Journal of Sound and Vibration (2002) 257(2), 301–319

doi:10.1006/jsvi.5060, available online at http://www.idealibrary.com on

DYNAMIC ANALYSIS OF LARGE-DIAMETER SAGGED

CABLES TAKING INTO ACCOUNT FLEXURAL RIGIDITY
Y. Q. Ni
Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, Hung Hom,
Kowloon, Hong Kong. E-mail: [email protected]
J. M. Ko
Faculty of Construction and Land Use, The Hong Kong Polytechnic University, Hung Hom,
Kowloon, Hong Kong. E-mail: [email protected], web http://www.cse.polyu.edu.hk/
dynamics
and
G. Zheng
Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, Hung Hom,
Kowloon, Hong Kong. E-mail: [email protected]

(Received 11 May 2001, and in final form 16 January 2002)

For the purpose of developing a vibration-based tension force evaluation procedure

for bridge cables using measured multimode frequencies, an investigation on accurate
finite element modelling of large-diameter sagged cables taking into account
flexural rigidity and sag extensibility is carried out in this paper. A three-node
curved isoparametric finite element is formulated for dynamic analysis of bridge stay
cables by regarding the cable as a combination of an ‘‘ideal cable element’’ and a
fictitious curved beam element in the variational sense. With the developed finite
element formulation, parametric studies are conducted to evaluate the relationship
between the modal properties and cable parameters lying in a wide range covering most of
the cables in existing cable-supported bridges, and the effect of cable bending stiffness and
sag on the natural frequencies. A case study is eventually provided to compare the
measured natural frequencies of main cables of the Tsing Ma Bridge and the computed
frequencies with and without considering cable bending stiffness. The results show that
ignoring bending stiffness gives rise to unacceptable errors in predicting higher order
natural frequencies of the cables, and the proposed finite element formulation provides an
accurate baseline model for cable tension identification from measured multimode
frequencies.
# 2002 Elsevier Science Ltd. All rights reserved.

1. INTRODUCTION

Advances in modern construction technology have resulted in increasing application of

large-diameter and long-span structural cables in cable-supported bridges. As a result of
carrying both road and rail traffic, the Tsing Ma Bridge in Hong Kong, a suspension
bridge with the main span of 1377 m [1], has the most heavily loaded cables in the world.
The main cable section of the bridge is about 11 m in diameter after compacting. The
Akashi Kaikyo Bridge in Japan, which is the world’s longest suspension bridge with the

0022-460X/02/$35.00 # 2002 Elsevier Science Ltd. All rights reserved.

302 Y. Q. NI ET AL.

main span of 1991 m, also has the main cables of about 11 m diameter [2]. The effect of
bending stiffness in these large-diameter bridge cables is not negligible. With the Tatara
Bridge being a landmark, the construction of cable-stayed bridges is now entering a
new era, with main spans reaching 1000 m. The Stonecutters Bridge, currently
under design in Hong Kong, is a cable-stayed bridge with the main span of 1018 m
[3, 4]. The Tatara Bridge with 890 m main span, which is the existing longest cable-
stayed bridge in the world [5], has its longest stay cables of about 460 m in length.
A recent design trend for this bridge type is the multi-span cable-stayed bridges with three
or more towers [6, 7]. A critical problem of multi-span cable-stayed bridges is the
stabilization of the central tower(s), which has resulted in increasing application of
extremely long stabilizing cables. For example, the three-tower Ting Kau Bridge in Hong
Kong utilizes eight longitudinal stabilizing cables of 465 m long for strengthening the
slender central tower [8]. These long stay cables exhibit considerably large sags, and
consequently, the effect of sag–extensibility on the cable static and dynamic characteristics
is noticeable.
During the construction and service life of a cable-supported bridge, it is essential
to accurately define the cable forces, whose change due to degradation or other
factors will affect internal force distribution in the deck and towers and influence
the bridge alignment. As a result, the cable tension is an important index for assessing
overall structural health and condition of cable-supported bridges. Vibration
measurement has been one of the most widely used methods for in situ evaluation of
cable tension forces [9–12]. The engineering implementation of this technique is mostly
done based on the taut string theory. According to this theory, cable tension force
can be estimated from the measured fundamental natural frequency based on a
simple explicit relation. However, recent studies [13–15] indicated that identification
based on this simplified formula could not provide accurate tension force estimation for
bridge cables in many situations, and showed that cable sag–extensibility and bending
rigidity might have pronounced effects on the modal properties and tension force
identification of long-span bridge cables. In addition, it has been demonstrated [11, 16]
that the identification procedures based on multimode frequencies can provide more
accurate cable tension evaluation than the conventional approach using only fundamental
natural frequency. This needs a cable model that can accurately predict the relation
between the modal properties and cable parameters. The string theory cannot take into
account the sag effect [17], and even the modern cable theory does not consider the flexural
rigidity [18, 19].
In the present study, an accurate finite element model for structural cables which takes
into account flexural rigidity, sag–extensibility and spatial variability of dynamic tension is
formulated for the purpose of cable parameter identification from measured modal
properties. For the convenience of capturing the effects of sag–extensibility and flexural
rigidity, respectively, the formulation is made as a composition of two separable parts: a
pure sagged cable without flexural rigidity and a fictitious curved beam element with
only flexural rigidity. Parametric studies are conducted to examine the effects of
cable bending stiffness and sag–extensibility on modal properties over a wide parameter
range, and to compare the results from the proposed formulation and from the literature.
A case study is then presented to analyze the modal dynamic behaviors of a real
bridge cable with and without considering bending stiffness. The true modal parameters
of the cable are obtained by field ambient vibration measurements, and the predicted
and measured natural frequencies are compared to verify significant effect of the
bending stiffness on higher-mode frequencies and accuracy of the proposed
formulation.
 DYANAMIC ANALYSIS OF SAGGED CABLES 303
2. FORMULATION

