You are on page 1of 19

A parabolic cable element for

static analysis of cable structures


Wei-Xin Ren
Department of Civil Engineering, Central South University,
Changsha, Peoples Republic of China
Meng-Gang Huang
School of Civil & Environmental Engineering, Yonsei University,
Seoul, South Korea, and
Wei-Hua Hu
Department of Civil Engineering, University of Porto, Porto Codex, Portugal
Abstract
Purpose The purpose of this paper is to present a nite element formulation of enhanced two-node
parabolic cable element for the static analysis of cable structures.
Design/methodology/approach Unlike the assumed polynomial displacement interpolation
functions, the present approach uses the analytical cable dynamic stiffness matrix to obtain the
explicit expression of the static stiffness matrix of an inclined sagging cable by setting the frequency
at zero. The Newton-Raphson-based iterative method is used to obtain the solution.
Findings It is demonstrated that the present results agree well with those obtained from the
nonlinear analytical theory of a parabolic cable and previous reported methods in the literature.
Originality/value This paper proposes a two-node parabolic cable element. For comparable
accuracy with the truss element method, fewer numbers of such cable elements are needed.
Keywords Finite element analysis, Newton method, Physical properties of materials,
Strength of materials, Structural analysis, Design and theory
Paper type Research paper
1. Introduction
Cables are normally held in position by tension along their length and end forces at
supports. In recent decades, cables have been very popular structural members used in
tension structures such as cable-supported bridges, cable roofs, and guyed
towers/masts, etc. due to aesthetic appearance, efcient utilization of structural
materials and other notable advantages (Leonard, 1988). As the sagged cable structures
undergo a large displacement before attaining their equilibrium conguration under
pre-tensions, geometric nonlinearity should be taken into account in the analysis. The
static analysis of heavy parabolic cables under various types of applied loading dates
back to the middle of the nineteenth century and incases not amenable to a hand analysis
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/0264-4401.htm
This research is sponsored by the National Natural Science Foundation of China under Research
Grants No. 50378021 and 50678173. The rst author would like to acknowledge the nancial
support from the Program for New Century Excellent Talents (NCET-04-0612) in University,
Ministry of Education, Peoples Republic of China.
EC
25,4
366
Received 6 September 2007
Accepted 21 January 2008
Engineering Computations:
International Journal for
Computer-Aided Engineering and
Software
Vol. 25 No. 4, 2008
pp. 366-384
qEmerald Group Publishing Limited
0264-4401
DOI 10.1108/02644400810874967
even by the early 1890s. There exists an extensive literature on numerical methods for
analyzing various cable problems.
Ernst (1965) rst proposed a cable equivalent modulus of elasticity to approximately
account for the cable sag effect. In such a way, the cable sag decreases and the apparent
axial stiffness of the cable increases when the cable tension increases. Owing to its
simplicity, this approach has been adopted by several investigators (Chu and Ma, 1976;
Bruno and Grimaldi, 1985; Nazmy and Abdel-Ghaffar, 1990a, b), and has been proved to
be sufcient for some cases, e.g. for a short span cable-stayed bridge. Namely, when a
cable has relatively high stress and small length, the Ernst equivalent modulus
approach could achieve a good result. As shown in Ali and Abdel-Ghaffar (1995) and
Karoumi (1999), however, this approach results in a softer cable response not only
because it is only applicable for incremental method in which both drift-off error and
accumulative error are inevitable but also because while it takes account of the sag
effect, it does not include the stiffening effect due to large displacements.
In the nite element analysis of cable structures, most of nite element packages are still
lack of suitable cable elements. A sagging cable is often discretized by a series of straight
elements (common truss or beam elements) to model the curved geometry of cables
(Henghold and Russel, 1976; Abdel-Ghaffar and Khalifa, 1991; Ren et al., 2004; Ren and
Peng, 2005) and conventional displacement-based nite element procedure is implemented.
In such a case, the geometric nonlinearitydue to cable pre-tension may be included with the
help of the initial-stress stiffness matrix. However, the actual behavior of pre-tensioned
cables cannot be fully represented and the iteration procedure is sometimes divergent if the
stress stiffening is not properly dealt with. To better model the curved cables, several
straight cable elements were proposed (Ali and Abdel-Ghaffar, 1995; Gambhir and
Batchelor, 1977; Ni et al., 2002; Tang et al., 1997). Most of multi-node cable elements are
isoparametric and a higher-order displacement interpolation function is commonly used.
As a result, a better accuracy can be achieved to model a curved cable as straight elements.
The more natural and accurate curved cable element is to use an elastic catenary
element since the equilibriumcongurationof a hanging cable is a catenary innature. The
formulation of elastic catenary element was based on the exact analytical solutions for the
elastic catenary which was developed by OBrien (OBrien and Francis, 1964; OBrien,
1979). The explicit expression of catenary cable stiffness matrix was then developed by
Jayaraman and Knudson (1981). Several investigators have used the catenary cable
elements for engineering analysis (Peyrot and Goulois, 1979; Karoumi, 1996; Abbas and
Scordelis, 1994; Kim et al., 2002). However, the iteration procedure, as carried out in
Jayaraman and Knudson (1981), has to use the complicated catenary analytical solution
again for the relationships between the nodal forces and displacements, which defeats
somewhat the computational advantage of the numerical nite element method although
it can achieve an adequate accuracy.
The objective of this study is to present a nite element formulation of enhanced
two-node parabolic cable element for the static analysis of cable structures. The prole of a
free-hanging, uniform cable is assumed to be a quadratic rather than a catenary, which
requires that the ratio of sag to span be kept 1:8, or less. Usually, the uniformcables whose
geometry does not satisfy the above assumption are rarely used as the real structural
members supporting transverse loading. Unlike the assumed polynomial displacement
interpolation functions that are implemented in conventional displacement-based nite
element methods, the approach adopted in the present paper, hinges on the work reported
Parabolic cable
element
367
in Irvine (1981), and uses the analytical cable dynamic stiffness matrix (Veletsos and
Darbre, 1983; Kim and Chang, 2001; Starossek, 1991) to derive the explicit expression of
the static stiffness matrix, while setting the frequency in the cable dynamic stiffness
be zero. Associated with a Newton-Raphson-based iterative solution procedure, the
applicability, reliability and efciency of the proposed cable element formulation for
the static analysis of sagging cables have been demonstrated through a number of
numerical examples. For the comparable accuracy to the truss element method, fewer
number of such cable elements are needed.
2. Stiffness matrix of a parabolic cable element
2.1 Basic assumptions
The cable is assumedto be of uniformcross sectionandis made froma material of uniform
density which obeys Hookes elastic law. The exural rigidity of the cable is ignored so
that the cable can resist applied load only by tensions. Expansions and contractions of the
cross section, associatedwithchanges inthe lengthof the cable andthe effects of Poissons
ratio, are negligible. The prole of a free-hanging, uniform cable is assumed to be
a quadratic, which requires that the ratio of the sag to span be kept relatively small
(1:8, or less). In this case, the weight component parallel to the cable chord is neglected.
2.2 Cable dynamic stiffness
The cable dynamic stiffness functions are dened as time-independent relations between
boundary forces and boundary displacements of the cable as a part of a vibrating system.
Several investigators have derived the explicit expressions of the dynamic stiffness of a
free-hanging cable with different assumptions and approximations (Irvine, 1974; Irvine,
1981; Veletsos and Darbre, 1983; Kim and Chang, 2001; Starossek, 1991). In this study,
Starosseks dynamic stiffness matrix of an inclined sagging cable is adopted. To be
consistent with the derivation of explicit expression of the static stiffness matrix and to
discuss the physical meaning of the proposed two-node parabolic cable element, the key
issues of the cable dynamic stiffness matrix are briey addressed and some of parts are
rewrittentodemonstrate the contributionof eachcomponent of closed-formcable dynamic
stiffness.
Consider a uniform cable, which anchored on supports at the same level in static
equilibrium in a vertical plane as shown in Figure 1, the cable prole can be
represented by a parabola:
y
mgl
2
2H
x
l
2
x
l
_ _
2
_ _
1
where m is the cable mass per unit length and g is the gravitational acceleration; l is the
length of the chord; and H is the horizontal component of the cable force, which is
independent of x.
Figure 1.
Description of cable
vibration
d
l / 2
P(x,y,0)
w
u
v
P

