Mechanism and Machine Theory 80 (2014) 214229

Contents lists available at ScienceDirect

Mechanism and Machine Theory

journal homepage: www.elsevier.com/locate/mechmt

The initial value problem for the mathematical model of a

multiple-input multiple-output exible parallel mechanism
Yu Liu, Feng Gao
State Key Laboratory of Mechanical System and Vibration, Shanghai Jiao Tong University, Shanghai 200240, China

a r t i c l e

i n f o

Article history:
Received 19 March 2014
Received in revised form 28 April 2014
Accepted 2 May 2014
Available online 2 July 2014
Keywords:
FAST
Cable-net
Layer-wise substitution method
Longitude main cable
Latitude main cable
MIMO

a b s t r a c t
The cable-net support structure of the Five-hundred-meter Aperture Spherical Telescope
(FAST) is a multiple-input multiple-output (MIMO) flexible parallel mechanism. The
mathematical model of the cable-net structure is set up with the catenary equation, which
produces a large number of non-linear equations. Finding the initial value plays an important
role in the working state of the FAST. In this paper, a novel method is proposed to solve the
equations of the model. This method uses the solution of the (n 1)-th layer cable-net as the
initial value of the n-th layer cable-net. The n-layer cable-net equations are then solved
simultaneously and the tension of each cable and the coordinate of each node in the cable-net
can be obtained. With the introduction of the symmetry of cable-net, the number of unknown
parameters in the equations is reduced to about one-tenth of the whole cable-net. The initial
value of the cable-net support structure of the FAST is provided to show the validity of the
method. The results of this study are of practical value for the FAST's new type of flexible cable
driving mechanism design and control method.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
Based on approval from the National Development and Reform Commission, Chinese scientists have put forward an
innovation design for the largest single dish radio telescope [1,2], the Five-hundred-meter Aperture Spherical Telescope (FAST) in
a Karst depression in GuiZhou province [3]. A 3-D simulated image of the FAST is shown in Fig. 1. With the preliminary research
completed in the past few years, the FAST project is officially one of China's major scientific projects for the national 11th
Five-Year Plan. The main ring beam was built on the last day of 2013 as shown in Fig. 2.
The radius of the dish is 250 m and the main reflector, a spherical dish, has a radius of 300 m when it is in its tension state. A
paraboloid of revolution shaped dish is formed by adjusting the length of the driving cable with the diameter of opening of 300 m,
as shown in Fig. 3. The direction of the paraboloid of revolution is facing the tracking target when the observation starts. The feed
cabin can detect the signals when it moves on the focal surface under the control of the supporting cables. According to the
working principle of the FAST, the cable-net support structure of the FAST is a multiple-input multiple-output (MIMO) flexible
parallel mechanism. The input is the length of the driving cable and the output is the paraboloid of revolution on the spherical
dish when the FAST is in its working state.
The supporting structure of the main reflector surface is made of cables, as shown in Fig. 4. The lightweight cable-net structure
reflector for the FAST is a new driving mechanism. Because of its efficiency and esthetics, the cable-net structure has been widely
used in large-span roofs, bridges, and large-diameter antennae etc. [4,5]. Presently, there are some methods for solving the
cable-net structure problem. In some of the papers [6,7], the force density method was introduced and the main feature of this
Corresponding author at: Department of Mechanical Engineering, State Key Laboratory of Mechanical System and Vibration, Shanghai Jiao Tong University,
No. 800 Dong Chuan Road, Shanghai 200240, China.
E-mail address: fengg@sjtu.edu.cn (F. Gao).

http://dx.doi.org/10.1016/j.mechmachtheory.2014.05.003
0094-114X/ 2014 Elsevier Ltd. All rights reserved.

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

Active reflector
Main ring beam

Supporting cable
ed
Fe

215

in

cab

Cable net
Driving cable

Fig. 1. 3-D simulated model of the FAST.

