87_ftp | Control Theory | Physics & Mathematics

STRUCTURAL CONTROL AND HEALTH MONITORING

Struct. Control Health Monit. 2007; 14:109–143
Published online 11 July 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/stc.87
MATLAB-based seismic response control of a cable-stayed
bridge: cable vibration
Chin-Hsiung Loh*
,y,1
and Chia-Ming Chang
2
Department of Civil Engineering, National Taiwan University, Taiwan
SUMMARY
This paper presents the cable vibration problem of structural control for a cable-stayed bridge. The goal of
this study is not only to provide a test for development of strategies for control of cable vibration, but also
to examine the effect of cable vibration on control of the structure itself. Based on the detailed drawings of
the Gi-Lu bridge, a three-dimensional numerical model has been developed to represent complex behavior
of the full-scale bridge. Differential motion at multiple supports and cable flexibility are included in the
analysis. A MATLAB-based structural analysis tool has been developed that considers of geometrically
nonlinear behavior of beam elements that represent the cables. The dynamic characteristics of cables are
verified using field experiments, a commercial finite element code, and the MATLAB program. Evaluation
criteria are presented for the design problems that are consistent with the goal of control of seismic
response control of a cable-stayed bridge. Control devices are assumed to be installed either between the
deck and the end-abutment and/or between the deck and cables. Passive and active devices are used to
study behavior of the model. A comparison of the control responses using truss elements or using
geometrically nonlinear beam elements to represent the cable is also presented. Copyright # 2005 John
Wiley & Sons, Ltd.
KEY WORDS: cable-stayed bridge; active control algorithms; stayed cable; vibration
INTRODUCTION
In the past few decades, cable-stayed bridges have found wide application throughout the world.
The main spans of these bridges have reached a length of 900 m, leading to very long stayed
cables. The control of flexible bridge structures, such as cable-stayed bridges, is viewed as a
unique and challenging problem with many complexities in modeling, control design, and
implementation. Long-span cables are especially susceptible to vibration with large amplitude
under wind/rain loading or support excitations due to their high flexibility, relatively small mass
Received 5 January 2005
Revised 16 March 2005
Accepted 2 May 2005 Copyright # 2005 John Wiley & Sons, Ltd.
y
E-mail: lohc0220@ccms.ntu.edu.tw
1
Professor.
*Correspondence to: Chin-Hsiung Loh, Department of Civil Engineering, National Taiwan University, Taiwan.
2
Research Assistant.
Contract/grant Sponsor: Taiwan National Science Council; Contract Grant Number: NSC 93-2211-E-002-005
and very low inherent damping [1,2]. Therefore, cable-stayed bridges might be vulnerable to
dynamic loading such as earthquakes and strong wind loads. In 1993 Warnitchai et al. [3]
experimentally and analytically studied active tendon control of cable-stayed bridges, subjected
to a vertical sinusoidal force. They utilized a simple cable-supported cantilever beam as a model.
Up to now, many research efforts have been focused on the interaction of cables with the deck
and attenuation of the cable movement [4–6]. Other research efforts have been aimed at
understanding the overall dynamic behavior of cable-stayed bridges and developing finite
element models [7,8]. The working group on bridge control within the ASCE Committee on
Structural Control recently posted a first-generation benchmark structural control problem
based on the Cape Girardeau Bridge [7,8]. This problem focus on one-dimensional ground
acceleration applied in the longitudinal direction that is uniformly and simultaneously applied
at all supports. In the work of Moon et al. [9,10], a semi-active system for the benchmark bridge
employing MR dampers in conjunction with a LDG/clipped optimal control (LQG/MR) and a
sliding mode semi-active control system (SMC/MR) was studied. In addition, a hybrid base
isolation system employing semi-active control devices (using MR-dampers), often termed
‘smart’ dampers, has been suggested to control the damping force of bridge structure [11]. In the
design of modern feedback control systems, the question of uncertainties in the models and the
excitations, within the context of robust control, is becoming an important issue. For example,
an active scheme that uses active tendons via feedback of the states of the system for vibration
control of a cable-stayed bridge under seismic loads has also been proposed by Rodellar et al.
[12]. While different approaches have been adopted for robust control of civil engineering
structures, such as H
1
and related tools [13], neural networks and fussy logic [14], and sliding
mode control [15,16], the use of these approaches to control of cable-stayed bridges has been
limited so far.
In the ASCE phase II benchmark control problem, a three-dimensional evaluation model has
been developed to represent complex behavior of the multi-support and transverse excitations of
the Cape Girardeau Bridge [17]. The benchmark problem and a sample control design have been
made available in the form of a set of MATLAB equations.
In this research the control of cable-stayed bridge is studied. Based on the detailed information
of the Gi-Lu cable-stayed bridge in Taichung County, Taiwan, a three-dimensional numerical
model has been developed to represent the complex behavior of the full-scale bridge. Dynamic
behavior of the cable including sag geometry, pre-stress tension force, and nonlinear response are
considered in this analysis. The formulation of the cable stiffness is carried out both in ABAQUS
and in MATLAB equations for consistency. This paper is focused on the following issues:
1. A three-dimensional evaluation model is developed in a realistic way to represent complex
behavior of a full-scale cable-stayed bridge. Nonlinear beam elements are used to represent
each cable. A comparison of the dynamic response of the bridge is made by using a truss
model or nonlinear beam element model for the cables.
2. A comparison of the control effectiveness is discussed for: (a) using active control strategy
by employing actuators between the bridge deck and the end-abutments; or (b) using both
viscous dampers between the cables and the bridge deck and actuators between the deck
and the end-abutments.
3. Based on the different evaluation criteria the control effectiveness between two different
control algorithms, active and hybrid, is examined.
4. A MATLAB-based model for control of a cable-stayed bridge is developed.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 110
MODELING OF CABLE-STAYED BRIDGE
The cable-stayed bridge used for this study is the Gi-Lu bridge, located at Nantou County,
Taiwan. It is a modern, pre-stressed concrete cable-stayed bridge, which crosses the
Juosheui River in Taiwan. As shown in Figure 1(a) the bridge has a single pylon constructed
with reinforced concrete (58 m above the deck), two rows of harped cables (68 cables in total),
Figure 1. (a) Top view and side view of Gi-Lu bridge; and (b) cross-section of Gi-Lu bridge.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 111
and a streamline-shape single box girder, as shown in Figure 1(b). With depth 2.75 m and
width 24 m, the box girder rigidly connects with the pylon and spans 120 m to each side span.
17 pairs of near parallel pre-stressed steel cables extending from each side of the pylon
were developed. The material properties of the bridge are shown in Table I. The sag of each
cable is shown in Table II. Cables 1 and 2 indicate the shortest cables and cables 33 and 34 are
the longest cables. R and L indicate the left-hand and right-hand sides of the pylon. The
vibration frequency of each cable was also estimated from vibration test (Table III). The density
of the cable ranges from 0.410 to 0.613 kg/cm and the design cable force ranges from 220 to
290 ton.
Dynamic characteristics of the cable
A three-dimensional finite element model of the Gi-Lu bridge was developed in MATLAB.
A linear evaluation model is used in this cable-stayed bridge model. The finite element
model employs beam elements and rigid links. Due to large deformation of the cables
and their sag, geometric nonlinearity of beam elements needs to be considered in this
analysis. Analytical solutions of geometrically nonlinear beam element are available for
several special cases; for instance, a fixed beam subjected to concentrated loads.
Generally, conventional linear two-dimensional beam elements cannot consider the effect
of axial stresses due to large bending deformation. It is necessary to represent an
appropriate nonlinear behavior of beam. In this study, by using an energy method a
geometrically nonlinear beam element was generated that includes the nonlinear terms
plus the terms from the conventional linear beam element. Therefore, a geometric stiffness
matrix can be derived based on energy methods from Euler-Bernoulli equation as shown
follows [18]:
F
ext
¼ ðK
linear
þ K
nonlinear
Þu ð1Þ
Table I. Material properties of Gi-Lu bridge.
Element Young’s
modulus (N/m
2
)
Area (m
2
) Moment of
inertia (m
4
)
Poisson’s
ratio
Density
(kg/m
3
)
Main girder (deck) 2:77 Â 10
10
12.27 9.77, 334.55 0.2 3202
Plate (deck) 2:77 Â 10
10
4 334.55, 334.55 0.2 0.001
Side span 2:77 Â 10
10
8.121 6.66, 84.29 0.2 2400
Pier 2:77 Â 10
10
8.121 6.66, 84.29 0.2 2400
Cable (1) 2:04 Â 10
11
0.005163 0.515 Â10
À5
, 0.515 Â10
À5
0 8000
Cable (2) 2:04 Â 10
11
0.006 0.45 Â10
À5
, 0.45 Â10
À5
0 8000
Cable (3) 2:04 Â 10
11
0.00756 0.67 Â10
À5
, 0.67 Â10
À5
0 8000
Pylon
Ã
2:77 Â 10
10
8.67–0.07 6.85–63.63, 14.18–69.02 0.2 2400
Ã
Varying cross-section.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 112
Table II. Left-hand side four columns indicate the sag of each cable (middle node) and right-hand side
four columns indicate the 1st natural frequency of each cable (R** means west-side cables, L** means east-
side cables).
Sag (maximum) (m) 1st natural frequency (Hz)
0.51839 R33 L33 0.51839 0.75823 R33 L33 0.75823
0.52236 R31 L31 0.52236 0.75513 R31 L31 0.75513
0.4723 R29 L29 0.4723 0.79146 R29 L29 0.79146
0.29116 R27 L27 0.29116 0.99956 R27 L27 0.99956
0.23382 R25 L25 0.23382 1.111 R25 L25 1.111
0.23022 R23 L23 0.23022 1.1161 R23 L23 1.1161
0.16688 R21 L21 0.16688 1.3042 R21 L21 1.3042
0.13945 R19 L19 0.13945 1.4196 R19 L19 1.4196
0.091567 R17 L17 0.091567 1.7411 R17 L17 1.7411
0.072009 R15 L15 0.072009 1.9493 R15 L15 1.9493
0.070109 R13 L13 0.070109 1.9581 R13 L13 1.9581
0.058809 R11 L11 0.058809 2.1138 R11 L11 2.1138
0.049284 R9 L9 0.049284 2.2756 R9 L9 2.2756
0.033323 R7 L7 0.033323 2.7106 R7 L7 2.7106
0.022885 R5 L5 0.022885 3.1737 R5 L5 3.1737
0.018332 R3 L3 0.018332 3.3889 R3 L3 3.3889
0.011206 R1 L1 0.011206 4.0191 R1 L1 4.0191
Pylon &
Pier No.2
Abutment
(Pier No.1)
(Lu-Ku side)
Abutment
Pier No.3
(Gi-Lu side)
120 m
120 m
0.010133 R2 L2 0.010133 4.0193 R2 L2 4.0193
0.016358 R4 L4 0.016358 3.3891 R4 L4 3.3891
0.020243 R6 L6 0.020243 3.1738 R6 L6 3.1738
0.029262 R8 L8 0.029262 2.7108 R8 L8 2.7108
0.043072 R10 L10 0.043072 2.2758 R10 L10 2.2758
0.051267 R12 L12 0.051267 2.114 R12 L12 2.114
0.060964 R14 L14 0.060964 1.9582 R14 L14 1.9582
0.062502 R16 L16 0.062502 1.9493 R16 L16 1.9493
0.07934 R18 L18 0.07934 1.741 R18 L18 1.741
0.12064 R20 L20 0.12064 1.4193 R20 L20 1.4193
0.14422 R22 L22 0.14422 1.3037 R22 L22 1.3037
0.19876 R24 L24 0.19876 1.1149 R24 L24 1.1149
0.20176 R26 L26 0.20176 1.11 R26 L26 1.11
0.25108 R28 L28 0.25108 0.99808 R28 L28 0.99808
0.40785 R30 L30 0.40785 0.78728 R30 L30 0.78728
0.45105 R32 L32 0.45105 0.75068 R32 L32 0.75068
0.44722 R34 L34 0.44722 0.75464 R34 L34 0.75464
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 113
T
a
b
l
e
I
I
I
.
C
o
m
p
a
r
i
s
o
n
o
n
t
h
e
e
s
t
i
m
a
t
e
d
c
a
b
l
e
v
i
b
r
a
t
i
o
n
f
r
e
q
u
e
n
c
i
e
s
f
r
o
m
v
i
b
r
a
t
i
o
n
m
e
a
s
u
r
e
m
e
n
t
,
f
r
o
m
A
B
A
Q
U
S
a
n
d
f
r
o
m
B
c
o
d
e
u
s
i
n
g
n
o
n
l
i
n
e
a
r
b
e
a
m
e
l
e
m
e
n
t
s
.
A
n
a
l
y
z
e
n
a
t
u
r
a
l
f
r
e
q
u
e
n
c
y
o
f
c
a
b
l
e
s
(
R
1
,
R
1
7
,
R
3
4
)
R
1
(
s
h
o
r
t
e
s
t
)
R
3
4
(
l
o
n
g
e
s
t
)
E
l
e
m
e
n
t
s
1
0
0
0
1
5
5
0
1
0
0
2
0
0
E
l
e
m
e
n
t
s
1
0
0
0
1
5
5
0
1
0
0
M
e
t
h
o
d
A
B
A
Q
U
S
M
A
T
L
A
B
M
A
T
L
A
B
M
A
T
L
A
B
M
A
T
L
A
B
M
e
t
h
o
d
M
E
A
S
U
R
E
A
B
A
Q
U
S
M
A
T
L
A
B
M
A
T
L
A
B
M
A
T
L
A
B
M
o
d
e
H
z
H
z
H
z
H
z
H
z
M
o
d
e
H
z
H
z
H
z
H
z
H
z
1
4
.
1
2
4
.
1
3
0
1
4
.
1
2
3
8
4
.
1
2
3
7
4
.
1
2
3
7
1
0
.
7
6
9
0
.
7
6
9
0
.
7
5
1
1
9
0
.
7
5
9
7
4
0
.
7
5
9
5
6
2
8
.
4
3
8
.
1
6
8
7
8
.
1
5
6
6
8
.
1
5
6
8
8
.
1
5
6
9
2
1
.
4
9
9
1
.
4
9
6
1
.
4
8
8
5
1
.
5
0
6
2
1
.
5
0
5
4
3
1
2
.
9
7
1
2
.
0
2
7
1
2
.
0
2
3
1
2
.
0
2
5
1
2
.
0
2
6
3
2
.
2
4
8
2
.
2
4
7
2
.
2
2
6
6
2
.
2
5
7
6
2
.
2
5
6
5
4
1
7
.
9
3
1
5
.
6
2
2
1
5
.
6
7
1
1
5
.
6
8
2
1
5
.
6
8
5
4
2
.
9
9
7
2
.
9
9
9
2
.
9
5
5
2
3
.
0
0
6
7
3
.
0
0
5
2
5
2
3
.
2
6
1
8
.
8
8
4
1
9
.
0
7
8
1
9
.
1
0
9
1
9
.
1
1
7
5
3
.
7
5
1
3
.
7
5
5
3
.
6
6
8
9
3
.
7
5
3
3
.
7
5
1
2
2
1
.
7
6
3
2
2
.
2
4
5
2
2
.
3
1
2
2
.
3
2
7
6
4
.
5
1
1
4
.
5
1
4
4
.
3
6
0
2
4
.
4
9
5
7
4
.
4
9
3
8
2
4
.
2
3
6
2
5
.
1
8
9
2
5
.
3
0
7
2
5
.
3
3
8
7
5
.
2
7
9
5
.
2
7
8
5
.
0
2
0
1
5
.
2
3
4
2
5
.
2
3
2
1
R
1
7
(
m
i
d
d
l
e
)
8
6
.
0
4
4
6
.
0
4
7
5
.
6
3
7
6
5
.
9
6
7
7
5
.
9
6
5
7
E
l
e
m
e
n
t
s
1
0
0
0
1
5
5
0
1
0
0
9
6
.
8
3
3
6
.
8
2
2
6
.
2
0
3
6
6
.
6
9
5
6
6
.
6
9
4
M
e
t
h
o
d
M
E
A
S
U
R
E
A
B
A
Q
U
S
M
A
T
L
A
B
M
A
T
L
A
B
M
A
T
L
A
B
1
0
7
.
5
9
1
7
.
6
0
4
6
.
7
0
4
1
7
.
4
1
7
1
7
.
4
1
6
2
M
o
d
e
H
z
H
z
H
z
H
z
H
z
1
1
8
.
3
8
6
8
.
3
9
3
7
.
1
3
2
8
.
1
3
1
6
8
.
1
3
2
1
1
.
7
3
2
7
1
.
7
3
2
5
1
.
7
3
1
8
1
.
7
2
0
8
1
.
7
2
0
2
1
2
9
.
1
7
6
9
.
1
9
7
.
4
7
3
9
8
.
8
3
8
4
8
.
8
4
0
7
2
3
.
4
6
2
8
3
.
4
6
2
9
3
.
4
5
4
3
.
4
3
4
2
3
.
4
3
3
1
3
9
.
9
7
8
9
.
9
9
6
7
.
7
2
4
8
9
.
5
3
6
8
9
.
5
4
2
1
3
5
.
1
9
2
3
5
.
2
0
0
1
5
.
1
5
8
7
5
.
1
3
4
9
5
.
1
3
3
1
1
4
1
0
.
8
0
2
1
0
.
8
1
7
.
8
8
1
1
0
.
2
2
6
1
0
.
2
3
5
4
6
.
9
4
9
6
6
.
9
4
3
5
6
.
8
3
4
9
6
.
8
1
6
6
.
8
1
3
9
1
5
1
1
.
6
3
1
1
1
.
6
3
1
5
.
7
6
1
1
0
.
9
0
6
1
0
.
9
2
1
5
8
.
7
0
4
3
8
.
6
9
6
8
.
4
7
0
6
8
.
4
7
1
1
8
.
4
6
9
5
1
6
1
2
.
4
5
5
1
2
.
4
7
1
5
.
7
6
6
1
1
.
5
7
5
1
1
.
5
9
7
6
1
0
.
4
7
6
1
0
.
4
5
9
1
0
.
0
5
2
1
0
.
0
9
4
1
0
.
0
9
5
1
7
1
3
.
3
0
3
1
3
.
3
2
1
5
.
8
2
9
1
2
.
2
3
3
1
2
.
2
6
4
7
1
2
.
2
5
4
1
1
2
.
2
3
6
1
1
.
5
6
4
1
1
.
6
7
9
1
1
.
6
8
5
1
8
1
4
.
1
6
7
1
4
.
1
8
1
5
.
8
4
6
1
2
.
8
8
1
2
.
9
2
2
8
1
4
.
0
3
0
3
1
4
.
0
2
8
1
2
.
9
9
1
3
.
2
2
1
3
.
2
3
5
1
9
1
5
.
0
2
1
1
5
.
0
5
1
5
.
9
3
4
1
3
.
5
1
4
1
3
.
5
7
1
9
1
5
.
8
4
8
7
1
5
.
8
3
7
1
4
.
3
1
2
1
4
.
7
1
3
1
4
.
7
4
3
2
0
1
5
.
9
0
7
1
5
.
9
3
1
5
.
9
6
7
1
4
.
1
3
6
1
4
.
2
1
1
0
1
7
.
6
4
9
4
1
7
.
6
6
5
1
5
.
5
0
9
1
6
.
1
5
2
1
6
.
2
0
6
2
1
1
6
.
8
2
1
1
6
.
8
3
1
4
.
7
4
4
1
4
.
8
3
9
1
1
1
9
.
5
3
5
9
1
9
.
5
1
5
1
6
.
5
5
7
1
7
.
5
3
4
1
7
.
6
2
1
2
2
1
7
.
7
3
3
1
7
.
7
4
1
5
.
3
3
8
1
5
.
4
5
8
1
2
2
1
.
3
9
8
5
2
1
.
3
8
7
1
7
.
4
2
8
1
8
.
8
5
6
1
8
.
9
8
7
2
3
1
8
.
6
3
5
1
8
.
6
7
1
5
.
9
1
6
1
6
.
0
6
7
N
o
t
e
:
(
a
)
R
1
,
R
1
7
,
a
n
d
R
3
4
i
n
d
i
c
a
t
e
t
h
e
s
h
o
r
t
e
s
t
,
m
i
d
d
l
e
,
a
n
d
l
o
n
g
e
s
t
c
a
b
l
e
s
,
r
e
s
p
e
c
t
i
v
e
l
y
;
(
b
)
t
r
a
n
s
l
a
t
i
o
n
a
l
a
n
d
t
r
a
n
s
v
e
r
s
e
d
e
g
r
e
e
s
-
o
f
-
f
r
e
e
d
o
m
o
f
e
a
c
h
c
a
b
l
e
n
o
d
e
a
r
e
l
o
c
k
e
d
.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 114
or
F
x1
F
y1
M
z1
F
x2
F
y2
M
z2
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
¼
EA
L
0
12EI
L
3
0
6EI
L
2
4EI
L
À
EA
L
0 0
EA
L
0 À
12EI
L
3
À
6EI
L
2
0
12EI
L
3
0
6EI
L
2
2EI
L
0 À
6EI
L
2
4EI
L
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
þ N
0
0
6
5L
0
1
10
2L
15
0 0 0 0
0 À
6
5L
À
1
10
0
6
5L
0
1
10
À
L
30
0 À
1
10
2L
15
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
x
1
y
1
f
1
x
2
y
2
f
2
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
ð1aÞ
where E is the Young’s modulus, A is the cross-sectional area, I is the moment inertia, L is the
length of the element, F is the force, M is the moment, x and y are the translational deformation
of both ends of the element, and j is the rotational deformation of both ends of the element.
Subscripts (1 and 2) indicate the two ends of the beam, the first term of the stiffness matrix is the
linear stiffness matrix, the second term is due to nonlinearity, and N is related to the axial force
due to large deformation:
N ¼
EA
L
ðL
0
À LÞ ð2Þ
in which L
0
is the actual length along the center line of the beam. However, to deal with the
complex behavior of cables, the computational procedures must be modified. In development of
the initial cable stiffness is important to obtain the actual pre-tensioned values (not designed
values) and to include correct element properties (modulus of elasticity, moment inertia of cross-
sectional area, length of taut cables, and Poisson’s ratio). The procedures for developing the
initial cable stiffness are shown below:
1. Assume that Poisson’s ratio equals zero.
2. Choose how many elements to use in the formulation, and transform Equation (1) into a
three dimensional nonlinear beam element.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 115
3. Use the initial length to generate the stiffness matrix where N equals the pre-tensioned
value.
4. Divide the external force (self-weight of cables) into an appropriate number of equal parts
n.
5. Use Newton–Raphson (see Figure 1) iterations to solve ½KŠfDg ¼ fF
ext
g by appropriate
conditions of convergence.
6. Calculate the present length due to large lateral deformation and re-formulate the cable
stiffness matrix where N is still the pre-tensioned value.
7. Repeat Step 5 until the total steps equal to n (in Step 4).
From these procedures, stiffness matrices of the cables can satisfy effects of sag and measured
natural frequencies.
In order to evaluate the dynamic characteristics of the stayed-cable in this model, the
vibration frequencies of each stayed-cable in the Gi-Lu bridge were examined from three
different approaches: (1) experiments; (2) ABAQUS; and (3) MATLAB equations. For example,
for the longest cables (33 and 34), it is described as: (i) angle of elevation 268; (ii) designed length
126.42 m; (iii) length density 47.9 kg/m; (iv) Young’s modulus 1:83 Â 10
11
N=m
2
; (v) cross
sectional area 0.0060 m
2
; and (vi) gravity 9.81m/sec
2
. The FEM model developed using
ABAQUS has 1000 beam elements and considers the cable structure to be subjected to gravity
and to have fixed-end boundary conditions. The cable force and moment of inertia were
identified by matching the 21 vibration frequencies of the cables to those obtained from the field
experiment. Through an optimally iterative process, the identified cable force is 1:66 Â10
6
N;
and the identified moment of inertia is 5:0 Â 10
À6
m
4
: This information is used to formulate
equations in MATLAB for the stiffness matrix of the cable system. The vibration frequency of
each cable predicted by MATLAB is verified using field experiments. The configuration of
Cables R1, R17 and R34 is shown in Figure 2, and the identified cable force is only 91.4%
(1:66 Â10
6
=1:81 Â 10
6
N) of the force estimated by string vibration theory. Comparison
between the identified vibration frequencies of the cable R1, R17 and R34 using MATLAB and
ABAQUS is shown in Table III. The vibration modes of cable R34 (longest) and cable L17 are
shown in Figure 3. To catch the dynamic characteristics of the cable 1000 beam elements are
used in ABAQUS to analyze the detail of the dynamic characteristics of a single cable, but on
Figure 2. Configuration of Cable-R34 and the identified.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 116
the contrary, only 9 nodes were used to form a single cable which is sufficient to involve the low-
frequency dynamic behavior (0.7–4 Hz) as in relating to the whole bridge system.
DESCRIPTION OF FINITE ELEMENT MODEL
The finite element model of the cable-stayed bridge, as shown in Figure 4, has a total of 1009
nodes. The pylon is modeled by 90 nodes with 540 degrees of freedom (DOF) (25 nodes are
above the deck, 5 nodes are below the deck, and 60 nodes are near the anchors). The deck is
modeled by 729 nodes with 4374 DOFs. Pier 2 (under the pylon) is modeled by 5 nodes
(including the node attached to the ground). Pier 1 (North side) and pier 3 (South side) have 4
nodes for each (including the node attached to the ground) with a total of 56 DOFs. The cable is
Figure 3. The first six mode shapes: (a) cable R34; and (b) cable R17.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 117
modeled with 9 nodes for each cable and total of 612 nodes for 64 cables with 3672 DOFs. The
mass matrix is formed by a lumped mass approach. The proportional damping formulation is
used in each element and then to form the full damping matrix is formed with the same
procedure as is used to form the stiffness matrix. The damping ratio for deck, pylon and piers
(including all supports) is assumed to be 5% and for cables is assumed to be 1%. Constraints are
applied to restrain the both ends of the deck (boundary conditions). All DOFs at the bottom of
both piers are fixed. As for the boundary condition of the bridge structure, all translational
DOFs and torsion DOF at the side spans connected to the embankment are fixed. The
transverse and vertical degrees of freedom at Piers 1 and 3, and at the side spans of the bridge
deck deform consistently (i.e. y- and z-directions are constraint, and x-, f
x
-, f
y
-, f
z
}directions
are free to move).
EVALUATION MODEL
The control system of the cable-stayed bridge includes: evaluation model, sensor processor,
controller and control devices. Based on the above-mentioned finite element model the
formulation of the model is described first.
Problem formulation
Since a precise mathematical model for analyzing the dynamic behavior of cable-stayed bridges
is very complicated, appropriately reduced methods were used to formulate the equation of
motion for the Gi-Lu bridge. The reduced methods use static condensation and quasi-static
reduction. Consider the general equation of motion for a structural system subjected to seismic
loads
M
total
.
U þ C
total