2.1. BASIC ASSUMPTIONS

The finite element formulation derived in this section is based on a pure sagged cable
element and a fictitious curved beam element. The cable element is characterized by the
following idealizations:
1. the cable material is linearly elastic;
2. the cable is perfectly flexible so that it is capable of developing stresses only in the
direction normal to the cross-section;
3. the normal stress is uniform over the cross-sectional area;
4. though the displacement may be arbitrarily large, the strain is assumed to be small,
which means that the cross-sectional area does not change during deformation; and
5. the radius of the cable curvature is large with regard to its cross-section dimension.
The curved beam element is characterized by the following idealizations:
1. the material is linearly elastic;
2. a plane section originally normal to the neutral axis remains plane after deformation,
but does not necessarily remain normal to the neutral axis (Timoshenko beam’s
assumption); and
3. the curvature radius of the beam is large with regard to its cross-section dimension.

2.2. THREE-NODE CURVED ELEMENT OF SAGGED CABLE

Without losing generality, the cable static equilibrium profile is assumed in the x2y
plane as shown in Figure 1. This initial (static) configuration is defined by xðsÞ and yðsÞ;
here s denotes the arc length co-ordinate. Let L; E; A and m be the cable length, modulus
of elasticity, cross-sectional area and mass per unit length respectively. In static
equilibrium state, the cable is subjected to dead load and the static tension is HðsÞ: Due
to the geometric non-linearity of the cable, the static equilibrium configuration of the cable
is achieved through a non-linear iteration process by applying the dead load in a stepwise
fashion and continuously updating the geometry of the cable (including the tension force).
The Newton–Raphson procedure is used to iteratively calculate the dead-load deformed
shape with prescribed cable tension if known. The iteration starts from a straight profile

s=0

z y
Initial configuration

Dynamic configuration

s=L
x
Figure 1. Schematic of cable configuration.
304 Y. Q. NI ET AL.

with a presumed tension force. Following this approach, accurate cable profile in static
equilibrium state is finally obtained which may be catenary, parabolic or other shapes
depending on specific static loads.
After achieving the static equilibrium configuration through a non-linear static analysis,
the cable is subjected to the action of dynamic external forces px ðs; tÞ; py ðs; tÞ and pz ðs; tÞ:
The dynamic configuration of the cable is described by the displacement responses
uðs; tÞ; vðs; tÞ and wðs; tÞ measured from the position of static equilibrium in the x, y and z
directions respectively. Let U ¼ fuðs; tÞ vðs; tÞ wðs; tÞgT and P ¼ fpx ðs; tÞ py ðs; tÞ pz ðs; tÞgT :
By using the Lagrangian strain measure, the cable extensional strain due to dynamic loads,
ignoring flexural rigidity, can be expressed as
dXT @U 1 @UT @U
e ¼ e0 þ el ¼  þ ; ð1Þ
ds @s 2 @s @s
where X ¼ fxðsÞ yðsÞ 0gT is the static equilibrium co-ordinate vector.
The finite element formulation is derived from the Hamilton’s principle
Z t2 Z L Z t2 Z L
dH ¼ d ðQ V Þ ds dt þ dW ds dt ¼ 0 ð2Þ
t1 0 t1 0

in which Q is the kinetic energy density, V is the elastic strain energy density, and dW is
the virtual work density associated with the dead load, dynamic excitation and damping
force. They are expressed as
m @UT @U
Q¼ ; ð3aÞ
2 @t @t
EA 2
V ¼ Vi þ e þ HðsÞe; ð3bÞ
2
 
T @U
dW ¼ dU  q þ P c  ; ð3cÞ
@t
where Vi is the elastic strain energy density held in the initial (static) configuration, q is the
dead load vector existent in the initial state and c ¼ diag½cx cy cz  is the viscous damping
coefficient matrix.
Substituting equations (3) and (1) into equation (2) yields
Z t2 Z L "  2
m @UT @U EA dXT @U 1 @UT @U
dI ¼ d Vi þ
t1 0 2 @t @t 2 ds @s 2 @s @s
 T  Z t2 Z L  
dX @U 1 @UT @U T @U
H þ ds dt þ dU q þ P c ds dt ¼ 0: ð4Þ
ds @s 2 @s @s t1 0 @t
Here the displacement vector U is selected to fulfill the boundary conditions and initial
conditions. By integrating equation (4) by parts and accounting for the static equilibrium
configuration, we have
Z t2 Z L   T  
@2U @ dX @U 1 @UT @U dX @U
dI ¼ dUT m 2 EA þ þ
t1 0 @t @s ds @s 2 @s @s ds @s
  