(x + u, y + v, w)
EC
25,4
368
A sagging cable subjected to a slight disturbance is as shown in Figure 1. Considering
the equilibrium of an in-plane differential cable element as shown in Figure 2, the
equation of motion can be obtained as follow (Irvine, 1981):
H

2
v
x
2
h
d
2
y
dx
2
m

2
v
t
2
2
where the auxiliary quantity h is the dynamic increment in H which is independent of
the position co-ordinate, x. This is the consequence of the assumption that only the
normal components of the inertia force is of importance.
The elastic and geometric compatibility equation of the cable element is:
h
EA
ds
dx
_ _
3

u
x

dy
dx
v
x
3
where E represents Youngs modulus of elasticity for the cable, A represents its
cross-sectional area and ds is the original length of cable element. Equations (2) and (3) are
the fundamental equations controlling the behavior of in-plane free vibration of the cable.
The dynamic part of vertical component of total cable tension n, as dened in
Veletsos and Darbre (1983), can be derived as:
n h
dy
dx
H
v
x
4
By separating the variables, the cable vibration can be described by the products:
ux; t ~ uxe
ivt
5a
vx; t ~ vxe
ivt
5b
hx; t
~
hxe
ivt
5c
Substituting equations (5a)-(5c) into equations (2) and (3), the following equations can
be derived, respectively:
H

