November 10, 2007 13:42 WSPC/165-IJSSD 00243

International Journal of Structural Stability and Dynamics

Vol. 7, No. 4 (2007) 571–588

c World Scientific Publishing Company
by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.

GEOMETRIC NONLINEAR ANALYSIS OF CABLE

STRUCTURES WITH A TWO-NODE CABLE ELEMENT
BY GENERALIZED DISPLACEMENT CONTROL METHOD

Y. B. YANG∗ and JIUNN-YIN TSAY

Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

Department of Civil Engineering

National Taiwan University
Taipei, Taiwan 10617
∗[email protected]

Received 26 January 2007

Accepted 21 March 2007

This paper presents a two-node catenary cable element for the analysis of three-
dimensional cable-supported structures. The stiffness matrix of the catenary cable ele-
ment was derived as the inverse of the flexibility matrix, with allowances for selfweight
and pretension effects. The element was then included, along with the beam and truss
elements, in a geometric nonlinear analysis program, for which the procedure for com-
puting the stiffness matrix and for performing iterations was clearly outlined. With the
present element, each cable with no internal joints can be modeled by a single element,
even for cables with large sags, as encountered in cable nets, suspension bridges and
long-span cable-stayed bridges. The solutions obtained for all the examples are in good
agreement with the existing ones, which indicates the accuracy and applicability of the
element presented.

Keywords: Cable; cable-stayed bridge; catenary; geometric nonlinear analysis; suspension

bridge.

1. Introduction
Cable structures with elegant, slender outlook and high efficiency of structural
behavior in simple tension have been widely used as structures for spanning a long
distance or covering a large space, such as suspension bridges, cable-stay bridges
and suspension roofs, etc. To avoid the high construction cost for infrastructures
in deep water, modern long-span bridges with less numbers of piers are proposed
to cross the rivers and straits. So far, cable-stay bridges with a main span over
1,000 m (e.g. Sutong Changjiang Highway Bridge with a main span of 1,088 m) and
suspension bridges with a main span over 2,000 m (e.g. Messina Strait Bridge with
a main span of 3,300 m) are already under construction. Meanwhile, cables with
excellent performance, i.e. with higher strength and less weight, are under study
for future super-long span bridges.

571
 November 10, 2007 13:42 WSPC/165-IJSSD 00243

572 Y. B. Yang & J.-Y. Tsay

In lieu of the highly nonlinear behavior of the cable element, the effects
of flexibility and large displacements need to be considered in establishing the
by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.

equilibrium equations of the cable. Previously, a common approach for the analysis
of cable-supported structures is to model each cable as a truss element with an
equivalent modulus of elasticity or as a series of linear truss elements. However,
due to the continuous increase in the span of cable structures, such approaches
may not be accurate enough to simulate the nonlinear behavior of cables in such
structures. Thus, more accurate approaches that take into account the catenary
profile of the long-span cables should be developed.
The historical development of the theory for catenary cables is available in the
Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

book by Irvine.1 There are two types of cable elements, i.e. the cable-stayed element
with a shallow sag and the catenary element with a deep sag; the former can be
regarded as a limiting case of the latter. The geometry of a shallow sag cable element
can be treated as a parabolic profile, although the actual profile of a deep sag cable
is the elastic catenary configuration. In general, a shallow sag cable is defined as
one with a sag to span ratio of less than 1:8, according to Irvine1 and Abbas.2 The
shallow sag cables are those that can be seen in cable-stayed bridges, guyed masts
and suspension roofs, etc. whereas the deep sag cables are those used in suspension
bridges and mooring structures.
There are two major approaches for formulating the catenary cable element,
namely, the exact analytical method and finite element method. The cable element
derived in this study is based on the exact analytical expressions for the deformed
geometry of the elastic catenary given in Refs. 1–9. Furthermore, the flexibility
approach that does not require explicit description of the shape functions will be
adopted herein to derive the stiffness matrix. This element can be used to model
cables with larger sags in suspension bridges as well as cables with smaller sags
in cable-stayed bridges. Another finite element approach is to use the Lagrangian
functions for interpolating the geometry of the cable element, which have been
adopted in formulating two-node,10,11 three-node12–14 and four-node15 cable
elements.
The objective of this paper is to develop a three-dimensional two-node cable
element considering the geometric nonlinear effects for the analysis of suspension
bridges. The generalized displacement control (GDC) method developed by Yang
and Shieh16 is adopted for solving the large-displacement nonlinear response for its
general stability and efficiency. Furthermore, the cable element together with the
beam and truss elements presented by Yang and Kuo17 will be included in the Geo-
metric Nonlinear Analysis (GNA) program in Fortran codes for three-dimensional
cable-supported structures. The entire method of solution is built around an efficient
cable element subprogram. For given loads at specified joints and given positions
of the two ends of each cable, the subprogram determines the geometry, end forces,
tangent stiffness and sag of the cable. The reliability and efficiency of the element
will be demonstrated in the study of six examples.
November 10, 2007 13:42 WSPC/165-IJSSD 00243