@ @U @U
H þc P ds dt ¼ 0: ð5Þ
@s @s @t
An isoparametric curved element with three nodes is introduced to describe the cable.
As shown in Figure 2, the shape functions in the natural co-ordinate system are given by
N1 ¼ 12ð1 xÞ 12ð1 x2 Þ; N2 ¼ 1 x2 ; N3 ¼ 12ð1 þ xÞ 12ð1 x2 Þ ð6a2cÞ
 DYANAMIC ANALYSIS OF SAGGED CABLES 305

1 (x1j, y1j)

lj
y
2 2 (x2j, y2j)
s
x lj 3 (x3j, y3j)
2
z
(a)

ξ=1 ξ=0 ξ =+ 1

1 2 ξ 3
(b)

Figure 2. Three-node curved cable element: (a) physical co-ordinate; (b) natural co-ordinate.

and the co-ordinates and the displacement functions are expressed as

X X
x¼ Ni xi ; y ¼ Ni yi ð7a; bÞ
X X X
u¼ N i ui ; v ¼ N i vi ; w ¼ Ni w i ; ð8a cÞ
where xi ; yi (i ¼ 1; 2; 3) are the nodal co-ordinates; and the nodal displacement vector is
defined as
fUj g ¼ ffUgTj1 fUgTj2 fUgTj3 gT ¼ fu1j v1j w1j u2j v2j w2j u3j v3j w3j gT : ð9Þ
By rewriting equation (8) as
U ¼ fu v wgT ¼ ½N1 I N2 I N3 IffUgTj1 fUgTj2 fUgTj3 gT ¼ ½NfUj g ð10Þ
and substituting equations (7)–(10) into equation (1), we get the following expressions:
e0 ¼ ½B0 fUj g ¼ ½fB01 g fB02 g fB03 gfUj g; ð11aÞ

el ¼ ½Bl fUj g ¼ ½fBl1 g fBl2 g fBl3 gfUj g; ð11bÞ

fB0i g ¼ J12 fx0 Ni0 y0 Ni0 0g; ð11cÞ

fBli g ¼ 2J1 2 fu0 Ni0 v0 Ni0 w0 Ni0 g; ð11dÞ

where J ¼ ds=dx and the prime denotes the derivative with respect to x:
Substituting equations (9)–(11) into equation (5), after some manipulation, yields
Z t2 X
dI ¼ . j g þ ½Cj fU
dfUj gT f½Mj fU ’ j g þ ½K0j þ K1j ðfUj gÞ
t1

þ K2j ðfUj g fUj gT ÞfUj g fPj gg dt ¼ 0 ð12Þ

from which the governing equation of motion of the element j is obtained as
. j g þ ½Cj fU
½Mj fU ’ j g þ ½K0j þ K1j ðfUj gÞ þ K2j ðfUj g fUj gT ÞfUj g ¼ fPj g ð13Þ
306 Y. Q. NI ET AL.

in which
Z þ1
½Mj  ¼ mJ ½NT ½N dx; ð14aÞ
1
Z þ1
½Cj  ¼ J ½NT ½c½N dx; ð14bÞ
1
Z þ1
fPj g ¼ J ½NT fPg dx; ð14cÞ
1
Z þ1 Z þ1
½K0j  ¼ EAJ ½B0 T ½B0  dx þ 2J1 H½N0 T ½N0  dx; ð14dÞ
1 1
Z þ1
½K1j  ¼ EAJ ð½Bl T ½B0  þ 2½B0 ½Bl Þ dx; ð14eÞ
1
Z þ1
½K2j  ¼ 2EAJ ½Bl T ½Bl  dx: ð14fÞ
1

The global equation of the cable is then obtained through assembling the element mass
matrix, damping matrix, stiffness matrix and nodal load vector by the standard assembly
procedure. It is noted that in equation (14) the stiffness matrix includes linear stiffness
term [K0], quadratically non-linear stiffness term [K1] and cubically non-linear stiffness
term [K2]. The present study only addresses linear problem of cable dynamics by ignoring
the non-linear stiffness terms. For non-linear dynamic analysis of cables in terms of the
above formulae, see reference [20].

2.3. FORMULATION FOR FLEXURAL RIGIDITY

The additional stiffness contribution due to the flexural rigidity of cable is
derived by assuming a fictitious curved beam. The curved beam element is same
as shown in Figure 2, but a new local co-ordinate system in terms of tangential and
normal axes is introduced for the convenience of formulation to relate displacements
with stress resultants. As illustrated in Figure 3, the displacements at any node i are
expressed as
fdgi ¼ fui vi wi ysi yti yzi gT ; ð15Þ
where ui is the in-plane displacement in the tangential direction, vi is the displacement in
the transverse direction, wi is the displacement in the z direction, ysi is the total rotation in
the tangential direction, yti is the angle of twist, and yzi is the total rotation of transverse
bending. Similar to equation (9), the displacement vector is expressed with isoparametric
interpolation functions as
fUg ¼ fu v w ys yt yz gT ¼ ½N1 I N2 I N3 IffdgT1 fdgT2 fdgT3 gT ¼ ½Nfdg: ð16Þ
The strain vector is written as
feg ¼ fkz kt a gvs gws g; ð17Þ
where kz and kt are the in-plane and out-of-plane curvature changes, respectively, a is the
cross-sectional torsion change, gvs and gws are the shear strains. The strain–displacement
 DYANAMIC ANALYSIS OF SAGGED CABLES 307