2
~ v
x
2
v
2
m~ v
8d
l
2
~
h 6
~
h
EA
ds
dx
_ _
3

~ u
x

dy
dx
~ v
x
7
where v is the natural frequency of the cable and d is the sag of the cable as shown in
Figure 1.
Figure 2.
In-plane differential cable
element
(a) At rest
V
H
V + dV
H
mgdx
V + n
H + h
V + dV + n + dn
H + h
mgdx
(b) In motion
Parabolic cable
element
369
To obtain the distinct dynamic stiffness matrix with denite physical explanations, the
derivation now proceeds separately for the applied horizontal and vertical boundary
dynamic displacements so that the symmetric and antisymmetric contributions can be
decomposed (Starossek, 1991).
2.2.1 Cable is subjected to horizontal boundary dynamic displacement u. When a
horizontal boundary dynamic displacement u is applied to the cable, as shown in Figure 3,
the vibrationcanbe dividedintotwo components: the symmetric andantisymmetric parts.
The dynamic stiffness coefcients of the symmetric components are dened as:
k
s
11

h
s
u

~
h
s
~ u
8a
k
s
21

n
s
u

~ n
s
~ u
8b
where
~
h
s
is the symmetric dynamic part of horizontal component of total cable tension.
Using equations (6) and (7), the expression of
~
h
s
can be obtained as:
~
h
s

EA
L
e
1
1 l
2
=v
2
b 2 1
~ u 9
in which L
e
is the effective length of the cable calculated by:
L
e

_
l
0
ds
dx
_ _
3
dx < l 1
1
8
q
y
l
H
_ _
2
_ _
l
2
is the characteristic cable parameter represented by:
l
2

mgl
H
_ _
2
EAl
HL
e
10
and b is a auxiliary parameter, dened as:
b
tanv=2
v=2
11
Then the dynamic stiffness coefcient k
s
11
(equation (8a)) can be obtained:
k
s
11

EA
L
e
1
1 l
2
=v
2
b 2 1
12
Similarly, from equation (4), the symmetric dynamic part of vertical component of total
cable tension ~ n
s
results in:
~ n
s

1
2
b 2 1
mgl
H
~
h
s
13
Figure 3.
Decomposition of
horizontal boundary
displacement u
(a) Symmetric component (b) Antisymmetric component
h
s
n
s
h
s
n
s
h
a
n
a
n
a
h
a
u f
1
u
f
3
u
f
4
u
f
2
u
2 u 2 u 2 u 2 u
EC
25,4
370
Subsequently, the corresponding dynamic stiffness coefcient can be obtained from
equations (8b), (9) and (13):
k
s
21

EA
L
e
1=2mgl=Hb 2 1
1 l
2
=v
2
b 2 1
14
Taking the antisymmetric part, as shown in Figure 3(b), into account, the dynamic
stiffness coefcients are now dened as:
k
a
11

h
a
u

~
h
a
~ u
15a
k
a
21

n
a
u

~ n
a
~ u
15b
When the boundary conditions of governing equation (6) are taken into account,
presupposing the existence of an antisymmetric solution for ~ vx and integrating
equation (7) result in a vanishing
~
h
a
. Hence, when a horizontal boundary dynamic
displacement u is applied to the cable, the antisymmetric dynamic stiffness coefcients
become:
k
a
11
0 16
k
a
21
0 17
2.2.2 Cable is subjected to vertical boundary dynamic displacement n. When a vertical
boundary dynamic displacement n is applied to the cable, the vibration contribution can be
also separated into the symmetric and antisymmetric parts, respectively, as shown in
Figure 4. The correspondingdynamic stiffness coefcients of symmetric part are denedas:
k
s
12

h
s
v

~
h
s
~ v
18a
k
s
22

n
s
v

~ n
s
~ v
18b
where
~
h
s
is the symmetric dynamic part of horizontal component of total cable tensiondue to
applied vertical boundary dynamic displacement n. Using equations (6) and (7), the
expression
~
h
s
can be derived as:
~
h
s

EA
L
e
1=2mgl=Hb 2 1
1 l
2
=v
2
b 2 1
~ v 19
Then the dynamic stiffness coefcient k
s
12
(equation (18a)) can be obtained as:
k
s
12

EA
L
e
1=2mgl=Hb 2 1
1 l
2
=v
2
b 2 1
20
which is exactly the same as k
s
21
(equation (14)).
Figure 4.
Decomposition of vertical
boundary displacement v
(a) Symmetric component (b) Antisymmetric component
f
4
v
f
1
v
v/2
v
f
3
v
f
2
v
h
s
n
s
h
s
n
s
h
a
n
a
n
a
h
a
v/2 v/2
v/2
Parabolic cable
element
371
Similarly, from equation (4), the symmetric dynamic part of vertical component of total
cable tension ~ n
s
due to applied vertical boundary dynamic displacement n results in:
~ n
s