U þ K
total
U ¼ P
ext
ð3Þ
where system matrices can be written as
M
total
¼
M
cdp
M
coup
M
T
coup
M
sp
_ _
C
total
¼
C
cdp
C
coup
C
T
coup
C
sp
_ _
K
total
¼
K
cdp
K
coup
K
T
coup
K
sp
_ _
ð3aÞ
Figure 4. Model for cable-stayed bridge (Gi-Lu bridge) with the span length of 318.9 m, the pylon height
of 58 m from deck, and the deck width of 24 m.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 118
where subscript cdp indicates DOFs of cables, deck and pylon, subscript sp indicates the DOFs
for support from piers and side spans and subscript coup indicates coupled terms between
supports and others; U is the displacement response vector; M
total
, C
total
, K
total
are the mass,
damping and stiffness matrices respectively of the whole structural system, and P
ext
is the vector
of external forces.
Model reduction (static condensation and quasi-static reduction)
The finite element formulation of the bridge model has a large number of degrees of freedom.
Static condensation is applied to reduce the redundant DOFs and retain the main DOFs. The
main DOFs is taken here to include the nodes of the main girder, pylon, piers, side span, and
cables including the nodes connecting the cables to the deck and pylon. The redundant DOFs
refers to the other nodes of the deck since these nodes are used to analyze static behavior. It is
assumed that the total displacement can be divided into two parts: active DOFs (main nodes)
and dependent DOFs (redundant nodes). Therefore, the equation for the static condensation
can be formed as
K
aa
K
ad
K
T
ad
K
dd
_ _
U
a
U
d
_ _
¼
P
a
0
_ _
ð4Þ
where the subscript a denotes the active DOFs and d denotes dependent DOFs. Then the total
displacement vector can be transformed into
U
a
U
d
_ _
¼
I
ÀK
À1
dd
K
T
ad
_ _
U
a
or
U
a
U
d
_ _
¼ ½TRŠU
a
ð5Þ
where [TR] is the transformation matrix of the static condensation. Consequently, the system
total mass, damping, and stiffness matrices can be changed as follows:
M
_
total
¼ TR
T
M
total
TR; C
_
total
¼ TR
T
C
total
TR; K
_
total
¼ TR
T
K
total
TR ð6Þ
To consider the seismic excitation of the multiple-supported system both quasi-static and
dynamic analyses must be employed. By using the transformation matrix of the static
condensation Equation (1) can be re-arranged as follows:
M
_
total
.
U
a
þ C
_
total

U
a
þ K
_
total
U
a
¼ TR
T
P
ext
ð7Þ
where the active DOFs of U
a
represent two components: U
a,cdp
, the DOFs from cables, deck
and pylon, and U
a,sp
, the DOFs from piers and side spans (supports). The total displacement
U
a,cdp
can also be separated into the displacement U
a,cdp
s
due to static application of the ground
motion, and the dynamic displacement U
a,cdp
d
relative to the quasi-static displacement. The
relationship between these displacement components is
U
a;cdp
U
a;sp
_ _
¼
U
s
a;cdp
U
a;sp
_ _
þ
U
d
a;cdp
0
_ _
ð8Þ
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 119
By neglecting the dynamic terms in Equation (7), the quasi-static solution can be obtained
from the following equation
K
a;cdp
K
a;coup
K
T
a;coup
K
a;sp
_ _
U
s
a;cdp
U
a;sp
_ _
¼
0
P
s
ext
_ _
ð9Þ
Solving Equation (9) for quasi-static displacement leads to
U
s
a;cdp
¼ ÀK
À1
a;cdp
K
a;coup
U
a;sp
or
U
s
a;cdp
¼ R
s
U
a;sp
ð10Þ
Substituting Equations (8) and (10) into (7), one can obtain the dynamic equation of motion
of the structure as follows:
M
_
a;cdp
.
U
d
a;cdp
þ C
_
a;cdp

U
d
a;cdp
þ K
_
a;cdp
U
d
a;cdp
¼ ÀðM
_
a;cdp
R
s
þ M
_
a;sp
Þ
.
U
a;sp
À ðC
_
a;cdp
R
s
þ C
_
a;sp
Þ

U
a;sp
ð11Þ
To solve for the dynamic response of a cable-stayed bridge the state-space formulation is
used. A condition for convergence of the discrete state space calculation is that the following
criteria must be satisfied:
DT4
p
o
ð12Þ
where o is the highest modal frequency (rad/sec) that is considered. In this study the first 400
modes of vibration are used for the analysis. Since the sampling time for the input ground
motion is 0.005 s, the convergence condition of Equation (12) is satisfied.
After the model has been reduced the equation of motion for the damped structural system
results from Equation (11):
%
M
a;cdp
.
Y þ
%
C
a;cdp

Y þ
%
K
a;cdp
Y ¼ ÀF
T
ðM
_
a;cdp
R
s
þ M
_
a;sp
Þ
.
U
a;sp
ÀF
T
ðC
_
a;sp
R
s
þ C
_
a;sp
Þ

U
a;sp
ð13Þ
where
Y ¼ FU
d
a;cdp
;
%
M
a;cdp
¼ F
T
M
_
a;cdp
F;
%
C
a;cdp
¼ F
T
C
_
a;cdp
F;
%
K
a;cdp
¼ F
T
K
_
a;cdp
F
and F is the modal matrix. A state space form for the cables, deck and pylon is as follows:
’ x ¼

Y
.
Y
_ _
¼ Ax þ Eu
ext
¼
0 I
À
%
M
À1
a;cdp
%
K
a;cdp
À
%
M
À1
a;cdp
%
C
a;cdp
_ _
Y

Y
_ _
þ
0 0
À
%
M
À1
a;cdp
F
T
ðM
_
a;cdp
R
s
þ M
_
a;sp
Þ À
%
M
À1
a;cdp
F
T
ðC
_
a;cdp
R
s
þ C
_
a;sp
Þ
_
_
_
_
.
U
a;sp