θsi s
y
t
R
ds
θzi
i θti
ui
vi
wi
x
z
Figure 3. Displacements at node i:

relation can be expressed as

@yz 1 @u @ys yt
kz ¼ þ ; kt ¼ ; ð18a; bÞ
@s R @s @s R
@yt ys @v @w
a¼ ; gvs ¼ yz ; gws ¼ þ ys ð18c2eÞ
@s R @s @s
in which R is the curvature radius of the element. It should be noted that R is not a
constant for a sagged cable. It is calculated using the formula
½1 þ ðdy=dxÞ2 3=2
R¼ : ð19Þ
d2 y=d2 x
Combining equations (16) and (18) yields
feg ¼ ½Bfdg ¼ ½½B1  ½B2  ½B3 fdg ð20Þ
in which ½Bi  (i ¼ 1; 2; 3) is expressed as
2 0 3
Ni 0 0 0 0 RNi0
6 7
6 0 0 0 RNi0 JNi 07
16 6
7
7:
½Bi  ¼ 0 0 0 JNi RNi0 0 ð21Þ
RJ 66
7
7
4 0 RNi0 0 0 0 RJNi 5
0 0 RNi0 JRNi 0 0
Figure 4 shows the stress resultants at node i: The stress strain relation is given by
fsg ¼ fMz Mt T Vz Vs gT ¼ diag½EIz EIs GJs bGA bGAfeg ¼ ½Dfeg: ð22Þ
With equations (20) and (22), the additional element stiffness matrix due to cable
flexural rigidity is derived in a similar way as
Z þ1
½Ka  ¼ J ½BT ½D½B dx: ð23Þ
1

The additional stiffness matrix given in equation (23) is obtained by referring to the
local co-ordinate system. It should be transformed into the element stiffness relation in the
global x2y2z co-ordinate system before performing assembly to obtain overall stiffness
matrix. Likewise, the element stiffness matrix given in equation (14), with 9 9 dimension,
only accommodates the translational degrees of freedom. It should be expanded in the
assembly process as an 18 18 matrix to cater for the rotation degrees of freedom.
308 Y. Q. NI ET AL.

Msi s
y t
R
ds
Mzi
i Ti
Pi
Vsi
Vzi
x
z
Figure 4. Stress resultants at node i:

3. PARAMETRIC STUDIES

The proposed formulation has been encoded into a versatile finite element program. In
this section, parametric studies are conducted to evaluate the effect of bending stiffness
and sag–extensibility on natural frequencies, and the relation between the modal
properties and cable parameters for a wide parameter range. A numerical verification is
first carried out by comparing the computed results by the present method with the
analytical results available in literature. Mehrabi and Tabatabai [15] formulated a
differential equation for solution of free vibration of suspended cables by use of finite
difference technique. This approximate formula accounted for cable bending stiffness and
sag–extensibility, but was based on the assumption of flat sag and invariability of dynamic
tension along cable length. The following dimensionless parameters have been adopted to
characterize the bending stiffness and sag–extensibility respectively:
rffiffiffiffiffiffi  
Hh 2 LEA mgL 2
x¼L ; l ¼ ; ð24a; bÞ
EI Hh Le Hh
where Hh is the horizontal component of cable static tension force, I is the moment of
inertia of cable cross-section, and
Z L  3 "   #
ds 1 mgL 2
Le ¼ dx ffi L 1 þ : ð25Þ
0 dx 8 Hh

Four suspended cables with the same length of 100 m but different sag–extensibility ðl2 Þ
and bending-stiffness ðxÞ parameters are analyzed. Table 1 shows the parameters of the
four cables. Cable 1 ðl2 ¼ 079; x ¼ 6055Þ has a moderate sag and a low bending stiffness;
Cable 2 ðl2 ¼ 5070; x ¼ 3027Þ has a large sag and an average bending stiffness; Cable 3
ðl2 ¼ 141; x ¼ 505Þ has a moderate sag and a high bending stiffness; Cable 4
ðl2 ¼ 5070; x ¼ 505Þ has a large sag and a high bending stiffness. Modal properties of
the four cables are evaluated by the proposed finite element formulation. The static
profiles of the cables are assumed as parabolas. Sixty equi-length cable elements are used
in the computation. Table 2 presents a comparison of predicted natural frequencies of the
first two in-plane modes obtained by the taut string theory, the finite difference formula
and the present method. It is observed that for all the four cases the results by the present
method coincide well with those by the finite difference formula. Both the methods take
into account sag–extensibility and bending stiffness. It is found from the table that the
computed natural frequencies from the taut string equation (ignoring sag–extensibility and
bending stiffness) are quite different from those calculated by the present method and the
 DYANAMIC ANALYSIS OF SAGGED CABLES 309
Table 1
Material and geometric parameters of four cables
2
Cable no. l x m (kg/m) g (N/kg) L (m) Hh (106 N) E (Pa) A (m2) Js (m4)
1 079 6055 4000 98 1000 290360 15988e+10 78507e 03 49535e 06
2 5070 3027 4000 98 1000 072590 17186e+10 76110e 03 46097e 06
3 141 505 4000 98 1000 2613254 20826e+13 78633e 03 49204e 06
4 5070 505 4000 98 1000 072590 47834e+08 27345e 01 59506e 03