Geometric Nonlinear Analysis of Cable Structures 573

2. Two-Node Cable Element

The catenary cable element to be presented in this paper is based on the explicit
by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.

solution to the differential equations and boundary conditions for a cable with
elastic catenary. Using the present approach, each cable with no internal joints
can be represented by a single cable element. Compared with the conventional
approaches based on the concept of equivalent modulus or by representing a cable
by a series of linear truss elements, the present approach is featured by the fact
that fewer number of degrees of freedom is required, the sag of geometry shape is
exactly taken into account, and the pretension force can be calculated with the aid
of the unstressed cable length. With respect to derivation of the stiffness matrix
Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

for the catenary element, the related analytical expressions were derived mainly by
Irvine.1
The cable element considered in this paper is assumed to be perfectly flexible
and linearly elastic, for which Hooke’s law applies, with the self-weight uniformly
distributed along the length of the curve. Moreover, the temperature effect on the
cable is also included in updating the uniform load on the cable.
Figure 1(a) shows an infinitesimal segment of the cable in the initial and strained
configurations, in terms of the Cartesian coordinates x and y and the Lagrangian
coordinate p for the deformed shape. The Lagrangian approach is adopted to sim-
plify the derivation, as the coordinate is measured along the deformed length of the
cable. Let L0 denote the unstressed length of the cable, which need not necessarily
be longer than the chord length of the cable. With reference to the free boy diagram
in Fig. 1(b), the horizontal and vertical equilibrium conditions of the cable are
dx
T = H, (1)
dp
dy s
T =V −W , (2)
dp L0

(a) (b)
Fig. 1. Cable configurations: (a) infinitesimal segment and (b) forces acting on a strained cable
profile.
 November 10, 2007 13:42 WSPC/165-IJSSD 00243

574 Y. B. Yang & J.-Y. Tsay

where T is tension of the cable, H is horizontal component of the tension, which is

constant everywhere since no longitudinal loads are acting on the cable, and V is
by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.

the vertical reaction of the support. Furthermore, W (= mgL0 ) is the self-weight of

the cable, m the mass per unit length, g the acceleration of gravity, and s the length
of the portion of the cable measured from the support. Let ds and dp respectively
denote the length of an infinitesimal segment of the cable in the initial (unstressed)
and deformed (stressed) configuration. The stress σ of the cable can be related to
the strain by Hooke’s law as


dp − ds
σ=E , (3)
Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

ds
where E is Young’s modulus.
From the equilibrium equations in Eqs. (1) and (2), the cable tension T can be
solved as

2
2
s
T (s) = H + V − W , (4)
L0

by noting that dp2 = dx2 + dy 2 . The coordinates x(s) and y(s) can be integrated
from Eqs. (1) and (2), respectively, as

  
Hs HL0 −1 V −1
V −W L0
s
x(s) = + sinh − sinh , (5)
EA W H H
 


2  
Ws 2
Ws V s HL0  V V − L0 
y(s) = − +  1+ − 1+  , (6)
EA W 2L0 W H H

where A is the cross-sectional area of the cable in the unstressed profile.

