You are on page 1of 9

- 1 -

http://www.ivypub.org/AE
Architectural Engineering
February 2014, Volume 2, Issue 1, PP.1-9
Numerical Solutions of Various Cables and
Their Applications
Yuemin Yang
The First Highway Survey and Design Institute of China, 710075, China
Email: yuemin_yang@163.com
Abstract
This article deeply analyzes the curve form, deflection and deformation of the length given cable under various loads by the method
of discretization, solution of nodal equilibrium, and iterative solution of the whole system balance. The curves of all other kinds of
cables and their mechanical properties are also discussed. The solution of the cable is adopted to analyze the effect of the cable's sag
on the non-linearity of cable-stayed bridge. Besides, the solutions are applied to determine the ideal curves of arch bridges.
Keywords: Analysis of Main Cables; Cable's Sag; I deal Curves of Arch Bridges
1 INTRODUCTION
The most important step in cable supported structure design, especially the suspension bridge, cable stayed bridge
and cableway, is the nonlinear analysis of the structure under various loads. Then how to exactly present the
mechanical properties of its cable, tower and beam becomes essential. Because the beams and towers are stiff
elements whose relations between displacement and applied load are well known, the objective of this paper is to
find an exact formulation of the flexible cable's mechanical properties. Besides, their applications are also discussed.
For example, since the solution of the cable is no bending result for flexible structures, hence, it can be adopted to
determine the ideal curves of arch bridges, and the structures alike.
2 MAIN CABLE'S MECHANICAL PROPERTIES AND ITS DEFORMATION
We divide the cable into n segments along x direction. Each segment has the length:
cos
i i i
dx ds o =
(1)

i
i
i
dy
tg
dx
o =
(2)
The weight, q(x
i
) ds
i
, of each segment can be equivalently represented by two neighbouring nodal joint loads. When
n is large enough, the discretized curve is identical to the original smooth one. Each segment is assumed as straight
line. Starting from point A, its equilibrium equations are:
1 1
cos N H o =

(3)

1
1 1 1
1
1
sin ( )
2
cos
A
dx
N V q x o
o
=

(4)

1 1
1
1
1
2 cos
A
V q dx
tg
H H
o
o
=


(5)
If we define the internal axial force of the cable as N
i
, to a generic point i, we have:
1 1
cos cos
i i i i
N N H o o

= =

(6)

1 1
sin sin
i i i i i i
N N P Q o o

=

(7)

1
1
1
1 1
( ) ( )
2 cos 2 cos
i i
i i i
i i
dx dx
Q q x q x
o o

= +

(8)
- 2 -
http://www.ivypub.org/AE
FIG.1 DISCRETE MODELING OF THE CABLE
where P
i
is the applied concentrated load at point i. From Eqs.(6) (7) (8), we obtain:

1 1
1
1
( ) ( ) 1 1
2 cos 2 cos
i i i i i
i i
i i
P q x dx q x dx
tg tg
H H H
o o
o o

=

(9)
If Eq.(9) is divided by 0 dx ,we get the differential equation of the cable
[1] [3]
:

2
( )
( ) 1
i
i
P q x
y x x y
H H
o '' ' = +


(10)
Eq.(9) is a discrete form of Eq.(10). ( ) x o is Dirac function. Eqs.(5) and (9) can be simplified as:

sin cos
i i i i
a b o o = +

(11)
in which

1 1
1
1
( ) 1
2 cos
i i i
i i
i
P q x dx
a tg
H H
o
o

=

(12)

( ) 1
2
i i
i
q x dx
b
H
=

(13)
When 1 i = , we have P
1
=V
A
,
0
0 o = ,

q(x
0
)=0. The solution of (11) gives:


2 2 2 2
2
( 1)( 1)
cos
( 1)
i i i i i i
i
i
a b a b a b
a
o
+ +
=
+
(14)

arcsin( cos )
i i i i
a b o o = +

(15)
From Eqs.(5) and (9), it is clear that when V
A
and H are given all the unknown quantities in Eq.(9) can be obtained in
recursion. The coordinate y
i
of each point is:

1 i i i i
y y dx tgo
+
= + 1 1 i n =

(16)
We begin with starting estimates for V
A
and the horizontal force H by assuming the cable in the shape of a triangle,
where the cable is straight, as shown in Fig.2. The length of cable AC is L
0
/2.