Table 2
Comparison of computed frequencies of in-plane modes (Hz)
String theory Finite difference formula Present method
2
Cable no. l x First Second First Second First Second
mode mode mode mode mode mode
1 079 6055 0426 0852 0440 0853 0441 0854
2 5070 3027 0213 0426 0428 0464 0421 0460
3 141 505 1278 2556 1399 2679 1400 2682
4 5070 505 0213 0426 0447 0464 0438 0461

Table 3

Parameters of two sets of cables

Cable set 1 Cable set 2
2 2
A (m ) Hh (N) d=L f1s (Hz) A (m ) Hh (N) d=L f1s (Hz)
7 5 2
Minimum value 9788 10 01031 1/23 750 05810 9575 10 9846 10 1/2375 01817
Maximum value 014258 3236 106 1/110 85284 14258 3236 1010 1/11 26969

finite difference formula, indicating a considerable influence of sag–extensibility and

bending stiffness in these cases.
In order to relate the modal properties with cable parameters, the relation surfaces of
dimensionless frequencies versus l2 and x are obtained for a wide range of structural
parameters. Two cable sets are considered. The fixed cable parameters are L ¼ 100 m and
E ¼ 200 GPa.The change ranges of other parameters are given in Table 3 that result in
00014l2410 000 and 104x4260 for both cable sets. The volume mass density of set 1 is
kept as a constant value of r ¼ m=A ¼ 786 103 kg/m3, while the density of set 2 is
appropriately altered to produce same ranges of l2 and x for sets 2 and 1. The main
difference between the two sets is the range of their sag-to-span ratio d=L: Both the finite
difference formula and the present method are used to compute the relation surfaces. The
cable is divided into 50 elements in the finite element solution and 100 elements in the finite
difference formula, with totally 101 nodes used for each method. With the computed
natural frequency fn of the nth mode, the corresponding dimensionless frequency is
defined as
fn
f%n ¼ ; ð26Þ
f1s
310 Y. Q. NI ET AL.

Figure 5. Relation surfaces of two in-plane fundamental modes obtained by finite difference formula: (a) first
symmetric mode (cable sets 1 and 2); (b) first antisymmetric mode (cable sets 1 and 2).

Figure 6. Relation surfaces of two in-plane fundamental modes obtained by the present method: (a) first
symmetric mode (cable set 1); (b) first antisymmetric mode (cable set 1); (c) first symmetric mode (cable set 2); (d)
first antisymmetric mode (cable set 2).

where f1s is the fundamental frequency of the corresponding stretched string which is
obtained from
rffiffiffiffiffiffiffiffiffiffi
1 Hh A
f1s ¼ : ð27Þ
2L m
Figure 5 shows the relation surfaces of dimensionless frequencies versus l2 and x for the
first symmetric and antisymmetric in-plane modes obtained by the finite difference
formula. Figure 6 gives the corresponding relation surfaces obtained by the present
 DYANAMIC ANALYSIS OF SAGGED CABLES 311

Figure 7. Comparison of relation surfaces of higher order in-plane modes obtained by two methods:
(a) seventh symmetric mode by finite difference formula (cable sets 1 and 2); (b) seventh symmetric mode by the
present method (cable set 2); (c) eighth symmetric mode by finite difference formula (cable sets 1 and 2); (d) eighth
symmetric mode by the present method (cable set 2).

method. By comparing the two figures, it is seen that the results for the two fundamental
modes obtained by the approximate finite difference formula and by the present method
coincide very well with each other throughout the concerned ranges of l2 and x for cable
set 1 (with low sag-to-span ratio). For cable set 2 (with high sag-to-span ratio), a good
agreement is still achieved except for a small region with very high l2 and very low x where
the relation surfaces display a slight difference with each other. Figure 7 provides a
comparison of the relation surfaces of two high order modes obtained by the two methods.
It is observed that in these cases the results deviate significantly from each other in the
range with small x: The deviation of the approximate finite difference formula from the
finite element results increases with the mode order. Figure 8 illustrates the relation
surfaces of dimensionless frequency versus l2 and x for high order in-plane modes of cable
set 2 obtained by the present method. Figure 9 shows the relation surfaces for out-of-plane
modes of cable set 2 using the present method. It is obvious from Figures 8 and 9 that the
high order frequencies of the in-plane as well as out-of-plane modes depend on both
parameters l2 and x; although the change rate of the frequencies along the parametric axis
of l2 is much smaller than that along the parametric axis of x: This observation is different
from that made in reference [15]. These differences are attributed to the fact that the finite
difference formula was derived on the assumptions of parabolic static profile, flat sag and
spatial invariability of dynamic tension, while the present method eschews these
assumptions. The above studies conclude that when high order modal properties are
required in an inverse problem, a precise finite element model is more acceptable in regard
to its accuracy, versatility and reliability.
312 Y. Q. NI ET AL.

Figure 8. Relation surfaces of high order in-plane modes obtained by the present method for cable set 2:
(a) second symmetric mode; (b) second antisymmetric mode; (c) third symmetric mode; (d) third antisymmetric
mode; (e) fourth symmetric mode; (f) fourth antisymmetric mode; (g) eighth symmetric mode; (h) eighth
antisymmetric mode.
 DYANAMIC ANALYSIS OF SAGGED CABLES 313

Figure 9. Relation surfaces of out-of-plane modes obtained by the present method for cable set 2: (a) first
symmetric mode; (b) first antisymmetric mode; (c) second symmetric mode; (d) second antisymmetric mode;
(e) third symmetric mode; (f) third antisymmetric mode; (g) eighth symmetric mode; (h) eighth antisymmetric
mode.
314 Y. Q. NI ET AL.