U
a;sp
_ _
ð14Þ
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 120
Figure 5. (a) The first six vibration modes of the cable-stayed bridge. Its corresponding modal frequencies
are 0.51487, 0.73124, 0.75022, 0.75069, 0.75393, and 0.75464 Hz, respectively; (b) the 31st (f ¼ 1:506 Hz),
64th (f ¼ 1:4456 Hz), 97th (f ¼ 1:756 Hz), 102nd (f ¼ 1:8941 Hz) and 115th (f ¼ 2:0378 Hz) vibration
modes of the cable stayed bridge; and (c) first six mode shapes of cable-stayed bridge using truss elements
to simulate the cable behavior (the vibration frequencies of the first six modes are: 0.42152, 0.44952,
0.80013, 1.3695, 1.4455, and 1.5871 Hz).
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 121
Through static condensation the final reduced model for the pylon and the deck reduce from
90 to 85 nodes and from729 to 243 nodes, respectively; the number of nodes for the pier and
cable remain the same.
Based on this formulation the mode shapes of the cable-stayed bridge are calculated. Figure 5(a)
shows the first six vibration modes of the cable-stayed bridge. The corresponding modal
frequencies are 0.51487, 0.73124, 0.75022, 0.75069, 0.75393, and 0.75464 Hz, respectively. It is
found that the first fundamental mode is mainly dominated by the vibration of deck, and that
the following five modes are dominated by vibration of the cables. In addition, the 31st
(f ¼ 1:506 Hz), 64th (f ¼ 1:4456 Hz), 97th (f ¼ 1:756 Hz), 102th (f ¼ 1:8941 Hz) and 115th
(f ¼ 2:0378 Hz) vibration modes of the cable stayed bridge are also shown in Figure 5(b),
contributions from both cable and deck vibration to these higher modes are observed. Since the
implementation of geometric stiffness matrix to simulate the cable using beam element, then the
sag behavior of cable can be truly reflected. Besides, the cable vibration in transverse direction
and the pre-stressed force in the cable can also be used. Figure 5(c) shows the calculated lowest
six vibration modes of the cable-stayed bridge that are calculated using truss element to simulate
the cable (the vibration frequencies of the first six modes are: 0.42152, 0.44952, 0.80013, 1.3695,
1.4455, and 1.5871 Hz). Therefore, a significant difference between the bridge vibration modes is
obtained if the cable is modeled using truss elements instead of dynamics of geometrically
nonlinear beam elements. A comparison of the acceleration and displacement response in the
longitudinal, transverse and vertical directions at the top of the pylon is shown in Figure 6. It
can be observed that larger displacement and acceleration responses occur in the transverse
Figure 5. Continued.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 122
Figure 6. Comparison of estimated acceleration and displacement response at the top of pylon using truss
elements for cables and nonlinear beam elements for cables.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 123
direction at the top of pylon when the truss model is used for the cable; by contrast, the
longitudinal direction at the top of the pylon has the most significant responses when using the
nonlinear beam elements. Figure 7 compares the acceleration and displacement response at
the northern end of the bridge deck when using truss elements for the cables (blue line) and
nonlinear beam elements for the cables (red line). It is found that use of the truss model for the
cables induces larger acceleration response at both ends of the bridge deck. For estimation of
transverse direction at both ends of the bridge, a larger displacement is observed by using the
nonlinear beam element model for the cables. It is believed that use of nonlinear beam elements
to model for the cables is a reasonable and necessary approach for predicting behavior of the
stayed cables. Therefore, in what follows only the modeling of the bridge cables using the
nonlinear beam elements for the cables is used to evaluate the control effectiveness. It is believed
that using truss elements to model the cable for the cable-stayed bridge may induce significant
bias on the response calculation.
CONTROL SYSTEM DESIGN
In this study a controller is designed to serve as an active control system for the equivalent linear
cable-stayed bridge model. Readings from acceleration and displacement transducers are fed
back to the control algorithm. The sample control system employs a total of 12 hydraulic
actuators located at both ends of the bridge between the deck and the top of each pier (6
actuators for each pier); actuators are oriented to apply forces longitudinally (X-direction). It is
assumed that each actuator can provide up to 1500 kN.
A total of 19 sensors are used to collect the response of the bridge and are used for control
purposes. The locations of these sensors are specified as follows:
(a) four sensors: acceleration and displacement sensors at the northern end of the deck (in
both longitudinal (x) and transverse (z) directions;
(b) four sensors: acceleration and displacement sensors at the southern end of the deck (in
both longitudinal (x) and transverse (z) directions;
(c) four sensors: acceleration and displacement sensors at the middle node of the deck
(in both longitudinal (x) and transverse (z) directions;
(d) four sensors: acceleration and displacement sensors at the top of the pylon (in both
longitudinal (x) and transverse (z) directions;
(e) three sensors: displacement sensors located at the top of all three piers in the longitudinal
direction.
Both sensors from bridge and cables are measured. In the present study the cable sensors
sometimes are and sometimes are not connected to the controller. This is done to study the effect
of the control on the bridge as well as on the vibration of the cables.
CONTROL DESIGN PROBLEM
A test bed for development of effective control strategies for a cable-stayed bridge has also been
developed with emphasis on nonlinear behavior of the stayed-cable. This sample active control
design study defines the excitation, evaluated model, devices, sensors, and control algorithms. A
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 124
Figure 7. Comparison of estimated acceleration and displacement response at the northern end of the
bridge deck using truss elements for cables and nonlinear beam elements for cables.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 125
SIMULINK block diagram for this active control study is shown in Figure 8. It is patterned for
SIMULINK block diagram provided by Dyke et al. [7] and Caicedo et al. [17]. A detailed
description of the SIMULINK block diagram is now given.
Excitation
The earthquake record from the Chi-Chi earthquake can be selected as the excitation to this
cable-stayed bridge. Each earthquake record contains three-dimensional acceleration and
velocity data (from integration of acceleration data). To consider the spatial variation of
earthquake excitation, uniform input with a phase delay was specified (4 km/s apparent wave
velocity was assumed).
Evaluation model
The evaluation model in SIMULINK contains two input ports and six output ports (Figure 9).
The excitation input port includes nine acceleration values and nine velocity values to describe
the multi-support problem in three-dimensional and the time-delayed excitation. The input port
of the control force is developed from the design of actuators or dampers. The evaluated output
serves to determine the control efficiency of various control criteria. The output port of sensors
gives the predicted responses of the deck or the pylon (or piers). At the same time, the output of
Figure 8. Modified SIMULINK block diagram with active control.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 126
sensors is used to estimate the control forces which include a Kalman filter estimator for
feedback control. The output port of cable sensors gives the responses of the cables. If the
control algorithm focuses on the cables, these cable sensors can also be used to determine the
feedback control. The device sensors provide the responses of actuators or dampers between
connections. The output port of tension estimates the variation of cable tensile force using the
deformation of cables. The cable deformations are selected to consider the dimensionless
responses at the middle node of each cable.
Sensor processor
The sensor processor converts the responses of the bridge to the voltage signal. The converter
for each sensor has a range of Æ10 V: Each of the measured responses contains a noise level
with an rms value of 0.01 V. Measurement noises are modeled with Gaussian rectangular pulses
that have a pulse width equal to the integration step. Figure 10 shows the SIMULINK block of
the sensor processor.
Controller
The controller contains a signal converter and a force estimator. The signal converter mainly
transforms the sensors signal into real responses with a constraint. The constraint within the
signal converter has an upper bound (þ10 V) and a lower bound (À10 V). In regard to the force
estimator, it follows with H
2
control algorithm and Kalman estimator for use of actuators.
First, an appropriate design model must be developed. The design model is formed from the
evaluation model by choosing suitable modes of the system. In this case, in order to use
actuators to control the responses between the deck and piers (end-abutments), the piers (1 and 3)
or abutments are selected to apply the control devices. The control design model contains 44
modes that significantly affect responses of the deck and piers. Originally, the first 100 modes of
Figure 9. SIMULINK block: evaluation model.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 127
the whole bridge system were used to obtain the control gain using H
2
control theory. In real
applications, it may waste much time to calculate the optimal control forces if the matrix of the
control gain has a large dimension. However, there are some modes from the first 100 modes
which do not directly and strongly contribute to control the system. Therefore, these modes with
minor contributions in control gain are eliminated and remain the 44 modes which can be used
to generate the optimal control force.
The measured outputs focus on the sensor outputs (responses of the deck and the pylon). The
state-space (discrete form) system of the control design model is represented as follows
x½n þ1Š ¼ A
dcon
x½nŠ þ B
dcon
u½nŠ þ E
dcon
. x
g
½nŠ
’ x
g
½nŠ
_ _
ð15Þ
y
s
½nŠ ¼ C
dcon
x½nŠ þ D
dcon
. x
g
½nŠ
’ x
g
½nŠ
_ _
þ F
dcon
u½nŠ ð16Þ
where x is the state vector, and y is a vector of sensor outputs. The controller employs an H
2
control algorithm to estimate the appropriate control force. To obtain the optimal control force,
the external disturbances, such as the ground excitations, are assumed to be independent with
respect to the control force. First, the objective function is defined as
J
2
¼

k!1
k¼k
0
x
T
½kŠQx½kŠ þ u
T
½kŠRu½kŠ ð17Þ
where R is a weighting matrix related to the optimal control force, Q is a weighting matrix
related to the system of the design model, and k
0
is the initial time. Through the computation of
the variation method, the optimal control force can be obtained as
A
T
dcon
P½kŠA
dcon
À A
T
dcon
P½kŠB
dcon
ð2R þ B
T
dcon
P½kŠB
dcon
Þ
À1
B
T
dcon
P½kŠA
dcon
þ 2Qj
k!1
¼ 0 ð18Þ
u½kŠ ¼ Àð2R þ B
T
dcon
PB
dcon
Þ
À1
B
T
dcon
PA
dcon
x½kŠ ¼ Gx½kŠ ð19Þ
Figure 10. SIMULINK block: sensor processor.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 128
where Equation (18) is the Riccati equation, P in Equation (19) is the solution of Equation (18),
and G is the optimal control gain. Although the control gain can be obtained from Equation
(19), the optimal control force needs the full state vector of the design model. Therefore, the full
state vector must be computed from the Kalman estimator, given by
x
_
½k þ1Š ¼ A
dcon
x
_
½kŠ þ B
dcon
u½kŠ þ Lðy
s
½kŠ À C
dcon
x
_
½kŠ À F
dcon
u½kŠÞ ð20Þ
where L is the Kalman gain solved from the Ricatti equation. The optimal control force
combined with the estimator (Equation 20) can be obtained as
u½kŠ ¼ Gx
_
½kŠ ð21Þ
and the completed controller can be expressed as
x
_
½k þ 1Š ¼ ðA
dcon
þ B
dcon
G À LC
dcon
À LD
dcon
GÞ x
_
½kŠ þ Ly
s
½kŠ ð22Þ
u½k þ1Š ¼ Gx
_
½k þ 1Š ð23Þ
A SIMULINK block form of the controller is shown in Figure 11.
Control devices
The block of control devices, as shown in Figure 12, contains a converter for the actuators
(forces to voltages) and a real output force. The converter for the actuators transforms
command forces into corresponding voltages for single actuators. The real output force is
indicated by the force from single actuator multiplied by number of control devices. The
connection between the command forces and the output forces is limited by bounds on the
capacity of the device, which is 1500 kN per actuator and the minimum force is À1500 kN per
actuator.
Two kinds of control devices are employed for the vibration Gi-Lu Bridge. The first type of
control device is the actuator, and the other kind of control device is the viscous damper. Two
strategies are used to arrange the locations of the control devices:
Strategy 1: actuators are placed at the top of the Piers 1 and 3 in connection with the deck. Six
actuators are placed between the deck and piers at each location. The voltage command for each
actuator with respect to the output control force is 10 V per 1500 kN. The limitation of actuators
between the control force and the desired force is bounded from 1500 to À1500 kN.
Figure 11. SIMULINK block: controller.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 129
Strategy 2: besides using the Strategy 1 dampers are also implemented between cables and the
bridge deck (which is placed between cables and the deck with 1 piece at each cable). Each
damper is connected at the lowest nodal point of the cable relative to the deck. Based on the
uncontrolled responses of the Gi-Lu Bridge to six specified excitations, four types of viscous
dampers are selected. The damping coefficient of all types of dampers is: 120 000 (Ns/m) for
cable 1 through 10, 42 857 (Ns/m) for cable 11 and 12, 30 000 (Ns/m) for cable 13 and 14, and
7500 (Ns/m) for cable 15 through 17, respectively. The input used to drive the dampers is the
relative velocity response between the connecting cables and the deck.
Figure 13 shows the SIMULINK block of the strategy 2 control devices. Figure 14 shows the
schematic diagram of the locations of actuators and viscous dampers for control strategies 1 and
2. On the practical implementation of the dampers in stayed cable will discuss later.
Figure 13. SIMULINK block: control devices (case B).
Figure 12. SIMULINK block: control devices (Strategy 1).
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 130
Evaluation criteria
To evaluate the performance of the control algorithm used in this study, the shear forces and
moments in the pylon and piers induced by earthquake excitation must be considered. Thus,
these evaluation criteria must consider the ability of the controller to reduce the peak responses,
the normalized responses over the entire time record, and the control requirements. For this
purpose, a set of 45 criteria have been employed to evaluate the ability of each control strategy.
Shear forces and moments at each pier, including the base of the pylon, are considered in the
evaluation criteria. Other criteria contain the displacements at the top node of the pylon, the
middle node on the deck, and the north and south ends of the deck. Additionally, the behavior
of the cable tension force and the deformation of the middle node of each cable are important to
observe. Because the earthquake is assumed to have three-dimensional components at a
specified incidence angle, several criteria are evaluated in both the X (longitudinal) and Z
(transverse) directions. These evaluation criteria are listed and shown in Table V.
CASE STUDY OF CONTROL EVALUATION
The basic problem for control of the Gi-Lu cable-stayed bridge in this study focuses on
modeling of the stayed cables. Both truss and nonlinear beam elements are selected to model the
cable. The effect of cable modeling on the seismic response of cable-stayed bridge is examined
first. Earthquake ground motion data from the Chi-Chi earthquake from recorded station
TCU089 is selected as the excitation (PGA in the EW direction is 244 gal, in the NS direction
324 gal and in vertical direction 190 gal). In this study a uniform input motion with three
components (two horizontal and one vertical component) is used. Figure 15 shows the
acceleration response spectrum of the recorded data. Before the evaluation of control
effectiveness of this bridge is considered response of uncontrolled case using different models
is examined first.
As discussed earlier there are two cases of control devices in this study, control strategy 1:
using actuators only and control strategy 2: using actuators at both ends of the deck and
Figure 14. (a) Location of actuators between deck and abutment; and (b) location of viscous damper
between deck and cable.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 131
Table IV. Comparison of the control effectiveness using two different control systems: Case A: putting
actuators between the abutments (both ends) and the bridge deck; Case B: besides the actuators (in Case A)
viscous dampers were also added on the cables.
Case A Case B
J
1
1.044 0.848
J
2
0.937 0.921
J
3
0.926 0.825
J
4
1.0 0.999
J
5
0.999 0.999
J
6
1.0 0.999
J
7
1.067 0.882
J
8
1.0 0.998
J
9
1.0 0.999
J
10
1.0 0.999
J
11
1.0 0.999
J
12
0.921 0.879
J
13
0.922 0.868
J
14
0.911 0.694
J
15
1.0 0.999
J
16
1.010 0.846
J
17
0.956 0.755
J
18
1.011 0.844
J
19
0.950 0.750
J
20
0.999 0.999
J
21
1.0 1.001
J
22
0.999 0.999
J
23
0.995 0.789
J
24
0.999 1.030
J
25
1.0 0.999
J
26
1.0 1.004
J
27
1.0 0.999
J
28
0.721 0.443
J
29
0.970 0.756
J
30
0.721 0.444
J
31
0.999 1.030
J
32
0.995 0.789
J
33
0.802 0.807
J
34
0.814 0.750
J
35
1.0 0.999
J
36
0.878 0.894
J
37
0.893 0.739
J
38
0.999 1.001
J
39
0.878 0.773
J
40
0.872 0.700
J
41
1.0 0.999
J
42
0.961 0.811
J
43
0.888 0.783
J
44
0.999 1.001
J
45
1.0 1.0
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 132
attaching viscous dampers to the cables. Figure 16 compares the displacement and acceleration
responses at the north end of the deck for the uncontrolled case and the case of using control
strategy 1. The proposed control devices are shown to effectively reduce the longitudinal
displacement response of the deck. If control strategy 2 is used, an improved level of control can
be observed, as shown in Figure 17. This is the case because viscous dampers are added between
deck and cables in control strategy 2, which affords a significant reduction in the response of the
cables as is clearly shown in Figure 18.
Based on the proposed 45 evaluation criteria the performance indices are evaluated for the
controlled and uncontrolled cases. On the left of Figure 19(a) shows the evaluation criteria with
respect to different indices for control strategy 1. The majority of the evaluation criteria have
values less than one, except for indices J
1
(longitudinal shear force at the first pier) and J
7
(longitudinal shear force at the base of the pylon). This means that by putting actuators at both
ends of the deck, the pylon base shear force is larger than for uncontrolled case. The right hand
side of Figure 19(a) shows the estimated value of the normalized displacement at the midpoint
of each cable for control case A. For control strategy 2 the evaluation criteria and the estimated
normalized displacement at the midpoint of each cable is shown in Figure 19(b). It is observed
that control strategy 2 provides an improved level of control effectiveness not only for the
response of the bridge itself but also for the stayed cables. Most of the displacement at the
midpoint of the cable is reduced after control. Table IV compares the normalized displacement
(with respect to the uncontrolled case) of the evaluation criteria for both control cases A and B.
Figure 20 shows the acceptable cable tension between the provided ranges after the control (for
control strategy 2).
Figure 15. Plot of acceleration response spectrum of ground motion data collected from Station TCU089,
Chi-Chi earthquake.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 133
DISCUSSIONS ON CONTROL STRATEGY
In control strategy 2 dampers were attached between cable and the bridge deck. Because only
eight nonlinear beam elements were modeled for cable and the lowest nodal point was used to
connect the damper to the bridge deck, considering the height of the lowest nodal point of the
cable it is impractical to implement the damper for such a location (almost 6.9 m above the deck
level) under control strategy 2. Therefore an alternative control strategy, control strategy 3, was
used to implement the dampers for the cable. Instead of installing the damper between the cable
and the bridge deck, forty dampers are placed between the deck and cables through cable 1 to
cable 10 at each side. Remaining dampers are placed at the middle of the cable connecting two
neighboring cables through cable 10 to cable 17 at each side, as shown in Figure 21. In this case,
three types of MR dampers are selected to control cable vibration. Type I of MR dampers is
used to position between cable 4 to cable 9 and the deck. Type II MR dampers are used between
cables 1–3 and the deck, between cable 10 and the deck, and between cable 10–12. Type III MR
dampers are positioned between cables 13–17.
Figure 16. Comparison of the estimated displacement and acceleration responses between uncontrolled
and controlled case using strategy 1: (a) longitudinal responses at northern end of the deck; (b) longitudinal
responses at southern end of the deck; and (c) longitudinal responses at the top of pylon.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 134
The model of the MR damper for use in structural control of this study combines the
modified bi-viscous model and the bilinear model [19]. The command voltage is sent to the
modified bi-viscous model with the states as input and then generates the corresponding force.
The voltage command ranged from 0 to 1.2 V. A constraint function of voltage with a second-
order polynomial function is used to obtain the lower bound of the force. If the corresponding
force is greater than the lower bound, the output force is still the force generated by the modified
bi-viscous model. On the contrary, the output force is changed to obtain the bilinear model.
There are three types of MR dampers including: (1) 120 mm stroke, 350 mm/s maximum
velocity, and 30 kN force capacity; (2) 120 mm stroke, 1050 mm/s maximum velocity, and
30 kN force capacity; and (3) 240 mm stroke, 3500 mm/s maximum velocity, and 30 kN force
capacity.
Figure 22 shows the comparison on the 45 evaluation criteria of three different control
systems. It is found that J
1
, J
2
, J
3
, and J
7
are larger for strategy 3 using the input motion from
TCU089. But the other indices are all smallest for using strategy 3 and the control of cable
Figure 17. Comparison on the estimated displacement and acceleration responses between uncontrolled
and controlled case of strategy 2: (a) longitudinal responses at northern end of the deck; (b) longitudinal
responses at southern end of the deck; and (c) longitudinal responses at the top of pylon.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 135
Figure 18. Responses of cable (EN-1): (a) longitudinal displacement; (b) vertical displacement; (c)
transverse displacement; (d) longitudinal acceleration; (e) vertical acceleration; and (f) transverse
acceleration.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 136
vibration the strategy 3 provides much better results than others. Figure 23 shows the
comparison on the cable vibration. It is also found that with the control strategy 3 the vibration
of cable was significantly reduced.
Figure 19. Plot of evaluation criteria with respect to different index, and the normalized displacement of
middle point of all cables: (a) for case A control; and (b) for case B control. (Note: The number of D
indices for 1–17: R1, R3,. . ., R31, R33; for 18–34: R2, R4,. . ., R32, R34; for 35–51: L1, L3,. . ., L31, L33;
for 52–68: L2, L2,. . ., L32, L34).
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 137
T
a
b
l
e
V
.
S
u
m
m
a
r
y
o
f
e
v
a
l
u
a
t
i
o
n
c
r
i
t
e
r
i
a
.
J
m
¼
m
a
x
m
;
t
j
F
p
m
x
ð
t
Þ
j
F
0
x
;
m
m
a
x
_
_
;
m
¼
1
3
F
p
m
x
i
n
d
i
c
a
t
e
s
t
h
e
l
o
n
g
i
t
u
d
i
n
a
l
(
x
-
d
i
r
e
c
t
i
o
n
)
s
h
e
a
r
f
o
r
c
e
a
t
m
t
h
p
i
e
r
;
F
0
x
;
m
m
a
x
i
s
t
h
e
m
a
x
i
m
u
m
u
n
c
o
n
t
r
o
l
l
e
d
s
h
e
a
r
f
o
r
c
e
a
t
m
t
h
p
i
e
r
;
m
¼
1
(
n
o
r
t
h
p
i
e
r
)
,
2
(
p
i
e
r
b
e
l
o
w
p
y
l
o
n
)
,
3
(
s
o
u
t
h
p
i
e
r
)
.
J
n
¼
m
a
x
m
;
t
j
F
p
m
z
ð
t
Þ
j
F
0
z
;
m
m
a
x
_
_
;
n
¼
4
2
6
F
p
m
z
i
n
d
i
c
a
t
e
s
t
h
e
t
r
a
n
s
v
e
r
s
e
(
z
-
d
i
r
e
c
t
i
o
n
)
s
h
e
a
r
f
o
r
c
e
a
t
m
t
h
p
i
e
r
;
F
0
z
;
m
m
a
x
i
s
t
h
e
m
a
x
i
m
u
m
u
n
c
o
n
t
r
o
l
l
e
d
s
h
e
a
r
f
o
r
c
e
a
t
m
t
h
p
i
e
r
.
J
7
¼
m
a
x
t
j
F
d
x
ð
t
Þ
j
F
0
d
x
m
a
x
_
_
F
d
x
i
s
t
h
e
l
o
n
g
i
t
u
d
i
n
a
l
(
x
-
d
i
r
e
c
t
i
o
n
)
s
h
e
a
r
f
o
r
c
e
a
t
t
h
e
d
e
c
k
l
e
v
e
l
o
f
t
h
e
p
y
l
o
n
;
F
0
d
x
m
a
x
i
n
d
i
c
a
t
e
s
t
h
e
m
a
x
i
m
u
m
u
n
c
o
n
t
r
o
l
l
e
d
s
h
e
a
r
f
o
r
c
e
a
t
t
h
e
d
e
c
k
l
e
v
e
l
o
f
t
h
e
p
y
l
o
n
.
J
8
¼
m
a
x
t
j
F
d
z
ð
t
Þ
j
F
0
d
z
m
a
x
_
_
F
d
z
i
s
t
h
e
t
r
a
n
s
v
e
r
s
e
(
z
-
d
i
r
e
c
t
i
o
n
)
s
h
e
a
r
f
o
r
c
e
a
t
t
h
e
d
e
c
k
l
e
v
e
l
o
f
t
h
e
p
y
l
o
n
;
F
0
d
z
m
a
x
i
n
d
i
c
a
t
e
s
t
h
e
m
a
x
i
m
u
m
u
n
c
o
n
t
r
o
l
l
e
d
s
h
e
a
r
f
o
r
c
e
a
t
t
h
e
d
e
c
k
l
e
v
e
l
o
f
t
h
e
p
y
l
o
n
.
J
l
¼
m
a
x
m
;
t
M
p
m
x
ð
t
Þ
M
0
x
;
m
m
a
x
_
_
;
l
¼
9
1
1
M
p
m
x
i
n
d
i
c
a
t
e
s
t
h
e
l
o
n
g
i
t
u
d
i
n
a
l
(
x
-
d
i
r
e
c
t
i
o
n
)
m
o
m
e
n
t
a
t
m
t
h
p
i
e
r
;
M
0
x
;
m
m
a
x
i
s
t
h
e
m
a
x
i
m
u
m
u
n
c
o
n
t
r
o
l
l
e
d
m
o
m
e
n
t
a
t
m
t
h
p
i
e
r
.
J
k
¼
m
a
x
m
;
t
M
p
m
z
ð
t
Þ
M
0
z
;
m
m
a
x
_
_
;
k
¼
1
2
2
1
4
M
p
m
z
i
n
d
i
c
a
t
e
s
t
h
e
t
r
a
n
s
v
e
r
s
e
(
z
-
d
i
r
e
c
t
i
o
n
)
m
o
m
e
n
t
a
t
m
t
h
p
i
e
r
;
M
0
x
;
m
m
a
x
i
s
t
h
e
m
a
x
i
m
u
m
u
n
c
o
n
t
r
o
l
l
e
d
m
o
m
e
n
t
a
t
m
t
h
p
i
e
r
.
J
1
5
¼
m
a
x
t
M
d
x
ð
t
Þ
M
0
d
x
m
a
x
_
_
M
d
x
i
n
d
i
c
a
t
e
s
t
h
e
l
o
n
g
i
t
u
d
i
n
a
l
(
x
-
d
i
r
e
c
t
i
o
n
)
m
o
m
e
n
t
a
t
t
h
e
d
e
c
k
l
e
v
e
l
o
f
t
h
e
p
y
l
o
n
;
M
0
d
x
m
a
x
i
s
t
h
e
m
a
x
i
m
u
m
u
n
c
o
n
t
r
o
l
l
e
d
m
o
m
e
n
t
a
t
t
h
e
s
a
m
e
l
o
c
a
t
i
o
n
.
J
1
6
¼
m
a
x
t
M
d
z
ð
t
Þ
M
0
d
z
m
a
x
_
_
M
d
z
i
n
d
i
c
a
t
e
s
t
h
e
t
r
a
n
s
v
e
r
s
e
(
z
-
d
i
r
e
c
t
i
o
n
)
m
o
m
e
n
t
a
t
t
h
e
d
e
c
k
l
e
v
e
l
o
f
t
h
e
p
y
l
o
n
;
M
0
d
z
m
a
x
i
s
t
h
e
m
a
x
i
m
u
m
u
n
c
o
n
t
r
o
l
l
e
d
m
o
m
e
n
t
a
t
t
h
e
s
a
m
e
l
o
c
a
t
i
o
n
.
J
i
¼
n
o
r
m
m
;
t
ð
j
F
p
m
x
ð
t
Þ
j
Þ
F
0
x
;
m
n
o
r
m
_
_
;
i
¼
1
7
2
1
9
F
0
x
;
m
n
o
r
m
i
s
t
h
e
n
o
r
m
e
d
v
a
l
u
e
o
f
u
n
c
o
n
t
r
o
l
l
e
d
s
h
e
a
r
f
o
r
c
e
s
i
n
t
h
e
l
o
n
g
i
t
u
d
i
n
a
l
(
x
-
d
i
r
e
c
t
i
o
n
)
d
i
r
e
c
t
i
o
n
a
t
m
t
h
p
i
e
r
;
w
h
e
r
e
n
o
r
m
ð