By applying the boundary conditions of x(0) = y(0) = 0 and x(L0 ) = l, y(L0 ) =
h, the horizontal and vertical projection lengths l, h of the cable can be obtained
as follows:



HL0 H −1 V −1 V −W
l= + sinh − sinh , (7)
EA q H H


L0 W 1  2  
h= V − + H + V 2 − H 2 + (V − W )2 , (8)
EA 2 q
where q = W /L0 , or
W
q= , (9)
L0 (1 + α∆T )
in consideration of the temperature effect. Here α is the thermal coefficient of
expansion and ∆T is the temperature rise from the design temperature. As can
be seen, both the projection lengths l and h are functions of the horizontal and
November 10, 2007 13:42 WSPC/165-IJSSD 00243

Geometric Nonlinear Analysis of Cable Structures 575

vertical tension components H and V , i.e. l = f (H, V ) and h = g(H, V ). The total
differentials of the projection lengths l and h in Eqs. (7) and (8) can be expressed as
by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.

 
∂f ∂f
     
dl dH  ∂H ∂V 
= [F ] =  dH . (10)
dh dV  ∂g ∂g  dV
∂H ∂V
The coefficients fij in the flexibility matrix [F ] are obtained directly as the differ-
entials of l and h, namely,



Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

∂f L0 1 V V −W
f11 = = + sinh−1 − sinh−1
∂H EA q H H
 
1 V V −W
+ −√ + , (11a)
q H2 + V 2 H + (V − W )2
2

 
∂f 1 H H
f12 = = √ − , (11b)
∂V q H2 + V 2 H 2 + (V − W )2
√  
∂g 1 H2 + V 2 H 2 + (V − W )2
f21 = = −
∂H q H H
 
1 −V 2 H (V − W )2 H
+ √ + , (11c)
qH 2 H2 + V 2 H 2 + (V − W )2
 
∂g L0 1 V V −W
f22 = = + √ − . (11d)
∂V EA q H2 + V 2 H 2 + (V − W )2

It was recommended by Abbas2 that the values of f12 and f21 as given above should
be replaced by their average, i.e.

1
f¯12 = f¯21 = (f12 + f21 ). (11e)
2

In fact, it was demonstrated by Luo et al.9 that the aforementioned two coefficients
f12 and f21 are equal.
For a three-dimensional cable element, as indicated in Fig. 2, there are three
force components acting at each of the two ends. The three-dimensional flexibility
matrix [F ]3×3 in the local coordinates can be obtained as a direct extension from
the in-plane matrix [F ]2×2 plus an out-of-plane coefficient f33 (Jayaraman and
Kundson,7 Luo et al.9 ), as given below:



l L0 1 −1 V −1 V −W
f33 = = + sinh − sinh , (11f)
H EA q H H
 November 10, 2007 13:42 WSPC/165-IJSSD 00243

576 Y. B. Yang & J.-Y. Tsay

by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

Fig. 2. Three-dimensional cable element.

with the other out-of-plane coefficients known to be zero,

f13 = f31 = f23 = f32 = 0. (11g)

The stiffness matrix [K] of the cable element can be obtained directly as the inverse
of the flexibility matrix, namely,
 
[k] −[k]
[K]6×6 = , (12)
−[k] [k]

where
 −1
f11 f12 0
 
[k] = [F ]−1
3×3 = f21 f22 0  . (13)
0 0 f33
Obviously, the stiffness matrix of the cable element is symmetric.
The cable element has been developed in the stretched plane under the resultant
loads, such as the dead load and wind load, etc. The transformation rule should
be applied to updating the stiffness matrix from the local coordinate system to the
global coordinate system. For the case where only the gravity loads are imposed on
the cable, the transformation matrix [T ] for the stiffness matrix is9
 
[t] [0]
[T ]6×6 = , (14)
[0] [t]

where
 
cos θ 0 sin θ
 
[t] =  0 1 0 , (15)
− sin θ 0 cos θ
and θ is the projection angle shown in Fig. 2.
November 10, 2007 13:42 WSPC/165-IJSSD 00243

Geometric Nonlinear Analysis of Cable Structures 577

3. Procedure for Computing the Cable Stiffness Matrix

Before assembling the cable elements with the beam and truss elements in the global
by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.

coordinate system, the following iterative procedure suggested by O’Brien and

Francis3 for obtaining the cable element stiffness matrix should first be performed.