1
2
b 2 1
mgl
H
~
h
s
2
H
4l
v
2
b~ v 21
Subsequently, the corresponding dynamic stiffness coefcient can be obtained from
equations (18b), (19) and (21):
k
s
22
2
EA
4L
e
mgl=H
2
v
2
=l
2
b l
2
=v
2
b 2 1
_
1 l
2
=v
2
b 2 1
22
For the antisymmetric part due to applied vertical boundary dynamic displacement n,
the corresponding dynamic stiffness coefcients of antisymmetric part as shown in
Figure 4(b) are dened as:
k
a
12

h
a
v

~
h
a
~ v
23a
k
a
22

n
a
v

n
a
~ v
23b
They can be derived in the same way as follows:
k
a
12
0 24
k
a
22

EA
L
e
1
l
2
b
mgl
H
_ _
2
25
2.2.3 Local dynamic stiffness matrix of a cable. Superposing the symmetric and
antisymmetric contributions due to both applied horizontal and vertical boundary
displacements, the dynamic stiffness coefcients of the sagging cable can be obtained
as follows.
For the vibration induced by horizontal boundary displacement:
k
d
1
k
s
11
k
a
11
26a
k
d
2
k
s
21
k
a
21
26b
For the vibration induced by vertical boundary displacement:
k
d
3
k
s
12
k
a
12
27a
k
d
4
k
s
22
k
a
22
27b
Consider a two-node parabolic cable element in local coordinates as shown in Figure 5.
The local dynamic stiffness matrix [k
d
] is then dened by:
{f
d
} k
d
{u
d
} 28
where the dynamic nodal force and dynamic nodal displacement vectors in local
coordinates are:
EC
25,4
372
{f
d
}
f
1
f
2
f
3
f
4
_

_
_

_
; {u
d
}
u
1
u
2
u
3
u
4
_

_
_

_
29
and local dynamic stiffness matrix is represented by:
k
d

k
d
1
k
d
3
2k
d
1
2k
d
3
k
d
2
k
d
4
2k
d
3
2k
d
4
2k
d
1
2k
d
2
k
d
1
k
d
3
2k
d
2
2k
d
4
k
d
2
k
d
4
_

_
_

_
30
It is clearly shown that the local dynamic stiffness matrix [k
d
] of the cable is a
symmetric matrix since k
d
2
k
d
3
.
2.3 Local static stiffness matrix of a cable
By carrying out the limit operation letting the frequency v trend to zero in the above
local dynamic stiffness coefcients, the corresponding static stiffness coefcients of a
sagging cable can be obtained:
k
s
1
k
d
1

v!0

EA
l
1
1 1=12l
2
31a
k
s
2
k
d
2

v!0
0 31b
k
s
3
k
d
3

v!0
0 31c
k
s
4
k
d
4

v!0

H
l
31d
So the corresponding local static stiffness matrix [k
s
] of a parabolic cable element can
be represented by a symmetric matrix:
Figure 5.
Parabolic cable element in
local coordinates
x
y
f
4
f
3
f
2
f
1
u
4
u
3
u
2
u
1
Parabolic cable
element
373
k
s

k
s
1
sym:
0 k
s
4
2k
s
1
0 k
s
1
0 2k
s
4
0 k
s
4
_

_
_

_
32
It is demonstrated that the horizontal (longitudinal) static problemis mainly governed by
the Ernst model, while the vertical (transversal) model differs from a more common
pre-stressed beamelement by a different coefcient in the k
s
4
term. The introduced termis
expected to produce better convergent performance in the static analysis of sagging cable.
3. Formulation of a parabolic cable element for static analysis
3.1 Formulation of a parabolic cable element
If the static stiffness matrix of a parabolic cable element in local coordinates as shown
in Figure 5 is [k
s
], the static equilibrium of the element results in:
{f} k
s
{u} 33
in which:
{f}
f
1
f
2
f
3
f
4
_

_
_

_
and {u}
u
1
u
2
u
3
u
4
_

_
_

_
34
are the vectors of the nodal static force and static displacement in local coordinates,
respectively. According to the previous assumptions of a parabolic cable, the weight
component parallel to the cable chord is neglected, so the nodal force vector of the cable
element in local coordinates subjected to its own weight has the form of:
{ f
0
}
0
mgl
2
0
mgl
2
_ _
T
Consider a two-node parabolic cable element in global coordinates as shown in Figure 6
where the nodal static force and static displacement vectors are represented by:
Figure 6.
Parabolic cable element in
global coordinates
F
2
F
1
U
2
U
1
F
4
F
3
U
4
U
3
Y
X
a
EC
25,4
374
{F}
F
1
F
2
F
3
F
4
_