4. CASE STUDIES

4.1. TSING MA BRIDGE CABLES

The Tsing Ma Bridge, as shown in Figure 10, is a double deck suspension bridge with a
main span of 1377 m and an overall length of 2160 m [1]. The two main cables of the bridge
are 36 m apart and have a cross-section of 1.1 m in diameter after compacting. The central
span deck and the Ma Wan side span deck are suspended at 18 m intervals by hangers to
the main cables. The Tsing Yi side span deck is instead supported from the ground by
three concrete piers spaced at 72 m centers. As a result, the main cables on the Tsing Yi
side span are free cables without bridge deck being suspended. The modal properties of the
Tsing Ma Bridge in different construction stages (including the erection completion stage)
have been measured through a series of ambient vibration survey [21]. One stage under
measurement is the freely suspended cable stage. In this stage, only the tower–cable system
was erected but none of deck segments has been hoisted into position. The main cables on
all the three spans were free cables in this stage. The modal parameters of the main cables
on the main span and the Tsing Yi side span in this stage are analyzed by the present
method and compared with the measurement results for verification. The cable length and
sag are 13978 and 1125 m for the suspended main span cable, and 3291 and 57 m for
the inclined Tsing Yi side cable. The horizontal component of the tension force is
122 642 kN for both the cables. The main span cable is partitioned into 77 elements and
the Tsing Yi side span cable is partitioned into 17 elements. The computation is conducted
by assuming the cable supports as pinned ends and fixed ends respectively. Tables 4 and 5
list the natural frequencies of the first three in-plane and out-of-plane modes of the two
cables. It is seen that the computed natural frequencies agree favorably with the
measurement results.
After completion of the bridge construction, a total of about 300 sensors, including four
accelerometers on the main cables, have been permanently installed on the Tsing Ma

23m 76.5m 355.5m SPAN = 1377m 300m

72m 72m 72m 72m

206.4m 206.4m

78.58m

Anchorage
Anchorage
Ma Wan Island Tsing Yi Island

Figure 10. Elevation of Tsing Ma Bridge.

Table 4
Natural frequencies of main span cable in freely suspended cable stage (Hz)
Out-of-plane modes In-plane modes
Mode no. First Second Third First Second Third
Computed: pinned ends 00522 01040 01557 01008 01471 02081
Computed: fixed ends 00528 01052 01578 01020 01488 02091
Measurement 00530 01050 01560 01020 01430 02070
 DYANAMIC ANALYSIS OF SAGGED CABLES 315
Table 5
Natural frequencies of Tsing Yi side span cable in freely suspended cable stage (Hz)
Out-of-plane modes In-plane modes
Mode no. First Second Third First Second Third
Computed: pinned ends 02352 04696 07154 03527 04693 07216
Computed: fixed ends 02450 04946 07534 03569 04943 07593
Measurement 02360 04770 07400 03430 04780 07310

12.0
Present method
10.0 Pure cable model
Frequency (Hz)

Measurement
8.0

6.0

4.0

2.0

0.0
0 2 4 6 8 10 12 14 16 18 20
(a) Mode order

12.0
Present method
10.0 Pure cable model
Frequency (Hz)

Measurement
8.0

6.0

4.0

2.0

0.0
0 2 4 6 8 10 12 14 16 18 20
(b) Mode order

Figure 11. Comparison of computed and measured natural frequencies of Tsing Yi side span cable in erection
completion stage: (a) frequencies of in-plane modes; (b) frequencies of out-of-plane modes.

Bridge for structural health and condition monitoring [22]. The tension forces of the Tsing
Yi side span free cables can be therefore monitored by means of vibration measurement.
In order to examine the effect of cable bending stiffness on the modal properties, the
natural frequencies of the Tsing Yi side span free cable in the erection completion stage are
predicted using the present method (considering cable bending stiffness) as well as the pure
cable model (ignoring cable bending stiffness), and then compared with the measurement
results in the same stage from ambient vibration survey. The cable length and sag in this
stage are 3315 and 18 m. The horizontal component of the tension force is 405 838 kN.
The cable is divided into 200 elements in computation. Figure 11 shows a comparison of
the natural frequencies of the first 19 in-plane and out-of-plane modes obtained by the
present method, the pure cable model and ambient vibration measurement. Because the
316 Y. Q. NI ET AL.

1.8
accelerometer 8th mode 16th mode
1.2
Mode shape
0.6

0.0

_ 0.6

_1.2
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Normalized cable length

Figure 12. Deployment of accelerometer on cable for ambient vibration measurement.

accelerometer for ambient vibration measurement was located close to a modal node of
the eighth and 16th modes as shown in Figure 12, the natural frequencies of these two
modes could not be measured with high fidelity. It is observed from Figure 11 that the
predicted natural frequencies by the present method agree very well with the measurement
results, whereas the natural frequencies predicted using the pure cable model gradually
deviate from the measured values with the increase of mode order. The maximum error is
as large as 30% for the first 20 modes. This example proves again the significant influence
of cable bending stiffness on higher-mode frequencies and the necessity of using an
accurate model for multimode-based cable tension and parameter identification.

