −Q2 0 0∗ ∗ −Q3 0 h
2KT
2 BT
∗ ∗ ∗ −I h2F T
∗ ∗ ∗ ∗ −h2R−1
< 0, (3.23)
3.2. NOMINAL SYSTEMS 39
R M1 M2
∗ Z1 Z2
∗ ∗ Z3
≥ 0, (3.24)
where
Ω11 := P (A + BK1) + (A + BK1)T P + Q1 + µ2I + MT
1 + M1 +h
2Z1,
Ω12 := Q2 −MT1 + M2 +
h
2Z2,
Ω22 := Q3 −Q1 −MT2 −M2 +
h
2Z3.
In order to show the less conservatism of Proposition 5, we take a well known
example from [22].
Example 1. Consider a linear system with a time delay
x(t) =
[−2 00 −0.9
]x(t) +
[−1 0−1 −1
]x(t− h) (3.25)
References [10, 28, 62, 65, 69] calculated the maximum admissible upper bound
(MAUB) of delay h, and the obtained results were all the same, i.e. 4.4721. However,
Proposition 5 can achieve a much larger MAUB, i.e. 5.7175, which sufficiently shows
the less conservatism of the new delay-dependent stability criterion by employing
the new LKF (3.22).
Similarly, we can also design a state feedback controller based on Proposition 5.
For simplicity, we just give the same scheme as Proposition 3 and the other design
scheme like Proposition 4 is omitted.
Proposition 6. For some given positive scalars µ and h, system (3.4) with K1 =
Y1X−1 and K2 = Y2X
−1 is asymptotically stable if there exist matrices X > 0, Q =[Q1 Q2
QT2 Q3
]> 0, R > 0 and Y1, Y2, M1, M2, Z1, Z2, Z3 of appropriate dimensions such
that
Ω11 Ω12 BY2 F h2(XAT + Y T
1 BT ) µX
∗ Ω22 −Q2 0 0 0∗ ∗ −Q3 0 h
2Y T
2 BT 0∗ ∗ ∗ −I h
2F T 0
∗ ∗ ∗ ∗ −h2R 0
∗ ∗ ∗ ∗ ∗ −I
< 0, (3.26)
40 CHAPTER 3. STATE FEEDBACK CONTROL
2X − R M1 M2
∗ Z1 Z2
∗ ∗ Z3
≥ 0. (3.27)
where
Ω11 := AX + BY1 + XAT + Y T1 BT + Q1 + MT
1 + M1 +h
2Z1,
Ω12 := Q2 − MT1 + M2 +
h
2Z2,
Ω22 := Q3 − Q1 − MT2 − M2 +
h
2Z3.
Proposition 6 provides another design scheme based on a new delay-dependent
stability condition. It is interesting that, for the offshore structure, the obtained
controller (3.3) by the Controller Design from Proposition 6 is more effective than
Ks1 by the Controller Design from Proposition 3. To show this, using the same
setting as the Controller Design from Proposition 3, that is, h = 0.1 and µ = 0.1,
the controller gains are given by
Ks3 :
K1 = [−1.3364 −0.3302 3.3623 1.5286 0.7710 −1.1746]× 103
K2 = [5.1730 1.0975 −75.1301 −1.4988 2.7732 6.7567]
(3.28)
The responses of the offshore structure via controller (3.3) with Ks3 are shown in
Figures 3.19 - 3.24. (The simulation diagram of the Controller Design from Propo-
sition 6 is similar to that of the Controller Design from Proposition 3. It’s omitted.)
Comparing with Figures 3.7 - 3.12 by the Controller Design from Proposition 3, we
find that
À the oscillation amplitudes of the jacket platform for ω = 0.5773 are almost the
same as those by the Controller Design from Proposition 3;
Á the oscillation amplitudes for ω = 0.5773 are little worse than those by the
Controller Design from Proposition 3, but control results are also good; and
 the controller gain is greatly reduced;
3.2. NOMINAL SYSTEMS 41
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
First floor − Improved State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.19: The displacement of the first floor via the controller (3.3) with (3.28)for ω = 1.8
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
Second floor − Improved State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.20: The displacement of the second floor via the controller (3.3) with (3.28)for ω = 1.8
42 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
Third floor − Improved State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.21: The displacement of the third floor via the controller (3.3) with (3.28)for ω = 1.8
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
First floor − Improved State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.22: The displacement of the first floor via the controller (3.3) with (3.28)for ω = 0.5773
3.2. NOMINAL SYSTEMS 43
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
Second floor − Improved State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.23: The displacement of the second floor via the controller (3.3) with (3.28)for ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
Third floor − Improved State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.24: The displacement of the third floor via the controller (3.3) with (3.28)for ω = 0.5773
44 CHAPTER 3. STATE FEEDBACK CONTROL
which are enough to illustrate the validity and superiority of the Controller Design
from Proposition 6.
3.3 Uncertain Systems
In the previous section, we discussed the controller design for a nominal offshore
structure system and three schemes have been derived to reduce the internal oscilla-
tions. However, when the system matrices are subject to uncertainties, the obtained
results fail to this situation. As is well known, depending on the uncertainty type, we
can have different results. In this section, we will consider two types of uncertainties,
one is a norm-bounded uncertainty and the other polytopic uncertainty.
3.3.1 A Norm-Bounded Uncertainty
For the norm-bounded uncertainty, it is assumed that
[A(t) B(t)] = [A B] + LG(t)[Ea Eb] (3.29)
where A,B,L, Ea and Eb are constant matrices with appropriate dimensions, and
G(t) is a time-varying real matrix with Lebesgue-measurable elements satisfying
GT (t)G(t) ≤ I, ∀t. (3.30)
The closed-loop system of (3.1) with (3.29) connecting (3.3) is then given by
x(t) = [(A + BK1) + LG(t)(Ea + EbK1)]x(t)
+ [BK2 + LG(t)EbK2]x(t− h) + Fg(x, t),
x(t) = φ(t), t ∈ [−h, 0],
(3.31)
Based on Lemma 2 (see Appendix A), we have
Proposition 7. For some given positive scalars µ and h, system (3.31) with (3.30)
is robustly stable if there exist matrices X > 0, Q > 0, R > 0, and Y1, Y2, M1, M2,
3.3. UNCERTAIN SYSTEMS 45
Z1, Z2, Z3 of appropriate dimension such that (3.14) and
Ξ11 Ξ12 F hXAT + hY T1 BT µX λL (EbY1 + EaX)T
∗ Ξ22 0 hY T2 BT 0 0 Y T
2 ETb
∗ ∗ −I hF T 0 0 0∗ ∗ ∗ −hR 0 λhL 0∗ ∗ ∗ ∗ −I 0 0∗ ∗ ∗ ∗ ∗ −λI 0∗ ∗ ∗ ∗ ∗ ∗ −λI
< 0, (3.32)
where Ξ11, Ξ12 and Ξ22 are defined in (3.11). Moreover, the controller gains are
given by K1 = Y1X−1 and K2 = Y2X
−1.
Proof. According to Proposition 3, if (3.11) and (3.14) holds, system (3.4) is asymp-
totically stable. Replacing A and B in (3.11) with A + LG(t)Ea and B + LG(t)Eb,
respectively, we find that (3.11) for system (3.31) with (3.30) is equivalent to the
following
Ξ + H1G(t)N1 + HT1 GT (t)NT
1 < 0 (3.33)
where
H1 = [LT 0 0 hLT 0]T
N1 = [EbY1 + EaX EbY2 0 0 0]
and Ξ is defined in (3.11). By Lemma 2 (see Appendix A), (3.33) holds for any G(t)
satisfying (3.30) if and only if there exists a positive number λ > 0 such that
Ξ + λH1HT1 + λ−1NT
1 N1 < 0. (3.34)
Applying the Schur complement to (3.34) yields (3.32). This completes the proof.
Similarly, the following result is straightforward from Proposition 6.
Proposition 8. For some given positive scalars µ and h, system (3.31) with (3.30)
is robustly stable if there exist matrices X > 0, Q1 > 0, Q2 > 0, Q3 > 0, R1 > 0,
46 CHAPTER 3. STATE FEEDBACK CONTROL
R2 > 0 and Y1, Y2, M1, M2, Z1, Z2, Z3 of appropriate dimensions such that (3.27)
and
Ω11 Ω12 BY2 F ρ15 µX λL ρ18
∗ Ω22 −Q2 0 0 0 0 0∗ ∗ −Q3 0 h
2Y T
2 BT 0 0 Y T2 ET
b
∗ ∗ ∗ −I h2F T 0 0 0
∗ ∗ ∗ ∗ −h2R 0 h
2λL 0
∗ ∗ ∗ ∗ ∗ −I 0 0∗ ∗ ∗ ∗ ∗ ∗ −λI 0∗ ∗ ∗ ∗ ∗ ∗ ∗ −λI
< 0, (3.35)
where
ρ15 =h
2(XAT + Y T
1 BT ),
ρ18 = (EbY1 + EaX)T .
and Ω11, Ω12, Ω22 are defined in (3.26). Moreover, the controller gains are given by
K1 = Y1X−1 and K2 = Y2X
−1.
3.3.2 Simulation Results
0 10 20 30 40 50 60 70 80 90 100−2.5
−2
−1.5
−1
−0.5
0
0.5
1
1.5
2
2.5First floor − Uncontrolled System with Uncertainty
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.25: The displacement of the first floor when no control is used to the systemwith norm-bounded uncertainties and ω = 1.8
3.3. UNCERTAIN SYSTEMS 47
0 10 20 30 40 50 60 70 80 90 100−3
−2
−1
0
1
2
3Second floor − Uncontrolled System with Uncertainty
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.26: The displacement of the second floor when no control is used to thesystem with norm-bounded uncertainties and ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−3
−2
−1
0
1
2
3Third floor − Uncontrolled System with Uncertainty
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.27: The displacement of the third floor when no control is used to thesystem with norm-bounded uncertainties and ω = 1.8
48 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8First floor − Uncontrolled System with Uncertainty
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.28: The displacement of the first floor when no control is used to the systemwith norm-bounded uncertainties and ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1Second floor − Uncontrolled System with Uncertainty
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.29: The displacement of the second floor when no control is used to thesystem with norm-bounded uncertainties and ω = 0.5773
3.3. UNCERTAIN SYSTEMS 49
0 10 20 30 40 50 60 70 80 90 100−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1Third floor − Uncontrolled System with Uncertainty
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.30: The displacement of the third floor when no control is used to thesystem with norm-bounded uncertainties and ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4First floor − Robust State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.31: The displacement of the first floor via the controller (3.3) with (3.36)for ω = 1.8
50 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5Second floor − Robust State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.32: The displacement of the second floor via the controller (3.3) with (3.36)for ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5Third floor − Robust State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.33: The displacement of the third floor via the controller (3.3) with (3.36)for ω = 1.8
3.3. UNCERTAIN SYSTEMS 51
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8First floor − Robust State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.34: The displacement of the first floor via the controller (3.3) with (3.36)for ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8Second floor − Robust State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.35: The displacement of the second floor via the controller (3.3) with (3.36)for ω = 0.5773
52 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8Third floor − Robust State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.36: The displacement of the third floor via the controller (3.3) with (3.36)for ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4First floor − Improved Robust State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.37: The displacement of the first floor via the controller (3.3) with (3.37)for ω = 1.8
3.3. UNCERTAIN SYSTEMS 53
0 10 20 30 40 50 60 70 80 90 100−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5Second floor − Improved Robust State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.38: The displacement of the second floor via the controller (3.3) with (3.37)for ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5Third floor − Improved Robust State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.39: The displacement of the third floor via the controller (3.3) with (3.37)for ω = 1.8
54 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8First floor − Improved Robust State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.40: The displacement of the first floor via the controller (3.3) with (3.37)for ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8Second floor − Improved Robust State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.41: The displacement of the second floor via the controller (3.3) with (3.37)for ω = 0.5773
3.3. UNCERTAIN SYSTEMS 55
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8Third floor − Improved Robust State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.42: The displacement of the third floor via the controller (3.3) with (3.37)for ω = 0.5773
Now, for the offshore structure system, suppose the system matrices are subject
to norm-bounded uncertainties. Then we will show, via numerical simulation, the
effectiveness of the proposed methods - Propositions 7 and 8. The system matrices
are assumed to be of the form (3.29) with
L = [0 0 0.001 0 0 0.01]T ,
Ea = [1 0 0 0 0 0],
Eb = 0.
The wave frequency ω is also separated to two cases: 1.8 rps and 0.5773 rps, and
h = 0.1, µ = 0.1.
Figures 3.25 - 3.30 plot the states of the uncertain system without control. It’s
easy to see that the oscillation amplitudes are different from those of the system
with no uncertainty, see Figures 3.1 - 3.6, especially for the responses for ω = 1.8
(see Figures 3.25 - 3.27). That is, the uncertainty brings much influence to internal
oscillations of the offshore structures, so it is of much significance in practice to take
the uncertainty into account.
56 CHAPTER 3. STATE FEEDBACK CONTROL
By the Controller Design from Proposition 7 deriving from Propositions 7, the
controller gains can be obtained as
Ks4 :
K1 = [−0.2555 −0.1584 1.6036 0.1897 0.1050 −0.1216]× 104
K2 = [0.3045 0.0542 −1.4487 0.0115 0.0394 0.1280]× 103
(3.36)
Figures 3.31 - 3.36 show all the states of the uncertain system by the Controller
Design from Proposition 7. (The simulation diagram is shown in Appendix C.2.)
Comparing with Figures 3.25 - 3.30, we find that the responses of the platform are
controlled in reasonable scopes instead of devastatingly drastic oscillations. In fact,
it is clear from Figures 3.31 - 3.33 that the amplitudes of the platform for ω = 1.8
are about only 14% of those shown in Figures 3.25 - 3.27. For ω = 0.5773, from
Figures 3.34 - 3.36, the internal oscillations induced by wave and the uncertainty are
also reduced, the amplitudes of the displacement of three floors are decreased from
1.5 ft (in Figure 3.28) to 1.2 ft (in Figure 3.34), from 1.6 ft (in Figure 3.29) to 1.3
ft (in Figure 3.35) and from 1.8 ft (in Figure 3.30) to 1.4 ft (in Figure 3.36), peak
to peak, respectively. These figures attest the effectiveness of the design control
technique in reducing the uncontrolled responses.
On the other hand, the Controller Design from Proposition 8 deriving from
Propositions 8 yields the controller gains as
Ks5 :
K1 = [−1.6666 −0.4307 5.8123 2.0646 0.6845 −1.4603]× 103
K2 = [89.7419 5.3661 −703.9537 −11.9643 40.3774 83.3679]
(3.37)
The state responses of the system under study via a memory controller by the
Controller Design from Proposition 8 are shown in Figures 3.37 -3.40, from which, we
can see that the internal oscillations are almost the same as those by the Controller
Design from Proposition 7. However, it is clear that the gain Ks5 is much smaller
than Ks4, which means the Controller Design from Proposition 8 is more efficient
and economical than the Controller Design from Proposition 7 in the sense of energy
3.3. UNCERTAIN SYSTEMS 57
consumption. (The simulation diagram of the Controller Design from Proposition
8 is similar to that of the Controller Design from Proposition 7. It’s omitted.)
3.3.3 A Polytopic Uncertainty
For the polytopic uncertainty, (A(t), B(t)) in (3.1) is assumed to be bounded by a
given convex bounded polyhedral domain Σ, i.e.
(A(t), B(t)) ∈ Σ :=
q∑
j=1
ξj(Aj, Bj),
q∑j=1
ξj = 1, ξj ≥ 0
(3.38)
The closed-loop system of (3.1) with (3.38) under controller (3.3) becomes
x(t) = [A(t) + B(t)K1]x(t) + B(t)K2x(t− h) + Fg(x, t),
x(t) = φ(t), t ∈ [−h, 0],(3.39)
Following the ideas in [50, 51], we choose a parameter-dependent Lyapunov-
Krasovskii functional candidate as
V (xt) =
q∑j=1
[xT (t)ξjPjx(t) +
∫ t
t−h
xT (s)ξjQjx(s)ds
+
∫ 0
−h
dθ
∫ t
t+θ
xT (s)ξjRjx(s)ds]
4=
q∑j=1
Vj(xt). (3.40)
where xt = x(t + α), α ∈ [−h, 0] and Pj > 0, Qj > 0, Rj > 0 are to be determined.
Then we have
Proposition 9. The system (3.1) with (3.38) is robustly stable if there exist matrices
Pj > 0, Qj > 0, Rj > 0, M1j,M2j, Z1j, Z2j, Z3j and N of appropriate dimensions,
such that
Υj :=
Υ11j Υ12j Υ13j NT F∗ Υ22j NT BjK2 NT F∗ ∗ Υ33j 0∗ ∗ ∗ −I
< 0 (3.41)
and
Rj M1j M2j
∗ Z1j Z2j
∗ ∗ Z3j
≥ 0, (3.42)
58 CHAPTER 3. STATE FEEDBACK CONTROL
hold for j = 1, 2, · · · , q, where
Υ11j = Qj + µ2I + MT1j + M1j + hZ1j + NT (Aj+BjK1) + (Aj+BjK1)
T N,
Υ12j = Pj −NT + (Aj + BjK1)T N,
Υ22j = hRj −NT −N,
Υ13j =−MT1j + M2j + hZ2j + NT BjK2,
Υ33j =−MT2j −M2j + hZ3j.
Proof. Taking the derivative of V (xt) with respect to time t along the trajectory of
system (3.39) yields
V (xt) =
q∑j=1
Vj(xt) (3.43)
where
Vj(xt) = 2xT (t)ξjPjx(t) + xT (t)ξjQjx(t)− xT (t− h)ξjQjx(t− h)
+ hxT (t)ξjRjx(t)−∫ t
t−h
xT (s)ξjRjx(s)ds (3.44)
By Lemma 3 (see Appendix A), for any matrices M1j,M2j and Z1j, Z2j, Z3j, if (3.42)
is true, then so is the following
−∫ t
t−h
xT (s)ξjRjx(s)ds
≤ ξj
[x(t)
x(t− h)
]T [MT
1j + M1j −MT1j + M2j
∗ −MT2j −M2j
] [x(t)
x(t− h)
]
+ hξj
[x(t)
x(t− h)
]T [Z1j Z2j
∗ Z3j
] [x(t)
x(t− h)
], (3.45)
Similar to [28], for any N with appropriate dimensions, it is clear from (3.39) that
the following is true
0 = 2xT (t)NT [−x(t) + (A(t) + B(t)K1)x(t) + B(t)K2x(t− h) + Fg(x, t)]
Noting that (3.38), we have
0 = 2
q∑j=1
ξjxT (t)NT [−x(t) + (Aj + BjK1)x(t) + BjK2x(t− h) + Fg(x, t)] (3.46)
3.3. UNCERTAIN SYSTEMS 59
In addition, from (3.2)
0 ≤ µ2xT (t)x(t)− gT (x, t)g(x, t) (3.47)
Substituting (3.44), (3.45), (3.46), (3.47) into (3.43) gives
V (xt) ≤q∑
j=1
ξjψT (t)Υjψ(t)
where Υj is defined in (3.41) and
ψT (t) = [ xT (t) xT (t) xT (t− h) gT (x, t) ]
Thus, if (3.41) and (3.42) hold for j = 1, 2, · · · , q, then there exists a δ > 0 such
that V (xt) ≤ −δ ‖x(t)‖2 < 0 for x(t) 6= 0, which guarantees that system (3.39) is
robustly stable for polytopic uncertainty (3.38). This completes the proof.