1. For each cable, the following are given: elastic modulus E, cross-sectional area A,
unit weight W , coordinates of the two ends, and the unstressed cable length L0 .
Based on these data, the initial horizontal and vertical projection length of l0 ,
h0 and the parameter q can be computed. It follows that the initial (trial) values
for the reaction forces H0 and V0 can be computed as5 :
Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

ql0
H0 = , (16)
2λ0
 
q cosh(λ0 )
V0 = − |h0 | + L0 + W. (17)
2 sinh(λ0 )
An initial value of 0.2 is recommended for λ0 when the unstressed length L0 of
the cable is shorter than the chord length, i.e. (l02 + h20 )1/2 , and an arbitrarily
large value of 106 is adopted for λ0 when l0 is zero in the case of vertical hangers.
For the other cases, the following is adopted5 :

2 1/2
L0 − h20
λ0 = 3 −1 . (18)
l02
2. With the input of the aforementioned initial data, one can obtain new values for
l and h from Eqs. (7) and (8). Let ∆l and ∆h denote the differences between
the initial and computed projection lengths,

∆l = l0 − l, ∆h = h0 − h. (19)

When the differences ∆l and ∆h are greater than preset tolerance, the coefficients
of the flexibility matrix [F ] can be calculated from Eq. (11), for given values of
E, A, L0 , q and assumed reaction forces H and V . The cable stiffness matrix [K]
can then be calculated as the inverse of the flexibility matrix [F ], as indicated
by Eqs. (12) and (13). Let {F } denote the reaction forces, i.e. {F } = {H, V }T ,
and {d∆} the differences of the projection lengths, i.e. {d∆} = {∆l, ∆h}T . The
reaction forces {F } can be updated as follows:

{F }i+1 = {F }i + [K]i {d∆}i , (20)

or

H i+1 = H i + k11
i
× ∆li + k12
i
× ∆hi , (21a)
V i+1 = V i + k21
i
× ∆li + k22
i
× ∆hi . (21b)

Then the above procedure can be repeated to compute new l and h until the
differences ∆l and ∆h are smaller than the preset tolerances.
 November 10, 2007 13:42 WSPC/165-IJSSD 00243

578 Y. B. Yang & J.-Y. Tsay

by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

Fig. 3. In-plane cable element.

3. Once the tolerance requirements are satisfied, meaning that the values computed
for the reaction forces H and V are exactly the ones desired, we can proceed to
calculate the in-plane end reaction forces F1 to F4 (see Fig. 3) and tension forces
T1 and T2 using the following relations:

F1 = −H, F2 = W − V, F3 = H, F4 = V, (22)
 
T1 = F12 + F22 , T2 = F32 + F42 . (23)

Then we can determine the final coefficients of the flexibility matrix [F ] using
Eq. (11) and the associated cable stiffness matrix [K] as the inverse of the flexi-
bility matrix, as indicated by Eqs. (12) and (13). As can be seen from Eq. (23),
the tension forces T1 and T2 are functions of the end reaction force F1 to F6 for
three-dimensional structures.
4. As long as the cable stiffness matrix and end reaction forces are calculated, the
in-plane strained cable length L can be computed as


1 2 F4 + T2
L = L0 + F4 T2 + F2 T1 + F1 ln . (24)
2EAq T1 − F2

Accordingly, the sag y of the cable as indicated in Fig. 4 can be computed as

follows1 :
ql2
y = ȳ , (25)
H cos θ
where
1  ε 
ȳ = x̄(1 − x̄) 1 + (1 − 2x̄) , (26)
2 6
and x̄ = x/l, ε = ql sin θ/H.
November 10, 2007 13:42 WSPC/165-IJSSD 00243

Geometric Nonlinear Analysis of Cable Structures 579

by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

Fig. 4. The sag y of the inclined cable.

4. Procedure for Geometric Nonlinear Analysis

The catenary cable element as derived in the preceding section will be incorpo-
rated with the other two kinds of elements, i.e. the beam and truss elements, in
the Geometric Nonlinear Analysis (GNA) program. The beam and truss element
adopted herein are those derived by Yang and Kuo17 using the updated Lagrangian
approach.
Although the frame elements, i.e. beam and truss elements, may exhibit soften-
ing behavior under compression, there is no postbuckling phenomenon for cables as
they are always under tension. As a matter of fact, a cable may get stiffened due
to an increase of the tension forces. In studying the stability of long-span bridges
with high pylons, the Generalized Displacement Control (GDC) method devised
by Yang and Shieh16 will be used, instead of the Newton–Raphson method, for its
general numerical stability in treating problems with limit points and snap-back
points, although such problems may not be encountered in the analysis of tension
structures, such as cable-supported structures.
In an incremental-iterative nonlinear analysis, we shall typically deal with fol-
lowing structural equation for the jth iterative step of the ith incremental step:
i
[Kj−1 ]{∆Uji } = λij {P̂ } + {Rj−1
i
}, (27)
i
where [Kj−1 ] denotes the tangent stiffness matrix, {P̂ } denotes the reference loads,
i i
λj the load parameter, {Rj−1 } the unbalanced forces, and {∆Uji } the displacement
increments generated during the current iterative step. By the GDC method, the
load parameter λij is allowed to vary during each iterative step. The unbalanced
 November 10, 2007 13:42 WSPC/165-IJSSD 00243