2 2
0
1 1
( ) ( )
2 2
f L a =

(17)

FIG. 2 ESTIMATING OF THE STARTING VALUES
x
y
H H
A
B
y
V
V
A B
Q
1
+Q p
2
2
+Q p
i
i
+Q
p
i+1 i+1 +Q p
k k
n
k
a
H
A V
Q
i
1 N

1
P
i
i

N
i
Ni-1
i-1

i
A
1
f
+Q
a
x
y
H
C
B
H
V V
A B
A
f

- 3 -
http://www.ivypub.org/AE
According to the balance of the whole cable in Fig.2, here all the loads on the cable are left out, we get:
) ( )
cos
(
1
i
i i
i A
x a
dx q
P
a
V + =



(18)

A A
H V ctg H | = =

(19)

2 f
tg
a
| =

(20)

0 k
L L L A =

(21)
where L
0
is the given length of the cable, and L
k
is the cable's length in

k

th iteration:

2 2
1 1
( ) ( )
k i i i i
L y y x x

= +


(22)
If we define the angle of the resultant of H and V
A
as with respect to horizontal line, the iterative sequences for
k

and H
k
are given as follows
[1] [5]
:


) (
1
a
y
arctg
n
k k
=




(23)


1
0
(1 )
k k
L
H H
L

A
= +

(24)
Substituting into Eqs.(5) and (9) for the starting values V
A
and H, we get the coordinate y
n
of point B and the length
L
k
of the cable. When y
n
and L
k
are put into (21), (23), (24), we get new values V
A
and H. The iteration is carried on
until the changes between two following iterations are small enough or less than the given tolerance values. The
iterations (23) and (24) converge so fast that after 4~5 repeats the error between two following iterations is less than
10
-5
,

whereas the iteration is started with any initial values larger than zero. To any ratio f/a, the numerical solution
of Eq.(9) is almost the same as the analytical solution of Eq.(10). Elastic extension of the cable is given by

= A
i i
i i
EA
dx N
L
o cos

(25)
If the given length of free cable is L
0
, and assuming that the cable can't extend under load, we define this state of the
cable as state I. Then we define the state of the cable with extension as state II, and its length is L. Since we only
consider geometric non-linearity of the cable, the cable's extension is only related to its internal force in state II. But,
so far the cable's length in state II is an unknown, the problem can't be solved for. Therefore, we need first to take the
calculated extension of Eq.(25) in state I,

L+L
0
, as an approximate length of state II. Namely, the cable's length in
state II is taken as L+L
0
. With the length assumption of

L+L
0
as L
0
of Eq.(24) in state II, we can carry on the
analysis of the problem obtaining a new extension

L of the cable in state II. If the new extension

L is equal to the
assumed one, that is the answer we are looking for. Otherwise, we still need to take the recent extension

L+L
0
as
cable's length L
0
of Eq.(24) in state II, and repeat the calculation until the requirement is satisfied. From context we
know that this second iteration is to satisfy the cable's deformation compatibility, and it converges even faster.
3 DETERMINATION OF VARIOUS CABLE FORMS
3.1 Curve Form of the Cable When its Span and Vertical Height Are Given
The main cable of a suspension bridge can be considered as a combination of two symmetric half cables. The form
of the half cable is then the same as a tilted cable. To a tilted cable, the derivation of its equilibrium is exactly the
same as above. The setting up of equilibrium equations starts from point A. The iterative variable is the horizontal
force H at point A. The iterative sequence of H goes as
[5]
:


1
0
(1 )
k k
df
H H
f

= + (26)

0 n
df y f = (27)
where y
n
is the vertical coordinate of point B during the iterative process, and f
0
is the designed height of point B.
- 4 -
http://www.ivypub.org/AE
FIG.3 TILTED CABLE MODELLING
3.2 Curve Form of the Cable When H I s Given
In order to keep balance of the tower, side span cable must offer the same horizontal force as the middle span, i.e.,
the horizontal force in the side span cable is given. We may take the vertical load P at point A as variable in the
iteration. The balance of the whole cable requires that