The following proposition provides a controller design method for system (3.39)
with polytopic uncertainty (3.38).
Proposition 10. The system (3.39) with (3.38) is robustly stable if there exist
matrices Pj > 0, Qj > 0, Rj > 0, Mij, Zkj, N and Yi of appropriate dimensions,
where i = 1, 2, j = 1, 2, ...q, k = 1, 2, 3, such that the following LMIs hold for
j = 1, 2, ...q :
Υ(j) :=
Υ11 Υ12 Υ13 F µNT
∗ Υ22 BjY2 F 0
∗ ∗ Υ33 0 0∗ ∗ ∗ −I 0∗ ∗ ∗ ∗ −I
< 0, (3.48)
Rj M1j M2j
∗ Z1j Z2j
∗ ∗ Z3j
≥ 0, (3.49)
where
Υ11 = Qj + MT1j + M1j + hZ1j
60 CHAPTER 3. STATE FEEDBACK CONTROL
+AjN + BjY1 + (AjN + BjY1)T ,
Υ12 = Pj − N + (AjN + BjY1)T ,
Υ22 = hRj − N − NT ,
Υ13 = −MT1j + M2j + hZ2j + BjY2,
Υ33 = −MT2j − M2j + hZ3j.
Moreover, the controller gains are given by K1 = Y1X−1 and K2 = Y2X
−1.
Proof. From Proposition 9, if (3.41) and (3.42) hold, system (3.39) with (3.38) is
robustly stable. In addition, from (3.41), noting that N is invertible. Pre- and
post-multiplying both sides of matrix inequality (3.41) by
diag(
NT)−1
,(NT
)−1,
(NT
)−1, I
,
and its transpose, respectively, yields
Υ(j) =
Υ11 Υ12 Υ13 F∗ Υ22 BjK2N
−1 F∗ ∗ Υ33 0∗ ∗ ∗ −I
< 0 (3.50)
where
Υ11 =(NT
)−1QjN
−11 + µ2
(NT
)−1N−1 +
(NT
)−1MT
1jN−1
+(NT
)−1M1jN
−1 + h(NT
)−1Z1jN
−1
+ (Aj + BjK1)N−1 +
(NT
)−1(Aj + BjK1)
T ,
Υ12 =(NT
)−1PjN
−1 −N−1 +(NT
)−1(Aj + BjK1)
T ,
Υ22 = h(NT
)−1RjN
−1 −N−1 − (NT
)−1,
Υ13 = − (NT
)−1MT
1jN−1 +
(NT
)−1M2jN
−1
+ h(NT
)−1Z2jN
−1 + BjK2N−1,
Υ33 = − (NT
)−1MT
2jN−1 − (
NT)−1
M2jN−1 + h
(NT
)−1Z3jN
−1.
Meanwhile, pre- and post-multiplying both sides of matrix inequality (3.42) by
diag(
NT)−1
,(NT
)−1,
(NT
)−1
,
3.3. UNCERTAIN SYSTEMS 61
and its transpose, respectively, yields
(NT
)−1RjN
−1(NT
)−1M1jN
−1(NT
)−1M2jN
−1
∗ (NT
)−1Z1jN
−1(NT
)−1Z2jN
−1
∗ ∗ (NT
)−1Z3jN
−1
≥ 0.
Let
(NT
)−1QjN
−1 = Qj,(NT
)−1PjN
−1 = Pj,(NT
)−1RjN
−1 = Rj,(NT
1
)−1Z1jN
−1 = Z1j,(NT
)−1Z2jN
−1 = Z2j,(NT
)−1Z3jN
−1 = Z3j,(NT
)−1M1jN
−1 = M1j,(NT
)−1M2jN
−1 = M2j,N−1 = N , K1N
−1 = Y1,K2N
−1 = Y2.
Applying the Schur complement to (3.50), we obtain (3.49) and (3.48). This com-
pletes the proof.
3.3.4 Simulation Results
0 10 20 30 40 50 60 70 80 90 100−4
−3
−2
−1
0
1
2
3
4First floor − Uncontrolled System with Polytopic Type Uncertainty
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.43: The displacement of the first floor when no control is used to the systemwith polytopic uncertainties and ω = 1.8
Let’s consider the closed-loop system (3.39) with
A(t) = A + ∆A, B(t) = B, (3.51)
62 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−4
−3
−2
−1
0
1
2
3
4Second floor − Uncontrolled System with Polytopic Type Uncertainty
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.44: The displacement of the second floor when no control is used to thesystem with polytopic uncertainties and ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−4
−3
−2
−1
0
1
2
3
4Third floor − Uncontrolled System with Polytopic Type Uncertainty
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.45: The displacement of the third floor when no control is used to thesystem with polytopic uncertainties and ω = 1.8
3.3. UNCERTAIN SYSTEMS 63
0 10 20 30 40 50 60 70 80 90 100−1.5
−1
−0.5
0
0.5
1
1.5First floor − Uncontrolled System with Polytopic Type Uncertainty
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.46: The displacement of the first floor when no control is used to the systemwith polytopic uncertainties and ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−1.5
−1
−0.5
0
0.5
1
1.5Second floor − Uncontrolled System with Polytopic Type Uncertainty
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.47: The displacement of the second floor when no control is used to thesystem with polytopic uncertainties and ω = 0.5773
64 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−1.5
−1
−0.5
0
0.5
1
1.5Third floor − Uncontrolled System with Polytopic Type Uncertainty
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.48: The displacement of the third floor when no control is used to thesystem with polytopic uncertainties and ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4First floor − Robust State Feedback Control (Polytopic Type Uncertainties)
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.49: The displacement of the first floor via the controller (3.3) with (3.52)for ω = 1.8 (polytopic uncertainties)
3.3. UNCERTAIN SYSTEMS 65
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4Second floor − Robust State Feedback Control (Polytopic Type Uncertainties)
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.50: The displacement of the second floor via the controller (3.3) with (3.52)for ω = 1.8 (polytopic uncertainties)
0 10 20 30 40 50 60 70 80 90 100−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5Third floor − Robust State Feedback Control (Polytopic Type Uncertainties)
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.51: The displacement of the third floor via the controller (3.3) with (3.52)for ω = 1.8 (polytopic uncertainties)
66 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8Third floor − Robust State Feedback Control (Polytopic Type Uncertainties)
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.52: The displacement of the first floor via the controller (3.3) with (3.52)for ω = 0.5773 (polytopic uncertainties)
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8Second floor − Robust State Feedback Control (Polytopic Type Uncertainties)
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.53: The displacement of the second floor via the controller (3.3) with (3.52)for ω = 0.5773 (polytopic uncertainties)
3.3. UNCERTAIN SYSTEMS 67
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8Third floor − Robust State Feedback Control (Polytopic Type Uncertainties)
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.54: The displacement of the third floor via the controller (3.3) with (3.52)for ω = 0.5773 (polytopic uncertainties)
where A,B defined in (2.6) and
∆A =
0 0 0 0 0 00 0 0 0 0 0δ1 0 0 0 0 00 0 0 0 0 00 δ2 0 0 0 00 0 0 0 0 0
with
δ1 ∈[ −0.09, 0.09
], δ2 ∈
[ −0.15, 0.15].
Clearly, (3.39) with (3.51) is a system with polytopic uncertainty, with four
vertices as (Aj, B) (j = 1, 2, 3, 4), where
A1 =
0 1 0 0 0 0−3.3235 −0.0212 0.0184 0.0030 −5.3449 −0.8819−0.09 0 0 1 0 00.0184 0.0030 −118.1385 −0.1117 5.3468 0.8822
0 −0.15 0 0 0 1−0.0114 −0.0019 0.0114 0.0019 −3.3051 −0.5454
;
68 CHAPTER 3. STATE FEEDBACK CONTROL
A2 =
0 1 0 0 0 0−3.3235 −0.0212 0.0184 0.0030 −5.3449 −0.8819−0.09 0 0 1 0 00.0184 0.0030 −118.1385 −0.1117 5.3468 0.8822
0 0.15 0 0 0 1−0.0114 −0.0019 0.0114 0.0019 −3.3051 −0.5454
;
A3 =
0 1 0 0 0 0−3.3235 −0.0212 0.0184 0.0030 −5.3449 −0.8819
0.09 0 0 1 0 00.0184 0.0030 −118.1385 −0.1117 5.3468 0.8822
0 −0.15 0 0 0 1−0.0114 −0.0019 0.0114 0.0019 −3.3051 −0.5454
;
A4 =
0 1 0 0 0 0−3.3235 −0.0212 0.0184 0.0030 −5.3449 −0.8819
0.09 0 0 1 0 00.0184 0.0030 −118.1385 −0.1117 5.3468 0.8822
0 0.15 0 0 0 1−0.0114 −0.0019 0.0114 0.0019 −3.3051 −0.5454
;
As previously presented, two cases of the wave frequency are taken into consid-
eration as 1.8 rps and 0.5773 rps, and the value of µ is 0.1. In the following, we
compare the simulation result when u(t) = 0 with that via memory controller (3.3)
with h = 0.1.
Figures 3.43 - 3.48 show the states of the system with polytopic type uncertainty
when no control is added. For ω = 1.8, the displacements of the three floors are
between −4 ft and 4 ft (see Figures 3.43 - 3.45). Figures 3.46 - 3.48 show that the
amplitudes of the responses are about -2.6 ft peak to peak. Recalling Figures 3.1 -
3.6, we can see how the uncertainty badly degrades the performance of the offshore
structure system. However, the memory state feedback controller (3.3) based on
Proposition 10 can be used to effectively control this uncertain system. To see this,
the controller gains are given as
Ks6 :
K1 = [−0.2266 −0.0659 1.1847 0.2477 0.0463 −0.1982]× 104
K2 = [−18.5225 −12.7426 169.8950 16.7815 −6.8169 −19.5826]
(3.52)
3.4. STATE FEEDBACK H∞ CONTROL 69
The state responses of the controlled system are demonstrated in Figures 3.49 -
3.54. They are the displacements of the three floors for ω = 1.8 and ω = 0.5773 in
sequence. (The simulation diagram of the Controller Design from Proposition 10 is
similar to that of the Controller Design from Proposition 8 except the definitions
of fcn. See Appendix C.3 for their definitions.) From Figures 3.49 - 3.51, we find
that responses of these floors, for ω = 1.8, oscillate between -0.3 ft and 0.35 ft,
between -3.2 ft and 3.9 ft, and between -0.35 ft and 0.41 ft, respectively, which are
merely about 10% of the corresponding responses of the uncontrolled system (see
Figures 3.43 - 3.45). The amplitudes of the displacement for ω = 0.5773 shown
in Figures 3.52 - 3.54 are reduced to 1.2 ft, 1.3 ft and 1.4 ft peak to peak, which
obtain 40% decrease compared with Figures 3.46 - 3.48. Under the actions of the
designed controller, the maximum amplitudes of the three floors are reduce into an
acceptable scope successfully.
3.4 State Feedback H∞ Control
In this section, we focus on an H∞ control problem for the offshore steel jacket
platform. The system can be described as follows:
x(t) = A(t)x(t) + B(t)u(t) + Fg(x, t) + Bwω(x, t),
z(t) = Czx(t),
x(t) = φ(t), t ≤ 0
(3.53)
where x(t) ∈ Rn is the state vector, u(t) ∈ Rm is the control, z(t) ∈ Rr is the con-
trolled output, w(t) ∈ Rl is the external disturbance, which belongs to L2[0, +∞),
g(x, t) ∈ Rp is the nonlinear self-excited hydrodynamic force vector, which is uni-
formly bounded and satisfies the following cone-bounding constraint:
‖g(x, t)‖ ≤ µ ‖x(t)‖ , (3.54)
where µ is a positive scalar; φ(t) denotes an initial condition; F, Bw, and Cz are
constant real matrices with appropriate dimensions, A(t) and B(t) are system ma-
70 CHAPTER 3. STATE FEEDBACK CONTROL
trices, which may be time-varying, but are known to belong to a certain compact
set Ω, that is,
(A(t), B(t)) ∈ Rn×(n+m) ⊂ Ω
The goal of the section is to develop an H∞ controller
u(t) = K1x(t) + K2x(t− h), (3.55)
where K1 and K2 are constant matrices, which are to be determined, delay h is
assumed to be constant satisfying h > 0, such that
(i) the following closed-loop system
x(t) = (A(t) + B(t)K1)x(t) + B(t)K2x(t− h)
+ Fg(x, t) + Bwω(t),
z(t) = Czx(t),
x(t) = φ(t), t ≤ 0
(3.56)
with ω(t) = 0 is asymptotically stable; and
(ii) under the condition φ(t) = 0, the H∞ performance
‖z(t)‖2 ≤ γ ‖ω(t)‖2
of the closed-loop system (3.56) is guaranteed for all nonzero ω(t) ∈ L2[0, +∞) and
a prescribed γ > 0.
In what follows, first, we will consider an H∞ controller design problem for the
nominal case of system (3.56). Then, the robust H∞ state feedback control issue of
system (3.56) with uncertainty will be investigated.
3.4.1 An H∞ Control for Nominal Systems
In this subsection, we aim at designing an H∞ controller for the nominal system. In
this case, let A(t) ≡ A and B(t) ≡ B for all t ≥ 0. Then, the resulting closed-loop
system can be described as
3.4. STATE FEEDBACK H∞ CONTROL 71
x(t) = (A + BK1)x(t) + BK2x(t− h)
+ Fg(x, t) + Bwω(t),
z(t) = Czx(t),
x(t) = φ(t), t ≤ 0
(3.57)
Choose a Lyapunov-Krasovskii functional candidate as
V (xt) = xT (t)Px(t) +
∫ t
t−h
xT (s)Qx(s)ds +
∫ 0
−h
dθ
∫ t
t+θ
xT (s)Rx(s)ds,
where P, Q, R ∈ Rn×n, P > 0, Q > 0 and R > 0.
Proposition 11. For some given positive scalars µ, γ and h, system (3.57) is
asymptotically stable and satisfies ‖z(t)‖2 ≤ γ ‖ω(t)‖2 for all nonzero ω(t) ∈ L2[0, +∞),
if there exist matrices P > 0, Q > 0, R > 0 and M1, M2, Z1, Z2, Z3 of appropriate
dimensions such that
z :=
z11 z12 PF PBw h(A + BK1)T
∗ z22 0 0 hKT2 BT
∗ ∗ −I 0 hF T
∗ ∗ ∗ −γ2I hBTw
∗ ∗ ∗ ∗ −hR−1
< 0, (3.58)
R M1 M2
∗ Z1 Z2
∗ ∗ Z3
≥ 0, (3.59)
where
z11 = P (A + BK1) + (A + BK1)T P + Q
+µ2I + MT1 + M1 + hZ1 + CT
z Cz,
z12 = PBK2 −MT1 + M2 + hZ2,
z22 = −(1− d)Q−MT2 −M2 + hZ3.
Proof. Firstly, we consider the asymptotic stability of system (3.57) with ω(t) = 0.
When ω(t) = 0, system (3.57) is degraded to system (3.4). According to Proposition
1, if the linear matrix inequalities (3.5) and (3.6) hold, system (3.4) is asymptotically
stable. Therefore, if (3.5) and (3.6) hold, system (3.57) with ω(t) = 0 is asymptot-
ically stable. Clearly, (3.6) is the same as (3.59), and (3.58) implies (3.5) by the
72 CHAPTER 3. STATE FEEDBACK CONTROL
Schur complement. Then, if (3.58) and (3.59) hold, system (3.57) with ω(t) = 0
is asymptotically stable. So, in the following, we only need to prove the H∞ per-
formance ‖z(t)‖2 ≤ γ ‖ω(t)‖2 is guaranteed for all nonzero ω(t) ∈ L2[0, +∞) under
zero initial condition. For this, taking the derivative of V (xt) with respect to time
t along the trajectory of system (3.57) yields
V (xt) =2xT (t)Px(t) + xT (t)Qx(t)− xT (t− h)Qx(t− h)
+ hxT (t)Rx(t)−∫ t
t−h
xT (s)Rx(s)ds (3.60)
Introducing a new vector
ψT (t) =[
xT (t) xT (t− h) gT (x, t) ωT (t)].
by using Lemma 3 (see Appendix A) and noting that (3.54), after simple manipu-
lation, we have
V (xt) + zT (t)z(t)− γ2ωT (t)ω(t) ≤ ψT (t)[Θ + hΓT RΓ
]ψ(t) (3.61)
where
Θ :=
z11 z12 PF PBw
∗ z22 0 0∗ ∗ −I 0∗ ∗ ∗ −γ2I
,
Γ :=[
A + BK1 BK2 F Bw
]
with z11, z12, z22 being defined in (3.58).
Obviously, if (3.58) is feasible, then applying the Schur complement gives Θ +
hΓT RΓ < 0, which leads to
V (xt) + zT (t)z(t)− γ2ωT (t)ω(t) ≤ 0. (3.62)
Integrating both sides of (3.62) from 0 to ∞ yields
∫ ∞
0
[zT (t)z(t)− γ2ωT (t)ω(t)
]dt ≤ V (xt)|t=0 − V (xt)|t=∞.
3.4. STATE FEEDBACK H∞ CONTROL 73
Under zero initial condition φ(t) = 0, one has V (xt)|t=0 = 0, thus
∫ ∞
0
[zT (t)z(t)− γ2ωT (t)ω(t)
]dt ≤ 0,
which means ‖z(t)‖2 ≤ γ ‖ω(t)‖2 . This completes the proof.