Þ
¼




















t
f
t
¼
t
0

ð
t
Þ
2
_
:
J
j
¼
n
o
r
m
m
;
t
ð
j
F
p
m
z
ð
t
Þ
j
Þ
F
0
z
;
m
n
o
r
m
_
_
;
J
¼
2
0
2
2
2
F
0
z
;
m
n
o
r
m
i
s
t
h
e
n
o
r
m
e
d
v
a
l
u
e
o
f
u
n
c
o
n
t
r
o
l
l
e
d
s
h
e
a
r
f
o
r
c
e
s
i
n
t
h
e
t
r
a
n
s
v
e
r
s
e
(
z
-
d
i
r
e
c
t
i
o
n
)
d
i
r
e
c
t
i
o
n
a
t
m
t
h
p
i
e
r
.
J
2
3
¼
n
o
r
m
t
ð
j
F
d
x
ð
t
Þ
j
Þ
F
0
d
x
n
o
r
m
_
_
F
0
d
x
n
o
r
m
i
s
t
h
e
n
o
r
m
e
d
v
a
l
u
e
o
f
u
n
c
o
n
t
r
o
l
l
e
d
s
h
e
a
r
f
o
r
c
e
s
o
f
t
h
e
p
y
l
o
n
i
n
t
h
e
l
o
n
g
i
t
u
d
i
n
a
l
(
x
-
d
i
r
e
c
t
i
o
n
)
d
i
r
e
c
t
i
o
n
a
t
t
h
e
d
e
c
k
l
e
v
e
l
.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 138
J
2
4
¼
n
o
r
m
t
ð
j
F
d
z
ð
t
Þ
j
Þ
F
0
d
z
n
o
r
m
_
_
F
0
d
z
n
o
r
m
i
s
t
h
e
n
o
r
m
e
d
v
a
l
u
e
o
f
u
n
c
o
n
t
r
o
l
l
e
d
s
h
e
a
r
f
o
r
c
e
s
o
f
t
h
e
p
y
l
o
n
i
n
t
h
e
t
r
a
n
s
v
e
r
s
e
(
z
-
d
i
r
e
c
t
i
o
n
)
d
i
r
e
c
t
i
o
n
a
t
t
h
e
d
e
c
k
l
e
v
e
l
.
J
w
¼
n
o
r
m
m
;
t
ð
j
M
p
m
x
ð
t
Þ
j
Þ
M
0
x
;
m
n
o
r
m
_
_
;
ð
w
¼
2
5
2
2
7
Þ
M
0
x
;
m
n
o
r
m
i
s
t
h
e
n
o
r
m
e
d
v
a
l
u
e
o
f
u
n
c
o
n
t
r
o
l
l
e
d
m
o
m
e
n
t
s
i
n
t
h
e
l
o
n
g
i
t
u
d
i
n
a
l
(
x
-
d
i
r
e
c
t
i
o
n
)
d
i
r
e
c
t
i
o
n
a
t
m
t
h
p
i
e
r
.
J
s
¼
n
o
r
m
m
;
t
ð
j
M
p
m
z
ð
t
Þ
j
Þ
M
0
z
;
m
n
o
r
m
_
_
;
ð
s
¼
2
8
2
3
0
Þ
M
0
z
;
m
n
o
r
m
i
s
t
h
e
n
o
r
m
e
d
v
a
l
u
e
o
f
u
n
c
o
n
t
r
o
l
l
e
d
m
o
m
e
n
t
s
i
n
t
h
e
t
r
a
n
s
v
e
r
s
e
(
z
-
d
i
r
e
c
t
i
o
n
)
d
i
r
e
c
t
i
o
n
a
t
m
t
h
p
i
e
r
.
J
3
1
¼
n
o
r
m
t
ð
j
M
d
x
ð
t
Þ
j
Þ
M
0
d
x
n
o
r
m
_
_
M
0
d
x
n
o
r
m
i
s
t
h
e
n
o
r
m
e
d
v
a
l
u
e
o
f
u
n
c
o
n
t
r
o
l
l
e
d
m
o
m
e
n
t
s
o
f
t
h
e
p
y
l
o
n
i
n
t
h
e
l
o
n
g
i
t
u
d
i
n
a
l
(
x
-
d
i
r
e
c
t
i
o
n
)
d
i
r
e
c
t
i
o
n
a
t
t
h
e
d
e
c
k
l
e
v
e
l
.
J
3
2
¼
n
o
r
m
t
ð
j
M
d
z
ð
t
Þ
j
Þ
M
0
d
z
n
o
r
m
_
_
M
0
d
z
n
o
r
m
i
s
t
h
e
n
o
r
m
e
d
v
a
l
u
e
o
f
u
n
c
o
n
t
r
o
l
l
e
d
m
o
m
e
n
t
o
f
t
h
e
p
y
l
o
n
i
n
t
h
e
t
r
a
n
s
v
e
r
s
e
(
z
-
d
i
r
e
c
t
i
o
n
)
d
i
r
e
c
t
i
o
n
a
t
t
h
e
d
e
c
k
l
e
v
e
l
.
J
3
3
¼
m
a
x
t
j
x
p
1
x
ð
t
Þ
j
x
0
x
;
1
m
a
x
_
_
;
J
3
4
¼
m
a
x
t
j
x
p
1
y
ð
t
Þ
j
x
0
y
;
1
m
a
x
_
_
;
J
3
5
¼
m
a
x
t
j
x
p
1
z
ð
t
Þ
j
x
0
z
;
1
m
a
x
_
_
x
:
l
o
n
g
i
t
u
d
i
n
a
l
,
y
:
v
e
r
t
i
c
a
l
,
z
:
t
r
a
n
s
v
e
r
s
e
S
u
b
s
c
r
i
p
t
(
1
)
i
n
d
i
c
a
t
e
s
t
h
e
n
o
r
t
h
e
n
d
o
f
t
h
e
d
e
c
k
d
i
s
p
l
a
c
e
m
e
n
t
s
a
t
t
h
e
n
o
r
t
h
e
n
d
o
f
t
h
e
d
e
c
k
l
e
v
e
l
(
a
t
t
h
e
t
o
p
o
f
P
i
e
r
1
)
,
a
n
d
x
0
;
1
m
a
x
i
s
t
h
e
m
a
x
i
m
u
m
d
i
s
p
l
a
c
e
m
e
n
t
o
f
t
h
e
u
n
c
o
n
t
r
o
l
l
e
d
r
e
s
p
o
n
s
e
.
J
3
6
¼
m
a
x
t
j
x
p
2
x
ð
t
Þ
j
x
0
x
;
2
m
a
x
_
_
;
J
3
7
¼
m
a
x
t
j
x
p
2
y
ð
t
Þ
j
x
0
y
;
2
m
a
x
_
_
;
J
3
8
¼
m
a
x
t
j
x
p
2
z
ð
t
Þ
j
x
0
z
;
2
m
a
x
_
_
S
u
b
s
c
r
i
p
t
(
2
)
i
n
d
i
c
a
t
e
s
t
h
e
m
i
d
d
l
e
n
o
d
e
o
f
t
h
e
d
e
c
k
d
i
s
p
l
a
c
e
m
e
n
t
s
a
t
t
h
e
m
i
d
d
l
e
n
o
d
e
o
f
t
h
e
d
e
c
k
l
e
v
e
l
(
a
t
t
h
e
b
o
t
t
o
m
o
f
t
h
e
p
y
l
o
n
)
.
J
3
9
¼
m
a
x
t
j
x
p
3
x
ð
t
Þ
j
x
0
x
;
3
m
a
x
_
_
;
J
4
0
¼
m
a
x
t
j
x
p
3
y
ð
t
Þ
j
x
0
y
;
3
m
a
x
_
_
;
J
4
1
¼
m
a
x
t
j
x
p
3
z
ð
t
Þ
j
x
0
z
;
3
m
a
x
_
_
S
u
b
s
c
r
i
p
t
(
3
)
i
n
d
i
c
a
t
e
s
t
h
e
s
o
u
t
h
e
r
n
e
n
d
o
f
t
h
e
d
e
c
k
d
i
s
p
l
a
c
e
m
e
n
t
s
a
t
t
h
e
s
o
u
t
h
e
r
n
e
n
d
o
f
t
h
e
d
e
c
k
l
e
v
e
l
(
a
t
t
h
e
t
o
p
o
f
p
i
e
r
3
)
.
J
4
2
¼
m
a
x
t
j
x
t
p
x
ð
t
Þ
j
x
0
x
;
t
p
m
a
x
_
_
;
J
4
3
¼
m
a
x
t
j
x
t
p
y
ð
t
Þ
j
x
0
y
;
t
p
m
a
x
_
_
;
J
4
4
¼
m
a
x
t
j
x
t
p
z
ð
t
Þ
j
x
0
z
;
t
p
m
a
x
_
_
S
u
b
s
c
r
i
p
t
(
t
p
)
i
n
d
i
c
a
t
e
s
t
h
e
t
o
p
o
f
t
h
e
p
y
l
o
n
d
i
s
p
l
a
c
e
m
e
n
t
s
a
t
t
h
e
t
o
p
o
f
t
h
e
p
y
l
o
n
.
J
4
5
¼
m
a
x
t
;
v
ð
F
c
o
n
;
v
ð
t
Þ
Þ
F
c
o
n
;
c
a
p
a
c
i
t
y
_
_
;
F
c
o
n
,
c
a
p
a
c
i
t
y
=
t
h
e
c
a
p
a
c
i
t
y
o
f
c
o
n
t
r
o
l
d
e
v
i
c
e
s
T
h
e
m
a
x
i
m
u
m
c
o
n
t
r
o
l
f
o
r
c
e
n
o
r
m
a
l
i
z
e
d
t
o
t
h
e
c
a
p
a
c
i
t
y
o
f
c
o
n
t
r
o
l
d
e
v
i
c
e
s
(
c
a
p
a
c
i
t
y
o
f
c
o
n
t
r
o
l
d
e
v
i
c
e
s
i
s
a
s
s
u
m
e
d
t
o
b
e
1
5
0
0
k
N
)
.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 139
CONCLUSIONS
This paper presents a detailed study of structural control for a seismically excited cable-
stayed bridge. The goal of this paper is to develop an evaluation model for control that portrays
salient features of the structural system, particularly the sag cables and their pre-tension
forces. To this end, a MATLAB-based computer program has been developed to simulate a
three-dimensional cable-stayed bridge; the code includes the nonlinear beam elements that
represent the sag cables. Both active control and hybrid-control algorithms are given to
illustrate some of the design challenges of the problem. Evaluation criteria are presented for the
design problems that are consistent with the goals of seismic response control of a cable-stayed
bridge.
Figure 20. Variation of cable tension force for un-controlled case and control case B.
MR Damper with
constant Voltage
VE Dampers