4.2. TING KAU BRIDGE CABLES

Due to large flexibility, relatively small mass and extremely low damping, structural
cables are susceptible to vibration in dynamic conditions. For example, unexpectedly large
oscillation occurring in bridge stay cables under specific combinations of wind and rain
has been observed in a number of cable-stayed bridges worldwide [23]. This has resulted in
increasing application of passive and semi-active dampers in cable-stayed bridges for cable
vibration mitigation. For a bridge cable attached with dampers, the existing analytical or
approximate formulae are difficult to accurately identify the tension force due to their
inability in dealing with damper stiffness effect. The present finite element method does
not suffer from this restriction. Modal analysis of the damper-attached Ting Kau bridge
cables by the present method is provided here as an example.
As shown in Figure 13, the Ting Kau Bridge is a multi-span cable-stayed bridge with
three monoleg towers supporting two main spans of 448 and 475 m and two side spans of
127 m each [8]. In this bridge, eight longitudinal stabilizing cables with a length of up to
465 m have been used to strengthen the slender central tower. Passive dampers have been
installed between the cables and deck near the lower ends in the cable planes. The dampers
were connected perpendicular to the cables at the location of 19.2 m cable length measured
from the lower ends. The length and sag of the longitudinal stabilizing cable are 4649 m
and 83 m respectively. The horizontal component of the cable tension force is 23915 kN.
The cable is divided into 200 elements in computation. Figure 14 illustrates the relation
diagrams of the cable natural frequency versus spring (damper) stiffness for the first six in-
plane modes. The maximum frequency discrepancy with and without considering damper
 DYANAMIC ANALYSIS OF SAGGED CABLES 317
127m 448m 475m 127m
Central Tower
Ting KauTower 201.45m Tsing Yi Tower
173.53m 164.74m

Figure 13. Elevation of Ting Kau Bridge.

0.276 0.404

Frequency (Hz)
0.274
Frequency (Hz)

0.398

0.272 0.392

0.270 0.386

0.268 0.380
1E _3 1E _ 1 1E+1 1E+3 1E+5 1E+7 1E+9 1E _ 3 1E _ 1 1E+1 1E+3 1E+5 1E+7 1E+9
(a) Spring stiffness (kN/m) (b) Spring stiffness (kN/m)

0.610 0.805
Frequency (Hz)

Frequency (Hz)

0.600 0.795

0.590 0.785

0.580 0.775

0.570 0.765
1E _ 3 1E _1 1E+1 1E+3 1E+5 1E+7 1E+9 1E _ 3 1E _1 1E+1 1E+3 1E+5 1E+7 1E+9
(c) Spring stiffness (kN/m) (d) Spring stiffness (kN/m)

1.010 1.210
Frequency (Hz)

Frequency (Hz)

0.995 1.195

0.980 1.180

0.965 1.165

0.950 1.150
1E _ 3 1E _1 1E+1 1E+3 1E+5 1E+7 1E+9 1E _ 3 1E_ 1 1E+1 1E+3 1E+5 1E+7 1E+9
(e) Spring stiffness (kN/m) (f) Spring stiffness (kN/m)

Figure 14. Relation diagrams of natural frequency versus spring (damper) stiffness for Ting Kau bridge
stabilizing cable: (a) first mode; (b) second mode; (c) third mode; (d) fourth mode; (e) fifth mode; (f) sixth mode.

stiffness is, respectively, 222, 435, 428, 435, 434 and 435% for the six modes. This
means that if each natural frequency of the six modes is used to evaluate tension force
from the taut string equation, the maximum identification error of tension force stemming
from the damper stiffness effect will be 444, 870, 856, 870, 868 and 870% respectively.
318 Y. Q. NI ET AL.

This also demonstrates the importance of a versatile finite element model for tension force
evaluation of cables connected with dampers. In addition, it is possible and feasible to
simultaneously identify the cable tension force and damper stiffness by use of a precise
finite element model and measured multimode frequencies.

5. CONCLUSIONS

In this study a three-dimensional finite element formulation is developed for dynamic

analysis of large-diameter structural cables. Only some basic assumptions have been
introduced in the finite element modelling and several constraints posed in the
conventional cable dynamic analysis have been relaxed in the present study. The proposed
formulation is suited for both suspended and inclined cables with arbitrary static profiles,
and allows for the consideration of cable flexural rigidity, sag–extensibility, spatial
variability of dynamic tension, boundary conditions, lumped masses and intermediate
springs and/or dampers. This formulation provides a good baseline model for accurate
identification of cable tension force and other structural parameters based on the
measurement of multimode frequencies.
Parametric studies are made to evaluate the effect of cable bending stiffness and sag–
extensibility on modal properties, and the relation between the natural frequencies and
cable parameters for a wide parameter range. The results show that the cable bending
stiffness contributes a considerable influence on the natural frequencies when the tension
force is relatively small, and affects the higher-mode frequencies more significantly than
the lower-mode frequencies. A comparison study of the computed and measured natural
frequencies of the Tsing Ma bridge cables shows that taking into account bending stiffness
is necessary for large-diameter bridge cables to obtain an accurate prediction of the
natural frequencies. The predicted higher-mode frequencies for such cables without
considering bending stiffness may significantly deviate from the true values. The case study
of the Ting Kau bridge cables demonstrates the effect degree of the stiffness of attached
dampers on the cable modal properties and on the tension identification accuracy. It is
concluded that the tension forces of long-span large-diameter bridge cables can be
accurately evaluated from vibration measurement only when a precise model accounting
for cable bending stiffness, sag–extensibility and other constraints is utilized in the
identification procedure.