Proposition 11 provides a bounded real lemma for system (3.57), which can
ensure that the system is not only asymptotical stability but also of a prescribed
disturbance attenuation level γ. In order to solve out the controller gains K1 and
K2, similar to proof procedure of Propositions 2 and 3, we have
Proposition 12. For some given positive scalars µ, γ and h, system (3.57) is
asymptotically stable and satisfies ‖z(t)‖2 ≤ γ ‖ω(t)‖2 for all nonzero ω(t) ∈ L2[0,∞),
if there exist matrices X > 0, Q > 0, R > 0 and Y, M1, M2, Z1, Z2, Z3 of appro-
priate dimensions such that LMIs
z11 z12 F Bw h(XAT + Y T1 BT ) µX XCT
z
∗ z22 0 0 hY T2 BT 0 0
∗ ∗ −I 0 hF T 0 0∗ ∗ ∗ −γ2I hBT
w 0 0∗ ∗ ∗ ∗ −hR 0 0∗ ∗ ∗ ∗ ∗ −I 0∗ ∗ ∗ ∗ ∗ ∗ −I
< 0 (3.63)
2X − R M1 M2
∗ Z1 Z2
∗ ∗ Z3
≥ 0. (3.64)
where
z11 = XAT + Y T1 BT + AX + BY1 + Q + MT
1 + M1 + hZ1,
z12 = BY2 − MT1 + M2 + hZ2,
z22 = −Q− MT2 − M2 + hZ3.
Moreover, the controller gains are given by K1 = Y1X−1 and K2 = Y2X
−1.
74 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−2.5
−2
−1.5
−1
−0.5
0
0.5
1
1.5
2
2.5First floor − Uncontrolled System with Disturbance
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.55: The displacement of the first floor when no control is used to the systemunder disturbance and ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−3
−2
−1
0
1
2
3Second floor − Uncontrolled System with Disturbance
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.56: The displacement of the second floor when no control is used to thesystem under disturbance and ω = 1.8
3.4. STATE FEEDBACK H∞ CONTROL 75
0 10 20 30 40 50 60 70 80 90 100−3
−2
−1
0
1
2
3Third floor − Uncontrolled System with Disturbance
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.57: The displacement of the third floor when no control is used to thesystem under disturbance and ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1First floor − Uncontrolled System with Disturbance
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.58: The displacement of the first floor when no control is used to the systemunder disturbance and ω = 0.5773
76 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1Second floor − Uncontrolled System with Disturbance
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.59: The displacement of the second floor when no control is used to thesystem under disturbance and ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1Third floor − Uncontrolled System with Disturbance
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.60: The displacement of the third floor when no control is used to thesystem under disturbance and ω = 0.5773
3.4. STATE FEEDBACK H∞ CONTROL 77
0 10 20 30 40 50 60 70 80 90 100−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
First floor − Hinf
State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.61: The displacement of the first floor via the controller (3.55) with (3.65)for ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.25
−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
Second floor − Hinf
State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.62: The displacement of the second floor via the controller (3.55) with(3.65) for ω = 1.8
78 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−0.25
−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
Third floor − Hinf
State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.63: The displacement of the third floor via the controller (3.55) with (3.65)for ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
First floor − Hinf
State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.64: The displacement of the first floor via the controller (3.55) with (3.65)for ω = 0.5773
3.4. STATE FEEDBACK H∞ CONTROL 79
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
Second floor − Hinf
State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.65: The displacement of the second floor via the controller (3.55) with(3.65) for ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
Third floor − Hinf
State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.66: The displacement of the third floor via the controller (3.55) with (3.65)for ω = 0.5773
80 CHAPTER 3. STATE FEEDBACK CONTROL
3.4.2 Simulation Results
Now, we will show the validity of the proposed method for the offshore structure sys-
tem when there exists an external disturbance. For simulation, the wave frequency
is still set to be 1.8 or 0.5773 rps. Let µ = 0.1, γ = 3 and
Bw =[
1 0 0.1 0 0 0]T
,
Cz =[
0 0.5 0 0 0 0]T
.
When u(t) = 0, i.e. no control is added to the offshore platform, in this case,
Figures 3.55 - 3.60 depict the response of all the states of the system. The displace-
ments for ω = 1.8 are drawn in Figures 3.55 - 3.57, respectively, and in Figures 3.57
- 3.58 for ω = 0.5773. It’s easy to see that the external disturbance significantly
influences the amplitudes of the system. However, under memory controller (3.55)
with h = 0.1, the system can be effectively controlled. In fact, by Proposition 12,
the controller gains are given by
Ks7 :
K1 = [−0.3770 −0.2423 4.6490 0.4048 0.0828 −0.1934]× 104
K2 = [273.8290 −30.0347 −812.6438 21.0510 15.0692 42.9539]× 103
(3.65)
The state responses of the controlled system are shown in Figures 3.61 - 3.66.
(The simulation diagram is described in Appendix C.4.) It is clear from Figures
3.61 - 3.63 that the amplitudes of displacement for ω = 1.8 are reduced to 0.4
ft, 0.45 ft and 0.5 ft peak to peak, which are only about 10% of the uncontrolled
responses of the system. From Figures 3.64 - 3.66, we find that the responses for for
ω = 0.5773 oscillate between -0.35 ft and 0.35 ft, between -0.38 ft and 0.4 ft, and
between -0.4 ft and 0.42 ft, respectively, which are decreased about 45%. Therefore,
the simulation results have shown that the obtained controller has well reduced the
internal oscillations of the offshore structure subject to the external disturbance.
3.4. STATE FEEDBACK H∞ CONTROL 81
3.4.3 An Improved Controller Design Scheme
As shown in the previous section, by employing the new LKF (3.22), we can obtain a
less conservative delay-dependent criterion. It is interesting that, based on this new
condition, the resulting controller can act on the offshore structure more efficiently.
In what follows, we also extend this LFK (3.22) to construct an H∞ controller for
the offshore structures.
By choosing the LFK (3.22), an improved delay-dependent stability criteria can
be formulated as follows.
Proposition 13. For some given positive scalars µ, γ and h, system (3.57) is
asymptotically stable and satisfies ‖z(t)‖2 ≤ γ ‖ω(t)‖2 for all nonzero ω(t) ∈ L2[0,∞),
if there exist matrices X > 0,
[Q1 Q2
QT2 Q3
]> 0, R > 0 and Y1, Y2, Mi, Zj (i = 1, 2,
j = 1, 2, 3) of appropriate dimensions such that
ℵ11 ℵ12 BY2 F Bwh2(XAT + Y T
1 BT ) µX XCTz
∗ ℵ22 −Q2 0 0 0 0 0∗ ∗ −Q3 0 0 h
2Y T
2 BT 0 0∗ ∗ ∗ −I 0 h
2F T 0 0
∗ ∗ ∗ ∗ −γ2I h2BT
w 0 0∗ ∗ ∗ ∗ ∗ −h
2R 0 0
∗ ∗ ∗ ∗ ∗ ∗ −I 0∗ ∗ ∗ ∗ ∗ ∗ ∗ −I
< 0, (3.66)
XR−1X M1 M2
∗ Z1 Z2
∗ ∗ Z3
≥ 0, (3.67)
where
ℵ11 :=AX + BY1 + XAT + Y T1 BT + Q1 + MT
1 + M1 +h
2Z1,
ℵ12 :=Q2 − MT1 + M2 +
h
2Z2,
ℵ22 :=Q3 − Q1 − MT2 − M2 +
h
2Z3.
Proof. Since the proof is similar to that of Proposition 11, it is thus omitted.
Noting that
XR−1X ≥ 2X − R,
82 CHAPTER 3. STATE FEEDBACK CONTROL
then a sufficient condition, which is in the form of LMIs, can be obtained in the
following.
Proposition 14. For some given positive scalars µ, γ and h, system (3.57) is
asymptotically stable and satisfies ‖z(t)‖2 ≤ γ ‖ω(t)‖2 for all nonzero ω(t) ∈ L2[0,∞),
if there exist matrices X > 0,
[Q1 Q2
QT2 Q3
]> 0, R > 0 and Y1, Y2, Mi, Zj (i = 1, 2,
j = 1, 2, 3) of appropriate dimensions such that (3.66) and
2X − R M1 M2
∗ Z1 Z2
∗ ∗ Z3
≥ 0.
Moreover, the controller gains are given by K1 = Y1X−1, and K2 = Y2X
−1.
3.4.4 Simulation Results
0 10 20 30 40 50 60 70 80 90 100−0.25
−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
First floor − Improved Hinf
State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.67: The displacement of the first floor via the controller (3.55) with (3.68)for ω = 1.8
Here, we compare the obtained performances of the offshore structure by two
control schemes: the Controller Design from Proposition 12 and the Controller
Design from Proposition 14. Under the same parameter values as the Controller
Design from Proposition 12, the achieved controller gains by the Controller Design
3.4. STATE FEEDBACK H∞ CONTROL 83
0 10 20 30 40 50 60 70 80 90 100−0.25
−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
Second floor − Improved Hinf
State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.68: The displacement of the second floor via the controller (3.55) with(3.68) for ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.25
−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
Third floor − Improved Hinf
State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.69: The displacement of the third floor via the controller (3.55) with (3.68)for ω = 1.8
84 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
First floor − Improved Hinf
State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.70: The displacement of the first floor via the controller (3.55) with (3.68)for ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
Second floor − Improved Hinf
State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.71: The displacement of the second floor via the controller (3.55) with(3.68) for ω = 0.5773
3.4. STATE FEEDBACK H∞ CONTROL 85
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
Third floor − Improved Hinf
State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.72: The displacement of the third floor via the controller (3.55) with (3.68)for ω = 0.5773
from Proposition 14 are given by
Ks8 :
K1 = [−0.8389 −0.3904 5.7361 0.2219 0.0080 −0.3967]× 104
K2 = [0.6931 0.0413 −2.8987 0.0654 0.0795 0.2371]× 103
(3.68)
The simulation results of the offshore structure by the Controller Design from
Proposition 14 (Ks8) are shown in Figures 3.67 - 3.72. Comparing Figures 3.67 -
3.72 with Figures 3.61 - 3.66, we find internal oscillations of the three floors by the
Controller Design from Proposition 14 are almost the same as the Controller Design
from Proposition 12. (The simulation diagram is similar to that of the Controller
Design from Proposition 12. So it’s omitted.)
However, we calculate the achieved minimum H∞ performance values γ by both
the Controller Design from Proposition 12 and the Controller Design from Proposi-
tion 14, and the obtained values γ for various h are listed in Table 3.1.
From the table, the Controller Design from Proposition 14 can achieve much less
minimum γ than the Controller Design from Proposition 12. Therefore, the Con-
troller Design from Proposition 14 not only reduces internal oscillations effectively,
86 CHAPTER 3. STATE FEEDBACK CONTROL
Table 3.1: Comparison results about minimum of γ for various h.
h 0.1 0.2 0.3 0.4 0.5 0.6Proposition 12 0.56023 0.61417 0.66017 1.00800 - -Proposition 14 0.53918 0.56018 0.58746 0.61414 0.63837 0.66016
but also ensures much less H∞ performance value. That is, the Controller Design
from Proposition 14, indeed, is an improvement over the Controller Design from
Proposition 12.
3.4.5 An H∞ Control for Uncertain Systems
In this subsection, we consider a robust H∞ control problem. For two types of
uncertainties, namely norm-bounded uncertainty and polytopic uncertainty, robust
H∞ controllers are designed, which can not only ensure the asymptotical stability
and guarantee a prescribed H∞ level as well for the offshore structures.
3.4.6 A Norm-Bounded Uncertainty
For the norm-bounded uncertainty, it is assumed that
[A(t) B(t)] = [A B] + LG(t)[Ea Eb]
where the meanings of A,B,L, Ea, Eb and G(t) are the same as those in (3.29). The
closed-loop system is then given by
x(t) =[A+BK1+LG(t)(Ea+EbK1)]x(t)+[BK2+LG(t)EbK2]x(t− h)
+ Fg(x, t) + Bwω(t),
z(t) =Czx(t),
x(t) =φ(t), t ∈ [−h, 0]
(3.69)
Based on Lemma 2 (see Appendix A) and following the proofs of Propositions
12 and 14, the results below are not difficult to obtain. The proofs str omitted.
3.4. STATE FEEDBACK H∞ CONTROL 87
Proposition 15. For some given positive scalars µ, γ and h, system (3.69) is
asymptotically stable and satisfies ‖z(t)‖2 ≤ γ ‖ω(t)‖2 for all nonzero ω(t) ∈ L2[0,∞),
if there exist λ > 0 and matrices X > 0, Q > 0, R > 0 and Y, M1, M2, Z1, Z2, Z3
of appropriate dimensions such that
z11 z12 F Bw σ1 µX XCTz λL σ2
∗ z22 0 0 h2Y T
2 BT 0 0 0 Y T2 ET
b
∗ ∗ −I 0 h2F T 0 0 0 0
∗ ∗ ∗ −γ2I h2BT
w 0 0 0 0∗ ∗ ∗ ∗ −h
2R 0 0 h
2λL 0
∗ ∗ ∗ ∗ ∗ −I 0 0 0∗ ∗ ∗ ∗ ∗ ∗ −I 0 0∗ ∗ ∗ ∗ ∗ ∗ ∗ −λI 0∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ −λI
< 0, (3.70)
2X − R M1 M2
∗ Z1 Z2
∗ ∗ Z3
≥ 0, (3.71)
where
σ1 := h(XAT + Y T1 BT ),
σ2 := (EbY1 + EaX)T ,
and z11, z12, z22 are defined in (3.63). Furthermore, the controller gains are given
by K1 = Y1X−1 and K2 = Y2X
−1.
Proposition 16. For some given positive scalars µ, γ and h, system (3.69) is
asymptotically stable and satisfies ‖z(t)‖2 ≤ γ ‖ω(t)‖2 for all nonzero ω(t) ∈ L2[0,∞),
if there exist matrices X > 0,
[Q1 Q2
QT2 Q3
]> 0, R > 0 and Y1, Y2, Mi, Zj (i = 1, 2,
j = 1, 2, 3) of appropriate dimensions such that
ℵ11 ℵ12 BY2 F Bw σ3 µX XCTz λL σ4
∗ ℵ22 −Q2 0 0 0 0 0 0 0∗ ∗ −Q3 0 0 h
2Y T
2 BT 0 0 0 Y T2 ET
b
∗ ∗ ∗ −I 0 h2F T 0 0 0 0
∗ ∗ ∗ ∗ −γ2I h2BT
w 0 0 0 0∗ ∗ ∗ ∗ ∗ −h
2R 0 0 h
2λL 0
∗ ∗ ∗ ∗ ∗ ∗ −I 0 0 0∗ ∗ ∗ ∗ ∗ ∗ ∗ −I 0 0∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ −λI 0∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ −λI
< 0 (3.72)
88 CHAPTER 3. STATE FEEDBACK CONTROL
2X − R M1 M2
∗ Z1 Z2
∗ ∗ Z3
≥ 0. (3.73)
where
σ3 := h(XAT + Y T1 BT ),
σ4 := (EbY1 + EaX)T ,
and ℵ11, ℵ12, ℵ22 are defined in (3.66). Moreover, the controller gains are given by
K1 = Y1X−1, and K2 = Y2X
−1.
3.4.7 Simulation Results
0 10 20 30 40 50 60 70 80 90 100−2.5
−2
−1.5
−1
−0.5
0
0.5
1
1.5
2
2.5First floor − Uncontrolled Uncertain System under Disturbance
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.73: The displacement of the first floor of the uncontrolled system withnorm-bounded uncertainties under disturbance and ω = 1.8
Now, we show the validity of Propositions 15 and 16. Consider system (3.57)
with the parameters: µ = 0.1, γ = 3, h = 0.1 and
Bw =[
1 0 0.1 0 0 0]T
,
Cz =[
0 0.5 0 0 0 0],
L =[
0 0 0.001 0 0 0.01]T
,
3.4. STATE FEEDBACK H∞ CONTROL 89
0 10 20 30 40 50 60 70 80 90 100−3
−2
−1
0
1
2
3Second floor − Uncontrolled Uncertain System under Disturbance
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.74: The displacement of the second floor of the uncontrolled system withnorm-bounded uncertainties under disturbance and ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−3
−2
−1
0
1
2
3Third floor − Uncontrolled Uncertain System under Disturbance
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.75: The displacement of the third floor of the uncontrolled system withnorm-bounded uncertainties under disturbance and ω = 1.8
90 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1First floor − Uncontrolled Uncertain System under Disturbance
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.76: The displacement of the first floor of the uncontrolled system withnorm-bounded uncertainties under disturbance and ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1Second floor − Uncontrolled Uncertain System under Disturbance
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.77: The displacement of the second floor of the uncontrolled system withnorm-bounded uncertainties under disturbance and ω = 0.5773
3.4. STATE FEEDBACK H∞ CONTROL 91
0 10 20 30 40 50 60 70 80 90 100−1.5
−1
−0.5
0
0.5
1
1.5Third floor − Uncontrolled Uncertain System under Disturbance
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.78: The displacement of the third floor of the uncontrolled system withnorm-bounded uncertainties under disturbance and ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100
−0.25
−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
First floor − Robust Hinf
State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.79: The displacement of the first floor via the controller (3.55) with (3.74)for ω = 1.8
92 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
Second floor − Robust Hinf
State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.80: The displacement of the second floor via the controller (3.55) with(3.74) for ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
Third floor − Robust Hinf
State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.81: The displacement of the third floor via the controller (3.55) with (3.74)for ω = 1.8
3.4. STATE FEEDBACK H∞ CONTROL 93
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
First floor − Robust Hinf
State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.82: The displacement of the first floor via the controller (3.55) with (3.74)for ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
Second floor − Robust Hinf
State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.83: The displacement of the second floor via the controller (3.55) with(3.74) for ω = 0.5773
94 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
Third floor − Robust Hinf
State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.84: The displacement of the third floor via the controller (3.55) with (3.74)for ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
First floor − Improved Robust Hinf
State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.85: The displacement of the first floor via the controller (3.55) with (3.75)for ω = 1.8
3.4. STATE FEEDBACK H∞ CONTROL 95
0 10 20 30 40 50 60 70 80 90 100−0.25
−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
Second floor − Improved Robust Hinf
State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.86: The displacement of the second floor via the controller (3.55) with(3.75) for ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
Third floor − Improved Robust Hinf
State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.87: The displacement of the third floor via the controller (3.55) with (3.75)for ω = 1.8
96 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
First floor − Improved Robust Hinf
State Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.88: The displacement of the first floor via the controller (3.55) with (3.75)for ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
Second floor − Improved Robust Hinf
State Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.89: The displacement of the second floor via the controller (3.55) with(3.75) for ω = 0.5773
3.4. STATE FEEDBACK H∞ CONTROL 97
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
Third floor − Improved Robust Hinf
State Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.90: The displacement of the third floor via the controller (3.55) with (3.75)for ω = 0.5773
Ea =[
1 0 0 0 0 0];
The other parameter values can be found in Chapter 2.