Figure 21. Schematic diagram of the location of dampers in the middle point of cables (Line 10–17) and
between deck and cable (Line 1–10).
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 140
Through this study the following conclusions are made:
(a) A comparison is made between simulation results using either truss elements or nonlinear
beam elements to represent dynamic behavior of the stayed cable of the Gi-Lu bridge.
Significant differences in the dynamic characteristics of cable-stayed bridge are estimated
using different models. To form the stiffness matrix of the cable the truss element can only
consider the linear system with the basic properties (Young’s modulus, cross-sections and
length of elements) for formulation. On the contrary, the nonlinear beam elements not only
use the conventional linear stiffness system (partially same as truss elements), but also
include the geometrically nonlinear beam elements to form the stiffness matrix and the self-
weight of the cable and the pre-tension force in the cable can be implemented. Using data
from field tests it is believed that the nonlinear beam elements to simulate the stayed cable
in a realistic manner.
(b) Comparison on the responses of the cable-stayed bridge, as shown in Figures 6 and 7,
cannot prove that using the nonlinear beam element to represent the cable is more
reasonable than using the truss element. But there are some defects associated with using
the truss element for cable, e.g., using truss elements cannot obtain the bridge out-plane
responses and underestimates the cable tensile force.
(c) The simulation results show that use of the actuators as control devices between the bridge
deck and both end-abutments is not very effective for controlling the vibration of the bridge
deck, pylon, and the stayed cables.
(d) Instead of using an active control method, hybrid control devices are applied to the
structure to control response due to seismic excitation. Because nonlinear beam elements
are used to simulate the stayed cable, it is a relatively simple matter to add dampers between
cables and the deck. It is shown that adding viscous dampers to each cable can provide
Figure 22. Comparison of evaluation criteria of three different control systems.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 141
good control effectiveness, that not only the reduces the deck and pylon response, but also
the response of the cables.
(e) With implementation of extra viscous dampers between the deck and the stayed cables, the
displacement at the top of the pylon (J
42
and J
43
) in both longitudinal and transverse
directions and shear force at the bottom of pylon (J
7
) and at the top of the pier 3 (J
3
) in the
longitudinal direction are significantly reduced. For practical consideration the control
strategy 3 can also provide good control effectiveness in most of the evaluation criteria.
(f) In this study the inherent damping ratio for cable is assumed 1% and with the
supplementation of damping ratio from the viscous damper in the cable (1.3–20% from
damper) the damping ratio for cable is ranging from 2.3% (longest cable) to 21% (shortest
cable).
(g) A formulation of the structural model and a simple control design has been made in the
form of a set of MATLAB equations. This code can provide another type of benchmark
Figure 23. Comparison of the normalized displacement at the middle point of all cables: (a) North-West;
(b) South-West; (c) North-East; and (d) South-East sides.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
C.-H. LOH AND C.-M. CHANG 142
structural control problem for cable-stayed bridges. The numerical results from use of a
hybrid control system look promising; however, a number of practical aspects not
considered here need to be tackled in further studies in order to gain a deeper evaluation of
the potential effectiveness for mitigation of vibration.
ACKNOWLEDGEMENT
The authors wish to express their thanks to Dr Z. K. Lee (Associate Research Fellow of NCREE) to
provide experimental data of Gi-Lu bridge for this study and the support from National Science Council
under Grant NSC93-2211-E-002-005 is also acknowledged.
REFERENCES
1. Virlogcux M. Cable vibration in cable-stayed bridge. In: Larsen A, Esdahl S (eds) Bridge Aerodynamics. Balkema:
Rotterdam, 1998, 213–233.
2. Pacheco BM, Fujino Y, Sulekh A. Estimation curves for modal damping in stay cables with viscous dampers. ASCE
Journal of Structural Engineering 119(6);1961–1979.
3. Warnitchai P, Fujino Y, Pacheco BM, Agret R. Experimental study on active tendon control of cable-stayed
bridges. Earthquake Engineering and Structural Dynamics 1993; 22(2):93–111.
4. Fujino Y, Warnitchai P, Pacheco BM. Active stiffness control of cable vibration. ASCE Journal of Applied
Mechanics 1993; 60:948–953.
5. Fujino Y, Susumpow T. An experimental study on active control of planner cable vibration by axial support motion.
Earthquake Engineering and Structural Dynamics 1994; 23:1283–1297.
6. Gattulli V, Paolone A. Planar motion of a cable-supported beam with feedback controlled action. Journal of
Intelligent Material Systems and Structures 1997; 8:767–774.
7. Dyke JS, Caicedo JM, Turan G, Bergman LA, Hague D. Benchmark control problem for seismic response of cable-
stayed bridge. http://wusceel.cive.wustl.edu/quake/benchmark/2000
8. Schemmann AG, Smith HA. Vibration control of cable-stayed bridge. Earthquake Engineering and Structural
Dynamics 1998; 27:811–843.
9. Moon SJ, Bergman LA, Voulgaris PG. Application of MR-dampers to control of a cable-stayed bridge subject to
seismic excitation. Technical Report, University of Illinois at Urbana-Champaign, 2001.
10. Moon SJ, Bergman LA, Voulgaris PG. Sliding mode control of cable-stayed bridge subjected to seismic excitation.
ASCE Journal of Engineering Mechanics 2003; 129:71–77.
11. Nagarajaiah S, Sahasrabudhe S, Iyer I. Seismic response of sliding isolated bridges with smart dampers subjected to
near source ground motion. Proceedings of the 14th Analysis and Computational Speciality Conference, Philadelphia,
2000; CD-ROM.
12. Rodellar J, Manosa V, Monroy C. An active tendon control scheme for cable-stayed bridges with model
uncertainties and seismic excitation. Journal of Structural Control 2002; 9:75–94.
13. Schmitendorf WE, Jabbari F, Yang JN. Robust control techniques for building under earthquake excitations.
Earthquake Engineering and Structural Dynamics 1994; 23:539–552.
14. Kim DH, Lee IW. Neuro-control of seismically excited steel structures through sensitivity evaluation scheme.
Earthquake Engineering and Structural Dynamics 2001; 30:1361–1377.
15. Luo N, Rodeller J, de la Sen M. Composite robust active control of seismically excited structures with actuator
dynamics. Earthquake Engineering and Structural Dynamics 27:301–311.
16. Yang JN, Wu JC. Sliding mode control for nonlinear and hysteretic structures. ASCE Journal of Engineering
Mechanics 1995; 121:1330–1339.
17. Caicedo JM, Dyke SJ, Moon SJ, Bergman LA, Turan G, Hague S. Phase II benchmark control problem for seismic
response of cable-stayed bridges. Journal of Structural Control 2003; 10:137–168.
18. Przemieniecki SP. Theory of Matrix Structural Analysis. McGraw-Hill: New York, 1968.
19. Loh CH, Chang CM. Vibration control assessment of ASCE benchmark model of cable-stayed bridge. Journal of
Structural Control and Health Monitoring 2005.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143
MATLAB-BASED SEISMIC RESPONSE CONTROL 143

110

C.-H. LOH AND C.-M. CHANG

and very low inherent damping [1,2]. Therefore, cable-stayed bridges might be vulnerable to dynamic loading such as earthquakes and strong wind loads. In 1993 Warnitchai et al. [3] experimentally and analytically studied active tendon control of cable-stayed bridges, subjected to a vertical sinusoidal force. They utilized a simple cable-supported cantilever beam as a model. Up to now, many research efforts have been focused on the interaction of cables with the deck and attenuation of the cable movement [4–6]. Other research efforts have been aimed at understanding the overall dynamic behavior of cable-stayed bridges and developing finite element models [7,8]. The working group on bridge control within the ASCE Committee on Structural Control recently posted a first-generation benchmark structural control problem based on the Cape Girardeau Bridge [7,8]. This problem focus on one-dimensional ground acceleration applied in the longitudinal direction that is uniformly and simultaneously applied at all supports. In the work of Moon et al. [9,10], a semi-active system for the benchmark bridge employing MR dampers in conjunction with a LDG/clipped optimal control (LQG/MR) and a sliding mode semi-active control system (SMC/MR) was studied. In addition, a hybrid base isolation system employing semi-active control devices (using MR-dampers), often termed ‘smart’ dampers, has been suggested to control the damping force of bridge structure [11]. In the design of modern feedback control systems, the question of uncertainties in the models and the excitations, within the context of robust control, is becoming an important issue. For example, an active scheme that uses active tendons via feedback of the states of the system for vibration control of a cable-stayed bridge under seismic loads has also been proposed by Rodellar et al. [12]. While different approaches have been adopted for robust control of civil engineering structures, such as H1 and related tools [13], neural networks and fussy logic [14], and sliding mode control [15,16], the use of these approaches to control of cable-stayed bridges has been limited so far. In the ASCE phase II benchmark control problem, a three-dimensional evaluation model has been developed to represent complex behavior of the multi-support and transverse excitations of the Cape Girardeau Bridge [17]. The benchmark problem and a sample control design have been made available in the form of a set of MATLAB equations. In this research the control of cable-stayed bridge is studied. Based on the detailed information of the Gi-Lu cable-stayed bridge in Taichung County, Taiwan, a three-dimensional numerical model has been developed to represent the complex behavior of the full-scale bridge. Dynamic behavior of the cable including sag geometry, pre-stress tension force, and nonlinear response are considered in this analysis. The formulation of the cable stiffness is carried out both in ABAQUS and in MATLAB equations for consistency. This paper is focused on the following issues: 1. A three-dimensional evaluation model is developed in a realistic way to represent complex behavior of a full-scale cable-stayed bridge. Nonlinear beam elements are used to represent each cable. A comparison of the dynamic response of the bridge is made by using a truss model or nonlinear beam element model for the cables. 2. A comparison of the control effectiveness is discussed for: (a) using active control strategy by employing actuators between the bridge deck and the end-abutments; or (b) using both viscous dampers between the cables and the bridge deck and actuators between the deck and the end-abutments. 3. Based on the different evaluation criteria the control effectiveness between two different control algorithms, active and hybrid, is examined. 4. A MATLAB-based model for control of a cable-stayed bridge is developed.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143

MATLAB-BASED SEISMIC RESPONSE CONTROL

111

MODELING OF CABLE-STAYED BRIDGE The cable-stayed bridge used for this study is the Gi-Lu bridge, located at Nantou County, Taiwan. It is a modern, pre-stressed concrete cable-stayed bridge, which crosses the Juosheui River in Taiwan. As shown in Figure 1(a) the bridge has a single pylon constructed with reinforced concrete (58 m above the deck), two rows of harped cables (68 cables in total),

Figure 1. (a) Top view and side view of Gi-Lu bridge; and (b) cross-section of Gi-Lu bridge.
Copyright # 2005 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:109–143

85–63.02 Poisson’s ratio 0. Dynamic characteristics of the cable A three-dimensional finite element model of the Gi-Lu bridge was developed in MATLAB. 2007. Struct.121 8. 0. The vibration frequency of each cable was also estimated from vibration test (Table III). 14. 0. The finite element model employs beam elements and rigid links.515  10À5 0.75 m and width 24 m. Analytical solutions of geometrically nonlinear beam element are available for several special cases.613 kg/cm and the design cable force ranges from 220 to 290 ton. Element Main girder (deck) Plate (deck) Side span Pier Cable (1) Cable (2) Cable (3) Pylonà à Varying cross-section.45  10À5. as shown in Figure 1(b). 14:109–143 .07 Moment of inertia (m4) 9.2 0.515  10À5.2 0.55 334.55. In this study.005163 0. Control Health Monit. a geometric stiffness matrix can be derived based on energy methods from Euler-Bernoulli equation as shown follows [18]: Fext ¼ ðKlinear þ Knonlinear Þu ð1Þ Copyright # 2005 John Wiley & Sons.67–0. CHANG Table I. 334.2 0.001 2400 2400 8000 8000 8000 2400 and a streamline-shape single box girder. Therefore.77. 0. It is necessary to represent an appropriate nonlinear behavior of beam.410 to 0.006 0. by using an energy method a geometrically nonlinear beam element was generated that includes the nonlinear terms plus the terms from the conventional linear beam element. 84.66. Ltd.67  10À5 6. The density of the cable ranges from 0.2 Density (kg/m3) 3202 0. Cables 1 and 2 indicate the shortest cables and cables 33 and 34 are the longest cables. 84.27 4 8. for instance. the box girder rigidly connects with the pylon and spans 120 m to each side span.66.29 6. With depth 2.45  10À5 0. Material properties of Gi-Lu bridge.-M. The sag of each cable is shown in Table II.2 0 0 0 0. Young’s modulus (N/m2) 2:77  1010 2:77  1010 2:77  1010 2:77  1010 2:04  1011 2:04  1011 2:04  1011 2:77  1010 Area (m2) 12. A linear evaluation model is used in this cable-stayed bridge model. R and L indicate the left-hand and right-hand sides of the pylon.-H.63.55 6. 17 pairs of near parallel pre-stressed steel cables extending from each side of the pylon were developed. geometric nonlinearity of beam elements needs to be considered in this analysis. Due to large deformation of the cables and their sag.18–69. The material properties of the bridge are shown in Table I. Generally. LOH AND C. 334. a fixed beam subjected to concentrated loads. conventional linear two-dimensional beam elements cannot consider the effect of axial stresses due to large bending deformation.112 C.67  10À5.121 0.29 0.00756 8.

111 1.19876 0.029262 0.033323 0.2756 2.3889 4.40785 0.78728 0.99808 0.062502 0. Control Health Monit.070109 0.45105 0.1138 2.75464 Abutment Pier No.1161 1.741 1. Left-hand side four columns indicate the sag of each cable (middle node) and right-hand side four columns indicate the 1st natural frequency of each cable (R** means west-side cables.20176 0.020243 0.043072 0.114 1.75513 0.44722 1st natural frequency (Hz) 0.1737 3.9493 1.1) (Lu-Ku side) Copyright # 2005 John Wiley & Sons.4193 1.060964 0. Ltd.091567 0.1738 2.51839 0.7106 3.0193 3.3042 1.16688 0.19876 0.011206 0.022885 0.75823 0.010133 0.9493 1.1149 1.020243 0. 2007.44722 R33 R31 R29 R27 R25 R23 R21 R19 R17 R15 R13 R11 R9 R7 R5 R3 R1 R2 R4 R6 R8 R10 R12 R14 R16 R18 R20 R22 R24 R26 R28 R30 R32 R34 L33 L31 L29 L27 L25 L23 L21 L19 L17 L15 L13 L11 L9 L7 L5 L3 L1 L2 L4 L6 L8 L10 L12 L14 L16 L18 L20 L22 L24 L26 L28 L30 L32 L34 0.7411 1.9582 1.51839 0.9581 2.029262 0.4196 1. Struct.072009 0.016358 0.23382 0.99808 0.058809 0.1149 1.25108 0.79146 0.051267 0.2758 2.75068 0.99956 1.11 0.3889 4.99956 1.9581 2.9493 1.114 1. 14:109–143 .033323 0.3891 3.018332 0.049284 0.16688 0.058809 0.23382 0.022885 0.13945 0.111 1.3037 1.072009 0.3891 3.9582 1.4196 1.75513 0.75464 R33 R31 R29 R27 R25 R23 R21 R19 R17 R15 R13 R11 R9 R7 R5 R3 R1 R2 R4 R6 R8 R10 R12 R14 R16 R18 R20 R22 R24 R26 R28 R30 R32 R34 L33 L31 L29 L27 L25 L23 L21 L19 L17 L15 L13 L11 L9 L7 L5 L3 L1 L2 L4 L6 L8 L10 L12 L14 L16 L18 L20 L22 L24 L26 L28 L30 L32 L34 0.7108 2.3037 1.25108 0.3 (Gi-Lu side) 120 m Pylon & Pier No.12064 0.062502 0.14422 0. Sag (maximum) (m) 0.010133 0.011206 0.75068 0.091567 0.0191 4.29116 0.2 120 m Abutment (Pier No.29116 0.4193 1.75823 0.2758 2.MATLAB-BASED SEISMIC RESPONSE CONTROL 113 Table II.1738 2.40785 0.051267 0.3042 1.11 0.2756 2.52236 0.1161 1.060964 0.9493 1.13945 0.52236 0.20176 0.7108 2.1737 3.018332 0.23022 0.0193 3.070109 0.7411 1.049284 0.07934 0.043072 0.14422 0.4723 0.78728 0.741 1.23022 0.1138 2.7106 3.016358 0. L** means eastside cables).07934 0.0191 4.79146 0.12064 0.45105 0.4723 0.