0 = + H f a P M
BP

(28)
where M
BP
is the moment about B produced by all the loads on the cable. It is easy to verify the increment of M
BP
is a
smaller amount during the iteration. The differentiation of (28) gives out:

df
a
H
dP =

(29)
The iterative relation of variable P develops in the way
[5]
:

df
a
H
P P
k k
=
1

(30)
The setting up of equilibrium is the same as above, which is started from point A. Under the same conditions, we
obtain exactly the same result by Eqs. (26) and (30).
3.3 Curve Form of A Length Given Tilted Cable
When the length of a cable is given, the equilibrium equations for the cable are set up in the same manner as above.
The iterative variables and their iterative relations are
[1] [5]
:


1
0
(1 )
k k
L
H H
L

A
= +

(31)

1
k
k k
H
P P df
a

=

(32)
in which
0 k
L L L A = .
3.4 I terative Solution of the Cable With a Horizontal Restriction At its Middle
In order to increase the stiffness of the cable, a horizontal restriction is often imposed on a middle point k of the cable
[8]
.
Let the cable's length of point k's left side be L
L0
, and k's right side be L
R0
.The governing equations of point k are:

1 1
cos cos
k k k k k
N H N o o

+ =

(33)

k k k k k k
Q P N N =
1 1
sin sin o o

(34)
Let:

cos
R K K
H N o =
1 1
cos
L k k
H N o

=

(35)
Here point k could be any generic point. The constant horizontal force in the cable of k's left is H
L
, and that of k's right
y
x
H A
B
H
A
H
P
N
Q
1
1
P +Q
k k
P
+Q
k-1 k-1
P
+Q
2 2
P
+Q
n
n
P
+Q
1
a

1
f
- 5 -
http://www.ivypub.org/AE
is H
R
.H
K
is the horizontal restriction force on point k. From (33) and (34) we obtain:

1 1
sin
k k k k
k
R R R
N P Q
tg
H H H
o
o

=

(36)
Here Eq.(36) can also be written as the form of Eq.(11). Thus

1 1 1 1
1
sin ( ) 1
2 cos
k k k k k
k
R R R k
N P q x dx
a
H H H
o
o

=

(37)

( ) 1
2
k k
k
R
q x dx
b
H
=

(38)
The starting value of H
L
is the same as Eqs.(18) (19). Then the starting value of H
R
can be achieved, as in Fig.2, by the
consideration of whole cable's balance:

( ) 1
( )
cos
i i
B i i
q x dx
V P x
a |
= +


(39)

R B B
H ctg V H = =

(40)
The iterative sequences are
[1] [5]
:

1
( )
n
k k
y
arctg
a
| |

=

(41)

, 1
0
(1 )
L
Lk L k
L
L
H H
L

A
= +

(42)

, 1
0
(1 )
R
Rk R k
R
L
H H
L

A
= +

(43)
where

0 L Lk L
L L L A =
;
0 R Rk R
L L L A =

(44)

L
Lk,
L
L0;
L
Rk,
L
R0
are the calculated lengths and assumed lengths of the cables of k's left and right sides, respectively. The
iteration has a great accuracy and converges fast as well. The horizontal restriction force

H
k
is the difference between
H
L
and H
R
. According to the discussion above, we may find that the influence on the vertical displacement produced by
horizontal restriction of the hanging members of the suspension bridge in the example of reference [1] is about
0.4661%, which can be ignored thoroughly. This can also be achieved by an observation that the horizontal force H in
the main cable with a general ratio f/a is usually bigger than the dead load including the main cable and the deck. If we
compare the horizontal restriction of the hanging members with the dead load of entire bridge, it is clear that the
horizontal restriction is meaningless and negligible, so as stated by the digital result above.
All the solutions of various cables above are typical applications of the shooting method in Numerical Analysis.
4 EFFECT OF CABLE'S SAG ON THE NON-LINEARITY OF CABLE STAYED BRIDGE
FIG. 4 MODELS OF STRAIGHT AND SAGGED CABLES
To discuss the effect of cable's sag, we consider two cable structures with same properties as shown in Fig.4 (a) and
u

f
a
P
0
A
B
C
(a) (b)
a
P
0
A
C
f
B
- 6 -
http://www.ivypub.org/AE
Fig.4 (b). Two cables have the same free length. Member AC in both structures only acts as a horizontal restriction
[8]
.
In Fig.4 (a), we assume the cable to be straight. In the analysis of the cable in Fig.4 (a), the exact formulation of out-off-
balance force is adopted, i.e., the deformation compatibility requirement. The extending amount of the cable is:


o | sin sin
a a
L = A

(45)
Where:

f
a
tg = o
a
tg
f u
| =
+

(46)
The internal force of the cable is:

EA
N L
L
= A

(47)
The out-off-balance force at joint A writes
[2]
:


0
cos R P N | =

(48)
The solution to Fig.4 (a) is carried out by N-R method where the remainder is Eq.(48). The numerical method given
in the paper, i.e., the iteration sequence Eq.(31), is adopted to find out the solution to Fig.4 (b). We assume: the
weight density of the cable =80kN/m
3
,

E=2.05x10
8
kPa
[8]
, and the axial force of the cable at joint A when A is
unmoved:

| |
0.011 1860
A
A
N
n
n
o

= =

(49)
where A is the area of the cable, []=1860MPa is the allowable stress of the cable, and n is safety factor. Then the
applied external load at A is:

0
cos
A
P N o =

(50)
In order to describe the deformation of the straight cable exactly, we equilibrate the weight qL of the cable by two
nodal joint loads at points A and B, i.e., the load at A is:

0
1
2
P P qL = +

(51)
where q=A. This formulation is to take the mean value of the cable's internal forces as the axial force of the
weightless straight cable. Although the mean value can predict the deformation very well, there are some errors in its
axial force at joints A and B. If we assume f=150.0m, a=300.0m, and n=3, the solutions for the cables in Fig.4 are
given in Table 1.
TABLE 1 EFFECT OF THE CABLE'S SAG

Straight cable Sagged Cable
Errors Straight cable by
N-R method
Displace. u at A 2.3416544m 2.2991289m 1.85% 2.327599m
Expansion L 1.0537385m 1.0540650m 0.03099% 1.047375m
Hor. Restric. H 6296.85404kN 6298.6041kN 0.0278% 6259.0640kN
Axial Force N
A
7062.2143kN 6998.2064kN 0.915% 7019.6983kN
It is clear from Table 1 that the expansions of two cables are almost the same. However, there are some errors in the
displacement u. Here u is measured under the reference f=150.0m.When P
0
is small enough, the height f of the
sagged cable is f150.0m.This means a negative displacement with respect to f=150.0m. Thus, some amount of the
load P
0
is transferred to overcome the weight sagging. The rest load produces the positive displacement u only. The
total displacement of the sagged cable, including the negative part, is much larger than that of the straight member.
The straight cable in Fig.4 (a) is the limit stage of the sagged cable. In engineering, however, we often take the
straight member as reference. Thus, the negative part has no meaning in analysis. The errors occurred to

H

and N
A
of
the straight cable arose out of the analysis model. These errors can't occur in practice. These errors should only be
taken as reference.
The cable in Table 1 is relative long and sagged heavy in engineering. Then the cable commonly used in engineering
- 7 -
http://www.ivypub.org/AE
will have less sag effect. If we set points A and B of the cable in Table 1 at same horizontal level, then the effect of
the sag on cable's length is about 7.8044x10
-3
%
[7]
with respect to the straight line AB. Hence we can think the lengths
are the same. Then, for the cables with f0, their lengths have less difference from straight lines. Thus, their
expansions are almost the same. It is clear that the effect of cable's sag can't be estimated by the expansion. Only the
difference between the vertical positive displacements can show the effect. To imitate the sag effect by a straight
member then, and when the load increment is large, i.e., the loading is started from zero, the modified elastic
modulus of the straight member may be larger than its original value. If a small load increment after some loading is
concerned, according to more detail analysis, we may use the tangent stiffness of the sagged cable, which is smaller
than that of the straight member. We should pay more attention to the variation of the cable's stiffness for different
loading during analysis. The last column in Table 1 is the results where the out-off-balance force at joint A reads
[2] [7]
:

( ) R P K u u =

(52)
where K(u) is the vertical stiffness of point A
[2]
:

2
( ) cos
EA
K u
L
| =

(53)
By the well known formulations of cable element, the horizontal length is
[3] [6]
:

1 1 0
( )
HL V V W
a sh sh
W H H


= +
0
HL
EA

(54)
the vertical height

+
(


+ + =
2 2 0
) ( 1 ) ( 1
H
W V
H
V
W
HL
f
0
1
( )
2
WL V
EA W


(55)
where

;
0 0 0
V qL P W P V
B
= + = + =
A B
H H H = =

(56)

If we let

W0 in Eqs.(54),(55), we get the solution of a straight cable:

0
0
cos
HL
a L
EA
o = +

(57)

0
0
sin
VL
f L
EA
o = +

(58)
where

V
tg
H
o =

(59)
Based on the data in Table 1, and letting a=300 in (54) and (57), we obtain the solutions for H. The substitution of H
into (55), (58) yields the height f:
Straight cable: H=6297.15172kN f=152.33445m
Sagged cable: H=6298.0001kN f=152.291842m
The results above are almost the same as in Table 1. The minute differences between them arose out of that the
different cable lengths are used in the calculation. The author believes that the extended length should be used in the
calculation of the cable's expansion because we usually use the extended length of the member in its stiffness. Indeed,
the difference between extended length and the original length of the cable is rather small, at most no more than

5
1860
0.9073%
2.05 10
E
o
c = = =


(60)
When a safety factor is considered, the extensionc of the cable in engineering is even smaller. Hence, in practice the
difference often is ignored. The author proposes that the straight model could be used to represent the flexible cable's
properties in the structure analysis. The effect of the cable's sag should be evaluated by the model in Fig.4 (b). Here
we keep the same P
0
in the global model of the structure as the one in the separated cable model in Fig.4 (b). The
differences of the element nodal forces between the global and the separated cable models are taken as extra loads to
- 8 -
http://www.ivypub.org/AE
be put on the same joints of the global structure in the iterative process. To a cable with large span, we should adopt
different elastic modulus according to loading increment. When the loading increment is small, we may adopt the
tangent elastic modulus, as Ernsts formula expressed. The results show that when the model in Fig.4 (b) is used to
modify the corresponding straight member in the entire model, the axial force, as in a cable stayed bridge deck, is a
little larger than unmodified. The modification has no effect on convergence of the iteration.
5 DETERMINATION OF IDEAL CURVES FOR ARCH BRIDGE
FIG. 5 MODELLING OF ARCH BRIDGE FIG. 6 SIDE SPAN
If we take the discrete elements with axial pressure as the analysis model of the arch bridge shown in Fig.5, the
model will behave as an unstable structure when the bending moment of each section vanishes
[4]
. There is no
solution to such a model in mathematics. If we turn upside down the arch bridge, and the pressing elements are
changed into stretching, then it is stable. There must exist a solution to such a problem. When the pressing model is
studied, we have to consider the flexural stiffness of the member. This implies that there must exist some bending
moment at each section. The moment can't vanish at all. If we think the arch bridge as an upside down cable, we can
easily find its exact solution by the ways given in the article. It can be proved that if we apply finite element method
to the arch bridge found by the method in the paper, we may find the bending moment at each section almost is zero.
The minute moment is thoroughly induced by angle restriction and axial deformation. There exists a huge horizontal
pushing force in the side span as shown in Fig.5. The side span can also be considered as an upside down cable, as
shown in Fig.6. The solution for Fig.6 can also be obtained by the way presented in the paper. Here we only need to
turn round the cable about line A-B from downside to upside. This is the ideal curve for arch bridge we are looking
for. If the iteration (30) can't converge, this means the horizontal pushing force H is too small. Then we should take
the iteration (26) to find out an appropriate H.