When u(t) = 0, Figures 3.73 - 3.78 depict the responses of all the states of
the uncertain system for different wave frequencies. Comparing with Figures 3.1 -
3.6, which are for the nominal system, we find that the responses of the uncertain
system become more drastic. For example, the amplitudes of the displacement for
ω = 0.5773 has increased by nearly 13% (see Figure 3.73 - 3.75). So, it’s necessary
to take measures to cripple such passive effects and constrain the oscillations of the
states.
However, based on Propositions 15 and 16, the obtained controllers can effec-
tively control the uncertain system, which are shown in the following.
By solving LMIs (3.70) and (3.71), the Controller Design from Proposition 15
gives the controller gains as
Ks9 :
K1 = [−0.1941 −0.1424 1.9398 0.3026 0.1145 −0.0999]× 104
K2 = [0.2640 0.0649 −2.0823 −0.0230 0.0326 0.1235]× 103
(3.74)
98 CHAPTER 3. STATE FEEDBACK CONTROL
The simulation diagram is shown in Appendix C.5. And the state responses of
system (3.57) under the Controller Design from Proposition 15 are shown in Figures
3.79 - 3.84. Comparing Figure 3.79 with Figure 3.73, we find that the curve, which
indicates the displacement of the first floor under control, oscillates in a range about
from -0.25 ft to 0.3 ft, not in the other range from -2.4 ft to 2.4 ft any longer when
no control is applied. Similarly sharp decrease occurs to the displacements of the
second and third floors (see Figures 3.80 and 3.81). Comparing with Figures 3.76
- 3.78, their oscillation amplitudes in Figures 3.82 - 3.84 are reduced from 1.7 ft to
1.2 ft, from 1.8 ft to 1.4 ft and from 1.9 ft to 1.5 ft peak to peak, respectively, which
obtain nearly 30% reduction. These results show that the servo under the Controller
Design from Proposition 15 motivates the AMD to vibrate in some scope, so that
the internal oscillations of the offshore structure can be restrained successfully.
On the other hand, by LMIs (3.72) and (3.73), the Controller Design from Propo-
sition 16 gives the controller gains as
Ks10 :
K1 = [−0.9159 −0.3360 6.4051 0.2540 0.0617 −0.3493]× 104
K2 = [0.8114 0.0912 −2.8218 0.0571 0.0593 0.2468]× 103
(3.75)
The state responses under the Controller Design from Proposition 16 are shown in
Figures 3.85 - 3.90. Comparing Figures 3.85 - 3.90 with Figures 3.79 - 3.84, we
find the internal oscillations of the offshore structure under the Controller Design
from Proposition 16 are slightly weaker than those under the Controller Design
from Proposition 15. Taking the displacement of the first floor for ω = 1.8 as an
example, the amplitude is 0.35 ft peak to peak under the Controller Design from
Proposition 16, but 0.55 ft under the Controller Design from Proposition 15. (The
simulation diagram of the Controller Design from Proposition 16 is similar to that
of the Controller Design from Proposition 15, so it’s omitted.)
As a byproduct, we compare the achieved H∞ performance by the Controller
Design from Proposition 15 with that by the Controller Design from Proposition 16.
3.4. STATE FEEDBACK H∞ CONTROL 99
For different values of h, the obtained minimum H∞ levels, γ, both by the Controller
Design from Proposition 15 and by the Controller Design from Proposition 16, are
listed in Table 3.2, from which, the Controller Design from Proposition 16 can
achieve better H∞ performance than the Controller Design from Proposition 15.
Table 3.2: The achieved minimum values of γ for various h
h 0.1 0.2 0.3 0.4 0.5 0.6Proposition 15 0.56067 0.61502 0.66156 1.01410 - -Proposition 16 0.53937 0.56061 0.58807 0.61505 0.63947 0.66150
3.4.8 A Polytopic Uncertainty
For the polytopic uncertainty, (A(t), B(t)) in (3.56) is assumed to be bounded by
a given convex bounded polyhedral domain Σ, i.e.
(A(t), B(t)) ∈ Σ :=
q∑
j=1
ξj(Aj, Bj),
q∑j=1
ξj = 1, ξj ≥ 0
(3.76)
System (3.53) with (3.76) under controller (3.55) becomes
x(t) = [A(t) + B(t)K1]x(t) + B(t)K2x(t− h) + Fg(x, t) + Bwω(t),
z(t) = Czx(t),
x(t) = φ(t), t ∈ [−h, 0]
(3.77)
Based on a parameter Lyapunov-Krasovskii functional approach, we have the
following conclusion. The proof is similar to that in Propositions 9 and 10.
Proposition 17. For some given positive scalars µ, γ and h, system (3.77) with
polytopic uncertainty (3.38) is asymptotically stable and satisfies ‖z(t)‖2 ≤ γ ‖ω(t)‖2
for all nonzero ω(t) ∈ L2[0,∞), if there exist matrices Pj > 0, Qj > 0, Rj > 0,
100 CHAPTER 3. STATE FEEDBACK CONTROL
M1j, M2j, Z1j, Z2j, Z3j, N and Y1, Y2 of appropriate dimensions such that
Υ11 Υ12 Υ13 F Bw µNT NT CTz
∗ Υ22 BjY2 F Bw 0 0
∗ ∗ Υ33 0 0 0 0∗ ∗ ∗ −I 0 0 0∗ ∗ ∗ ∗ −γ2I 0 0∗ ∗ ∗ ∗ ∗ −I 0∗ ∗ ∗ ∗ ∗ ∗ −I
< 0 (3.78)
Rj M1j M2j
∗ Z1j Z2j
∗ ∗ Z3j
≥ 0 (3.79)
hold for j = 1, 2, · · · , q, where
Υ11 =Qj + MT1j + M1j + hZ1j + AjN + BjY1 + (AjN + BjY1)
T ,
Υ12 =Pj − N + ε(AjN + BjY1)T ,
Υ22 =hRj − εN − εNT ,
Υ13 =− MT1j + M2j + hZ2j + BjY2,
Υ33 =− MT2j − M2j + hZ3j.
Moreover, the controller gains are given by K1 = Y1X−1 and K2 = Y2X
−1.
3.4.9 Simulation Results
Consider the closed-loop system (3.77) with the following parameter matrices
A(t) = A + ∆A, B(t) = B,
where A,B defined in (2.6) and
∆A =
0 0 0 0 0 00 0 0 0 0 0δ1 0 0 0 0 00 0 0 0 0 00 δ2 0 0 0 00 0 0 0 0 0
with
δ1 ∈[ −0.09, 0.09
], δ2 ∈
[ −0.15, 0.15].
3.4. STATE FEEDBACK H∞ CONTROL 101
0 10 20 30 40 50 60 70 80 90 100−4
−3
−2
−1
0
1
2
3
4First floor − Uncontrolled Polytopic Type Uncertain System under Disturbace
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.91: The displacement of the first floor of the uncontrolled system withpolytopic type uncertainties under disturbance and ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−4
−3
−2
−1
0
1
2
3
4Second floor − Uncontrolled Polytopic Type Uncertain System under Disturbace
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.92: The displacement of the second floor of the uncontrolled system withpolytopic type uncertainties under disturbance and ω = 1.8
102 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−4
−3
−2
−1
0
1
2
3
4Third floor − Uncontrolled Polytopic Type Uncertain System under Disturbace
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.93: The displacement of the third floor of the uncontrolled system withpolytopic type uncertainties under disturbance and ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1First floor − Uncontrolled Polytopic Type Uncertain System under Disturbace
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.94: The displacement of the first floor of the uncontrolled system withpolytopic type uncertainties under disturbance and ω = 0.5773
3.4. STATE FEEDBACK H∞ CONTROL 103
0 10 20 30 40 50 60 70 80 90 100−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1Second floor − Uncontrolled Polytopic Type Uncertain System under Disturbace
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.95: The displacement of the second floor of the uncontrolled system withpolytopic type uncertainties under disturbance and ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−1.5
−1
−0.5
0
0.5
1
1.5Third floor − Uncontrolled Polytopic Type Uncertain System under Disturbace
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.96: The displacement of the third floor of the uncontrolled system withpolytopic type uncertainties under disturbance and ω = 0.5773
104 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−0.25
−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
First floor − Robust Hinf
State Feedback Control (Polytopic Type)
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.97: The displacement of the first floor via the controller (3.55) with (3.80)for ω = 1.8 (polytopic uncertainties)
0 10 20 30 40 50 60 70 80 90 100−0.25
−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
Second floor − Robust Hinf
State Feedback Control (Polytopic Type)
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.98: The displacement of the second floor via the controller (3.55) with(3.80) for ω = 1.8 (polytopic uncertainties)
3.4. STATE FEEDBACK H∞ CONTROL 105
0 10 20 30 40 50 60 70 80 90 100−0.25
−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
Third floor − Robust Hinf
State Feedback Control (Polytopic Type)
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.99: The displacement of the third floor via the controller (3.55) with (3.80)for ω = 1.8 (polytopic uncertainties)
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
First floor − Robust Hinf
State Feedback Control (Polytopic Type)
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 3.100: The displacement of the first floor via the controller (3.55) with (3.80)for ω = 0.5773 (polytopic uncertainties)
106 CHAPTER 3. STATE FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
Second floor − Robust Hinf
State Feedback Control (Polytopic Type)
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 3.101: The displacement of the second floor via the controller (3.55) with(3.80) for ω = 0.5773 (polytopic uncertainties)
0 10 20 30 40 50 60 70 80 90 100
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
Third floor − Robust Hinf
State Feedback Control (Polytopic Type)
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 3.102: The displacement of the third floor via the controller (3.55) with (3.80)for ω = 0.5773 (polytopic uncertainties)
3.4. STATE FEEDBACK H∞ CONTROL 107
Clearly, (3.77) with (3.51) is a system with polytopic uncertainty, with four vertices
as (Aj, Bj) (j = 1, 2, 3, 4), where Bj = B and
A1 =
0 1 0 0 0 0−3.3235 −0.0212 0.0184 0.0030 −5.3449 −0.8819−0.09 0 0 1 0 00.0184 0.0030 −118.1385 −0.1117 5.3468 0.8822
0 −0.15 0 0 0 1−0.0114 −0.0019 0.0114 0.0019 −3.3051 −0.5454
;
A2 =
0 1 0 0 0 0−3.3235 −0.0212 0.0184 0.0030 −5.3449 −0.8819−0.09 0 0 1 0 00.0184 0.0030 −118.1385 −0.1117 5.3468 0.8822
0 0.15 0 0 0 1−0.0114 −0.0019 0.0114 0.0019 −3.3051 −0.5454
;
A3 =
0 1 0 0 0 0−3.3235 −0.0212 0.0184 0.0030 −5.3449 −0.8819
0.09 0 0 1 0 00.0184 0.0030 −118.1385 −0.1117 5.3468 0.8822
0 −0.15 0 0 0 1−0.0114 −0.0019 0.0114 0.0019 −3.3051 −0.5454
;
A4 =
0 1 0 0 0 0−3.3235 −0.0212 0.0184 0.0030 −5.3449 −0.8819
0.09 0 0 1 0 00.0184 0.0030 −118.1385 −0.1117 5.3468 0.8822
0 0.15 0 0 0 1−0.0114 −0.0019 0.0114 0.0019 −3.3051 −0.5454
.
Let h = 0.1, γ = 3, µ = 0.1, and
Bw =[
1 0 0.1 0 0 0]T
,
Cz =[
0 0.5 0 0 0 0].
When u(t) = 0, that is, no control is added into the system, Figures 3.91 -
3.96 demonstrate the responses of all the states of the system with polytopic type
uncertainty. Comparing with Figures 3.1 - 3.6, we can see the duplex influences
of ploytopic type uncertainty and disturbance to the system. For example, the
amplitude of the displacement of the first floor for ω = 1.8 has increased by nearly
40% (see Figure 3.91); and even the displacement of the first floor for ω = 0.5773
108 CHAPTER 3. STATE FEEDBACK CONTROL
has also increased by about 12% (see Figure 3.94). Therefore, it’s necessary to
take measures to weaken the effect brought by uncertainty and disturbance, and to
reduce the internal oscillations induced by wave forces. For this aim, a robust H∞
state feedback controller based on Proposition 17 is proposed. The gains are given
by
Ks11 :
K1 = [−7.1284 −0.9221 8.8308 −0.1684 −0.6949 −2.7038]× 104
K2 = [−247.1548 −79.4067 534.3352 −6.2850 −29.7022 −116.5531]
(3.80)
The state responses of the system with polytopic uncertainties by controller
(3.55) with Ks11 are shown in Figures 3.97 - 3.102. (The simulation diagram of the
Controller Design from Proposition 17 is similar to that of the Controller Design
from Proposition 15 except the definitions of fcn. See Appendix C.6 for details.)
Compared with Figures 3.91 - 3.96, the oscillation amplitudes of the three floors
for ω = 1.8 and ω = 0.5773 decrease sharply. In fact, from Figures 3.91 - 3.93,
for ω = 1.8, the amplitudes of the three floors are only 6% of those in Figures
3.97 - 3.99). Also, from Figures 3.100 - 3.102, it’s easy to see that the oscillation
amplitudes of the displacements reduce from 1.7 ft to 1 ft, from 1.9 ft to 1.1 ft and
from 2 ft to 1.2 ft peak to peak, respectively. All these have illustrated that the
controller (3.3) with Ks11 can reduce the internal oscillations and ensure the safety
and comfort of the offshore structure with polytopic uncertainty.
3.5 Conclusion
In order to effectively reduce the internal system oscillations, when the system states
are adopted as feedback, a memory state feedback approach is, for the first time,
proposed in this chapter for an offshore steel jacket platform. Based on Lyapunov-
Krasovskii stability theory, some delay-dependent robust stability criteria for the
closed-loop system have been obtained, where two classes of uncertainties, namely,
norm-bounded uncertainty and polytopic uncertainty, are taken into account. Then,
3.5. CONCLUSION 109
these criteria are used to design the desired controllers incorporating with some ma-
trix techniques. Simulation results sufficiently show the effectiveness of the proposed
memory controllers on the offshore structures. On the other hand, by introducing
a new Lyapunov-Krasovskii functional, an improved delay-dependent stability con-
dition is obtained, which is less conservative than those in the literature. Finally,
robust H∞ control for the offshore system has also been investigated, and an H∞
controller has been designed, which guarantees the system is asymptotically stable
with a prescribed H∞ performance.
Chapter 4
Dynamic Output FeedbackControl
In Chapter 3, the state feedback control for the offshore steel jacket platform was
considered when the system states are adopted as feedback. However, if the system’s
outputs are taken as feedback, we can design a dynamic output feedback controller
to reduce the internal systems oscillations, which motivates the study in this chap-
ter. Based on Lyapunov-Krasovskii stability theory and a project theorem, some
sufficient conditions for the existence of dynamic output feedback controllers will be
obtained, which are in terms of the solutions of a set of linear matrix inequalities.
Moreover, a robust H∞ performance for the system with norm-bounded uncertainty
will also be considered. The simulation results will be given to illustrate the validity
of the proposed method.
4.1 System Description
Let’s reconsider the offshore steel jacket platform with uncertainties as follows:
x(t) = (A + ∆A)x(t) + Bu(t) + Fg(x, t)
y(t) = Cx(t)
x(t) = φ(t), t ≤ 0
(4.1)
where x(t) ∈ Rn is the state vector, u(t) ∈ Rm is the control, y(t) ∈ Rq is the
measurement output, g(x, t) ∈ Rp is the nonlinear self-excited hydrodynamic force
112 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
vector, which is uniformly bounded and satisfies the following cone-bounding con-
straint:
‖g(x, t)‖ ≤ µ ‖x(t)‖ , (4.2)
where µ is a positive scalar; A, B, C, and F are constant real matrices with appro-
priate dimensions, ∆A represents norm-bounded uncertainty of the following form
4A(t) := LG(t)Ea, (4.3)
where G(t) is an uncertainty matrix bounded by I, i.e.
GT (t)G(t) ≤ I, ∀t, (4.4)
Assume that the outputs of system (4.1) are adopted as feedback, under which
we will design a dynamic output feedback controller as
ξ(t) = AKξ(t) + BKy(t− h)
u(t) = CKξ(t) + DKy(t− h)(4.5)
where ξ(t) ∈ Rk is the state of the controller, AK , BK , CK and DK are matrices to
be determined and h is a constant time-delay, such that the closed-system combined
by (4.1) and (4.5) is asymptotically stable.
Introducing an augmented vector
ϕ(t) := [xT (t) ξT (t)]T
then the closed-loop system combined by (4.1) and (4.5) is given by
ϕ(t) = (A0 + ∆A0)ϕ(t) + A1Eϕ(t− h) + ET Fg(x, t) (4.6)
where
A0 :=
[A BCK
0 AK
], ∆A0 :=
[∆A 00 0
], A1 :=
[BDKCBKC
], E :=
[In×n
0n×k
]T
. (4.7)
Let
K :=
[DK CK
BK AK
]
4.2. NOMINAL SYSTEMS 113
and
A00 :=
[A 00 0
], B00 :=
[B 00 I
], C00 :=
[C0
], D00 =
[0 00 I
]. (4.8)
Then, the matrices A0, A1, ∆A0 can be restated as
A0 = A00 + B00KD00,
A1 = B00KC00,
∆A0 = ET LG(t)EaE. (4.9)
In the following, firstly, we will discuss the dynamic output feedback control
for the nominal system of (4.1), then the obtained results will be extended to suit
for the uncertain system (4.1). Finally, the H∞ control issue via dynamic output
feedback controller will be also investigated.
4.2 Nominal Systems
In this section, we will focus on the nominal system of (4.1), in this case, assume
∆A ≡ 0. The resulting closed-loop system of (4.6) is given by
ϕ(t) = A0ϕ(t) + A1Eϕ(t− h) + ET Fg(x, t) (4.10)
4.2.1 Stability Analysis
Proposition 18. For given positive scalars µ and h, system (4.10) is asymptotically
stable if there exist real matrices P > 0, Q > 0 and R > 0 of appropriate dimensions
such that
Θ :=
Θ11 PA1 + ET R PET F hAT0 ET R
∗ −Q−R 0 hAT1 ET R
∗ ∗ −I hF T R∗ ∗ ∗ −R
< 0 (4.11)
where
Θ11 := PA0 + AT0 P + ET (Q−R + µ2I)E.