and R34 indicate the shortest.916 0.99 14.75974 1.9496 8.307 50 Elements 1000 Method Mode ABAQUS Hz ABAQUS MATLAB MATLAB MATLAB Hz Hz Hz Hz 0. R34 (longest) 100 MATLAB Hz 4.027 15. respectively.236 4.2036 6.846 15.999 3.4711 10.685 19.833 7.665 19.1301 8.078 22.906 11.4171 8.1687 12. Comparison on the estimated cable vibration frequencies from vibration measurement.8384 9.19 9.7512 4.455 13.631 12.733 18.4342 5.635 6.7043 10.1237 8.2266 2.2565 3.1238 8.433 5. from ABAQUS and from B code using nonlinear beam elements. R17.8487 17.881 15.511 5.278 0.755 4.026 15.4885 2.837 17.3602 5.6494 19.74 18.685 13. . Ltd. CHANG Method Mode MEASURE Hz 1 2 3 4 5 6 7 8 9 10 11 12 1.189 4.235 10.22 14.21 14.4695 10.028 15.2342 5.907 16.114 Table III.4162 8.499 2.387 1.743 16.496 2.7327 3.591 8.338 8 1000 ABAQUS MATLAB MATLAB MATLAB Hz Hz Hz Hz 11 12 13 14 15 16 17 18 19 20 21 22 23 1.303 14.9657 6.978 10.338 15.557 17.534 18.167 15.1566 12.856 1.3985 Struct.81 11.564 12.1316 8.5054 2.2576 3. LOH AND C.1587 6.067 Copyright # 2005 John Wiley & Sons.176 9.136 14.67 1 2 3 4 5 6 7 0.515 21.763 24.1237 8.327 25.052 11.1331 6.6376 6.132 7.248 2.83 17.75119 1.4938 5.766 15.023 15.921 11.934 15.43 12.575 12.744 15.095 11.4628 5.393 9.514 5.5062 2.93 23.621 18.8139 8.-M.18 15.839 15.0052 3.021 15.816 8.8407 9.4739 7.2321 5.571 14.802 11.751 4.1569 12.279 0.4957 5.604 8.4706 10.094 11.235 14. R34) R1 (shortest) 15 MATLAB MATLAB MATLAB Hz Hz Hz 4.245 25.132 8. and longest cables.884 21.713 16.88 13.7248 7.6689 4.233 12.93 16.454 5. 2007.761 15.12 8.509 16.1923 6.622 18.682 19.5368 10.769 1.7041 7.31 25.0067 3.047 6.694 7.769 1.753 4.26 R17 (middle) Elements C.0303 15.-H.1349 6.987 15 50 100 9 10 6.428 1.9552 3.9677 6.1568 12.458 16.696 10.821 17.05 15.922 13.0201 5. 1 2 3 4 5 4.236 14.47 13.2541 14.2001 6.4629 5.044 6.97 17.679 13.597 12. middle.5359 21. Control Health Monit.226 10.7202 3.9435 8.514 14.206 17.32 14.7208 3.997 3.7318 3.822 7. (b) translational and transverse degrees-of-freedom of each cable node are locked.5421 10.476 12.312 15.6956 7.63 12.75956 1.829 15.386 9. R17.967 Method Mode MEASURE Hz 200 Elements 1000 15 50 100 Analyze natural frequency of cables (R1.7325 3.671 19.152 17.025 15.8349 8. 14:109–143 Note: (a) R1.109 22.247 2.264 12.996 10.117 22.459 12.

and transform Equation (1) into a three dimensional nonlinear beam element. Copyright # 2005 John Wiley & Sons. the first term of the stiffness matrix is the linear stiffness matrix. A is the cross-sectional area. x and y are the translational deformation of both ends of the element. I is the moment inertia. F is the force. and j is the rotational deformation of both ends of the element. 14:109–143 . Control Health Monit. and N is related to the axial force due to large deformation: N¼ EA 0 ðL À LÞ L ð2Þ in which L0 is the actual length along the center line of the beam. the second term is due to nonlinearity. length of taut cables.MATLAB-BASED SEISMIC RESPONSE CONTROL 115 or 2 EA 6 L 3 6 2 6 12EI Fx1 6 0 7 6 6 L3 6 Fy1 7 6 7 6 6 6EI 7 6 6 6 Mz1 7 6 0 6 L2 7 6 6 7 ¼6 6 6 Fx2 7 6 EA 7 6À 6 0 7 6 L 6 Fy2 7 6 5 6 4 6 12EI 6 0 À 3 Mz2 6 L 6 4 6EI 0 L2 2 0 6 6 6 60 6 5L 6 6 6 1 60 6 10 þ N6 6 60 0 6 6 6 6 60 À 6 5L 6 4 1 0 10 3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 5 4EI L 0 À 6EI L2 2EI L EA L 0 0 12EI L3 6EI 4EI À 2 L L 3 72 3 7 x 7 1 76 7 76 7 7 6 y1 7 76 7 76 7 76 f 1 7 76 7 76 7 76 x2 7 76 7 76 7 7 4 y2 5 7 7 7 f2 2L 5 15 2L 15 0 1 10 L À 30 À 0 0 0 6 5L 1 À 10 ð1aÞ where E is the Young’s modulus. and Poisson’s ratio). In development of the initial cable stiffness is important to obtain the actual pre-tensioned values (not designed values) and to include correct element properties (modulus of elasticity. Assume that Poisson’s ratio equals zero. Struct. Ltd. moment inertia of crosssectional area. M is the moment. to deal with the complex behavior of cables. The procedures for developing the initial cable stiffness are shown below: 1. 2. L is the length of the element. Subscripts (1 and 2) indicate the two ends of the beam. However. 2007. Choose how many elements to use in the formulation. the computational procedures must be modified.

Repeat Step 5 until the total steps equal to n (in Step 4).9 kg/m. The configuration of Cables R1.116 C. the vibration frequencies of each stayed-cable in the Gi-Lu bridge were examined from three different approaches: (1) experiments. Use the initial length to generate the stiffness matrix where N equals the pre-tensioned value. 4. 14:109–143 . Divide the external force (self-weight of cables) into an appropriate number of equal parts n. The FEM model developed using ABAQUS has 1000 beam elements and considers the cable structure to be subjected to gravity and to have fixed-end boundary conditions. Configuration of Cable-R34 and the identified. and the identified cable force is only 91. LOH AND C. but on Figure 2. 5. The vibration modes of cable R34 (longest) and cable L17 are shown in Figure 3. 2007.0060 m2. and (vi) gravity 9. Control Health Monit. Calculate the present length due to large lateral deformation and re-formulate the cable stiffness matrix where N is still the pre-tensioned value. and (3) MATLAB equations. it is described as: (i) angle of elevation 268.-H. the identified cable force is 1:66  106 N. R17 and R34 is shown in Figure 2. To catch the dynamic characteristics of the cable 1000 beam elements are used in ABAQUS to analyze the detail of the dynamic characteristics of a single cable. Use Newton–Raphson (see Figure 1) iterations to solve ½KŠfDg ¼ fFext g by appropriate conditions of convergence. The cable force and moment of inertia were identified by matching the 21 vibration frequencies of the cables to those obtained from the field experiment. From these procedures. For example. (v) cross sectional area 0.-M.42 m.4% (1:66  106 =1:81  106 N) of the force estimated by string vibration theory.81m/sec2. (ii) designed length 126. CHANG 3. Comparison between the identified vibration frequencies of the cable R1. (iv) Young’s modulus 1:83  1011 N=m2 . 6. Through an optimally iterative process. The vibration frequency of each cable predicted by MATLAB is verified using field experiments. for the longest cables (33 and 34). Ltd. In order to evaluate the dynamic characteristics of the stayed-cable in this model. Copyright # 2005 John Wiley & Sons. R17 and R34 using MATLAB and ABAQUS is shown in Table III. (iii) length density 47. (2) ABAQUS. and the identified moment of inertia is 5:0  10À6 m4 : This information is used to formulate equations in MATLAB for the stiffness matrix of the cable system. Struct. stiffness matrices of the cables can satisfy effects of sag and measured natural frequencies. 7.

only 9 nodes were used to form a single cable which is sufficient to involve the lowfrequency dynamic behavior (0. Ltd. has a total of 1009 nodes. The pylon is modeled by 90 nodes with 540 degrees of freedom (DOF) (25 nodes are above the deck. and 60 nodes are near the anchors). The first six mode shapes: (a) cable R34. The deck is modeled by 729 nodes with 4374 DOFs. 5 nodes are below the deck.7–4 Hz) as in relating to the whole bridge system. and (b) cable R17. The cable is Copyright # 2005 John Wiley & Sons. the contrary. Pier 2 (under the pylon) is modeled by 5 nodes (including the node attached to the ground). 14:109–143 . DESCRIPTION OF FINITE ELEMENT MODEL The finite element model of the cable-stayed bridge. as shown in Figure 4. Struct.MATLAB-BASED SEISMIC RESPONSE CONTROL 117 Figure 3. Control Health Monit. Pier 1 (North side) and pier 3 (South side) have 4 nodes for each (including the node attached to the ground) with a total of 56 DOFs. 2007.

fx-. LOH AND C. sensor processor. the pylon height of 58 m from deck. and the deck width of 24 m.-H. The mass matrix is formed by a lumped mass approach.118 C. Constraints are applied to restrain the both ends of the deck (boundary conditions).-M. Model for cable-stayed bridge (Gi-Lu bridge) with the span length of 318. The transverse and vertical degrees of freedom at Piers 1 and 3. Control Health Monit.e. CHANG Figure 4.and z-directions are constraint. and x-. y. all translational DOFs and torsion DOF at the side spans connected to the embankment are fixed. appropriately reduced methods were used to formulate the equation of motion for the Gi-Lu bridge. Consider the general equation of motion for a structural system subjected to seismic loads . The damping ratio for deck. fz}directions are free to move). 2007. All DOFs at the bottom of both piers are fixed. pylon and piers (including all supports) is assumed to be 5% and for cables is assumed to be 1%. Ccoup Csp # Ktotal ¼ " Kcdp T Kcoup Kcoup Ksp # ð3aÞ Struct. 14:109–143 .9 m. ’ Mtotal U þ Ctotal U þ Ktotal U ¼ Pext ð3Þ where system matrices can be written as " # " Mcdp Mcoup Ccdp Mtotal ¼ Ctotal ¼ T T Mcoup Msp Ccoup Copyright # 2005 John Wiley & Sons. EVALUATION MODEL The control system of the cable-stayed bridge includes: evaluation model. The reduced methods use static condensation and quasi-static reduction. and at the side spans of the bridge deck deform consistently (i. controller and control devices. fy-. As for the boundary condition of the bridge structure. Ltd. Based on the above-mentioned finite element model the formulation of the model is described first. modeled with 9 nodes for each cable and total of 612 nodes for 64 cables with 3672 DOFs. Problem formulation Since a precise mathematical model for analyzing the dynamic behavior of cable-stayed bridges is very complicated. The proportional damping formulation is used in each element and then to form the full damping matrix is formed with the same procedure as is used to form the stiffness matrix.

deck and pylon. It is assumed that the total displacement can be divided into two parts: active DOFs (main nodes) and dependent DOFs (redundant nodes).sp Ua. the equation for the static condensation can be formed as " #" # " # Kaa Kad Ua Pa ð4Þ ¼ T Kad Kdd Ud 0 where the subscript a denotes the active DOFs and d denotes dependent DOFs. The main DOFs is taken here to include the nodes of the main girder. and Ua. the DOFs from piers and side spans (supports). Ktotal are the mass.cdp Ua. _ Ctotal ¼ TRT Ctotal TR. Consequently. pylon.MATLAB-BASED SEISMIC RESPONSE CONTROL 119 where subscript cdp indicates DOFs of cables. and Pext is the vector of external forces. Model reduction (static condensation and quasi-static reduction) The finite element formulation of the bridge model has a large number of degrees of freedom. Therefore. the DOFs from cables.cdp ¼ þ ð8Þ Ua. Then the total displacement vector can be transformed into " # " #   Ua I Ua or ð5Þ ¼ ½TRŠUa ¼ Ua À1 T Ud Ud ÀKdd Kad where [TR] is the transformation matrix of the static condensation. side span.cdp relationship between these displacement components is " # " s # " d # Ua. 2007. damping.cdp.sp 0 Copyright # 2005 John Wiley & Sons. 14:109–143 . deck and pylon. Ctotal.cdp Ua. Mtotal.cdp can also be separated into the displacement Us a.sp. the system total mass. piers. _ K total ¼ TRT Ktotal TR _ ð6Þ To consider the seismic excitation of the multiple-supported system both quasi-static and dynamic analyses must be employed. subscript sp indicates the DOFs for support from piers and side spans and subscript coup indicates coupled terms between supports and others. damping and stiffness matrices respectively of the whole structural system. U is the displacement response vector. The total displacement Ua. By using the transformation matrix of the static condensation Equation (1) can be re-arranged as follows: . Static condensation is applied to reduce the redundant DOFs and retain the main DOFs. and the dynamic displacement Ud relative to the quasi-static displacement. The a. Struct. and cables including the nodes connecting the cables to the deck and pylon. ’ M total U a þ Ctotal U a þ K total Ua ¼ TRT Pext _ _ _ ð7Þ where the active DOFs of Ua represent two components: Ua. Control Health Monit. and stiffness matrices can be changed as follows: M total ¼ TRT Mtotal TR. The redundant DOFs refers to the other nodes of the deck since these nodes are used to analyze static behavior.cdp due to static application of the ground motion. Ltd.

Ltd. the quasi-static solution can be obtained from the following equation " #" s # " # Ua.coup Ua.120 C.sp ext Solving Equation (9) for quasi-static displacement leads to s À1 Ua. % K a.cdp ÀM À1 FT ðMa.sp Rs þ C a.cdp ¼ FT M a.cdp % a.cdp % % a.cdp U d þ C a.cdp Y þ K a.-H.sp 5 þ4 ð14Þ _ _ _ _ ’ % a. Struct.cdp F. CHANG By neglecting the dynamic terms in Equation (7).cdp _ _ _ . ’ ¼ ÀðM a.sp ’ À FT ðC a. the convergence condition of Equation (12) is satisfied. a.coup 0 ¼ ð9Þ T Ka.sp ÞU a.cdp ’ a.cdp Rs þ C a.cdp Rs þ Ma. Since the sampling time for the input ground motion is 0.cdp Rs þ M a.sp Copyright # 2005 John Wiley & Sons. ’ % % % M a.sp ÞU a.sp À ðCa.sp Ps Ua. 14:109–143 . Control Health Monit. deck and pylon is as follows: " # " #" # ’ Y Y 0 I x¼ ¼ Ax þ Euext ¼ ’ % a.cdp Y Y 2 3" # .coup Ka.005 s.cdp Y þ C a.sp ÞU a. ÀM À1 K a. one can obtain the dynamic equation of motion of the structure as follows: d .cdp U d þ K a. A state space form for the cables.-M.cdp Y ¼ À FT ðM a. . In this study the first 400 modes of vibration are used for the analysis.cdp .cdp Rs þ M a.cdp Ka.cdp % ’ .cdp Ka.sp ÞU a. % C a. _ _ _ _ _ _ _ ð13Þ % M a.cdp ¼ Rs Ua. A condition for convergence of the discrete state space calculation is that the following criteria must be satisfied: p DT4 ð12Þ o where o is the highest modal frequency (rad/sec) that is considered.sp or s Ua. 0 0 U a.sp _ _ _ _ ð11Þ To solve for the dynamic response of a cable-stayed bridge the state-space formulation is used.cdp F.cdp Ka.cdp M a.cdp ¼ FT K a.cdp Ua.cdp ¼ ÀKa.cdp F and F is the modal matrix. 2007. LOH AND C.cdp Rs þ C a.sp ð10Þ Substituting Equations (8) and (10) into (7).cdp ¼ FT C a.sp Þ U a.sp Þ ÀM À1 FT ðC a.cdp ÀM À1 C a. After the model has been reduced the equation of motion for the damped structural system results from Equation (11): .sp where d Y ¼ FUa.

0. 102nd (f ¼ 1:8941 Hz) and 115th (f ¼ 2:0378 Hz) vibration modes of the cable stayed bridge. 0. Copyright # 2005 John Wiley & Sons. Control Health Monit. 64th (f ¼ 1:4456 Hz).75464 Hz.MATLAB-BASED SEISMIC RESPONSE CONTROL 121 Figure 5. 0.51487. and 1. and (c) first six mode shapes of cable-stayed bridge using truss elements to simulate the cable behavior (the vibration frequencies of the first six modes are: 0.75069.75393.5871 Hz).3695. 2007. 0. 1. 14:109–143 . respectively. and 0. 97th (f ¼ 1:756 Hz). Struct. 1. 0. (a) The first six vibration modes of the cable-stayed bridge. Its corresponding modal frequencies are 0. 0.42152.80013.75022.4455.73124. (b) the 31st (f ¼ 1:506 Hz).44952. Ltd.