580 Y. B. Yang & J.-Y. Tsay

i i
forces {Rj−1 } are computed as the differences between the applied loads {Pj−1 }
i
and resistant forces {Fj−1 } at the last iterative step, i.e.
by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.

i i i
{Rj−1 } = {Pj−1 } − {Fj−1 }. (28)
For convenience, the structural equations in Eq. (27) can be decomposed into two
parts as
i
[Kj−1 ]{∆Ûji } = {P̂ }, (29a)
i
[Kj−1 ]{∆Ūji } = {Rj−1
i
}. (29b)
Accordingly, the displacement increments {∆Uji } can be computed as
Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

{∆Uji } = λj {∆Ûji } + {∆Ūji }, (30)

where for the first iterative step, i.e. with j = 1, for each incremental step,
λi1 = λ11 |GSP|1/2 , (31)
and for the remaining iterative step, i.e. with j ≥ 2,
{∆Û1i−1 }T {∆Ūji }
λij = . (32)
{∆Û1i−1 }T {∆Ûji }
The Generalized Stiffness Parameter (GSP) is a parameter that can reflect the
change in stiffness of the structure, defined as
{∆Û11 }T {∆Û11 }
GSP = , (33)
{∆Û1i−1 }T {∆Û1i }
which starts with GSP = 1 for the first incremental step, i.e. for i = 1. The GSP
is negative only for the incremental steps “immediately after” the limit points,
while for the other load increments, it remains positive. Such a feature serves as a
signal for reversing the loading direction when passing the limit points.16 As can be
seen from Eq. (31), the load increment for each incremental step is automatically
adjusted according to the change in stiffness of the structure. The procedure for
performing an incremental-iterative nonlinear analysis based on the GDC method,
together with the GSP parameter, is outlined in Fig. 5.

5. Numerical Examples
Six examples will be studied to demonstrate the reliability and applicability of
the geometric nonlinear analysis program developed for cable-supported structures
using the proposed cable element. The results obtained will be compared with those
found in the literature.

Example 1. This example serves to demonstrate the convergence characteristics

of the present element by comparing the results obtained with those by the CBL1
element,7 truss element,19 and 3-node element.20 As shown in Fig. 6, a taut cable
November 10, 2007 13:42 WSPC/165-IJSSD 00243

Geometric Nonlinear Analysis of Cable Structures 581

by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

Fig. 5. Flow chart of incremental-iterative nonlinear analysis.

spanning 1,000 ft is supported at end points of the same elevation. The cable has a
unit weight of 3.16 lb per foot, a cross-sectional area of 0.0059 sq ft, and an effective
modulus of elasticity of 2.74 × 109 psf. The cable is subjected to a live load of 8
kips at 400 ft from the left support. The problem is to determine the displacement
at the loading point.
 November 10, 2007 13:42 WSPC/165-IJSSD 00243

Dây ngắn hơn nhịp hệ ft

582 Y. B. Yang & J.-Y. Tsay
by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.

Fig. 6. Taut cable under concentrated load (Example 1).

Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

-13.56
Displacement

Truss element
3-node element
(ft)

-13.54
Present CBL1 element
study
-13.52

4 8 12 16 20
No. of Elements

Fig. 7. Speed of convergence of elements (Example 1).

Table 1. Results and comparison for Example 1 (unit: ft).