Proof. Choose a Lyapunov functional candidate as
V (ϕt) = ϕT (t)Pϕ(t) +
∫ t
t−h
ϕT (s)ET QEϕ(s)ds
114 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
+ h
∫ 0
−h
ds
∫ t
t+s
ϕT (θ)ET REϕ(θ)dθ, (4.12)
where ϕt = x(t+α), α ∈ [−h, 0] and P > 0, Q > 0, R > 0 to be determined. Taking
the derivative of V (t, ϕt) with respect to t along the trajectory of (4.10) yields
V (ϕt) = 2ϕT (t)Pϕ(t)
+ ϕT (t)ET QEϕ(t)− ϕT (t− h)ET QEϕ(t− h)
+ h2ϕT (t)ET REϕ(t)− h
∫ t
t−h
ϕT (θ)ET REϕ(θ)dθ. (4.13)
By Lemma 1 (see Appendix A), one obtains
−h
∫ t
t−h
ϕT (θ)ET REϕ(θ)dθ
≤[
ϕ(t)Eϕ(t− h)
]T [ −ET RE ET R∗ −R
] [ϕ(t)
Eϕ(t− h)
]. (4.14)
Noting (4.2), we have
µ2ϕT (t)ET Eϕ(t)− gT (x, t)g(x, t) ≥ 0. (4.15)
Let
ηT (t) =[
ϕT (t) ϕT (t− h)ET gT (x, t)]. (4.16)
Combining (4.13) with (4.14) and (4.15), yields
V (t, xt) ≤ ηT (t)Φ + h2ΓT RΓ
η(t)
where
Φ :=
Θ11 PA1 + ET R PET F∗ −Q−R 0∗ ∗ −I
Γ :=[EA0 EA1 F
]
with Θ11 being defined in (4.11). If (4.11) is feasible, then Φ + h2ΓT RΓ < 0 by the
Schur complement. So there exists a scalar δ > 0 such that V (ϕt) ≤ −δ ‖ϕ(t)‖2 < 0
for ϕ(t) 6= 0, which guarantees the closed-loop system (4.10) is asymptotically stable.
This completes the proof.
4.2. NOMINAL SYSTEMS 115
Proposition 18 provides a delay-dependent stability criterion for the nominal
system (4.10), from which, however, the controller parameters cannot be directly
obtained. In order to solve out the controller parameters, we will follow the lines in
[30], and a parameterized controller design is presented in terms of a set of LMIs,
which is stated in the following.
4.2.2 Controller Design
To solve out the controller parameters, by substituting (4.9) into (4.11), after simple
algebraic manipulation, (4.11) can be rewritten as
Ω + ΣΠKΛT + ΛKT ΠT ΣT < 0 (4.17)
where
Ω :=
Ω11 ET R PET F hAT00E
T R∗ −Q−R 0 0∗ ∗ −I hF T R∗ ∗ ∗ −R
Σ := diagP, I, I, hR
Π :=[
BT00 0 0 BT
00ET
]T
Λ :=[
D00 C00 0 0]T
with
Ω11 := PA00 + AT00P + ET (Q−R + µ2I)E
From Lemma 4, inequality (4.17) is solvable for some K if and only if
Π⊥Σ−1ΩΣ−1(Π⊥)T < 0 (4.18)
Λ⊥Ω(Λ⊥)T < 0 (4.19)
where Π⊥ and Λ⊥ denote the orthogonal complements of Π and Λ, respectively. Let
W1 and [W T2 W T
3 ] be the orthogonal complements of CT and [BT BT ]T , respectively.
Then, Π⊥ and Λ⊥ can be chosen as
Π⊥ =
W T2 0 0 0 W T
3
0 0 I 0 00 0 0 I 0
T
116 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
Λ⊥ =
I 0 0 0 00 0 W T
1 0 00 0 0 I 00 0 0 0 I
T
. (4.20)
In order to simplify (4.18) and (4.19), P and P−1 can be partitioned as
P =
[Y NNT 4
], P−1 =
[X M
MT 4]
, (4.21)
where X, Y ∈ Rn×n, M , N ∈ Rn×k, 4 means irrelevant matrix, and positive definite
matrices X, Y satisfy [X II Y
]≥ 0. (4.22)
Substituting (4.8) and (4.21) into (4.18) and (4.19), then (4.18) is equivalent to
z :=
z11 W T
2 XR (W T2 + W T
3 )F∗ −Q−R 0∗ ∗ −I
< 0 (4.23)
where
z11 := W T2 [AX + XAT + X(Q−R + µ2I)X]W2
+ W T2 XAT W3 + W T
3 AXW2 − h−2W T3 R−1W3
and (4.19) is equivalent to
Ξ :=
Ξ11 RW1 Y F hAT R∗ W T
1 (−Q−R)W1 0 0∗ ∗ −I hF T R∗ ∗ ∗ −R
< 0 (4.24)
where
Ξ11 := Y A + AT Y + Q−R + µ2I
In summary, based on Proposition 18 and above analysis, an existence sufficient
condition of parameterized controllers of form (4.5) is obtained, which is stated as
Proposition 19. Let W1 and [W T2 W T
3 ]T be the orthogonal complements of CT and
[BT BT ]T , respectively. For given scalars µ > 0 and h > 0, the dynamic output
4.2. NOMINAL SYSTEMS 117
feedback control problem for system (4.10) is solvable if there exist real matrices
X > 0, Y > 0, Q > 0 and R > 0 of appropriate dimensions such that (4.22), (4.23)
and (4.24) are feasible.
Noting that Proposition 19 does not present the computation of the controller
itself, but the existence of parameterized controllers. Moreover, the obtained matrix
inequalities are not linear due to such terms as XQX inz11. In order to compute the
desired controller, first, we convert the non-convex feasibility problem of Proposition
19 into a nonlinear minimization problem subject to LMIs by employing a cone
complementary linearization algorithm in [4]. For this aim,
Pre- and post-multiplying both sides of z in (4.23) by T T := diagI, X, I, Iand its transpose, respectively, yields
T TzT =
z11 W T
2 XRX (W T2 + W T
3 )F∗ −XQX −XRX 0∗ ∗ −I
< 0 (4.25)
Introducing new matrix variables S > 0 and Z > 0 such that XRX ≥ S and
XQX ≥ Z, then since
[ −W T2 XRXW2 W T
2 XRX∗ −XRX
]
=
[W T
2
I
] [ −XRX XRXXRX −XRX
] [W2
I
]
=
[W T
2
I
] [I −I0 I
] [0 00 −XRX
] [I 0−I I
] [W2
I
]
≤[
W T2
I
] [I −I0 I
] [0 00 −S
] [I 0−I I
] [W2
I
]
=
[ −W T2 SW2 W T
2 S∗ −S
].
Therefore, if XRX ≥ S > 0 and XQX ≥ Z > 0, then (4.25) is implied byz11 W T
2 S (W T2 + W T
3 )F∗ −Z − S 0∗ ∗ −I
< 0 (4.26)
where
z11 :=W T2 [AX + XAT + X(Q + µ2I)X − S]W2
118 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
+ W T2 XAT W3 + W T
3 AXW2 − h−2W T3 R−1W3
By the Schur complement, (4.26) is equivalent to
z11 W T2 S (W T
2 + W T3 )F W T
2 X µW T2 X
∗ −Z − S 0 0 0∗ ∗ −I 0 0∗ ∗ ∗ −Q−1 0∗ ∗ ∗ ∗ −I
< 0 (4.27)
where
z11 :=W T2 [AX + XAT − S]W2 + W T
2 XAT W3
+ W T3 AXW2 − h−2W T
3 R−1W3
On the other hand, applying the Schur complement gives
XRX ≥ S ⇐⇒[
R X−1
X−1 S−1
]≥ 0 (4.28)
XQX ≥ Z ⇐⇒[
Q X−1
X−1 Z−1
]≥ 0 (4.29)
Let
R = R−1, X = X−1, S = S−1, Z = Z−1, Q = Q−1.
then a new sufficient condition for the existence of the dynamic output feedback
controller is derived as follows.
Proposition 20. Let W1 and [W T2 W T
3 ] be the orthogonal complements of CT and
[BT BT ]T , respectively. For given scalars µ > 0 and h > 0, the dynamic output
feedback control problem for system (4.10) is solvable if there exist real matrices
X > 0, Y > 0, Q > 0, R > 0, R > 0, X > 0, S > 0, S > 0, Z > 0, Z > 0 and
Q > 0 of appropriate dimensions such that (4.22), (4.24) and
z11 W T2 S (W T
2 + W T3 )F W T
2 X µW T2 X
∗ −Z − S 0 0 0∗ ∗ −I 0 0∗ ∗ ∗ −Q 0∗ ∗ ∗ ∗ −I
< 0 (4.30)
4.2. NOMINAL SYSTEMS 119
[R XX S
]≥ 0 (4.31)
[Q XX Z
]≥ 0 (4.32)
RR = I, XX = I, SS = I, ZZ = I, QQ = I (4.33)
where
z11 := W T2 [AX + XAT − S]W2 + W T
2 XAT W3 + W T3 AXW2 − h−2W T
3 RW3.
Clearly, matrix inequalities (4.22),(4.24), (4.30)-(4.32) are all linear on matrix
variables, while (4.33) are a set of equality constraints. Fortunately, these con-
straints can be converted into a nonlinear minimization problem subject to LMIs by
employing the Cone Complementary Problem (CCP) proposed in [4]. The key idea
of the CCP can be stated as: if the LMI
[S II S
]≥ 0
is feasible on the n × n matrix variables S > 0 and S > 0, then Tr(SS) ≥ n, and
Tr(SS) = n if and only if SS = I. According to the above CCP lines, the non-
convex feasibility problem formulated by (4.22),(4.24), (4.30)-(4.32) and (4.33) can
be considered as a minimization problem involving LMI conditions as
Nonlinear Minimization Problem Subject to LMIs
Minimize Tr(XX + RR + QQ + ZZ + SS)Subject to (4.22), (4.24), (4.30)-(4.32), and[
R II R
]≥ 0,
[X II X
]≥ 0,
[Q II Q
]≥ 0,
[Z II Z
]≥ 0,
[S II S
]≥ 0.
(4.34)
The iterative algorithm proposed in Algorithm 2.1 can be modified to solve the
above nonlinear minimization problem.
Now, we can present the computation of the desired controller. First, compute
some solutions by solving the above nonlinear minimization problem (4.34); second,
120 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
compute two full-column rank matrices M,N ∈ Rn×k such that
MNT = I −XY. (4.35)
Then, the unique solution P can be thus calculated from the following equation
[Y INT 0
]= P
[I X−1
0 MT
]. (4.36)
Finally, the controller parameter K =
[DK CK
BK AK
]can be easily obtained by solv-
ing LMI (4.11) with known P .
4.2.3 Simulation Results
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8First floor − Output Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 4.1: The displacement of the first floor when the output feedback controlleris used and ω = 1.8
In Chapter 3, we discussed the control problem of the offshore structure system.
When the system states are adopted as feedback, the system can be effectively con-
trolled by a memory state feedback controller. The internal oscillations are greatly
reduced to a suitable level. However, when the system outputs are considered as
feedback, we can exploit Proposition 20 to seek a dynamic output feedback con-
troller to reduce the internal oscillations of the system. Let h = 0.02 and µ = 0.1,
4.2. NOMINAL SYSTEMS 121
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8Second floor − Output Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 4.2: The displacement of the second floor when the output feedback controlleris used and ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8Third floor − Output Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 4.3: The displacement of the third floor when the output feedback controlleris used and ω = 1.8
122 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8First floor − Output Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 4.4: The displacement of the first floor when the output feedback controlleris used and ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8Second floor − Output Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 4.5: The displacement of the second floor when the output feedback controlleris used and ω = 0.5773
4.2. NOMINAL SYSTEMS 123
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8Third floor − Output Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 4.6: The displacement of the third floor when the output feedback controlleris used and ω = 0.5773
solving the nonlinear minimization problem (4.34) yields
X =
0.9739 −0.2793 0.0131 −0.3556 −0.1356 −0.4666−0.2793 6.3799 0.4450 0.5945 0.3318 2.13220.0131 0.4450 0.3279 −0.1595 0.0184 0.2631−0.3556 0.5945 −0.1595 37.1101 −0.2603 0.3984−0.1356 0.3318 0.0184 −0.2603 1.2820 −0.3217−0.4666 2.1322 0.2631 0.3984 −0.3217 2.0653
Y =
0.0002 −0.0000 −0.0000 0.0000 0.0159 0.0142−0.0000 0.0000 0.0000 −0.0000 −0.0056 −0.0013−0.0000 0.0000 0.0003 −0.0000 −0.0011 −0.00520.0000 −0.0000 −0.0000 0.0000 0.0002 0.00020.0159 −0.0056 −0.0011 0.0002 7.6689 0.87610.0142 −0.0013 −0.0052 0.0002 0.8761 2.6154
× 105
R =
34.3152 −2.1196 −2.4177 0.3390 7.6227 1.11032.1196 −10.8303 −3.4722 −0.4904 −1.7991 −3.4295−2.4177 −3.4722 19.9992 0.2347 −2.3994 −7.04510.3390 −0.4904 0.2347 0.3808 0.2910 −0.67807.6227 −1.7991 −2.3994 0.2910 43.4562 0.98181.1103 −3.4295 −7.0451 −0.6780 0.9818 26.8163
124 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
Q =
1.0733 −0.0616 −0.4189 −0.0096 0.3589 0.4931−0.0616 0.2663 −0.1686 −0.0206 −0.0827 −0.0228−0.4189 −0.1686 3.7791 0.0488 −0.2449 −0.6789−0.0096 −0.0206 0.0488 0.0278 −0.0049 −0.04170.3589 −0.0827 −0.2449 −0.0049 0.7836 0.36730.4931 −0.0228 −0.6789 −0.0417 0.3673 1.1980
Then, from (4.35) we obtain
M =
−0.1846 −0.0274 −0.4305 −0.2927 −0.3011 −0.77690.7583 −0.0636 −0.2661 −0.1514 0.5419 −0.18340.0862 −0.0212 −0.5942 −0.4188 −0.3304 0.59540.0320 −0.2172 0.5690 −0.7923 −0.0158 −0.01060.3395 0.8567 0.1906 −0.0770 −0.3235 −0.06210.5170 −0.4622 0.1753 0.2870 −0.6338 −0.0662
N =
−0.0464 0 0 0 0 00.0063 0.0054 0 0 0 00.0144 −0.0061 0.0001 0 0 0−0.0008 0.0001 −0.0000 0.0001 0 0−6.4494 −8.6021 0.0039 0.0203 −0.0456 0−7.5054 2.6504 −0.0100 −0.0043 0.0027 −0.0098
× 105
Finally, a desired output feedback controller is thus obtained as (4.5) with
AK =
−25.9976 2.3183 20.2489 −22.8328 −4.1473 4.973612.8569 −2.0863 −8.3747 9.4323 1.7165 −2.0621−34.1903 20.8718 −22.8694 35.8562 −4.5279 5.5525−34.4200 16.2495 −9.9625 16.9789 −5.1585 6.50804.4404 7.5349 −24.6161 34.1852 0.6353 0.252963.9004 −23.7410 −3.1931 −11.2750 7.7576 −17.4902
BK =
−0.0002 −0.0227 0.0014−0.0055 0.0097 −0.0019−1.0327 −1.1987 −0.0571−0.6641 −0.9038 −0.01820.2092 −0.5066 0.1348−6.4840 2.1548 −1.6073
CK =[ −4.0071 0.5979 3.1251 −3.5246 −0.6401 0.7672
]× 104
DK =[ −11.4847 −33.9770 −0.2274
].
See Appendix C.7 for the simulation diagram. Under the obtained output feedback
controller, the simulation results of the state response of the system are given in
Figures 4.1 - 4.6, from which, it is clear that the internal oscillations are effectively
controlled. For example, the amplitude of vibration of the third floor for ω = 1.8
4.3. UNCERTAIN SYSTEMS 125
has decreased by about 73% (compare Figure 4.3 with Figure 3.3). Figures 4.4 - 4.6
have plotted the states for ω = 0.5773. In a word, under the operation of design
control scheme, the responses of the platform have reduced greatly as compared
with the uncontrolled case.
4.3 Uncertain Systems
Now, we consider system (4.1) with norm-bounded uncertainty (4.3).
Based on Proposition 20, replacing A in (4.24) and (4.30) with (A + 4A), we
find that (4.24) and (4.30) for (4.1) are, respectively, equivalent to the following
conditions:
Ξ + Γ1G(t)Γ2 + ΓT2 GT (t)ΓT
1 < 0 (4.37)
and
z+ Γ3G(t)Γ4 + ΓT4 GT (t)ΓT
3 < 0. (4.38)
where
Γ1 : =[
LT Y 0 0 hLT R]T
,
Γ2 : =[
Ea 0 0 0],
Γ3 : =[
LT (W2 + W3) 0 0 0]T
,
Γ4 : =[
EaXW2 0 0 0].
Using Lemma 2 (see Appendix A), (4.37) and (4.38) hold for any G(t) satisfying
(4.4) if and only if there exist positive scalers λ > 0 and ε > 0 such that
Ξ + λΓ1ΓT1 + λ−1Γ2
T Γ2 < 0 (4.39)
and
z+ εΓ3ΓT3 + ε−1ΓT
4 Γ4 < 0. (4.40)
126 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
which, by the Schur complement, can be represented as
Ξ11 RW1 Y F hAT R Y L λETa
∗ W T1 (−Q−R)W1 0 0 0 0
∗ ∗ −I hF T R 0 0∗ ∗ ∗ −R hRL 0∗ ∗ ∗ ∗ −λI 0∗ ∗ ∗ ∗ ∗ −λI
< 0, (4.41)
z11 W T2 S α13 W T
2 X µW T2 X α16 W T
2 XETa
∗ −Z − S 0 0 0 0 0∗ ∗ −I 0 0 0 0∗ ∗ ∗ −Q 0 0 0∗ ∗ ∗ ∗ −I 0 0∗ ∗ ∗ ∗ ∗ −εI 0∗ ∗ ∗ ∗ ∗ ∗ −εI
< 0 (4.42)
where α13 = (W T2 + W T
3 )F, α16 = ε(W T2 + W T
3 )L, and Ξ11, z11 are defined in (4.24)
and (4.30), respectively.
Therefore, nonlinear minimization problem (4.34) can be modified to design a
robust controller for system (4.1), which is stated as
Minimize Tr(XX + RR + QQ + ZZ + SS)Subject to (4.22), (4.31), (4.32), (4.41), (4.42), and[
R II R
]≥ 0,
[X II X
]≥ 0,
[Q II Q
]≥ 0,
[Z II Z
]≥ 0,
[S II S
]≥ 0.
(4.43)
4.3.1 Simulation Results
For comparison, we set h = 0.02, µ = 0.1 and
L =[
0 0 0.001 0 0 0.01]T
,
Ea =[
1 0 0 0 0 0].