method lies in prescribing a force density coefficient and a force-to-length ratio for each cable element. The dynamic relaxation
method is applied to the elastic-plastic bending of circular plates in the large deflection [810] and nonlinear finite element
methods [1114]. All of these methods only provide the approximate solution. Furthermore, the nonlinear finite element method
has been adopted to research the cable-net supporting structure of the FAST when it is under the tension state and working state
[1517]. By virtue of the method, the working state of the FAST is achieved by repeating the inverse iteration method to solve for
the length of the drag cables. In this paper, the cable-net supporting structure is treated as a MIMO flexible parallel mechanism.
The mathematical model of the cable-net supporting structure is built with the catenary equation and the working state of the
FAST can be obtained by solving the nonlinear equations in real time and therefore the length of the drag cables can be solved
through a single step calculation. In previous work [18], the catenary equation is used to solve a simple cable-net structure and
because the number of the catenary is smaller, it does not require the initial value of the catenary equations. The supporting
cable-net of the FAST, however, is a very a large structure, and the mathematical model has a large number of non-linear
equations. Therefore, the initial value is very important for solving the non-linear equations. A new layer-wise substitution
method is different from other known methods. The method proposed is based on the exact solution of the catenary equations.
The use of the analytic expression of the catenary to solve or formulate other complex problems has been used previously by
some scholars [1924]. The differences between the aforementioned models and herein proposed one are as follows: there are
many more parameters of the herein model than the aforementioned models, and furthermore, the whole cable-net is controlled
by the driving cables of each node.
The cable-net is a spherical dish when it is in the tension state. The type of mesh is a geodesic mesh and the distance between
two adjacent nodes is about 11 m. The position of each node can be solved in the global frame. According to the characteristics of
the cable-net structure, the cable-net is divided into 31 layers from the center point to its boundary. The division method of the
cable-net will be described in detail in the following section. Since the practical length of the main cable is unknown (the main
cable refers to the cable between two adjacent nodes), the chord length between two adjacent nodes on the spherical dish can be
seen as the main cable length. Fig. 4 illustrates the cable-net support structure of the FAST.

Main ring beam

Fig. 2. Main ring beam of the FAST.

216

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

Fig. 3. The optical geometry of the FAST.

2. The formula of a cable in the local frame

The initial value of the cable-net is given by the weight of the reflectors and the force of the driving cables when it is at the
tension state under the gravity. In the tension state, the cable-net structure and each main cable under gravity follow the principle
of minimum potential energy. It is well known that the catenary is the equilibrium shape of a cable that is hanged between two
fixed points under its own weight. However, there are some different expressions about the catenary equation [25,26], but in
most cases the y-axis must pass through the lowest point of the curve that is symmetric with respect to the y-axis. In this paper
the local frame of the catenary is arbitrary. Fig. 5 shows a single three dimensional cable in the global frame OXYZ. Using the local
frame, oxyz, and coupled with the principle of minimum potential energy and the variation principle, the catenary equation can
be written as follows [27].
8
>
>
<

1
0
1
0

2
mg xc
1 @ 1
@
A
z
c cosh
A
mg
c1
>
>
:
y0

where m is the mass of the cable, g is the acceleration due to gravity, c1 is the cable shape parameter, and c2 and are the location
parameters.
Letting TA and TB be the tension at two endpoints, the equilibrium equation of the cable can be expressed by
(

T y T y mg 0
A

T x T x 0
where
(

T x T cos i
i A; B:
i
i
T y T sin i

Boundary support structure

Cable-net

Driving cable
Fig. 4. The cable-net structure.

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

217

Fig. 5. The catenary in the local and global frame.

The tangent slope at the two endpoints is

1
0

2
mg xi c
0
A i A; B
tan i zi sinh@
c1

and
1
0

8
>
mg xi c2
>
>
A
>
sin i tanh@
>
>
<
c1
1 i A; B:
0

>
>
mg xi c2
>
>
@
A
>
>
: cos i sech
c1

The components of the tension at the cable endpoints are

1
0

8
2
>
mg
x
c
>
i
i
i
>
A
>
T T  sech@
>
>
< x
c1
1 i A; B:
0

>
>
mg xi c2
> i
i
>
@
A
>
>
: T y T  tanh
c1

Thus, the tension at the cable endpoints can be expressed by

1
0

mg xi c2
A
cosh@

c1
i A; B
i
1


1
T mg
:
0

0
j B; A
mg x j c2
mg xi c2
A sinh@
A
sinh@
c1
c1

The cable length, L, is computed as

L l0 le lt

where l0 is the unstressed cable length, le is the increase of the length due to elastic deformation, and lt is the length increase
due to temperature. The increase of the length due to elastic deformation is given by
Z
le

T x
dx
L EA

218

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

where T(x) is the tension at each point of the cable which can be computed according to Eq. (8) and EA is the axial stiffness. After
some algebra, Eq. (9) can be rewritten as
0

1
1
0

mg xB c2
B
C
A
2 sinh@
B
C
B
C
c1
c1
B
1


1C
ln B
le
C:
0

0
2
2
C
EA B
mg xB c
mg xA c
B
C
A sinh@
AA
@ sinh@
1
1
c
c

10

It should be noted that in some applications the strain on the cables can be ignored, but the variation of length due to elasticity
should still be considered.
The increased length of the cable due to the temperature is written as
Z
lt

xdx

11

where is the coefficient of thermal expression and (x) is the temperature equation along the cable. It is assumed that the
temperature distribution along the cable is known or computed by other methods. However, a realistic assumption is to consider
both the coefficient of thermal expansion and the temperature constant along the cable. Under these circumstances, the increase
of the length due to temperature is
lt L

12

where L is the cable length given by Eq. (8).