_
_

_
; {U}
U
1
U
2
U
3
U
4
_

_
_

_
35
The relationship between local and global coordinates is associated with the coordinate
transformation matrix:
T
t 0
0 t
_ _
36
where:
t
cos a sin a
2sin a cos a
_ _
37
After all discrete cable elements are properly assembled together in the global
coordinates to form the global cable stiffness matrix, global nodal displacement vector
and global nodal load vector, the static analysis of cable structures can then be
performed by nding the solutions of the global parabolic cable element formulation.
3.2 Iteration procedures
As the derived analytical static cable stiffness matrix k
s
is cable tension and
deformation-dependent, an iterative procedure is required in the solution where the
element stiffness matrix should be updated in each iteration. According to equations
(31a)-(31d), the cable element tension increment dH caused by the horizontal elongation
dl of the element in the local coordinates can be calculated from:
dH
EA
l
1
1 l
2
=12
dl 38
This relationship is herein used to update the cable element stiffness matrix in the
incremental solution procedure. To realize that, a Newton-Raphson-based iterative
procedure is presented as shown in Figure 7. The owchart of solving steps for the
static analysis of proposed parabolic cable element formulation is shown in Figure 8.
4. Numerical examples
To demonstrate the applicability and accuracy of the proposed parabolic element, four
numerical examples are presented and the results are compared with the analytical
results obtained from the nonlinear theory of the parabolic cable and those results
reported in previous literatures.
4.1 Taut at cables
Irvine (1974) studied and tested the static responses of two types of taut at cables
subjected to a concentrate load. Two taut cables were anchored on supports at the
same level having a clear span of 91.5 cm as shown in Figure 9. One type of cable is
Parabolic cable
element
375
consisted of multiple twisted strands, whose properties are w
D
0:0553 N=m,
E 104 10
6
kN=m
2
, A 0:0113 1 cm
2
. Another type of cable is a piano wire, whose
properties are w
D
7:63 10
23
N=m, E 207 10
6
kN=m
2
, A 5:576 10
24
cm
2
.
Increments in load of 4.45 N were applied from 4.45 to 22.25 N at mid-span and the
corresponding deections were measured experimentally and calculated analytically.
A at, taut cable was chosen since its theory is closely related to the general theory of
parabolic cable and the obtained results can be well identied.
To check the formulation of proposed parabolic cable element, the deections of the
loaded point (mid-span) and the cable horizontal tensions are calculated where both of
the cables are modeled by two parabolic elements. Tables I and II compare the current
results with Irvines experimental results and the results fromboth nonlinear and linear
theories for the piano wire and the twisted strand cable, respectively. It can be clearly
seen that the results obtained fromcurrent cable element agree well with those fromthe
experiments and the nonlinear theory of parabolic cable. The linear theory does not
allowfor additional cable tension, nor stiffness fromthat tension, and therefore predicts
the higher deections. It is demonstrated that the proposed parabolic cable element is
efcient in modeling the cable geometric nonlinearity since only two elements are used.
4.2 Behavior of a a ying fox
This ying fox case study is concerned about the behavior of a hanging cable
subjected to a moving point load. The cable spans 100 m in its free-hanging position, as
shown in Figure 10, and is used to transport materials across a ravine. The problem is
often to nd the additional horizontal component of cable tension and the associated
deection under carrying load.
Two different ratios of cable sag to span, 1:50 and 1:10, are studied, respectively,
when a ying fox load of 20 kN is carried from one end to the mid-span. In the
calculation of proposed parabolic cable element, considering the symmetry, the cable
Figure 7.
Newton-Raphson iteration
Equilibrium Path
U
F
d F
1
1
d F
1
2
d F
1
0
dU
1
2
dU
1
0
dU
1
1
U
1
1
U
1
2
U
1
3
F
1
F
2
F
1
1
F
1
2
F
2
K
s0
1
K
s1
1
K
s2
1
U
1
U
2
F
1
EC
25,4
376
was again modeled with only two parabolic elements when the value of sag is d 2 m,
while four parabolic elements were used when the value of sag is d 10 m. The load at
each position was divided into eight incremental load steps resulting in increments in
load of 2.5 N.
Figure 8.
Solution owchart for the
static analysis of cable
structures
Compute the cable tension increment of the element (Eq.38)
1
1+ l
2
12
EA
j
i
dH dl
l
Compute the initial cable stiffness matrix [K
s
ini
]
Set j 0
1
j j
i i
dU dF
j+1 sj+1 j+1
i i
F U
j+1
i
dF F
i
j+1
i
dF
EPS
No
Output final results
Yes
Input: w
D
, E , A, l , EPS, initial cable tension H
ini
, global
coordinates, external load and load increment number n
i n
No
End
Begin
Yes
j j + 1 i i + 1
0, K
s0
1
K
s
ini
, H
1
0
H
ini
Set i 1 ,

1
]

1
]

1
]
Update cable stiffness matrix
sj+1
i
K

1
]

1
]

'

;
)

'

;
)
j +1
i
U

'

;
)

'

;
)

'

;
)

'

;
)

'

;
)

'

;
)
F
i

'

;
)
j
i
dU

'

;
)
U
1
0
K
sj
i

'

;
)
U
j
i
j
i
j+1 j
i i
H H dH +
j
i
K
i


i
F

'