Displacement (ft) Present Study CBL1 Element7 Truss Element19 3-Node Element20
Vertical −13.539 −13.537 −13.539 −13.539
Horizontal −0.075 −0.075 −0.075 −0.075

The cable was modeled by two cable elements each with unstressed length of
396.0 ft and 594.0 ft in this study. Concerning the rate of convergence, the displace-
ments computed for the point under the concentrated load are compared with those
by the other elements in Fig. 7 and Table 1. As can be seen, the results obtained
by using only two cable elements derived herein are most accurate. The same level
of accuracy can be achieved by other elements using much more elements.

Example 2. The second example is a freely suspended cable spanning 1,000 ft

between two supports at the same elevation, where the sag at the mid-span is 100 ft
(Fig. 8). The cable has a unit weight of 3.16 lb per foot, a cross-sectional area of
November 10, 2007 13:42 WSPC/165-IJSSD 00243

Geometric Nonlinear Analysis of Cable Structures 583

by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.

Fig. 8. Suspended cable under concentrated load (Example 2).

Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

0.85 sq in. and an effective modulus of elasticity of 19 × 106 psi. The problem is to
determine the cable displacements for an 8-kip live load placed at 400 ft from the
left support. The displacement at loaded point was studied by O’Brien and Francis3
and Jayaraman and Kundson.7

The cable was modeled by two cable elements each with unstressed length of
412.9 ft and 613.0 ft in this study. The displacements computed for the point under
the concentrated load are compared with the existing results in Table 2, from which
good agreement can be observed for the present results.

Example 3. The third example is the problem studied earlier by Peyrot and
Goulois.6 Consider an elastic cable of unstressed length L0 = 100, cross-sectional
area A = 1, modulus of elasticity E = 3 × 107 , weight per unit length W = 1,
thermal expansion coefficient α = 6.5×10−6, subjected to a temperature rise ∆T =
100. The problem being addressed here is that given the position of coordinates
of the two ends of the cable, the profile of the cable with end reactions can be
determined.

For this example, it is assumed that the left end is fixed at the coordinates
(0, 90), and the right end maintains a constant vertical elevation of 30 and is
subjected to controlled horizontal displacement increment of 20 units each until it
reaches the coordinates (100, 30). The cable was modeled by one cable element,
and the results obtained here are plotted in Fig. 9 and compared with Ref. 6 in
Table 3. As can be seen, most of the results obtained using a single cable element by
the present approach agree excellently with Ref. 6, except those for the coordinates
(100, 30). For the cases with displacements larger than the case for the coordinates
(80, 30), the cable behaves something like a stay cable. The results in the table
indicate that the tension force of the cable varies with the location of the anchoring

Table 2. Results and comparison for Example 2 (unit: ft).

Displacement (ft) Present Study Jayaraman & Kundson7 O’Brien & Francis3 Luo et al.9
Vertical −18.456 −18.458 −18.460 −18.457
Horizontal −2.819 −2.819 −2.820 −2.819
 November 10, 2007 13:42 WSPC/165-IJSSD 00243

584 Y. B. Yang & J.-Y. Tsay

by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

Fig. 9. Configuration and end forces of cable (Example 3).

Table 3. Results and comparison for Example 3.

Location Present Study Peyrot and Goulois6

Reaction Force H V H V
0.2 0.01 20.02 0.00 20.02
20 3.061 19.93 3.061 19.93
40 9.172 19.24 9.172 19.24
60 22.15 15.73 22.15 15.73
80 504.1 −328.9 504.0 −328.8
100 4,255,700 −2,553,400 4,170,000 −2,511,000

point at the right end. In other words, larger tension force is required to pull the
cable outward during the cable erection.

Example 4. The fourth example is the problem that was studied by Peyrot and
Goulois.6 As shown in Fig. 10, a three-dimensional cable net is composed of three
cables connected at a central node A, which in turn is pulled up by a vertical spring
of stiffness K = 1, 000 units anchored to a support that is free to roll horizontally.
The three cables each was modeled by one cable element and the spring by one
truss element anchored to the roller support.

The three cables have weights of 1, 2 and 2 units and unstressed lengths of
580, 510 and 510 units, respectively, identical cross-sectional area of 1 unit, and
a thermal coefficient of 6.5 × 10−6 unit, and an effective modulus of elasticity
November 10, 2007 13:42 WSPC/165-IJSSD 00243

Geometric Nonlinear Analysis of Cable Structures 585

by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

Fig. 10. Cable net: (a) Side view and (b) Top view (Example 4).