Ltd. LOH AND C. CHANG Figure 5.-H.3695. Therefore. 0. Based on this formulation the mode shapes of the cable-stayed bridge are calculated. Through static condensation the final reduced model for the pylon and the deck reduce from 90 to 85 nodes and from729 to 243 nodes.75022. The corresponding modal frequencies are 0. respectively.73124. Besides. transverse and vertical directions at the top of the pylon is shown in Figure 6.51487. 0. then the sag behavior of cable can be truly reflected. 64th (f ¼ 1:4456 Hz). Continued. Struct. the number of nodes for the pier and cable remain the same. It can be observed that larger displacement and acceleration responses occur in the transverse Copyright # 2005 John Wiley & Sons.80013. 102th (f ¼ 1:8941 Hz) and 115th (f ¼ 2:0378 Hz) vibration modes of the cable stayed bridge are also shown in Figure 5(b).44952. 1. and 0. 1. It is found that the first fundamental mode is mainly dominated by the vibration of deck. 0. Figure 5(c) shows the calculated lowest six vibration modes of the cable-stayed bridge that are calculated using truss element to simulate the cable (the vibration frequencies of the first six modes are: 0. 0. Since the implementation of geometric stiffness matrix to simulate the cable using beam element.75069. A comparison of the acceleration and displacement response in the longitudinal. In addition. 0. 0.-M.42152. contributions from both cable and deck vibration to these higher modes are observed. Control Health Monit. respectively. a significant difference between the bridge vibration modes is obtained if the cable is modeled using truss elements instead of dynamics of geometrically nonlinear beam elements.4455. 14:109–143 . and 1. 2007. the cable vibration in transverse direction and the pre-stressed force in the cable can also be used.122 C. Figure 5(a) shows the first six vibration modes of the cable-stayed bridge. the 31st (f ¼ 1:506 Hz).5871 Hz).75464 Hz. and that the following five modes are dominated by vibration of the cables. 97th (f ¼ 1:756 Hz).75393.

Copyright # 2005 John Wiley & Sons. Control Health Monit. Ltd. Struct. 2007. Comparison of estimated acceleration and displacement response at the top of pylon using truss elements for cables and nonlinear beam elements for cables. 14:109–143 .MATLAB-BASED SEISMIC RESPONSE CONTROL 123 Figure 6.

by contrast.-H. 14:109–143 . CONTROL DESIGN PROBLEM A test bed for development of effective control strategies for a cable-stayed bridge has also been developed with emphasis on nonlinear behavior of the stayed-cable. the longitudinal direction at the top of the pylon has the most significant responses when using the nonlinear beam elements. Therefore. In the present study the cable sensors sometimes are and sometimes are not connected to the controller.124 C. For estimation of transverse direction at both ends of the bridge. Control Health Monit. It is found that use of the truss model for the cables induces larger acceleration response at both ends of the bridge deck.-M. 2007. (d) four sensors: acceleration and displacement sensors at the top of the pylon (in both longitudinal (x) and transverse (z) directions. CONTROL SYSTEM DESIGN In this study a controller is designed to serve as an active control system for the equivalent linear cable-stayed bridge model. in what follows only the modeling of the bridge cables using the nonlinear beam elements for the cables is used to evaluate the control effectiveness. A Copyright # 2005 John Wiley & Sons. This sample active control design study defines the excitation. It is believed that using truss elements to model the cable for the cable-stayed bridge may induce significant bias on the response calculation. LOH AND C. It is believed that use of nonlinear beam elements to model for the cables is a reasonable and necessary approach for predicting behavior of the stayed cables. The sample control system employs a total of 12 hydraulic actuators located at both ends of the bridge between the deck and the top of each pier (6 actuators for each pier). (c) four sensors: acceleration and displacement sensors at the middle node of the deck (in both longitudinal (x) and transverse (z) directions. A total of 19 sensors are used to collect the response of the bridge and are used for control purposes. Ltd. Figure 7 compares the acceleration and displacement response at the northern end of the bridge deck when using truss elements for the cables (blue line) and nonlinear beam elements for the cables (red line). (b) four sensors: acceleration and displacement sensors at the southern end of the deck (in both longitudinal (x) and transverse (z) directions. This is done to study the effect of the control on the bridge as well as on the vibration of the cables. The locations of these sensors are specified as follows: (a) four sensors: acceleration and displacement sensors at the northern end of the deck (in both longitudinal (x) and transverse (z) directions. Readings from acceleration and displacement transducers are fed back to the control algorithm. and control algorithms. Struct. CHANG direction at the top of pylon when the truss model is used for the cable. a larger displacement is observed by using the nonlinear beam element model for the cables. It is assumed that each actuator can provide up to 1500 kN. evaluated model. Both sensors from bridge and cables are measured. (e) three sensors: displacement sensors located at the top of all three piers in the longitudinal direction. devices. sensors. actuators are oriented to apply forces longitudinally (X-direction).

Copyright # 2005 John Wiley & Sons. Struct. Ltd. 14:109–143 . Comparison of estimated acceleration and displacement response at the northern end of the bridge deck using truss elements for cables and nonlinear beam elements for cables. 2007. Control Health Monit.MATLAB-BASED SEISMIC RESPONSE CONTROL 125 Figure 7.

the output of Copyright # 2005 John Wiley & Sons. Each earthquake record contains three-dimensional acceleration and velocity data (from integration of acceleration data). The evaluated output serves to determine the control efficiency of various control criteria. The input port of the control force is developed from the design of actuators or dampers. To consider the spatial variation of earthquake excitation. The excitation input port includes nine acceleration values and nine velocity values to describe the multi-support problem in three-dimensional and the time-delayed excitation. LOH AND C. A detailed description of the SIMULINK block diagram is now given. At the same time. CHANG Figure 8.-M. Ltd.126 C. uniform input with a phase delay was specified (4 km/s apparent wave velocity was assumed). Excitation The earthquake record from the Chi-Chi earthquake can be selected as the excitation to this cable-stayed bridge. Control Health Monit. [17]. SIMULINK block diagram for this active control study is shown in Figure 8. The output port of sensors gives the predicted responses of the deck or the pylon (or piers).-H. 14:109–143 . Evaluation model The evaluation model in SIMULINK contains two input ports and six output ports (Figure 9). Struct. It is patterned for SIMULINK block diagram provided by Dyke et al. 2007. Modified SIMULINK block diagram with active control. [7] and Caicedo et al.

Figure 10 shows the SIMULINK block of the sensor processor. Measurement noises are modeled with Gaussian rectangular pulses that have a pulse width equal to the integration step. 14:109–143 . it follows with H2 control algorithm and Kalman estimator for use of actuators. Struct. SIMULINK block: evaluation model. Ltd. in order to use actuators to control the responses between the deck and piers (end-abutments). In this case. the first 100 modes of Copyright # 2005 John Wiley & Sons. The output port of tension estimates the variation of cable tensile force using the deformation of cables. If the control algorithm focuses on the cables. Control Health Monit. The device sensors provide the responses of actuators or dampers between connections. sensors is used to estimate the control forces which include a Kalman filter estimator for feedback control. these cable sensors can also be used to determine the feedback control.01 V. Originally. an appropriate design model must be developed. The converter for each sensor has a range of Æ10 V: Each of the measured responses contains a noise level with an rms value of 0. The design model is formed from the evaluation model by choosing suitable modes of the system. The cable deformations are selected to consider the dimensionless responses at the middle node of each cable. The control design model contains 44 modes that significantly affect responses of the deck and piers. The signal converter mainly transforms the sensors signal into real responses with a constraint. In regard to the force estimator. Controller The controller contains a signal converter and a force estimator. The constraint within the signal converter has an upper bound (þ10 V) and a lower bound (À10 V). First. 2007.MATLAB-BASED SEISMIC RESPONSE CONTROL 127 Figure 9. the piers (1 and 3) or abutments are selected to apply the control devices. Sensor processor The sensor processor converts the responses of the bridge to the voltage signal. The output port of cable sensors gives the responses of the cables.

2007. Therefore. the external disturbances. Ltd. SIMULINK block: sensor processor.-H. the whole bridge system were used to obtain the control gain using H2 control theory. To obtain the optimal control force. In real applications. x½n þ 1Š ¼ Adcon x½nŠ þ Bdcon u½nŠ þ Edcon ð15Þ xg ½nŠ ’ " ys ½nŠ ¼ Cdcon x½nŠ þ Ddcon xg ½nŠ . LOH AND C. there are some modes from the first 100 modes which do not directly and strongly contribute to control the system. The controller employs an H2 control algorithm to estimate the appropriate control force. ð18Þ ð19Þ Struct. The state-space (discrete form) system of the control design model is represented as follows " # xg ½nŠ . The measured outputs focus on the sensor outputs (responses of the deck and the pylon). and y is a vector of sensor outputs. Control Health Monit. and k0 is the initial time. the optimal control force can be obtained as AT P½kŠAdcon À AT P½kŠBdcon ð2R þ BT P½kŠBdcon ÞÀ1 BT P½kŠAdcon þ 2Qjk!1 ¼ 0 dcon dcon dcon dcon u½kŠ ¼ Àð2R þ BT PBdcon ÞÀ1 BT PAdcon x½kŠ ¼ Gx½kŠ dcon dcon Copyright # 2005 John Wiley & Sons. the objective function is defined as J2 ¼ k!1 X k¼k0 xT ½kŠQx½kŠ þ uT ½kŠRu½kŠ ð17Þ where R is a weighting matrix related to the optimal control force. such as the ground excitations. Q is a weighting matrix related to the system of the design model. Through the computation of the variation method. 14:109–143 . CHANG Figure 10. First. it may waste much time to calculate the optimal control forces if the matrix of the control gain has a large dimension.-M. However. xg ½nŠ ’ # þ Fdcon u½nŠ ð16Þ where x is the state vector. are assumed to be independent with respect to the control force. these modes with minor contributions in control gain are eliminated and remain the 44 modes which can be used to generate the optimal control force.128 C.

Struct. SIMULINK block: controller. Ltd. and G is the optimal control gain. Six actuators are placed between the deck and piers at each location. Although the control gain can be obtained from Equation (19). The connection between the command forces and the output forces is limited by bounds on the capacity of the device. The voltage command for each actuator with respect to the output control force is 10 V per 1500 kN. Two kinds of control devices are employed for the vibration Gi-Lu Bridge. P in Equation (19) is the solution of Equation (18). 14:109–143 . The limitation of actuators between the control force and the desired force is bounded from 1500 to À1500 kN. The converter for the actuators transforms command forces into corresponding voltages for single actuators. Control Health Monit. the full state vector must be computed from the Kalman estimator. where Equation (18) is the Riccati equation. Copyright # 2005 John Wiley & Sons. and the other kind of control device is the viscous damper. The real output force is indicated by the force from single actuator multiplied by number of control devices. Therefore. The optimal control force combined with the estimator (Equation 20) can be obtained as u½kŠ ¼ G x ½kŠ and the completed controller can be expressed as _ _ ð21Þ x ½k þ 1Š ¼ ðAdcon þ Bdcon G À LCdcon À LDdcon GÞ x ½kŠ þ Lys ½kŠ _ ð22Þ u½k þ 1Š ¼ G x ½k þ 1Š A SIMULINK block form of the controller is shown in Figure 11. which is 1500 kN per actuator and the minimum force is À1500 kN per actuator. as shown in Figure 12. Two strategies are used to arrange the locations of the control devices: Strategy 1: actuators are placed at the top of the Piers 1 and 3 in connection with the deck. given by _ x ½k þ 1Š ¼ Adcon x ½kŠ þ Bdcon u½kŠ þ Lðys ½kŠ À Cdcon x ½kŠ À Fdcon u½kŠÞ _ _ ð20Þ where L is the Kalman gain solved from the Ricatti equation. the optimal control force needs the full state vector of the design model.MATLAB-BASED SEISMIC RESPONSE CONTROL 129 Figure 11. contains a converter for the actuators (forces to voltages) and a real output force. The first type of control device is the actuator. 2007. Control devices _ ð23Þ The block of control devices.

On the practical implementation of the dampers in stayed cable will discuss later. 30 000 (N s/m) for cable 13 and 14. 42 857 (N s/m) for cable 11 and 12. LOH AND C. Strategy 2: besides using the Strategy 1 dampers are also implemented between cables and the bridge deck (which is placed between cables and the deck with 1 piece at each cable). and 7500 (N s/m) for cable 15 through 17. Figure 13. Struct. Figure 14 shows the schematic diagram of the locations of actuators and viscous dampers for control strategies 1 and 2. four types of viscous dampers are selected.-M. Figure 13 shows the SIMULINK block of the strategy 2 control devices. Based on the uncontrolled responses of the Gi-Lu Bridge to six specified excitations. SIMULINK block: control devices (case B). Ltd.-H. CHANG Figure 12. respectively. The damping coefficient of all types of dampers is: 120 000 (N s/m) for cable 1 through 10. Each damper is connected at the lowest nodal point of the cable relative to the deck. SIMULINK block: control devices (Strategy 1).130 C. 2007. 14:109–143 . Control Health Monit. Copyright # 2005 John Wiley & Sons. The input used to drive the dampers is the relative velocity response between the connecting cables and the deck.

Additionally. Earthquake ground motion data from the Chi-Chi earthquake from recorded station TCU089 is selected as the excitation (PGA in the EW direction is 244 gal. several criteria are evaluated in both the X (longitudinal) and Z (transverse) directions. These evaluation criteria are listed and shown in Table V. Control Health Monit.MATLAB-BASED SEISMIC RESPONSE CONTROL 131 Figure 14. the behavior of the cable tension force and the deformation of the middle node of each cable are important to observe. the middle node on the deck. For this purpose. these evaluation criteria must consider the ability of the controller to reduce the peak responses. Ltd. Struct. In this study a uniform input motion with three components (two horizontal and one vertical component) is used. 2007. The effect of cable modeling on the seismic response of cable-stayed bridge is examined first. and the north and south ends of the deck. Shear forces and moments at each pier. Other criteria contain the displacements at the top node of the pylon. 14:109–143 . CASE STUDY OF CONTROL EVALUATION The basic problem for control of the Gi-Lu cable-stayed bridge in this study focuses on modeling of the stayed cables. and the control requirements. (a) Location of actuators between deck and abutment. a set of 45 criteria have been employed to evaluate the ability of each control strategy. including the base of the pylon. Before the evaluation of control effectiveness of this bridge is considered response of uncontrolled case using different models is examined first. Evaluation criteria To evaluate the performance of the control algorithm used in this study. are considered in the evaluation criteria. in the NS direction 324 gal and in vertical direction 190 gal). Figure 15 shows the acceleration response spectrum of the recorded data. control strategy 1: using actuators only and control strategy 2: using actuators at both ends of the deck and Copyright # 2005 John Wiley & Sons. Both truss and nonlinear beam elements are selected to model the cable. and (b) location of viscous damper between deck and cable. Thus. As discussed earlier there are two cases of control devices in this study. the shear forces and moments in the pylon and piers induced by earthquake excitation must be considered. the normalized responses over the entire time record. Because the earthquake is assumed to have three-dimensional components at a specified incidence angle.

922 0.-M.783 1.999 1.0 0.999 0.999 1.999 0.911 1.814 1.0 0.878 0.921 0.010 0.750 0.999 0. Ltd.0 1.811 0.011 0.846 0.001 0.894 0.999 1.999 0.995 0.0 1. Struct.802 0. Control Health Monit.999 0.789 0.999 1.044 0. Case A J1 J2 J3 J4 J5 J6 J7 J8 J9 J10 J11 J12 J13 J14 J15 J16 J17 J18 J19 J20 J21 J22 J23 J24 J25 J26 J27 J28 J29 J30 J31 J32 J33 J34 J35 J36 J37 J38 J39 J40 J41 J42 J43 J44 J45 1.999 1.995 0.999 0. 2007. Case B: besides the actuators (in Case A) viscous dampers were also added on the cables. 14:109–143 .0 1.0 Copyright # 2005 John Wiley & Sons.848 0.0 1.721 0.999 0.970 0.030 0.444 1.694 0.0 Case B 0.755 0.807 0.950 0.739 1.999 0.893 0.998 0. Comparison of the control effectiveness using two different control systems: Case A: putting actuators between the abutments (both ends) and the bridge deck.0 1.132 C.0 0.0 1.956 1.882 0.872 1.773 0.0 0.961 0.888 0.844 0.999 0.001 0.926 1.700 0. LOH AND C.0 0.999 0.999 1.879 0.030 0.721 0.937 0.756 0.004 0.999 0.921 0.999 0.-H.878 0.067 1.750 0.0 0.0 1.868 0.999 0.789 1.001 1. CHANG Table IV.443 0.999 0.825 0.

Ltd. which affords a significant reduction in the response of the cables as is clearly shown in Figure 18. The right hand side of Figure 19(a) shows the estimated value of the normalized displacement at the midpoint of each cable for control case A. except for indices J1 (longitudinal shear force at the first pier) and J7 (longitudinal shear force at the base of the pylon). an improved level of control can be observed. Figure 20 shows the acceptable cable tension between the provided ranges after the control (for control strategy 2). The proposed control devices are shown to effectively reduce the longitudinal displacement response of the deck. Plot of acceleration response spectrum of ground motion data collected from Station TCU089. the pylon base shear force is larger than for uncontrolled case. as shown in Figure 17. 2007. Chi-Chi earthquake. This means that by putting actuators at both ends of the deck. 14:109–143 . Control Health Monit. Copyright # 2005 John Wiley & Sons. Struct. It is observed that control strategy 2 provides an improved level of control effectiveness not only for the response of the bridge itself but also for the stayed cables. The majority of the evaluation criteria have values less than one.MATLAB-BASED SEISMIC RESPONSE CONTROL 133 Figure 15. This is the case because viscous dampers are added between deck and cables in control strategy 2. On the left of Figure 19(a) shows the evaluation criteria with respect to different indices for control strategy 1. If control strategy 2 is used. Most of the displacement at the midpoint of the cable is reduced after control. Based on the proposed 45 evaluation criteria the performance indices are evaluated for the controlled and uncontrolled cases. Figure 16 compares the displacement and acceleration responses at the north end of the deck for the uncontrolled case and the case of using control strategy 1. attaching viscous dampers to the cables. For control strategy 2 the evaluation criteria and the estimated normalized displacement at the midpoint of each cable is shown in Figure 19(b). Table IV compares the normalized displacement (with respect to the uncontrolled case) of the evaluation criteria for both control cases A and B.