;
)
j+1
Parabolic cable
element
377
The calculated cable deection responses with different sags when a ying fox is
moved fromone end to the mid-span are shown in Figure 11, which achieves an excellent
agreement with the analytical predictions obtained from the nonlinear theory of
parabolic cable (Irvine, 1981). It can be observed that the response of the cable changes
Figure 9.
A taut at cable
91.5cm
P
Loads
Test results
(Irvine, 1974)
Nonlinear theory
(Irvine, 1974) Present results Linear theory
P (N) N H h N H h N H h n H
0 0 116 0 116 0 116.0 0 116
4.45 0.88 116 0.88 116 0.85 119.5 0.88 116
8.90 1.59 127 1.61 126 1.59 128.3 1.76 116
13.35 2.19 140 2.19 139 2.19 139.5 2.64 116
17.80 2.66 153 2.70 152 2.69 151.4 3.52 116
22.25 3.10 164 3.10 164 3.11 163.5 4.40 116
Note: Figure 9
Table I.
Comparison of deections
n (cm) and cable tensions
H h (N) for the piano
wire
Loads
Test results
(Irvine, 1974)
Nonlinear theory
(Irvine, 1974) Present results Linear theory
P (N) N H h n H h N H h n H
0 0 182 0 182 0 182.0 0 182
4.45 0.56 182 0.56 182 0.55 186.3 0.56 182
8.90 1.03 197 1.03 197 1.03 197.4 1.12 182
13.35 1.43 214 1.44 212 1.43 212.1 1.68 182
17.80 1.78 229 1.79 227 1.78 228.2 2.24 182
22.25 2.06 247 2.07 244 2.08 244.8 2.80 182
Note: Figure 9
Table II.
Comparison of deections
n (cm) and cable tensions
H h (N) for the strand
cable
Figure 10.
A ying fox
20 kN
100 m
Cable data: mg = 40 N / m, E = 150,000 MPa, A = 5 10
4
m
2
x
y
EC
25,4
378
from different sag. When the cable sag is relatively small, for instance d 2 m, the
deection reaches a maximum when the ying fox is hanged at mid-span, which is
almost the same as a beam. However, when the cable sag is relatively large, for instance
d 10 m, in most of cases, the cable deections are larger than the value when the ying
fox is hanged at the cable span center. That is because the catenary action of the cable is
effective when the cable sag is relatively large. With a limited number of proposed
parabolic cable elements, the static behavior of sagging cable can be well described.
4.3 Prestressed cable under uniformly distributed load
This example is presented to demonstrate the applicability of the proposed parabolic
cable element to the prestressed cable under uniformly distributed dead load.
An initially stressed cable subjected to uniform vertical external load w in addition
to the self-weight w
D
is shown in Figure 12 where the cable properties are given.
Figure 11.
Deection responses of the
ying fox cable
(see Figure 10) with
different sags
0 10 20 30 40 50
0
0.5
1
1.5
2
2.5
3
3.5
4
Analytical results [24]
Present results
d=10
d=2
D
e
f
l
e
c
t
i
o
n

a
t

t
h
e

p
o
i
n
t

o
f

t
h
e

f
o
x

(
m
)
Position of the flying fox
(m)
Figure 12.
Prestressed cable under
uniformly distributed load
A = 4.19 10
5
m
2
(0.065 in.
2
)
E = 138 10
6
kN/m
2
(20 10
6
lb/in.
2
)
w
D
=3.5 10
3
kN/m (0.02 lb/in.)
Initial stress = 138 10
3
kN/m
2
(20 10
3
lb/in.
2
)
254 m (10,000 in.)
w

Y
X
Parabolic cable
element
379
The horizontal line joining the supports has been used as the starting geometry for all
the idealizations considered here. For this problem, symmetry is used for discretization.
The calculated vertical displacements at the center of span are compared in Table III
with the results given by Jayaraman and Knudson (1981) as well as Tang et al. (1997).
As Jayaraman and Knudsons results were originally represented in the British units,
in addition to the use of System International (SI) units in Table III, the results in the
British units are also given in parentheses after the SI unit.
Figure 13 shows the variation of the central deections of the cable under increasing
uniformly distributed load applied statically where the stiffening of the system can be
clearly observed. The use of the proposed parabolic cable element is here again
advantageous as it uses only two elements to reach an excellent agreement with the
analytical results obtained from the nonlinear theory of parabolic cable (Irvine, 1981).
Displacements at mid-span (m)
Uniform load w (kN/m) Results
a
Results
b
Present results
3.5 10
23
(0.02 lb/in.) 23.34 (2131.63 in.) 23.34 (2131.60 in.) 23.34 (2131.60 in.)
10.5 10
23
(0.06 lb/in.) 25.95 (2234.19 in.) 25.87 (2231.00 in.) 25.96 (2234.49 in.)
17.5 10
23
(0.10 lb/in.) 27.44 (2292.79 in.) 27.32 (2288.00 in.) 27.44 (2293.10 in.)
24.5 10
23
(0.14 lb/in.) 28.54 (2336.03 in.) 28.41 (2331.00 in.) 28.54 (2336.22 in.)
31.5 10
23
(0.18 m lb/in.) 29.43 (2371.13 in.) 29.35 (2368.00 in.) 29.43 (2371.37 in.)
Note: Figure 12
Source:
a
Jayaraman and Knudson (1981);
b
Tang et al. (1997)
Table III.
Vertical displacement at
mid-span
Figure 13.
Variation of vertical
displacement at midspan
under uniformly
distributed load