6 CONCLUSIONS
Through the solution of nodal equilibrium, and iterative solution of the whole system balance, the cable's solution
satisfies its balance equation, all boundary conditions and the attached length restriction. So, its solution is exact in
mathematics. From the opinion of error analysis, the iterative solution for cable is in fact to distribute the final
accumulated errors to the discrete points on the cable by adjusting the initial values. The solution is very stable.
According to the discussion in the paper, we know the cable is a hardening member, and the straight member is its
limit stage. When the load increment is large, or the load is put on the structure from zero, the modified elastic
modulus of the imitating straight member may be larger than its original value, because the positive displacement is
smaller than the straight member. If a small load increment is concerned, according to more detail analysis, we may
use the tangent stiffness of the sagged cable in the analysis, which is smaller than that of the straight member.
If we set up the discrete model in terms of pressing elements to look for the ideal curve of an arch bridge, we can't
get an exact solution in mathematics. If we consider the arch bridge as an upside down cable, we can get its exact
solution.
REFERENCES
[1] Yue-min Yang. Geometrically Nonlinear Analysis of Suspension Bridge and Properties of its Main Cable. WOLD SCI-TECH
R&D. 31(4) (2009): 711-717
[2] Yue-min Yang. Comparison of Nonlinear Equations Solutions in Mechanics and Their Modification. WOLD SCI-TECH R&D.
A
B
H
H
y
y
i
i
- 9 -
http://www.ivypub.org/AE
33(3) (2011): 376-379 (in Chinese)
[3] Ahmadi-Kashani K, Bell A J. The Analysis of Cables Subject to Uniformly Distributed Loads. Engineering Structures,10 (1988):
174-184
[4] Bathe, Klaus-Jrgen. Numerical Methods in Finite Element Analysis. New Jersey: Prentice-Hall, Inc. Englewood Cliffs, 1976
[5] David Kincaid, Ward Cheney. Numerical Analysis-Mathematics of Scientific Computing. Thomson Learning, 2002
[6] Irvine H M. Cable Structures. The MIT Press, 1981
[7] Stricklin J A, etc. Static Geometric and Material Nonlinear Analysis-Advances in Computational Methods and Design. Huntsville,
Alabama: UAH Press, The Univ. of Alabama, 1972
[8] Niels JGimsing. Cable Supported BridgesConcept & Design, (2nd Edition). John Wiley & Sons Ltd, 1997
[9] HUANG Yan, LAN Wei-ren. STATIC ANALYSIS OF CABLE STRUCTURE. Applied Mathematics and Mechanics (English
Edition). 2006, 27(10):1425-1430
[10] Huu-Tai Thai, Seung-Eock Kim. Nonlinear Static and Dynamic Analysis of Cable Structures. Finite Elements in Analysis and
Design. Volume 47, Issue 3, March 2011, 237-246
[11] Wang Chunjiang, etc. A New Catenary Cable Element. International Journal of Space Structures. Volume 18, Number 4/
December 2003, 269-275
[12] Hiroyuki Sugiyama, etc. A Non-Incremental Nonlinear Finite Element Solution for Cable Problems. ASME 2003 International
Design Engineering Technical Conferences and Computers and Information in Engineering Conference. DETC2003/VIB-48321,
171-181
AUTHORS
Yuemin Yang, born in 1957, earned his bachelor degree of solid mechanics in Xian Institute of Metallurgy and Architecture
Engineering, Xian, China, in 1981. He was awarded master degree of structure engineering in Chong-Qing Institute of Architecture
Engineering in 1985. Finally, he got his doctor degree of bridge engineering in China Academy of Railway Sciences, Beijing, China, in
1995. His research area covers applied mechanics, dynamics of structures, especially the vibration of bridges.
He is now a member of bridge designing in The First Highway Survey and Design Institute of China. His favorite jog is the
designing of arch bridges. The largest stone arch bridge of a span 146m in the world, Danhe Bridge, is his work. Besides,
During the study for his doctor degree in Beijing, he finished an experiment of coupled vibration between vehicle and bridge, the first
experiment in China. By now, he has published the papers: Geometrically Nonlinear Analysis of Suspension Bridge and Properties of
its Main CableComparison of Nonlinear Equations Solutions in Mechanics and Their Modification, and Linear Multistep Numerical
Solutions for Vibration Equation, etc, in the journal WORLD SCI-TECH R&D. Other articles were also published in other journals.
Dr. Yang is a leading member of Shannxi Mechanics Academy. He used to be the leader of Bridge Designing Office in the institute. The
article, Linear Multistep Numerical Solutions for Vibration Equation, is awarded the hot paper by the journal. Now he is interested in
numerical analysis of engineering problems.

You might also like