E = 29 × 104 units; the aforementioned units are all consistent the physical sense.
Figure 10 shows all the necessary geometric data of the system. The cables are
subjected to a horizontal load of 1,000 units at the central node A with two tem-
perature cases, i.e. ∆T = 0, 100 units. The displacements computed for the three
directions of the central node are compared with existing results in Table 4. As can
be seen, good agreement has been achieved for the present results.

Example 5. The fifth example is a 12-node cable net with non-dimensional unit,
as shown in Fig. 11. All the nodal points of the net are assumed to lie in the
 November 10, 2007 13:42 WSPC/165-IJSSD 00243

586 Y. B. Yang & J.-Y. Tsay

Table 4. Results and comparison for Example 4.

Coyette & Peyrot &

by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.

Present Guisser14 Luo et al.9 Present Goulois6 Luo et al.9

Displacement ∆T = 0 ∆T = 0 ∆T = 0 ∆T = 100 ∆T = 100 ∆T = 100
x 26.083 26.110 26.114 26.443 26.473 26.474
y 40.394 40.420 40.421 41.116 41.135 41.143
z −2.848 −2.890 −2.889 −2.834 −2.874 −2.874
Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

Fig. 11. Plan view of cable system (Example 5).

Table 5. Results and comparison for Example 5.

Present Study Abbas2

Node x y z x y z
4 −0.014 −1.631 0.030 −0.014 −1.631 0.030
5 0.004 −1.359 −0.004 0.004 −1.358 −0.004
8 −0.063 −3.175 0.063 −0.063 −3.175 0.063
9 −0.030 −1.632 0.014 −0.030 −1.631 0.014

horizontal plan at equal spacing of 40 units. The cables in the net are orthogonal
to each other with the following common properties: A = 1 unit, W = 1 unit,
E = 29, 000 units and L0 = 40 units. All the boundary nodes with numbers 1, 2, 3,
6, 7, 10, 11 and 12 are fixed, whereas each of the internal nodes with numbers 4, 5,
8 and 9 has three translational degrees of freedom. A load of 1,000 units is applied
in the downward y-direction at node 8. The cable net was modeled by a total of
12 cable elements. The displacements computed for the internal nodes are compared
with those of Abbas2 in Table 5. Clearly, good agreement has been achieved for the
two sets of results.
November 10, 2007 13:42 WSPC/165-IJSSD 00243

Geometric Nonlinear Analysis of Cable Structures 587

Table 6. Results and comparison for Example 6.

Results Left TA (kN) Right TB (kN) Cable Length s (m) Sag f1/2 /L
by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.

Present study 7321.591 7104.359 576.616 1/101.236

Ren & Gu18 7331.219 7114.005 576.616 1/101.725

Example 6. The last example is to apply the present cable element for the
analysis of the world’s longest cable-stay bridge, i.e. the Sutong Changjiang high-
way bridge, in which the longest cable No. J34 with chord length of 576.488 m
is selected. Consider an elastic cable of unstressed length L0 = 574.805 m, cross-
sectional area A = 0.012046 m2 , modulus of elasticity E = 190 GPa, and weight
Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

per unit length W = 0.988 kN/m. In this study, it is assumed that the left end
is fixed at the coordinates (0, 220.564 m), and the right end at the coordinates
(532.626 m, 0). The cable was modeled by one cable element. The results computed
for the tension forces at the two ends, the cable length and the sag are compared
with those of Ren and Gu18 in Table 6. As can be seen, good agreement has been
achieved.

6. Concluding Remarks
A two-node catenary cable element is presented here for the study of three-
dimensional cable-supported bridges, for which the cables may exhibit large sags.
The stiffness matrix of the cable element was derived as the inverse of the flexibil-
ity matrix, with the effects of self weight and pretension duly taken into account.
This element is incorporated, together with the commonly used beam and truss
elements, in a general program for analyzing the geometric nonlinear behaviors of
cable-supported structures, including the pylons and bridge deck.
Because of the flexible characteristics of cable-supported structures, geomet-
ric nonlinear effect has to be considered in the analysis of such structures. The
algorithm adopted herein for performing the incremental-iterative analysis is the
Generalized Displacement Control (GDC) method, along with the general stiffness
parameter (GSP), by which the loads are not kept constant in the iterative steps
and general numerical stability can be maintained when passing the limit points
and snap-back points, though such points may not occur in the analysis of cable
structures.
The present cable element has been adopted in the solution of six cable struc-
tures involving large displacements. Through comparison with previous results, the
accuracy and applicability of the present cable element in simulating cables with
large sags are confirmed. With the present element, each cable of a cable-supported
structure can be modeled by a single or a small number of elements, even for cables
with rather large sags, as those encountered in the cable nets, suspension bridges
and long-span cable-stayed bridges.
 November 10, 2007 13:42 WSPC/165-IJSSD 00243