Comparing Figures 3.1 - 3.6 with Figures 3.25 - 3.30, we find that the states of the
system with uncertainty become more oscillatory than that of the system with no
uncertainty, that is, uncertainty indeed degrades the system performance. In the
following, assume that the outputs are adopted as feedback, we will design a robust
4.3. UNCERTAIN SYSTEMS 127
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8First floor − Robust Output Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 4.7: The displacement of the first floor when the robust output feedbackcontroller is used and ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8Second floor − Robust Output Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 4.8: The displacement of the second floor when the robust output feedbackcontroller is used and ω = 1.8
128 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1Third floor − Robust Output Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 4.9: The displacement of the third floor when the robust output feedbackcontroller is used and ω = 1.8
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8First floor − Robust Output Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 4.10: The displacement of the first floor when the robust output feedbackcontroller is used and ω = 0.5773
4.3. UNCERTAIN SYSTEMS 129
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8Second floor − Robust Output Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 4.11: The displacement of the second floor when the robust output feedbackcontroller is used and ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8Third floor − Robust Output Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 4.12: The displacement of the third floor when the robust output feedbackcontroller is used and ω = 0.5773
130 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
dynamic output feedback controller to reduce the oscillations. Solve the nonlinear
minimization problem (4.43), and also from (4.35), we have
X =
0.9819 −0.2382 0.0238 −0.3716 −0.1554 −0.4042−0.2382 6.5575 0.4362 1.4347 0.3511 2.15160.0238 0.4362 0.3181 −0.1429 0.0246 0.2561−0.3716 1.4347 −0.1429 35.2285 −0.2669 0.8984−0.1554 0.3511 0.0246 −0.2669 1.1753 −0.2491−0.4042 2.1516 0.2561 0.8984 −0.2491 1.9786
Y =
0.0002 −0.0000 −0.0000 0.0000 0.0071 0.0065−0.0000 0.0000 0.0000 −0.0000 −0.0026 −0.0005−0.0000 0.0000 0.0004 −0.0000 −0.0002 −0.00260.0000 −0.0000 −0.0000 0.0000 0.0001 0.00010.0071 −0.0026 −0.0002 0.0001 4.0692 0.26700.0065 −0.0005 −0.0026 0.0001 0.2670 1.3051
× 106
Q =
1.0335 −0.0693 −0.4557 −0.0160 0.3563 0.4699−0.0693 0.2377 −0.1407 −0.0244 −0.0960 −0.0379−0.4557 −0.1407 4.0913 0.0655 −0.2813 −0.7205−0.0160 −0.0244 0.0655 0.0388 −0.0088 −0.06090.3563 −0.0960 −0.2813 −0.0088 0.8355 0.37450.4699 −0.0379 −0.7205 −0.0609 0.3745 1.2169
R =
72.1236 −4.2052 −6.7008 0.7737 12.4381 1.1549−4.2052 22.4673 −4.9218 −1.5549 −2.8950 −7.4955−6.7008 −4.9218 40.1781 0.4039 −6.5749 −11.35530.7737 −1.5549 0.4039 1.4329 0.5770 −2.073212.4381 −2.8950 −6.5749 0.5770 87.3544 1.23881.1549 −7.4955 −11.3553 −2.0732 1.2388 53.6202
M =
−0.1665 −0.0607 −0.3505 −0.3426 −0.2964 −0.80030.7573 −0.0185 −0.2809 −0.0707 0.5477 −0.20570.0852 −0.0124 −0.5787 −0.5200 −0.2718 0.55990.1822 −0.2983 0.6476 −0.6769 −0.0181 −0.00250.3097 0.8618 0.2054 −0.0910 −0.3277 −0.05940.5123 −0.4051 0.0432 0.3752 −0.6562 −0.0124
N =
−0.0215 0 0 0 0 00.0025 0.0025 0 0 0 00.0073 −0.0032 0.0002 0 0 0−0.0004 0.0001 −0.0001 0.0001 0 0−2.6908 −4.5215 0.0007 0.0157 −0.0532 0−3.7994 1.3809 −0.0068 −0.0084 0.0032 −0.0117
× 106
4.4. OUTPUT FEEDBACK H∞ CONTROL 131
Then, the dynamic output feedback controller is given as
AK =
−47.6189 −2.6575 50.7682 −46.8550 −8.5812 11.055821.1351 −0.0906 −19.8367 18.2366 3.4111 −4.3565−60.4355 20.3676 −0.3215 3.6266 −7.3594 9.7559−91.0058 11.1410 52.5115 −48.5904 −14.3507 18.62284.2809 9.5237 −34.0269 38.1094 −1.1880 −1.6500
135.8845 −14.4186 −85.6415 65.9191 22.4684 −37.1408
,
BK =
0.0245 0.0163 0.0106−0.0271 −0.0003 −0.0060−1.3555 −1.5639 0.0810−1.1833 −1.4937 0.07430.5561 −0.9351 0.0981−12.6807 3.7234 −2.3869
,
CK =[ −7.4312 −0.1929 7.9140 −7.3066 −1.3380 1.7214
]× 104,
DK =[ −13.7838 30.8852 7.8082
]
By the dynamic output feedback controller, the displacement responses of the
three floors for ω = 1.8 are depicted in Figures 4.7 - 4.9. (See the simulation diagram
in Appendix C.8.) It’s easy to see that the amplitudes of displacement reduce by
about 70% (compare with Figures 3.25 and 3.27). Moreover, as shown in Figures
4.10 - 4.12, the amplitudes of vibration for ω = 0.5773 are 15% cut off (compare
with Figures 3.28 and 3.30). Finally, we find that the controller can process the
internal oscillations of offshore steel jacket platform positively.
4.4 Output Feedback H∞ Control
In this section, we consider an H∞ performance of system (4.1). The system is
restated as
x(t) = (A + ∆A)x(t) + Bu(t) + Fg(x, t) + Bwω(x, t)
y(t) = Cx(t)
z(t) = Czx(t)
x(t) = φ(t), t ≤ 0
(4.44)
where z(t) ∈ Rt is the controlled output, w(t) ∈ Rl is the external disturbance,
which belongs to L2[0, +∞), Bw, and Cz are constant real matrices with appropriate
132 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
dimensions. Introducing a augmented vector
ϕ(t) := [xT (t) ξT (t)]T
then the closed-loop system combined by (4.44) and (4.5) is given by
ϕ(t) = (A0 + ∆A0)ϕ(t) + A1Eϕ(t− h)
+ ET Fg(x, t) + ET Bwω(x, t)
z(t) = CzEϕ(t)
(4.45)
where ∆A0, A0, A1, E are defined in (4.7).
The aim below is to seek an H∞ output feedback controller (4.5) such that
(1) The closed-loop system (4.45) is asymptotically stable for any admissible un-
certainty satisfying (4.3); and
(2) Under zero initial condition, ‖z(t)‖2 ≤ γ ‖ω(t)‖2 is guaranteed for all nonzero
ω(t) ∈ L2[0, +∞), where γ > 0 is a prescribed scalar.
For this purpose, firstly, we discuss the H∞ control for the normal case of system
(4.45), and the obtained results are then extended to the uncertain case. Suppose
∆A = 0, then the normal system can be expressed as
ϕ(t) = A0ϕ(t) + A1Eϕ(t− h(t)) + ET Fg(x, t)
+ ET Bwω(t),
z(t) = CzEϕ(t).
(4.46)
Now, choose a Lyapunov-Krasovskii functional candidate as
V (xt) = ϕT (t)Pϕ(t) +
∫ t
t−h
ϕT (s)ET QEϕ(s)ds
+ h
∫ 0
−h
ds
∫ t
t+s
ϕT (θ)ET REϕ(θ)dθ (4.47)
where P > 0, Q > 0, R > 0, then we have
Proposition 21. For some given positive scalars µ, γ and h, system (4.46) is
asymptotically stable and satisfies ‖z(t)‖2 ≤ γ ‖ω(t)‖2 for all nonzero ω(t) ∈ L2[0,∞),
4.4. OUTPUT FEEDBACK H∞ CONTROL 133
if there exist real matrices P > 0, Q > 0 and R > 0 of appropriate dimensions such
that
Ψ =
Ψ11 PA1 + ET R PET D PET Bw hAT0 ET
∗ −Q−R 0 0 hAT1 ET
∗ ∗ −I 0 hF T
∗ ∗ ∗ −γ2I hBTw
∗ ∗ ∗ ∗ −R−1
< 0 (4.48)
where
Ψ11 = AT0 P + PA0 + ET (Q−R)E + µ2ET E + ET CT
z CzE.
Proof. From Proposition 18, if (4.11) holds, (4.10) is asymptotically stable. By the
Schur complement, (4.48) implies (4.11). Therefore, system (4.46) is asymptotically
stable if (4.48) holds. In the sequel, we only prove that ‖z(t)‖2 ≤ γ ‖ω(t)‖2 is
guaranteed for all nonzero ω(t) ∈ L2[0,∞) if (4.48) is feasible. To see this, taking
the time derivative of V (xt) in (4.47) along the trajectory of (4.46) yields
V (xt) = 2ϕT (t)Pϕ(t) + ϕT (t)ET QEϕ(t)− ϕT (t− h)ET QEϕ(t− h)
+ h2ϕT (t)ET REϕ(t)− h
∫ t
t−h
ϕT (θ)ET REϕ(θ)dθ. (4.49)
Define
ηT (t) =[
ϕT (t) ϕT (t− h)ET gT (x, t) ωT (t)].
By applying Lemma 1 (see Appendix A), we have
V (xt)− γ2ωT (t)ω(t) + zT (t)z(t) ≤ ηT (t)[Ψ + h2ΓT RΓ]η(t), (4.50)
where
Ψ :=
Ψ11 PA1 + ET R PET D PET Bw
∗ −Q−R 0 0∗ ∗ −I 0∗ ∗ ∗ −γ2I
Γ := [EA0 EA1 F Bw]
If matrix inequality (4.48) holds, then Ψ + h2ΓT RΓ < 0 by the Schur complement,
thus
V (xt)− γ2ωT (t)ω(t) + zT (t)z(t) < 0. (4.51)
134 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
Integrating both sides of (4.51) from 0 to ∞ yields
∫ ∞
0
[zT (s)z(s)− γ2ωT (s)ω(s)]ds < V (xt)|t=0 − V (xt)|t=∞. (4.52)
Under the zero initial condition, we have
∫ ∞
0
[zT (s)z(s)− γ2ωT (s)ω(s)]ds < 0. (4.53)
That is, ‖z(t)‖2 ≤ γ ‖ω(t)‖2 . This completes the proof.
For the uncertain system (4.45), based on Proposition 21, we also have
Proposition 22. For some given positive scalars µ, γ and h, system (4.45) is
asymptotically stable and satisfies ‖z(t)‖2 ≤ γ ‖ω(t)‖2 for all nonzero ω(t) ∈ L2[0,∞),
if there exist λ > 0, real matrices P > 0, Q > 0, and R > 0 of appropriate dimen-
sions such that
Ψ11 PA1 + ET R PET D PET Bw hAT0 ET PET L λET ET
a
∗ −Q−R 0 0 hAT1 ET 0 0
∗ ∗ −I 0 hF T 0 0∗ ∗ ∗ −γ2I hBT
w 0 0∗ ∗ ∗ ∗ −R−1 L 0∗ ∗ ∗ ∗ ∗ −λI 0∗ ∗ ∗ ∗ ∗ ∗ −λI
< 0 (4.54)
where Ψ11 is defined in (21).
Similar to Proposition 20, the H∞ controller design based on Propositions 21
and 22 can be converted into nonlinear minimization problems. For instance, robust
H∞ output feedback controller can be obtained by solving the following nonlinear
minimization problem as
Minimize Tr(XX + RR + QQ + ZZ + SS)Subject to (4.22), (4.31), (4.32), Σ1 < 0, Σ2 < 0, and[
R II R
]≥ 0,
[X II X
]≥ 0,
[Q II Q
]≥ 0,
[Z II Z
]≥ 0,
[S II S
]≥ 0, ε > 0, λ > 0.
(4.55)
4.4. OUTPUT FEEDBACK H∞ CONTROL 135
where
Σ1 :=
ρ1 RW1 Y F Y Bw hAT R Y L λETa
∗ W T1 (−Q−R)W1 0 0 0 0 0
∗ ∗ −I 0 hF T R 0 0∗ ∗ ∗ −γ2I hBT
wR 0 0∗ ∗ ∗ ∗ −R L 0∗ ∗ ∗ ∗ ∗ −λI 0∗ ∗ ∗ ∗ ∗ ∗ −λI
Σ2 :=
ρ2 W T2 S ρ3 ρ4 ρ5 W T
2 XETa W T
2 X µW T2 X W T
2 CZ
∗ −Z − S 0 0 0 0 0 0 0∗ ∗ −I 0 0 0 0 0 0∗ ∗ ∗ −γ2I 0 0 0 0 0∗ ∗ ∗ ∗ −εI 0 0 0 0∗ ∗ ∗ ∗ ∗ −εI 0 0 0∗ ∗ ∗ ∗ ∗ ∗ −Q 0 0∗ ∗ ∗ ∗ ∗ ∗ ∗ −I 0∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ −I
with
ρ1 := Y A + AT Y + Q−R + µ2I,
ρ2 := W T2 [AX + XAT − S]W2 + W T
2 XAT W3 + W T3 AXW2 − h−2W T
3 RW3,
ρ3 := (W T2 + W T
3 )F,
ρ4 := (W T2 + W T
3 )Bw,
ρ5 := ε(W T2 + W T
3 )L.
The controller parameters can be obtained similar to that proposed in the pre-
vious section and thus it is omitted.
4.4.1 Simulation Results
To show the validity of the proposed approach, let h = 0.02, µ = 0.1, γ = 2, and
Bw =[
1 0 0.1 0 0 0]T
,
Cz =[
0 0.5 0 0 0 0],
L =[
0 0 0.001 0 0 0.01]T
,
Ea =[
1 0 0 0 0 0].
136 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
First floor − Robust Hinf
Output Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 4.13: The displacement of the first floor when the robust H∞ output feedbackcontroller is used and ω = 1.8
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
Second floor − Robust Hinf
Output Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 4.14: The displacement of the second floor when the robust H∞ outputfeedback controller is used and ω = 1.8
4.4. OUTPUT FEEDBACK H∞ CONTROL 137
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
Third floor − Robust Hinf
Output Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 4.15: The displacement of the third floor when the robust H∞ output feed-back controller is used and ω = 1.8
0 10 20 30 40 50 60 70 80 90 100
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
First floor − Robust Hinf
Output Feedback Control
Time(sec)
Firs
t Flo
or R
espo
nse(
ft)
Figure 4.16: The displacement of the first floor when the robust H∞ output feedbackcontroller is used and ω = 0.5773
138 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
Second floor − Robust Hinf
Output Feedback Control
Time(sec)
Sec
ond
Flo
or R
espo
nse(
ft)
Figure 4.17: The displacement of the second floor when the robust H∞ outputfeedback controller is used and ω = 0.5773
0 10 20 30 40 50 60 70 80 90 100−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
Third floor − Robust Hinf
Output Feedback Control
Time(sec)
Thi
rd F
loor
Res
pons
e(ft)
Figure 4.18: The displacement of the third floor when the robust H∞ output feed-back controller is used and ω = 0.5773
4.4. OUTPUT FEEDBACK H∞ CONTROL 139
Figures 3.73 - 3.78 have showed the responses of all the states of the uncertain system
under the external disturbance (no control). In order to weaken the passive effects of
the uncertainty and the external disturbance, and constrain the internal oscillations
inducd by the wave forces, we design a robust H∞ output feedback controller under
the assumption that the outputs are adopted as feedback. By solving the nonlinear
minimization problem (4.55), and we obtain the following values:
X =
0.8920 −0.2880 0.0132 −0.3571 −0.0960 −0.6045−0.2880 6.4149 0.4542 0.0103 0.4285 2.35980.0132 0.4542 0.3101 −0.1612 0.0227 0.2586−0.3571 0.0103 −0.1612 35.1609 −0.2893 0.1598−0.0960 0.4285 0.0227 −0.2893 1.3604 −0.3514−0.6045 2.3598 0.2586 0.1598 −0.3514 2.3298
Y =
0.0001 −0.0000 −0.0000 0.0000 0.0049 0.0047−0.0000 0.0000 −0.0000 −0.0000 −0.0019 −0.0001−0.0000 −0.0000 0.0004 −0.0000 −0.0002 −0.00220.0000 −0.0000 −0.0000 0.0000 0.0001 0.00010.0049 −0.0019 −0.0002 0.0001 3.1825 0.22320.0047 −0.0001 −0.0022 0.0001 0.2232 1.0495
× 105
Q =
1.4201 −0.1067 −0.6063 −0.0118 0.5036 0.8241−0.1067 0.2436 −0.1599 −0.0211 −0.1548 −0.1000−0.6063 −0.1599 4.2173 0.0508 −0.2548 −0.7412−0.0118 −0.0211 0.0508 0.0348 0.0003 −0.04300.5036 −0.1548 −0.2548 0.0003 0.8050 0.44810.8241 −0.1000 −0.7412 −0.0430 0.4481 1.3537
R =
95.5147 −5.7053 −10.1872 0.7876 20.6250 5.8309−5.7053 23.5080 −5.2679 −1.3370 −4.2067 −7.4061−10.1872 −5.2679 39.7708 0.2020 −5.9512 −11.01280.7876 −1.3370 0.2020 1.3087 0.7261 −1.557920.6250 −4.2067 −5.9512 0.7261 88.8340 0.90665.8309 −7.4061 −11.0128 −1.5579 0.9066 52.7852
M =
−0.1896 0.0095 −0.3789 −0.2279 0.1773 0.85850.7648 −0.0119 −0.2059 −0.1664 −0.5674 0.15120.0770 −0.0151 −0.6787 −0.4336 0.3522 −0.4702−0.0382 −0.1938 0.5269 −0.8266 0.0054 0.00580.2963 0.8609 0.2071 −0.0809 0.3445 0.05470.5329 −0.4700 0.1814 0.2062 0.6354 0.1264
140 CHAPTER 4. DYNAMIC OUTPUT FEEDBACK CONTROL
N =
−0.0171 0 0 0 0 00.0014 0.0025 0 0 0 00.0067 −0.0029 0.0001 0 0 0−0.0004 0.0001 −0.0001 0.0001 0 0−2.4977 −4.0859 0.0044 0.0075 0.0225 0−3.3454 1.2286 −0.0016 −0.0065 −0.0032 0.0048
× 105
and the H∞ controller is given by
AK =
−55.1086 4.5678 31.6434 −39.9572 8.6959 −10.695925.0105 −2.9557 −13.0193 16.3788 −3.6237 4.4169−71.1069 25.0372 −6.3232 18.0134 10.9393 −13.0617−61.1169 13.9139 13.1385 −14.4612 9.7647 −11.756029.4671 −11.3510 13.1744 −30.4525 −9.7332 3.7473−176.8769 8.5685 96.8820 −84.4714 37.1686 −62.9122
BK =
0.0697 0.0159 0.0205−0.0337 −0.0023 −0.0082−2.3202 −1.9839 −0.1448−1.3075 −1.2734 −0.08151.9676 1.3790 0.473444.3579 −7.3372 9.3754
CK =[ −9.4697 1.0410 5.4371 −6.8718 1.4942 −1.8343
]× 104
DK =[
25.1348 37.7031 15.4304]
See the simulation diagram in Appendix C.9. Figures 4.13 - 4.18 show the state
responses of the controlled system. From Figures 4.13 - 4.15, for ω = 1.8, the
displacements have decreased by about 80% (See Figures 3.73 - 3.75). Compared
Figures 4.16 - 4.18 with Figures 3.76 - 3.78, the amplitudes of vibration for ω =
0.5773 achieve about 35% decrease. All these have shown that the proposed H∞
controller can effectively control the offshore structure system.