2 3 4 5 6 7 8 9 10
0
5
10
15
20
25
30
35
Displacement (m)
Initial equilibrium under w
D
Analytical results [24]
Present results
U
n
i
f
o
r
m

L
o
a
d

w

(
1
0

3
k
N
/
m
)
EC
25,4
380
The speed of convergence for the parabolic cable element is shown in Figure 14 with
the truss element-Link10 element provided in the ANSYS
w
(1999) commercial nite
element analysis package. The Link10 element is a tension-only truss element where
the stiffness is removed if the element goes into compression, thus it can simulate a
slack cable. The pre-tension of the cable is incorporated by the initial strain of the
element in the form of initial stress stiffness matrix. As a result, the element is
nonlinear and requires an iterative solution. It is clearly shown in Figure 14 that the
convergence of current cable elements is very rapid.
4.4 Uniformly distributed load on part of suspended cable
This example is focused on the behavior of the suspended cables under partly uniformly
distributed load. During the construction of suspension bridges, for example, the
segments of the deck are normally to be lifted into place stage by stage starting at
mid-span. Ineachconstruction stage it is required to determine the additional deections
at mid-span and the increments in the horizontal component of cable tension. The
long-span suspended cable considered and its properties are as shown in Figure 15. The
distributed weight of the deck segment per cable is w 58.4 kN/m. Two different ratios
of sag to span of 1:12 and 1:20 are studied, respectively. The cable is modeled by four
parabolic elements and the weight of the deck is divided into 30 increments. The
backstays are not included during calculation. It is assumed that the deck has no exural
stiffness during construction.
The calculated cable deection variation at mid-span during the erection of the bridge
deck is shown in Figure 16 with the analytical results obtained fromthe nonlinear theory
of parabolic sagging cable (Irvine, 1974). The catenary action of suspended cable is
clearly observed depending on the cable sag. When the ratio of sag to span is large, for
instance ratio of sag to span of 1:12, most of the deections at mid-span will exceed the
value when all the deck segments are placed on the main-span. The accuracy of the
proposed parabolic cable element is herein demonstrated again.
Figure 14.
Speed of convergence of
the elements
0 4 8 12 16 20 24
940
945
950
955
960
Parabolic element
Number of element
D
e
f
l
e
c
t
i
o
n

a
t

m
i
d
-
s
p
a
n

(
c
m
)
Truss element
Parabolic cable
element
381
5. Conclusions
Atwo-node parabolic cable element formulation has been proposed for the in-plane static
analysis of cable structures in this paper. The analytical static stiffness matrix of the
parabolic sagging cable has been derived from the corresponding analytical dynamic
stiffness matrix by carrying out the limit operation letting the frequency be zero. As the
derived analytical static cable stiffness matrix is cable tension or deformation-dependent,
an iterative procedure is required and the Newton-Raphson-based iterative procedure is
used to obtain the solutions. A number of numerical examples have demonstrated an
excellent agreement between the present results and the analytical results obtained from
the nonlinear theory of parabolic cable as well as the results reported in previous
literatures. To reach a comparable accuracy to the truss element method, a fewer
number of proposed cable elements are needed. The proposed parabolic cable element
formulationis expectedto carryout the static analysis of more challengingtensionedcable
structures such as cable-supported bridges, cable roofs, and guyed towers/masts, etc.
Figure 15.
Construction of the deck
proceeds of a suspension
bridge
a
A = 0.161 m
2
E = 180 10
6
kN/m
2
L = 915 m
w
D
= 4.4 kN/m
L
w
Figure 16.
Cable deection variation
at mid-span during
erection of the deck
segments
0 0.2 0.4 0.6 0.8 1
0
2
4
6
8
10
12
14
D
e
f
l
e
c
t
i
o
n

a
t

m
i
d
-
s
p
a
n

(
m
)
Bridge deck erected ratio
Analytical results [23]
Present results
L
a
Span sag ratio=1:20
Span sag ratio=1:12