-M. Control Health Monit. 2007.9 m above the deck level) under control strategy 2. Because only eight nonlinear beam elements were modeled for cable and the lowest nodal point was used to connect the damper to the bridge deck. three types of MR dampers are selected to control cable vibration. DISCUSSIONS ON CONTROL STRATEGY In control strategy 2 dampers were attached between cable and the bridge deck. was used to implement the dampers for the cable. Remaining dampers are placed at the middle of the cable connecting two neighboring cables through cable 10 to cable 17 at each side. forty dampers are placed between the deck and cables through cable 1 to cable 10 at each side. Comparison of the estimated displacement and acceleration responses between uncontrolled and controlled case using strategy 1: (a) longitudinal responses at northern end of the deck.134 C. Type I of MR dampers is used to position between cable 4 to cable 9 and the deck. Ltd. Type II MR dampers are used between cables 1–3 and the deck. 14:109–143 . CHANG Figure 16. Therefore an alternative control strategy. Copyright # 2005 John Wiley & Sons. Type III MR dampers are positioned between cables 13–17. and between cable 10–12. considering the height of the lowest nodal point of the cable it is impractical to implement the damper for such a location (almost 6. as shown in Figure 21. Struct. control strategy 3. Instead of installing the damper between the cable and the bridge deck. In this case. between cable 10 and the deck. LOH AND C.-H. and (c) longitudinal responses at the top of pylon. (b) longitudinal responses at southern end of the deck.

J3. and (3) 240 mm stroke. 3500 mm/s maximum velocity. Comparison on the estimated displacement and acceleration responses between uncontrolled and controlled case of strategy 2: (a) longitudinal responses at northern end of the deck. (b) longitudinal responses at southern end of the deck. Ltd. Struct. (2) 120 mm stroke. 1050 mm/s maximum velocity. and 30 kN force capacity. If the corresponding force is greater than the lower bound. Figure 22 shows the comparison on the 45 evaluation criteria of three different control systems. and J7 are larger for strategy 3 using the input motion from TCU089. On the contrary. A constraint function of voltage with a secondorder polynomial function is used to obtain the lower bound of the force. The voltage command ranged from 0 to 1. 2007.MATLAB-BASED SEISMIC RESPONSE CONTROL 135 Figure 17. and (c) longitudinal responses at the top of pylon. the output force is changed to obtain the bilinear model. But the other indices are all smallest for using strategy 3 and the control of cable Copyright # 2005 John Wiley & Sons. Control Health Monit. 350 mm/s maximum velocity. 14:109–143 . There are three types of MR dampers including: (1) 120 mm stroke. J2. It is found that J1. The command voltage is sent to the modified bi-viscous model with the states as input and then generates the corresponding force. the output force is still the force generated by the modified bi-viscous model.2 V. The model of the MR damper for use in structural control of this study combines the modified bi-viscous model and the bilinear model [19]. and 30 kN force capacity. and 30 kN force capacity.

Control Health Monit. Responses of cable (EN-1): (a) longitudinal displacement. (c) transverse displacement. and (f) transverse acceleration.-M. (b) vertical displacement.-H. (e) vertical acceleration. CHANG Figure 18. 2007. Copyright # 2005 John Wiley & Sons. LOH AND C. 14:109–143 . Struct. (d) longitudinal acceleration.136 C. Ltd.

. Figure 23 shows the comparison on the cable vibration. for 35–51: L1.. R33. It is also found that with the control strategy 3 the vibration of cable was significantly reduced... for 18–34: R2. L32. L31. R34. and (b) for case B control. R4. L33. 2007. Struct. . .. . L2. 14:109–143 . . for 52–68: L2.. R31. L3.. Control Health Monit.. Ltd. Plot of evaluation criteria with respect to different index. (Note: The number of D indices for 1–17: R1. L34). Copyright # 2005 John Wiley & Sons. R32. vibration the strategy 3 provides much better results than others.MATLAB-BASED SEISMIC RESPONSE CONTROL 137 Figure 19. .. R3. . and the normalized displacement of middle point of all cables: (a) for case A control. .

m ¼ 13 Fpmx indicates the longitudinal (x-direction) shear force at mth 0x. Fmax indicates the maximum uncontrolled shear force at the deck level of the pylon.t ðjFpmz ðtÞjÞ 0z. Jm ¼   maxm. 0x. Fpmz indicates the transverse (z-direction) shear force at mth pier. 0x. Mdz indicates the transverse (z-direction) moment at the deck level 0dz of the pylon.t ðjFpmx ðtÞjÞ . Fmax indicates the maximum uncontrolled shear force at the deck level of the pylon.-M. 3 (south pier).m Mmax  Jk ¼ maxm.m Fnorm is the normed value of uncontrolled shear forces in the transverse (z-direction) direction at mth pier. Fdx is the longitudinal (x-direction) shear force at the deck level of 0dx the pylon. where qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Ptf 2 normðÞ ¼ t¼t0 ðtÞ : 0z. 0dx Fnorm is the normed value of uncontrolled shear forces of the pylon in the longitudinal (x-direction) direction at the deck level. Mmax is the maximum uncontrolled moment at the same location.t Mpmx ðtÞ .t Mpmz ðtÞ 0z. C. LOH AND C. 2007.m Fnorm 0x. k ¼ 12214   maxm.m Fmax J7 ¼   maxt jFdx ðtÞj 0dx Fmax Copyright # 2005 John Wiley & Sons. Mdx indicates the longitudinal (x-direction) moment at the deck 0dx level of the pylon.t jFpmz ðtÞj .m Fmax is the maximum uncontrolled shear force at mth pier.m Mmax is the maximum uncontrolled moment at mth pier. Control Health Monit. Fdz is the transverse (z-direction) shear force at the deck level of 0dz the pylon. CHANG J15 ¼   maxt Mdx ðtÞ 0dx Mmax  J16 ¼ maxt Mdz ðtÞ 0dz Mmax i ¼ 17219 Ji ¼   normm. 2 (pier below pylon). Fmax is the maximum uncontrolled shear force at mth pier.138 Table V. 0x.m Mmax is the maximum uncontrolled moment at mth pier. 0x.m Fnorm . 0x. J8 ¼   maxt jFdz ðtÞj 0dz Fmax Jl ¼  . 14:109–143 J23 ¼ normt ðjFdx ðtÞjÞ 0dx Fnorm . J ¼ 20222 Jj ¼    normm. 0z. 0z. m ¼ 1 (north pier).m Mmax  Mpmz indicates the transverse (z-direction) moment at mth pier. Ltd.t jFpmx ðtÞj . Summary of evaluation criteria.-H.m Fmax n ¼ 426 Jn ¼   maxm. l ¼ 911 Mpmx indicates the longitudinal (x-direction) moment at mth pier.  Struct. Mmax is the maximum uncontrolled moment at the same location.m pier.m Fnorm is the normed value of uncontrolled shear forces in the longitudinal (x-direction) direction at mth pier.

Fcon.1 x0x.2 max     maxt jxp2y ðtÞj maxt jxp2z ðtÞj .3 x0z. J39 ¼ .v ðtÞÞ .m Mnorm is the normed value of uncontrolled moments in the longitudinal (x-direction) direction at mth pier. MATLAB-BASED SEISMIC RESPONSE CONTROL J36 ¼   maxt jxp2x ðtÞj x0x.v ðFcon.tp max Subscript (tp) indicates the top of the pylon displacements at the top of the pylon. J42 ¼       maxt jxtpx ðtÞj maxt jxtpy ðtÞj maxt jxtpz ðtÞj .1 x0z.m Mnorm ðw ¼ 25227Þ 0x. J43 ¼ .t ðjMpmx ðtÞjÞ . J37 ¼   maxt jxp1x ðtÞj Subscript (1) indicates the north end of the deck displacements at the north end of the deck level (at the top of Pier 1).J24 ¼   normt ðjFdz ðtÞjÞ 0dz Fnorm 0dz Fnorm is the normed value of uncontrolled shear forces of the pylon in the transverse (z-direction) direction at the deck level. z: transverse . 14:109–143 J45 ¼   maxt.3 max     maxt jxp3y ðtÞj maxt jxp3z ðtÞj . and x0. 0dx Mnorm is the normed value of uncontrolled moments of the pylon in the longitudinal (x-direction) direction at the deck level. Jw ¼   normm. Control Health Monit. 139 . J41 ¼ ¼ 0y. 0dz Mnorm is the normed value of uncontrolled moment of the pylon in the transverse (z-direction) direction at the deck level. Copyright # 2005 John Wiley & Sons.1 xmax max max x: longitudinal.m Mnorm is the normed value of uncontrolled moments in the transverse (z-direction) direction at mth pier.t ðjMpmz ðtÞjÞ .2 xmax max Subscript (2) indicates the middle node of the deck displacements at the middle node of the deck level (at the bottom of the pylon).capacity = the capacity of control devices Fcon.2 x0z.capacity The maximum control force normalized to the capacity of control devices (capacity of control devices is assumed to be 1500 kN).3 xmax max Subscript (3) indicates the southern end of the deck displacements at the southern end of the deck level (at the top of pier 3). J40   maxt jxp3x ðtÞj x0x. Ltd. 2007.1 is max the maximum displacement of the uncontrolled response. J31 ¼   normt ðjMdx ðtÞjÞ 0dx Mnorm J32 ¼   normt ðjMdz ðtÞjÞ 0dz Mnorm J33 ¼     maxt jxp1y ðtÞj maxt jxp1z ðtÞj . J44 ¼ 0x. y: vertical. J34 ¼ .tp 0y.tp xmax xmax x0z. 0x. Struct. 0z. J35 ¼ 0y. J38 ¼ 0y.m Mnorm ðs ¼ 28230Þ 0z. Js ¼   normm.

The goal of this paper is to develop an evaluation model for control that portrays salient features of the structural system. Ltd. MR Damper with constant Voltage • • • • VE Dampers Figure 21. Copyright # 2005 John Wiley & Sons. Struct. Schematic diagram of the location of dampers in the middle point of cables (Line 10–17) and between deck and cable (Line 1–10). 14:109–143 . CHANG Figure 20. the code includes the nonlinear beam elements that represent the sag cables. Evaluation criteria are presented for the design problems that are consistent with the goals of seismic response control of a cable-stayed bridge.-H. a MATLAB-based computer program has been developed to simulate a three-dimensional cable-stayed bridge. 2007.-M.140 C. particularly the sag cables and their pre-tension forces. To this end. Both active control and hybrid-control algorithms are given to illustrate some of the design challenges of the problem. Variation of cable tension force for un-controlled case and control case B. CONCLUSIONS This paper presents a detailed study of structural control for a seismically excited cablestayed bridge. Control Health Monit. LOH AND C.

It is shown that adding viscous dampers to each cable can provide Copyright # 2005 John Wiley & Sons. the nonlinear beam elements not only use the conventional linear stiffness system (partially same as truss elements). Significant differences in the dynamic characteristics of cable-stayed bridge are estimated using different models. Through this study the following conclusions are made: (a) A comparison is made between simulation results using either truss elements or nonlinear beam elements to represent dynamic behavior of the stayed cable of the Gi-Lu bridge. To form the stiffness matrix of the cable the truss element can only consider the linear system with the basic properties (Young’s modulus. but also include the geometrically nonlinear beam elements to form the stiffness matrix and the selfweight of the cable and the pre-tension force in the cable can be implemented. Struct. cross-sections and length of elements) for formulation.g. 14:109–143 . hybrid control devices are applied to the structure to control response due to seismic excitation. cannot prove that using the nonlinear beam element to represent the cable is more reasonable than using the truss element. Ltd.. 2007.MATLAB-BASED SEISMIC RESPONSE CONTROL 141 Figure 22. But there are some defects associated with using the truss element for cable. Comparison of evaluation criteria of three different control systems. (b) Comparison on the responses of the cable-stayed bridge. Using data from field tests it is believed that the nonlinear beam elements to simulate the stayed cable in a realistic manner. Because nonlinear beam elements are used to simulate the stayed cable. Control Health Monit. it is a relatively simple matter to add dampers between cables and the deck. (c) The simulation results show that use of the actuators as control devices between the bridge deck and both end-abutments is not very effective for controlling the vibration of the bridge deck. as shown in Figures 6 and 7. using truss elements cannot obtain the bridge out-plane responses and underestimates the cable tensile force. pylon. (d) Instead of using an active control method. and the stayed cables. On the contrary. e.

that not only the reduces the deck and pylon response. (e) With implementation of extra viscous dampers between the deck and the stayed cables. (g) A formulation of the structural model and a simple control design has been made in the form of a set of MATLAB equations.-M. the displacement at the top of the pylon (J42 and J43) in both longitudinal and transverse directions and shear force at the bottom of pylon (J7) and at the top of the pier 3 (J3) in the longitudinal direction are significantly reduced. but also the response of the cables. Ltd. Comparison of the normalized displacement at the middle point of all cables: (a) North-West. Control Health Monit. (c) North-East. CHANG Figure 23. good control effectiveness. (f) In this study the inherent damping ratio for cable is assumed 1% and with the supplementation of damping ratio from the viscous damper in the cable (1. This code can provide another type of benchmark Copyright # 2005 John Wiley & Sons.3–20% from damper) the damping ratio for cable is ranging from 2. 2007. Struct. For practical consideration the control strategy 3 can also provide good control effectiveness in most of the evaluation criteria.-H. and (d) South-East sides. LOH AND C. (b) South-West.142 C. 14:109–143 .3% (longest cable) to 21% (shortest cable).

Earthquake Engineering and Structural Dynamics 1994. 23:539–552. Yang JN. Susumpow T. Fujino Y. Agret R. Control Health Monit. however. 2007. Nagarajaiah S. Chang CM. 2000. 2001. Pacheco BM. Yang JN. ACKNOWLEDGEMENT The authors wish to express their thanks to Dr Z. Bergman LA. Caicedo JM. 12. Przemieniecki SP. Estimation curves for modal damping in stay cables with viscous dampers. Journal of Intelligent Material Systems and Structures 1997. Schmitendorf WE. Wu JC. Lee IW. Technical Report. 23:1283–1297.wustl. Rodellar J. 7. Fujino Y. Voulgaris PG. 9:75–94. Pacheco BM. Schemmann AG. Iyer I. Virlogcux M. Theory of Matrix Structural Analysis. Paolone A. Benchmark control problem for seismic response of cablestayed bridge. 13. 5. Turan G. Application of MR-dampers to control of a cable-stayed bridge subject to seismic excitation. Rodeller J. 2. Moon SJ. Robust control techniques for building under earthquake excitations. Caicedo JM. ASCE Journal of Engineering Mechanics 1995. 19. 18. 1998. An experimental study on active control of planner cable vibration by axial support motion. Sliding mode control for nonlinear and hysteretic structures. Planar motion of a cable-supported beam with feedback controlled action. Manosa V. Lee (Associate Research Fellow of NCREE) to provide experimental data of Gi-Lu bridge for this study and the support from National Science Council under Grant NSC93-2211-E-002-005 is also acknowledged. University of Illinois at Urbana-Champaign. ASCE Journal of Structural Engineering 119(6). 6. Phase II benchmark control problem for seismic response of cable-stayed bridges. Smith HA. Esdahl S (eds) Bridge Aerodynamics. Bergman LA. Warnitchai P. Sahasrabudhe S. Fujino Y. Bergman LA. Earthquake Engineering and Structural Dynamics 2001. Journal of Structural Control and Health Monitoring 2005. 17. Bergman LA. Philadelphia. Journal of Structural Control 2003. Jabbari F. Voulgaris PG. 8:767–774. 1968. Proceedings of the 14th Analysis and Computational Speciality Conference. REFERENCES 1.MATLAB-BASED SEISMIC RESPONSE CONTROL 143 structural control problem for cable-stayed bridges. Sulekh A. McGraw-Hill: New York. Monroy C. An active tendon control scheme for cable-stayed bridges with model uncertainties and seismic excitation. Composite robust active control of seismically excited structures with actuator dynamics. Turan G. 60:948–953. Struct. 15. CD-ROM. In: Larsen A. Earthquake Engineering and Structural Dynamics 1993. ASCE Journal of Applied Mechanics 1993. 27:811–843. Earthquake Engineering and Structural Dynamics 1998. http://wusceel. de la Sen M. Neuro-control of seismically excited steel structures through sensitivity evaluation scheme. Gattulli V. Dyke SJ. Ltd. 14:109–143 . Vibration control of cable-stayed bridge. Fujino Y. Experimental study on active tendon control of cable-stayed bridges. 22(2):93–111. 129:71–77. Earthquake Engineering and Structural Dynamics 27:301–311. 10. 16. 213–233. 9. Journal of Structural Control 2002. K. ASCE Journal of Engineering Mechanics 2003. Earthquake Engineering and Structural Dynamics 1994. a number of practical aspects not considered here need to be tackled in further studies in order to gain a deeper evaluation of the potential effectiveness for mitigation of vibration. 3. 14. Copyright # 2005 John Wiley & Sons. Moon SJ. Vibration control assessment of ASCE benchmark model of cable-stayed bridge.cive.edu/quake/benchmark/2000 8. Moon SJ. Pacheco BM. 4. The numerical results from use of a hybrid control system look promising. Cable vibration in cable-stayed bridge. Hague S. 121:1330–1339. 11. Seismic response of sliding isolated bridges with smart dampers subjected to near source ground motion. Kim DH. Balkema: Rotterdam. Loh CH. Sliding mode control of cable-stayed bridge subjected to seismic excitation. Warnitchai P. Hague D.1961–1979. 10:137–168. Dyke JS. Luo N. Active stiffness control of cable vibration. 30:1361–1377.

Sign up to vote on this title
UsefulNot useful