4.5 Conclusion
The design problem of dynamic output feedback control for the offshore steel jacket
platform under wave excitations has been studied. Based on Lyapunov-Krasovskii
stability theory and a project theorem, some sufficient conditions for the existence
of dynamic output feedback controllers has been obtained in terms of a set of linear
4.5. CONCLUSION 141
matrix inequalities. Furthermore, robust H∞ performance for the system with norm-
bounded uncertainties has also been investigated. Finally, the simulation results has
indicated that the proposed control schemes are capable of significantly reducing the
internal oscillations of the offshore steel jacket platform.
Chapter 5
Conclusion and Future Work
Conclusion
The control problem of offshore steel jacket platforms has been studied in this
thesis. In order to ensure the safety and comfort of the offshore structure, a mem-
ory control strategy has been proposed for the first time to reduce the internal
oscillations.
In Chapter 3, when the system states are adopted as feedback, we have designed
a memory state feedback controller. By using Lyapunov-Krasovskii functional the-
ory, some delay-dependent stability criteria have been first established, based on
which, and by combining with some linearization techniques, two controller design
approaches have then been given, which only depend on the solutions of a set of
LMIs. The simulation results have shown that such a controller can effectively reduce
the internal oscillations of the offshore structure subject to nonlinear wave-induced
forces.
In Chapter 4, when the system outputs are considered as feedback, a memory
dynamic output feedback controller has been addressed. By employing a projection
theorem and a cone complementary linearization approach, a memory dynamic out-
put feedback controller has been derived by solving a nonlinear minimization subject
to a set of LMIs. The obtained controller has been applied to an offshore structure
subject to nonlinear wave-induced forces, and the internal oscillations have been
well attenuated.
144 CHAPTER 5. CONCLUSION AND FUTURE WORK
In addition, an H∞ control for the offshore structure has been investigated for
the first time. By employing Lyapunov-Krasovskii theory, some delay-dependent
bounded real lemmas have been obtained, under which, via a memory H∞ controller,
the resulting closed-loop system is not only asymptotically stable but also with a
prescribed disturbance attenuation level. The simulation results have shown the
validity of the proposed method.
On the other hand, we proposed a new Lyapunov-Krasovskii functional to study
the stability for systems with time delay. With its help, a new delay-dependent
stability criterion has been obtained, which has been shown to be less conservative
than those in the literature. Furthermore, based on this criterion, a memory state
feedback controller for the offshore structure has also been designed. It has been
found that the controller, which is of less gain, can effectively ensure safety and
comfort of the offshore structure.
Future work
Accompanied by the sharply increase of energy need, offshore structures are de-
veloping into deeper and deeper ocean, which have to meet steeply growing menaces,
such as wave, wind, current, earthquake, etc. How these adverse factors effect the
offshore deep-water platforms needs to be further investigated in the future.
In this thesis, memory controller design is proposed to reduce the internal os-
cillations of an offshore structure. However, offshore structures are only a small
piece of a cake in structural vibration control area. How to extend this line to other
civil structures to constrain server vibration induced by environmental loads is a
challenging topic in our future work.
Finally, as we know, network based control has emerged as a topic of significant
interest in the control community in the recent years due to its more advantages,
such as low cost, high reliability and simple installation, etc. How to introduce the
network to an offshore structure and how to more effectively control this structure
145
are the main aim to study in the future.
Appendix A
Lemmas Referred
Lemma 1. [27] For any constant matrix W ∈ Rn×n,W = W T > 0, scalar γ > 0,
and vector function x : [−γ, 0] → Rn such that the following integration is well
defined, then
−γ
∫ 0
−γ
xT (t + ξ)Wx(t + ξ)dξ (A.1)
≤ (xT (t) xT (t− γ)
) ( −W WW −W
)(x(t)
x(t− γ)
). (A.2)
Lemma 2. [63]Given matrices Q = QT , H, E and R = RT > 0 of appropriate
dimensions,
Q + HFE + ET F T HT < 0
for all F satisfying F T F ≤ R, if and only if there exists some λ > 0 such that
Q + λHHT + λ−1ET RE < 0.
Lemma 3. [69]Let x(t) ∈ Rn be a vector-valued function with first-order continuous-
derivative entries. Then, the following integral inequality holds for any matrices X,
M1, M2 ∈ Rn×n, and Z ∈ R2n×2n, and a scalar function h := h(t) ≥ 0:
−∫ t
t−h
xT (s)Xx(s)ds ≤ ξT (t)Υξ(t) + hξT (t)Zξ(t)
where
Υ :=
[MT
1 + M1 −MT1 + M2
∗ −MT2 −M2
],
148 APPENDIX A. LEMMAS REFERRED
ξ(t) :=
[x(t)
x(t− h)
],
[X Y∗ Z
]≥ 0
with
Y :=[
M1 M2
].
Lemma 4. [14] Consider a symmetric matrix Ξ ∈ Rn×n and two matrices Π and Γ
with column dimension n. Then there exists a matrix Θ of compatible dimensions
such that
Ξ + ΠT ΘΓ + ΓT ΘT Π < 0
if and only if (Π⊥)T ΞΠ⊥ < 0 and (Γ⊥)T ΞΓ⊥ < 0, where Π⊥ and Θ⊥ denote the
orthogonal complements of Π and Θ, respectively.
Appendix B
Useful Theories
B.1 Stability of Time-Delay Systems
This part is borrowed from [19]. A time-delay system can be described by a retarded
functional differential equation as:
x(t) = f(t, xt), (B.1)
with
x(t) = φ(t), t ∈ [t0 − r, t0
], (B.2)
where x(t) ∈ Rn, f : R×C→ Rn and φ(·) is the initial condition function of (B.1).
x(θ) = x(t+θ), −r ≤ θ ≤ 0. (B.1) indicates that the derivative of the state variables
x at time t depends on t and x(ξ) for t− r ≤ ξ ≤ t.
B.1.1 Stability Concept:
For the system described by (B.1), the trivial solution x(t) = 0 is said to be stable
if for any t0 ∈ R and any ε > 0, there exists a δ = δ(t0, ε) > 0 such that ‖xt0‖c < δ
implies ‖x(t)‖ < ε for t ≥ t0.
• It is said to be asymptotically stable if it is stable, and for any t0 ∈ R and
any ε > 0, there exists a δa = δa(t0, ε) > 0 such that ‖xt0‖c < δa implies
limt→∞
x(t) = 0.
150 APPENDIX B. USEFUL THEORIES
• It is said to be uniformly stable if it is stable and δ = δ(t0, ε) can be chosen
independently of t0.
• It is uniformly asymptotically stable if it is uniformly stable and there exists
a δa > 0 such that for any η > 0, there exists a T = T (δa, η), such that
‖xt0‖c < δ implies ‖x(t)‖ < η for t ≥ t0 + T and t0 ∈ R.
• It is globally (uniformly) asymptotically stable if it is (uniformly) asymptoti-
cally stable and δa can be an arbitrarily large, finite number.
B.1.2 Lyapunov-Krasovskii Stability Theorem:
Suppose f : R × C → Rn in (B.1) maps (bounded sets in C) into a bounded sets
in Rn, and that u, v, w are continuous nondecreasing functions, where additionally
u(s) and v(s) are positive for s > 0, and u(0) = v(0) = 0. If there exists a continuous
differentiable functional V : R× C→ Rn such that
u(‖φ(0)‖) ≤ V (t, φ) ≤ v (‖φ(0)‖c) ,
and
V (t, φ) ≤ −w(‖φ(0)‖),
then the trivial solution of (B.1) is uniformly stable. If w(s) > 0 for s > 0, then it
is uniformly asymptotically stable. If, in addition, lims→∞
u(s) = ∞, then it is globally
uniformly asymptotically stable.
B.2 Linear Matrix Inequalities
This part is borrowed from [3]. A strict linear matrix inequality (LMI) has the
general form of
F (x) , F0 +m∑
i=1
xiFi > 0, (B.3)
where x =(
x1 x2 · · · xm
) ∈ Rm is a vector consisting of m variable, and the
symmetric matrices Fi = F Ti ∈ Rn, i = 0, 1, . . . , m are m+1 given constant matrices.
B.3. THE LMI TOOLBOX OF MATLAB 151
An LMI may also be nonstrict, where ” > ” is replaced by ” ≥ ”. Notice that the
variables appear linearly on the left hand side of the inequality. The basic LMI
problem is to find whether or not there exists an x ∈ Rm such that (B.3) is satisfied.
An LMI is an affine inequality constraint on the design variables. Specifications
such as regional pole placement, robust stability, LQG performance, or RMS gain
attenuation can be expressed as LMIs.
In many practical problems, the parameters may appear nonlinearly in their most
natural form. So some important techniques can be applied to transform them into
an LMI form. The popular one is as follows.
Schur Complement: For matrices A, B, C, the inequality
[A BBT C
]> 0
is equivalent to the following two inequalities
A > 0,
C −BT A−1B > 0.
B.3 The LMI Toolbox of Matlab
This part is borrowed from [15]. Commercial software of the LMI control toolbox is
a set of tools for use with Matlab. It provides state-of-the-art optimization routines
to solve linear matrix inequalities. It also includes specialized tools for LMI-based
analysis and design of control systems.
In the past few years, LMI solvers have emerged as powerful tools to solve convex
optimization problems that arise in many analysis and design applications, such as
control, identification, filtering, and structural design. While the toolbox emphasis
is on control, its LMI capabilities make it useful in any area where LMI techniques
are applicable.
The toolbox provides a fully integrated general-purpose environment for specify-
152 APPENDIX B. USEFUL THEORIES
ing and solving LMI problems. It offers a powerful and user-friendly environment for
developing LMI-based applications. The toolbox uses a structured representation
of LMI constraints that boosts efficiency and minimizes memory requirements.
The commands or functions which are often used in the research include:
• Setlimis: initialize the LMI description.
• Lmivar: define a new matrix variable.
• Lmiterm: specify the term content of an LMI.
• Feasp: compute a solution to the feasibility problem.
• Mincx: compute a solution to the optimization problem.
• Gevp: generalize eigenvalue minimization under LMI constraints.
Example 2. Find a symmetric matrix P satisfying the LMIs
AT1 P + PA1 < 0,
AT2 P + PA2 < 0,
AT3 P + PA3 < 0,
P > I. (B.4)
where
A1 =
[ −1 21 −3
], A2 =
[ −0.8 1.51.3 −2.7
], A3 =
[ −1.4 0.90.7 −2
].
The Matlab program can be written as follows to solve this question.
setlmis([])
P=lmivar(1,[2 1])
lmiterm([1 1 1 P],1,A1,’s’) % LMI #1
lmiterm([2 1 1 P],1,A2,’s’) % LMI #2
B.4. AN H∞ CONTROL 153
lmiterm([3 1 1 P],1,A3,’s’) % LMI #3
lmiterm([-4 1 1 P],1,1) % LMI #4: P
lmiterm([4 1 1 0],1) % LMI #4: I
lmis=getlmis
[tmin, xfeas]=feasp (lmis)
Though running this program, if it can be found that a tmin < 0, the LMIs (B.4)
are solvable. The answer is P =
[270.8 126.4126.4 155.1
].
B.4 An H∞ Control
H∞-norm:
‖G‖∞ , supω
σ(G(jω))
sup : the least upper bound
σ : the greatest singular value
H∞ Control Problem
This part is borrowed from [70]. Consider a system described by the block
diagram
The plant G and controller K are assumed to be real-rational and proper. It
will be assumed that state space models of G and K are available and that their
154 APPENDIX B. USEFUL THEORIES
realizations are assumed to be stabilizable and detectable. z, w, u, v are controlled
output, external disturbance, output vector, control vector, respectively.
Optimal Control: find all admissible controllers K(s) such that the close-loop
transfer function Twz(s) from the external disturbance w to controlled output z
satisfies that ‖Twz(s)‖∞ is minimized.
However, in practice it is often not necessary and sometimes even undesirable
to design an optimal controller, and it is usually much cheaper to obtain controllers
that are very close in the norm sense to the optimal ones, which will be called
suboptimal controllers.
Suboptimal H∞ Control: Given γ > 0, find all admissible controllers K(s),
if there are any, such that ‖Twz(s)‖∞ < γ.
Appendix C
Simulation Diagrams
C.1 The Simulation Diagram of the Controller
Design from Proposition 3
Figure C.1: The simulation diagram of the Controller Design from Proposition 3
In Figure C.1, Embedded MATLAB Function fct, dct is defined as follows:
function ftt = fct(u)
t=u(1);
x=u(2:7);
w=1.8;
if t==0
156 APPENDIX C. SIMULATION DIAGRAMS
ali=0;
a2i=0;
else
ali=x(2);
a2i=x(4);
end c2=1.694*cos(-w*t)-(-0.30419e-2*ali-0.44904e-2*a2i);
c3=4.25*cos(-w*t)+0.24-(-0.32902e-2*ali-0.15346e-2*a2i);
c4=11.67*cos(-w*t)+0.4-(-0.3445e-2*ali+0.34628e-2*a2i);
c6=1.694*cos(1.5708-w*t)-(-0.30419e-2*ali-0.44904e-2*a2i);
c7=4.25*cos(1.5708-w*t)+0.24-(-0.32902e-2*ali-0.15346e-2*a2i);
c8=11.67*cos(1.5708-w*t)+0.4-(-0.3445e-2*ali+0.34628e-2*a2i);
f2=808.071*abs(c2)*c2+6799.929*sin(-w*t);
f3=1247.282*abs(c3)*c3+26899.393*sin(-w*t);
f4=439.214*abs(c4)*c4+27017.722*sin(-w*t);
f6=808.071*abs(c6)*c6+6799.929*sin(1.5708-w*t);
f7=1247.282*abs(c7)*c7+26899.393*sin(1.5708-w*t);
f8=439.214*abs(c8)*c8+27017.722*sin(1.5708-w*t);
ft1=-0.30419e-2*f2-0.329e-2*f3-0.3445e-2*f4;
ft2=-0.30419e-2*f6-0.329e-2*f7-0.3445e-2*f8;
ft3=-0.44904e-2*f2-0.15346e-2*f3+0.34628e-2*f4;
ft4=-0.44904e-2*f6-0.15346e-2*f7+0.34628e-2*f8;
ftt=[0;ft1+ft2;0;ft3+ft4;0;0];% wave frequency
%end fct
function dtt = dct(u)
x=u(1:6); dt1=-0.30419e-2*x(1)-0.44904e-2*x(3);
dt2=-0.32902e-2*x(1)-0.15346e-2*x(3);
dt3=-0.3445e-2*x(1)+0.34628e-2*x(3); dtt=[dt1;dt2;dt3];
C.2. THE SIMULATION DIAGRAM OF THE CONTROLLER DESIGN FROMPROPOSITION 7 157
%end dct
Note: The definition of function fct, dct are also used in the later simulations
throughout this thesis.
C.2 The Simulation Diagram of the Controller
Design from Proposition 7
Figure C.2: The simulation diagram of the Controller Design from Proposition 7
In Figure C.2, Embedded MATLAB Function fcn is defined as follows:
function y = fcn(u,v,Gt)
A=[ 0 1 0 0 0 0
-3.3235 -0.0212 0.0184 0.0030 -5.3449 -0.8819
0 0 0 1 0 0
0.0184 0.0030 -118.1385 -0.1117 5.3468 0.8822
0 0 0 0 0 1
-0.0114 -0.0019 0.0114 0.0019 -3.3051 -0.5454 ];
B=[0;0.003445; 0;-0.00344628;0;0.00213];
158 APPENDIX C. SIMULATION DIAGRAMS
L=[0;0;0.001;0;0;0.01];
Ea=[1,0,0,0,0,0];
A2=L*Gt*Ea; %the norm-bounded uncertainties
A3=(A+A2);
K1 =1.0e+004 *[-0.2555 -0.1584 1.6036 0.1897 0.1050 -0.1216];
K2 =1.0e+003 *[0.3045 0.0542 -1.4487 0.0115 0.0394 0.1280]; %controller gains
y=(A3+B*K1)*u+B*K2*v;
G(t) is used to output a uniformly distributed random signal ranging between
-10 and 10.
Embedded MATLAB Functions fct, dct are the same as those described in Ap-
pendix C.1.
C.3 The Simulation Diagram of the Controller
Design from Proposition 10
See Figure C.2. Embedded MATLAB Function fcn in the Controller Design from
Proposition 10 is defined as follows:
function y = fcn(u,v,Gt)
A=[ 0 1 0 0 0 0
-3.3235 -0.0212 0.0184 0.0030 -5.3449 -0.8819
0 0 0 1 0 0
0.0184 0.0030 -118.1385 -0.1117 5.3468 0.8822
0 0 0 0 0 1
-0.0114 -0.0019 0.0114 0.0019 -3.3051 -0.5454 ];
B=[0;0.003445; 0;-0.00344628;0;0.00213];
L = [0;0;3;0;0;0];
Ea=[1,0,0,0,0,0];
L1 = [0;0;0;0;5;0];
C.4. THE SIMULATION DIAGRAM OF THE CONTROLLER DESIGN FROMPROPOSITION 12 159
Ea1=[0,1,0,0,0,0];
A2=L*Gt*Ea;
A3=L1*Gt*Ea1; %the polytopic uncertainties
K1 =1.0e+004 *[-0.2266 -0.0659 1.1847 0.2477 0.0463 -0.1982];
K2 =[-18.5225 -12.7426 169.8950 16.7815 -6.8169 -19.5826]; % h=0.1,
mu=0.1;contrller gains
y=(A+A2+A3+B*K1)*u+B*K2*v; %end fcn
G(t) is used to output a uniformly distributed random signal ranging between
-0.03 and 0.03.
Embedded MATLAB Functions fct, dct are the same as those described in Ap-
pendix C.1.
C.4 The Simulation Diagram of the Controller
Design from Proposition 12
Figure C.3: The simulation diagram of the Controller Design from Proposition 12
See Figure C.3, where w(t) is used to output a uniformly distributed random
signal ranging between -100 and 100, which is treated as the external disturbance.