EC
25,4
382
References
Abbas, S. and Scordelis, A. (1994), Nonlinear analysis of cable-stayed bridges, Proceedings of
the International Conference on Cable-Stayed and Suspension Bridges, Vol. 2, Deauville,
pp. 195-210.
Abdel-Ghaffar, A.M. and Khalifa, M.A. (1991), Importance of cable vibration in dynamics of
cable-stayed bridges, J. Engng. Mech., ASCE, Vol. 117, pp. 2571-89.
Ali, H.M. and Abdel-Ghaffar, A.M. (1995), Modeling the nonlinear seismic behavior of
cable-stayed bridges with passive control bearings, Comput. Struct., Vol. 54, pp. 461-92.
ANSYS
w
(1999), Users Manual, Swanson Analysis System, Houston, PA, revision 5.6.
Bruno, D. and Grimaldi, A. (1985), Nonlinear behaviour of long-span cable-stayed bridge,
Meccanica, Vol. 20, pp. 303-13.
Chu, K.H. and Ma, C.C. (1976), Nonlinear cable and frame interaction, J. Struct. Div., ASCE,
Vol. 102, pp. 569-89.
Ernst, H.J. (1965), Der E-modul von seilen unter beruecksichtigung des durchhanges,
Der Bauingenieur, Vol. 40, pp. 52-5.
Gambhir, M.L. and Batchelor, B.A. (1977), Finite element for 3-D pre-stressed cable nets,
International Journal for Numerical Methods in Engineering, No. 11, pp. 1699-718.
Henghold, W.M. and Russel, J.J. (1976), Equilibriumand natural frequencies of cable structures,
Comput. Sturct., Vol. 6, pp. 267-71.
Irvine, H.M. (1974), Studies in the statics and dynamics of simple cables systems, thesis,
California Institute of Technology, Pasadena, CA.
Irvine, H.M. (1981), Cable Structures, MIT Press, Cambridge, MA.
Jayaraman, H.B. and Knudson, W.C. (1981), A curved element for the analysis of cable
structures, Comput. Struct., Vol. 14, pp. 325-33.
Karoumi, R. (1996), Dynamic response of cable-stayed bridges subjected to moving vehicles,
Licentiate thesis, Department of Structural Engineering, Royal Institute of Technology,
Stockholm, TRITABKN Bulletin 22.
Karoumi, R. (1999), Some modeling aspects in the nonlinear nite element analysis of cable
supported bridges, Comput. Struct., Vol. 71, pp. 397-412.
Kim, H.K., Lee, M.J. and Chang, S.P. (2002), Non-linear shape-nding analysis of a self-anchored
suspension bridge, Engineering Structures, Vol. 24, pp. 1547-59.
Kim, J.H. and Chang, S.P. (2001), Dynamic stiffness matrix of an inclined cable, Engineering
Structures, Vol. 23, pp. 1614-21.
Leonard, J.W. (1988), Tension Structures, McGraw-Hill, New York, NY.
Nazmy, A.S. and Abdel-Ghaffar, A.M. (1990a), Three dimensional nonlinear static analysis of
cable-stayed bridges, Comput. Struct., Vol. 34, pp. 257-71.
Nazmy, A.S. and Abdel-Ghaffar, A.M. (1990b), Nonlinear earthquake-response analysis of
long-span cable-stayed bridges: theory, Earthquake Engrg. and Struct. Dyn., Vol. 19 No. 1,
pp. 45-62.
Ni, Y.Q., Ko, J.M. and Zheng, G. (2002), Dynamic analysis of large-diameter sagged cables taking
in account exural rigidity, Journal of Sound and Vibration, Vol. 257 No. 2, pp. 301-19.
OBrien, T. (1979), General solution of suspended cable structures, Comput. Struct., Vol. 10,
pp. 805-13.
OBrien, T. and Francis, A.J. (1964), Cable movement under two-dimensional loads, J. Struct.
Div., ASCE, Vol. 90, pp. 89-123.
Parabolic cable
element
383
Peyrot, A.H. and Goulois, A.M. (1979), Analysis of cable structures, Comput. Struct., Vol. 10,
pp. 805-13.
Ren, W.X. and Peng, X.L. (2005), Baseline nite element modeling of a large span cable-stayed
bridge through eld ambient vibration tests, Comput. Struct., Vol. 83 Nos 8/9, pp. 536-50.
Ren, W.X., Harik, I.E. and Blandford, G.E. (2004), Roebling suspension bridge: I. FE model and
free vibration response, J. Bridge Engrg., ASCE, Vol. 9 No. 2, pp. 119-26.
Starossek, U. (1991), Dynamic stiffness matrix of sagging cable, J. Engrg. Mech., ASCE,
Vol. 117 No. 12, pp. 2815-29.
Tang, J.M., Dong, M. and Qian, R.J. (1997), A nite element method with ve-node isoparametric
element for nonlinear analysis of tension structures, Chinese J. Computational Mechanics,
Vol. 14, pp. 108-13 (in Chinese).
Veletsos, A.S. and Darbre, G.R. (1983), Dynamic stiffness of parabolic cables, Earthquake
Engrg. Struct. Dyn., Vol. 11, pp. 367-401.
Corresponding author
Wei-Xin Ren can be contacted at: renwx@mail.csu.edu.cn
EC
25,4
384
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints

You might also like