ACKNOWLEDGMENTS

The work described in this paper was supported in part by a grant from the Research
Grants Council of the Hong Kong Special Administrative Region, China (Project No.
PolyU 5045/00E) and partially by a grant from the Hong Kong Polytechnic University
(Project No. G-V785).

REFERENCES

1. A. S. Beard 1995 Structural Engineering International 5, 138–140. Tsing Ma Bridge, Hong

Kong.
2. S. Saeki, Y. Fujino, K. Tada, M. Kitagawa and T. Kanazaki 1995 in Restructuring: America
and Beyond (M. Sanayei, editor), 237–252. New York: American Society of Civil Engineers.
Technological aspects of the Akashi Kaikyo Bridge.
 DYANAMIC ANALYSIS OF SAGGED CABLES 319
3. K. K. Ng, H. C. Wong, C. H. Hui and K. P.Wong 2000 in Civil Engineering in the 21st Century
(J. Song and G. Zhou, editors), 596–603. Beijing: Science Press. Stonecutters Bridge design
competition.
4. M. C. H. Hui 2001 in Proceedings of the 10th National Symposium on Structural Wind
Engineering, Guilin, China, 84–93. Stonecutters Bridge}deck section optimization.
5. M. Ito and T. Endo 1994 in Structures Congress XII (N. C. Baker and B. J. Goodno, editors),
677–682. New York: American Society of Civil Engineers. The Tatara Bridge}world’s longest
cable–stayed span.
6. M.-C. Tang 1995 in Restructuring: America and Beyond (M. Sanayei, editor), 455–458. New
York: American Society of Civil Engineers. Multispan cable-stayed bridges.
7. M. Virlogeux 2001 Structural Engineering International 11, 61–82. Bridges with multiple cable-
stayed spans.
8. R. Bergermann and M. Schlaich 1996 Structural Engineering International 6, 152–154. Ting
Kau Bridge, Hong Kong.
9. K. J. Kroneberger-Stanton and B. R. Hartsough 1992 Transactions of the ASAE 35,
341–346. A monitor for indirect measurement of cable vibration frequency and tension.
10. J. R. Casas 1994 Structural Engineering International 4, 235–240. A combined method for
measuring cable forces: the cable-stayed Alamillo Bridge, Spain.
11. J. C. Russell and T. J. Lardner 1998 American Society of Civil Engineers Journal of
Engineering Mechanics 124, 1067–1072. Experimental determination of frequencies and tension
for elastic cables.
12. A. Cunha and E. Caetano 1999 Experimental Techniques 23, 38–43. Dynamic measurements
on stay cables of cable-stayed bridges using an interferometry laser system.
13. H. Zui, T. Shinke and Y. Namita 1996 American Society of Civil Engineers Journal of Structural
Engineering 122, 651–656. Practical formulas for estimation of cable tension by vibration
method.
14. W.-H. P. Yen, A. B. Mehrabi and H. Tabatabai 1997 in Building to Last (L. Kempner Jr and
C. B. Brown, editors), 503–507. New York: American Society of Civil Engineers. Evaluation of
stay cable tension using a non-destructive vibration technique.
15. A. B. Mehrabi and H. Tabatabai 1998 American Society of Civil Engineers Journal of
Structural Engineering 124, 1313–1322. Unified finite difference formulation for free vibration of
cables.
16. G. Zheng, J. M. Ko and Y. Q. Ni 2001 in Smart Structures and Materials 2001: Smart Systems
for Bridges, Structures, and Highways (S. C. Liu, editor), Proceedings of SPIE, Vol. 4330,
511–522. Bellingham: The International Society for Optical Engineering. Multimode-based
evaluation of cable tension force in cable-supported bridges.
17. M. S. Triantafyllou 1984 Shock and Vibration Digest 16, 9–17. Linear dynamics of cables and
chains.
18. H. M. Irvine 1981 Cable Structures. Cambridge: The MIT Press.
19. U. Starossek 1994 Structural Engineering International 4, 171–176. Cable dynamics}a review.
20. J. M. Ko, G. Zheng and Y. Q. Ni 2000 in Proceedings of the International Conference on
Advanced Problems in Vibration Theory and Applications, Xi’an, China, 437–443. Periodically
forced vibration of nonlinear stay cables.
21. J. M. Ko, Y. Q. Ni and J. Y. Wang 2000 in Proceedings of the International Conference on
Advanced Problems in Vibration Theory and Applications, Xi’an, China, 285–291. Tsing Ma
suspension bridge: ambient vibration survey campaigns.
22. C. K. Lau, W. P. N. Mak, K. Y. Wong, W. Y. K. Chan and K. L. D. Man 1999 in Structural
Health Monitoring 2000 (F.-K. Chang, editor), 450–460. Lancaster: Technomic Publishing
Corporation. Structural health monitoring of three cable-supported bridges in Hong Kong.
23. R. W. Poston 1998 American Society of Civil Engineers Civil Engineering 68, 58–61. Cable-stay
conundrum.