160 APPENDIX C. SIMULATION DIAGRAMS
Embedded MATLAB Functions fct, dct are the same as those described in Ap-
pendix C.1.
C.5 The Simulation Diagram of the Controller
Design from Proposition 15
Figure C.4: The simulation diagram of the Controller Design from Proposition 15
The simulation diagram is depicted in Figure C.4, where Embedded MATLAB
Function fcn is defined as follows:
function y = fcn(u,v,Gt)
A=[ 0 1 0 0 0 0
-3.3235 -0.0212 0.0184 0.0030 -5.3449 -0.8819
0 0 0 1 0 0
0.0184 0.0030 -118.1385 -0.1117 5.3468 0.8822
0 0 0 0 0 1
-0.0114 -0.0019 0.0114 0.0019 -3.3051 -0.5454 ];
B=[0;0.003445; 0;-0.00344628;0;0.00213];
L=[0;0;0.001;0;0;0.01];
C.6. THE SIMULATION DIAGRAM OF THE CONTROLLER DESIGN FROMPROPOSITION 17 161
Ea=[1,0,0,0,0,0];
A2=L*Gt*Ea; %the norm-bounded uncertainties
A3=(A+A2);
K1 =1.0e+004 *[-0.1941 -0.1424 1.9398 0.3026 0.1145 -0.0999];
K2 =1.0e+003 *[0.2640 0.0649 -2.0823 -0.0230 0.0326 0.1235]; % h=0.1,
mu=0.1;contrller gains
y=(A3+B*K1)*u+B*K2*v; %end fcn
w(t) is used to output a uniformly distributed random signal ranging between
-100 and 100, which is treated as the external disturbance; G(t) is used to output a
uniformly distributed random signal ranging between -10 and 10.
Embedded MATLAB Functions fct, dct are the same as those described in Ap-
pendix C.1.
C.6 The Simulation Diagram of the Controller
Design from Proposition 17
See Figure C.4, where Embedded MATLAB Function fcn in the Controller Design
from Proposition 17 is described as follows:
function y = fcn(u,v,Gt)
A=[ 0 1 0 0 0 0
-3.3235 -0.0212 0.0184 0.0030 -5.3449 -0.8819
0 0 0 1 0 0
0.0184 0.0030 -118.1385 -0.1117 5.3468 0.8822
0 0 0 0 0 1
-0.0114 -0.0019 0.0114 0.0019 -3.3051 -0.5454 ];
B=[0;0.003445; 0;-0.00344628;0;0.00213];
L = [0;0;3;0;0;0];
Ea=[1,0,0,0,0,0];
162 APPENDIX C. SIMULATION DIAGRAMS
L1 = [0;0;0;0;5;0];
Ea1=[0,1,0,0,0,0];
A2=L*Gt*Ea;
A3=L1*Gt*Ea1; %the polytopic uncertainties
K1 =1.0e+004 *[-7.1284 -0.9221 8.8308 -0.1684 -0.6949 -2.7038];
K2 =[-247.1548 -79.4067 534.3352 -6.2850 -29.7022 -116.5531]; % h=0.1,
mu=0.1;contrller gains
y=(A+A2+A3+B*K1)*u+B*K2*v; %end fcn
w(t) is used to output a uniformly distributed random signal ranging between
-100 and 100, which is treated as the external disturbance; G(t) is used to output a
uniformly distributed random signal ranging between -0.03 and 0.03.
Embedded MATLAB Functions fct, dct are the same as those described in Ap-
pendix C.1.
C.7 The Simulation Diagram of the Output Feed-
back Control
The simulation diagram is depicted in Figure C.5.
C.8 The Simulation Diagram of the Robust Out-
put Feedback Control
The simulation diagram is depicted in Figure C.6, where Embedded MATLAB Func-
tion fcn is defined as follows:
function y = fcn(u,Ft)
A=[ 0 1 0 0 0 0
-3.3235 -0.0212 0.0184 0.0030 -5.3449 -0.8819
0 0 0 1 0 0
C.8. THE SIMULATION DIAGRAM OF THE ROBUST OUTPUTFEEDBACK CONTROL 163
Figure C.5: The simulation diagram of the output feedback control
Figure C.6: The simulation diagram of the robust output feedback control
164 APPENDIX C. SIMULATION DIAGRAMS
0.0184 0.0030 -118.1385 -0.1117 5.3468 0.8822
0 0 0 0 0 1
-0.0114 -0.0019 0.0114 0.0019 -3.3051 -0.5454 ];
L=[0;0;0.001;0;0;0.01];
Ea=[1,0,0,0,0,0];
A2=L*Ft*Ea; %the norm-bounded uncertainties
y=(A+A2)*u;%end fcn
F(t) is used to output a uniformly distributed random signal ranging between -10
and 10.
Embedded MATLAB Functions fct, dct are the same as those described in Ap-
pendix C.1.
C.9 The Simulation Diagram of the Output Feed-
back H∞ Control
Figure C.7: The simulation diagram of the output feedback H∞ control
The simulation diagram is shown in Figure C.7, where Embedded MATLAB
Functions fct, dct are the same as those described in Appendix C.1, and Embedded
C.9. THE SIMULATION DIAGRAM OF THE OUTPUT FEEDBACK H∞CONTROL 165
MATLAB Functions fcn is the same as that described in Appendix C.8. And w(t)
is used to output a uniformly distributed random signal ranging between -100 and
100. F(t) is used to output a uniformly distributed random signal ranging between
-10 and 10.
References
[1] M. Abdel-Rohman, Structural control of a steel jacket platform, Structural
Engineering and Mechanics, 4, (1996) 125-138.
[2] U. Baser and B. Kizilsac, Dynamic Output Feedback H∞ Control Problem for
Linear Neutral Systems, IEEE Transactions on Automatic Control, 52, (2007)
1113 - 1118.
[3] S. Boyd, L. E. Ghaoui, E. Feron and V. Balakrishnan, Linear Matrix Inequali-
ties in System and Control Theory. Society for Industrial and Applied Mathe-
matics, Philadelphia; 1994.
[4] L. El Chaoui, F. Oustry and M. AitRami, A cone complementarity linearization
algorithm for static output-feedabck and related problems, IEEE Transactions
on Automatic Control, 42, (1997) 1171-1176.
[5] W.-H. Chen and W. X. Zheng, On improved robust stabilization of uncertain
systems with unknown input delay, Automatica, 42, (2006) 1067-1072.
[6] J.-D. Chen, Robust H∞ output dynamic observer-based control of uncertain
time-delay systems, Chaos, Solitons and Fractals, 31, (2007) 391-403.
[7] H. H. Choi and M. J. Chung, LMI approach to H∞ controller design for linear
time-delay systems, Automatica, 33, (1997) 737-739.
168 REFERENCES
[8] E. Fridman, New Lyapunov-Krasovskii functionals for stability of linear re-
tarded and neutral type systems, Systems and Control Letters, 43, (2001) 309-
319.
[9] E. Fridman and U. Shaked, An improved stabilization method for linear time-
delay systems, IEEE Transactions on Automatic Control, 47, (2002) 1931-1937.
[10] E. Fridman and U. Shaked, Delay-dependent stability and H∞ control: constant
and time-varying delays, International Journal of Control, 76 (2003) 48-60.
[11] E. Fridman and U. Shaked, Parameter dependent stability and stablization of
uncertain time-delay systems, IEEE Transactions on Automatic Control, 48,
(2003) 861-866.
[12] E. Fridman, A. Seuret and J.-P. Richard, Robust sampled-data stabilization of
linear systems: an input delay approach, Automatica, 40, (2004) 1441-1446.
[13] Y. S. Fu, Z. H. Tian and S. J. Shi, Output feedback stabilization for a class
of stochastic time-delay nonlinear systems, IEEE Transactions on Automatic
Control, 50, (2005) 847-851.
[14] P. Gahinet and P. Apkarian, A linear matrix inequality approach to H∞ control,
International Journal of Robust and Nonlinear Control, 4, (1994) 421-448.
[15] P. Gahinet, A. Nemirovski, A. J. Laub, and M. Chilali, LMI Control Toolbox
for Use with MATLAB. The MathWorks, Inc. 3 Apple Hill Drive, Natick, MA;
1995.
[16] H. J. Gao, C. H. Wang and L. Zhao, Comments on ”An LMI-Based Approach
for Robust Stabilization of Uncertain Stochastic Systems with Time-Varying
Delays”, IEEE Transactions on Automatic Control, 48, (2003) 2073-2074.
REFERENCES 169
[17] K. Gu, Discretized LMI set in the stability problem of linear uncertain time-
delay systems. International Journal of Control, 68, (1997) 923-934.
[18] K. Gu, A further refinement of discretized Lyapunov functional method for the
stability of time-delay systems. International Journal of Control, 74, (2001)
967-976.
[19] K. Gu, V.L. Kharitonov and J. Chen, Stability of time-delay systems. Boston:
Birkhauser, Boston, USA.
[20] X.-P. Guan and C.-L. Chen, Delay-dependent guaranteed cost control for TCS
fuzzy systems with time-delays, IEEE Transactions Fuzzy System, 12, (2004)
236-249.
[21] Q.-L. Han, On delay-dependent stability for neutral delay-differential sys-
tems, International Journal of Applied Mathematics and Computer Science,
11, (2001) 965-976.
[22] Q.-L. Han, Robust stability of uncertain delay-differential systems of neutral
type, Automatica, 38, (2002) 719-723.
[23] Q.-L. Han, On robust stability of neutral systems with time-varying discrete
delay and norm-bounded uncertainty, Automatica, 40, (2004) 1087-1092.
[24] Q.-L. Han, A descriptor system approach to robust stability of uncertain neutral
systems with discrete and distributed delays, Automatica, 40, (2004) 1791-1796.
[25] Q.-L. Han, X. Yu and K. Gu, On computing the maximum time-delay bound for
stability of linear neutral systems, IEEE Transactions on Automatic Control,
49, (2004) 2281-2286.
[26] Q.-L. Han, On stability of linear neutral systems with mixed time-delays: a
discretized Lyapunov functional approach, Automatica, 41, (2005) 1209-1218.
170 REFERENCES
[27] Q.-L. Han, Absolute stability of time-delay systems with sector-bounded non-
linearity, Automatica, 41, (2005) 2171-2176.
[28] Y. He, M. Wu, J.-H. She, and G.-P. Liu, Parameter-dependent Lypunov func-
tional for stability of time-delay systems with polytopic-type uncertainties,
IEEE Transactions on Automatic Control, 49, (2004) 828-832.
[29] L.-R. Jesus and P. Allan E., Output feedback stabilizing controller for time-
delay systems, Automatica, 36, (2000) 613-617.
[30] E. T. Jeung, J. H. Kim and H. B. Park, H∞-output feedback controller de-
sign for linear systems with time-varying delayed state, IEEE Transactions on
Automatic Control, 43, (1998) 971-974.
[31] X. Jiang and Q.-L. Han, Delay-dependent robust stability for uncertain linear
systems with interval time-varying delay, Automatica, 41, (2006) 1059-1065.
[32] I. N. Kar, K. Seto and F. Foi, Multimode vibration control of a flexible structure
using H∞-base robust control, IEEE/ASME Transactions on Mechatronics, 5,
(2000) 23-31.
[33] H. Katayama and A. Ichikawa, H∞ control for discrete-time Takagi-Sugeno
fuzzy systems, International Journal of System Science, 33, (2002) 1099-1107.
[34] K. Kawano, Active control effects on dynamic response of offshore structures,
Proc. of the Third (1993) Int. Offshore and Polar Engineering Conf., (1993)
594-598.
[35] K. Kawano and K. Venkataramana, Seismic responds of offshore platform with
TMD, Proc. of 3rd ISOPE Conf. on Earthquake Engineering, 4, (1992) 2241-
2246.
REFERENCES 171
[36] S.-H. Song, J.-K. Kim, C.-H. Yim and H.-C. Kim, H∞ control of discrete-time
linear systems with time-varying delays in state, Automatica, 35, (1999) 1587-
1591.
[37] S. H. Kim, S. B. Choi, S. R. Hong and M. S. Han, Vibration control of a flexible
structure using a hybrid mount, International Journal of Mechanical Sciences,
46, (2004) 143-157.
[38] D. K. Kim, P. G. Park and J. W. Ko, Output-feedback H∞ control of systems
over communication networks using a deterministic switching system approach,
Automatica, 40, (2004) 1205-1212.
[39] O. M. Kwon and J. H. Park, On improved delay-dependent robust control for
uncertain time-delay systems, IEEE Transactions on Automatic Control, 49,
(2004) 1991-1995.
[40] V. Kolmanovskii and J. P. Richard, Stability of some linear systems with delays,
IEEE Transactions on Automatic Control, 44, (1999) 984-989.
[41] B. Lee and J. G. Lee, Robust stability and stabilization of linear delayed systems
with structured uncertainty, Automatica, 35, (1999) 1149-1154
[42] Y. S. Lee, Y. S. Moon, W. H. Kwon and P. G. Park, Delay-dependent robust
H∞ control for uncertain systems with a state delay, Automatica, 40, (2004)
65-72.
[43] Y. S. Lee, Y. S. Moon, W. H. Kwon and P. G. Park, Delay-dependent robust
H∞ control for uncertain systems with a state-delay, Automatica, 40, (2003)
65-72 .
[44] H. H. Lee, S.-H. Wong and R.-S. Lee, Response mitigation on the offshore
floating platform system with tuned liquid column damper, Ocean Engineering,
33, (2006) 1118-1142.
172 REFERENCES
[45] H. J. Li, S. Hu and C. Jakubiak, Active vibration control for offshore platform
subjected to wave loading, Journal of Sound and Vibration, 263, (2003) 709-
724.
[46] C. Lin, Q.-G. Wang and T. H. Lee, A less conservative robust stability test for
linear uncertain time-delay systems, IEEE Transactions on Automatic Control,
51, (2006) 87-91.
[47] M. Luo and W. Q. Zhu, Nonlinear stochastic optimal control of offshore plat-
forms under wave loading, Journal of Sound and Vibration, 296, (2006) 734-745.
[48] H. Ma, G. Y. Tang and Y. D. Zhao, Feedforward and feedback optimal control
for offshore structures subjected to irregular wave forces, Ocean Engineering,
33, (2006) 1105-1117.
[49] Y. S. Moon, P. Park, W. H. Kwon and Y. S. Lee, Delay-dependent robust sta-
bilization of uncertain state-delayed systems, International Journal of Control,
74, (2001) 1447-1455.
[50] M. C. de Oliveira, J. Bernussou and J. C. Geromel, A new discrete- time robust
stability condition, System Control Letter, 37, (1999) 261-265.
[51] M. C. de Oliveira, J. Bernussou and L. Hsu, LMI characterization of structural
and robust stability: the discrete-time case, Linear Algorithm Application, 296,
(1999) 27-38.
[52] P. Park, A delay-dependent stability criterion for systems with uncertain time-
invariant delays, IEEE Transactions on Automatic Control, 44, (1999) 876-877.
[53] K. C. Patil and R. S. Jangid, Passive control of offshore jacket platforms, Ocean
Engineering, 32, (2005) 1933-1949.
REFERENCES 173
[54] E. Reithmeier and G. Leitmann, Structural vibration control, Journal of the
Franklin Institute, 338, (2001) 203-223.
[55] J.-P. Richard, Time-delay systems: An overview of some recent advances and
open problems, Automatica, 39, (2003) 1667-1694.
[56] J. Suhardjo and A. Kareem, Feedback-feedforward control of offshore platforms
under random waves, Earthquake Engineering and Structural Dynamics, 30,
(2001) 213-235.
[57] X. Li and C. E. De Souza, Criteria for robust stability and stabilization of
uncertain linear systems with state delay, Automatica, 33, (1997) 1657-1662.
[58] C. E. de Souza and X. Li, Delay-dependent robust H∞ control of uncertain
linear state-delayed systems, Automatica, 35, (1999) 1313-1321.
[59] V. Suplin, E. Fridman and Uri Shaked, H∞ control of linear uncertain time-
delay systems - a projection approach, IEEE Transactions on Automatic Con-
trol, 51, (2006) 680-685.
[60] G. Y. Tang, H. Ma and B. L. Zhang, Feedforward and feedback optimal control
for discrete linear systems with disturbances , Control and Decision, 20, (2007)
633-644.
[61] W. Wang and G. Y. Tang, Feedback and feedforward optimal control for off-
shore jacket platforms, China Ocean Engineering, 18, (2004) 515-526.
[62] M. Wu, Y. He, J. H. She and G. P. Liu, Delay-dependent criteria for robust
stablity of time-varying delay systems, Automatica, 40, (2004) 1435-1439.
[63] L. Xie, Output feedback H∞ control of systems with parameter uncertainty,
International Journal of Control, 63, (1996) 741-750.
174 REFERENCES
[64] L. Xie, E. Fridman and U. Shaked, Robust control of disturbanced delay sys-
tems with application to combustion control, IEEE Transactions on Automatic
Control, 46, (2005) 1930-1935.
[65] S. Y. Xu and J. Jam, Improved delay-dependent stability criteria for time-delay
systems, IEEE Transactions on Automatic Control, 50, (2005) 384-387.
[66] S. Y. Xu, J. Lam and Y. Zou, New results on delay-dependent robust H∞
control for systems with time-varying delays, Automatica, 42, (2006) 343-348.
[67] D. Yue, Robust stabilization of uncertain systems with unknown input delay,
Automatica, 40, (2004) 331-336.
[68] D. Yue and Q.-L. Han, Delayed feedback control of uncertain systems with
time-varying input delay, Automatica, 41, (2005) 233-240.
[69] X. M. Zhang, M. Wu, J. H. She and Y. He, Delay-dependent stabilization of
linear systems with time-varying state and input delays, Automatica, 41, (2005)
1405-1412.
[70] K. Zhou, J. C. Doyle and K. Glover, Robust Optimal Control. Prentice-Hall,
Upper Saddle River, NJ; 1996.
[71] M. Zribi, N. Almutairi, M. Abdel-Rohman and M. Terro, Nonlinear and robust
control schemes for offshore steel jacket platforms, Nonlinear Dynamics, 35,
(2004) 61-80.
[72] http://www.esru.strath.ac.uk/EandE/Web sites/98-9/offshore/platintr.htm
(Access Date: March 15, 2006 5:30 p.m.).
[73] http://www.esru.strath.ac.uk/EandE/Web sites/98-9/offshore/steel.htm (Ac-
cess Date: March 16, 2006 2:33 p.m.).
REFERENCES 175
[74] http://www.naturalgas.org/naturalgas/extraction offshore.htm (Access Date:
2006 1:33 p.m.).