588 Y. B. Yang & J.-Y. Tsay

References
1. H. M. Irvine, Cable Structures (The MIT Press, Cambridge, MA, USA, 1981).
by AUCKLAND UNIVERSITY OF TECHNOLOGY on 01/29/19. Re-use and distribution is strictly not permitted, except for Open Access articles.

2. S. Abbas, Nonlinear Geometric Material and Time Dependent Analysis of Segmental

Erected Three-Dimensional Cable Stayed Bridges, PhD thesis, Department of Civil
Engineering, University of California, Berkeley (1993).
3. T. O’Brien and A. J. Francis, Cable movements under two-dimensional loads, J.
Struct. Div., ASCE 90, No. ST3 (1964) 89–123.
4. T. O’Brien, General solution of suspended cable problems, J. Struct. Div., ASCE 93,
No. ST1 (1967) 1–26.
5. A. H. Peyrot and A. M. Goulois, Analysis of flexible transmission lines, J. Struct.
Div., ASCE 104, No. ST5 (1978) 763–779.
Int. J. Str. Stab. Dyn. 2007.07:571-588. Downloaded from www.worldscientific.com

6. A. H. Peyrot and A. M. Goulois, Analysis of cable structures, Comput. Struct. 10

(1979) 805–813.
7. H. B. Jayaraman and W. C. Kundson, A curved element for the analysis of cable
structures, Comput. Struct. 14 (1981) 325–333.
8. R. Karoumi, Some modeling aspects in the nonlinear finite element analysis of cable
supported bridges, Comput. Struct. 71 (1999) 397–412.
9. X. H. Luo, R. C. Xiao and H. F. Xiang, Cable element based on exact analytical
expressions, J. Tongji Univ. (Nat. Sci.) 33(4) (2005) 445–450.
10. H. Ozdemir, A finite element approach for cable problems, Int. J. Solids Struct. 15
(1979) 427–437.
11. M. Yang and Z. Chen, The nonlinear finite element analysis for two-node catenary
element of cable structure based on UL formulation, China Civ. Eng. J. 36(8) (2003)
63–68.
12. W. M. Henghold and J. J. Russell, Equilibrium and natural frequencies of cable
structures (a nonlinear finite element approach), Comput. Struct. 6 (1976) 267–271.
13. H. M. Ali and A. M. Abdel-Ghaffar, Modeling the nonlinear seismic behavior of cable-
stayed bridges with passive control bearings, Comput. Struct. 54(3) (1995) 461–492.
14. J. P. Coyette and P. Guisser, Cable network analysis by a nonlinear programming
technique, Eng. Struct. 10 (January 1988) 41–46.
15. A. S. Nazmy and A. M. Abdel-Ghaffar, Three-dimensional nonlinear static analysis
of cable-stayed bridges, Comput. Struct. 34 (1990) 257–271.
16. Y. B. Yang and M. S. Shieh, Solution method for nonlinear problems with multiple
critical points, AIAA J. 28(12) (1990) 2110–2116.
17. Y. B. Yang and S. R. Kuo, Theory and Analysis of Nonlinear Framed Structures
(Prentice Hall, Singapore, 1994).
18. S. Y. Ren and M. Gu, Static analysis of cables’ configuration in cable-stayed bridges,
J. Tongji Univ. (Nat. Sci.) 33(5) (2005) 595–599.
19. W. C. Kundson, Static and dynamic analysis of cable net structures, Doctoral disser-
tation, University of California, Berkley, CA (1971).
20. M. Haase, Zur natürlichen Formulierung von Simplexelementen höherer Ordnung für
die Berechnung elastischer Membranschalen und Seikonstruktionen unter grossen Ver-
formungen, Doctoral dissertation, University of Stuttgart (1979).