All the above equations of the cable describe the behavior of the cable under its own weight in the local frame. Moreover, the
position of the single cable should be referred to in the global frame. The coordinate value of the cable in the local coordinate can
be described in the global frame using the coordinate transformation matrix
XRx

13

where X is the coordinate value in the global coordinate, x is the coordinate value in the local coordinate, and R is the coordinate
transformation matrix which transforms the vectors from local frame, oxyz, to the global frame, OXYZ.
2
2 3
X
cos
6 sin
6Y 7
6
6
7
X4 5 R4
Z
0
1
0

sin
cos
0
0

2 3
3
x
0 x1
6 7
0 y1 7
7X 6 0 7
4z5
1
0 5
1
0
1

14

Fig. 6. The model of the cable-net structure.

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

219

where is the angle between the x-axes of the local frame and global frame. According to the geometric relationship, the angle
is obtained
0

Y B Y A
B
C
arcsin@q
A
X B X A 2 Y B Y A 2

15

where XA,XB,YA and YB are the coordinate values of the endpoints of the cable in the global frame.
3. The catenary equations of the main cable in three dimensional cable-net structure
Fig. 6 illustrates the three dimensional cable-net structure subject to gravity, point loads, and the force of driving cables.
A complex cable-net has many cables with different unknown parameters that determine the shape of each cable. The cables
are linked together satisfying the following conditions. First, the main cables and nodes should satisfy static equilibrium. Second,
the cable-net structure under the action of some constraints should be added. Finally, the boundary of the cable-net structure
should be fixed. Furthermore, the global frame is established on the dish with the origin at the center. The origin of the local frame
is the projection point of the node on the XY plane of the global frame and the projection of a main cable on the XY plane of global
system is the xk axis of the local frame.
According to Eqs. (1) and (7), the number of unknown parameters of a main cable is determined to be five. Because the
position of the nodes in the cable-net structure is unknown, one node will bring three unknown parameters. Therefore, a more
complex cable-net structure requires establishing more nonlinear equations.
Let Lk be a main cable between two adjacent nodes, pi and pj, on the cable-net structure where the point, ok, is the projection of
the node, pi, on the XY plane of the global frame and also the origin of the local frame. The projection of the main cable, Lk, on the
XY plane is the x-axis of the local frame. The endpoints, A and B, of the main cable, Lk, meet the nodes, pi and pj, respectively. The
equations of the main cable at two endpoints can be expressed as


1
0
0
1
mk g 0c2k
1 @ 1
Ak A
zA
ck cosh@
mk g
c1k


1
0
0
1
mk g xB c2k
1 @ 1
@
A
ck cosh
k A:
zB
mk g
c1k

16

17

The cable length, Lk, can be obtained by means of the definite integral
Z
Lk

xB q
1 z2i dX
0

1


11
0
0
0
mk g xB c2k
mk g 0c2k
c1k @
A sinh@
AA:
sinh@

mk g
c1k
c1k

18

The explicit form of Eq. (9) applied to the cable length Lk is

0

1


1
0
mk g xB c2k
B
C
A
2 sinh@
B
C
1
1
B
C
c
k
c
B


1


1C
Lk lk k ln B
C Lk
0
0
2
2
C
EAk B
mk g xB ck
mi g 0ck
B
C
A sinh@
AA
@ sinh@
c1k
c1k
and Eqs. (18) and (19) are equal.
The equilibrium equations of the main cable Lk are


1


1
0
0
2
2
m
g
0c
m
g
x
c
k
k
k
k
B
A
A T Bk  tanh@
Amk g 0
T k  tanh@
c1k
c1k

A
Tk

1


1
0
0
2
2
mk g 0ck
mk g xB ck
B
@
A
@
A0
T k  sech
 sech
c1k
c1k

where xB

q
X B X A 2 Y B Y A 2 , XA, XB, YA and YB are the coordinate values in the global frame.

19

20

21

220

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

Generally, each node of the cable-net connects five or more main cables and they reach a state of equilibrium with the tension
of each main cable. The force analysis of the node is in the global frame, while the equilibrium of each main cable is solved in the
local frame. Therefore, when solving the equilibrium equations, the tension at the cable endpoints should be transformed into the
global frame.
The nodes of the cable-net structure are connected to each other by a main cable in order to ensure the uniformity of the local
coordinates on each node. The regulations are as follows: the projection of the central node of the cable-net on the plane is the
origin of five local coordinates; the node on the line, p0Di, is set up for four origins of local coordinates where others are set up
three origins of local coordinates. In Fig. 6, the point, pi, is a node of the cable-net structure with six arbitrary distributions of
forces defined in the global frame, OXYZ. Additionally, six cables and one driving cable are joined at the node, pi. The equilibrium
of forces at the node gives
X

i;m;n;u

i;m;n;u

T z

v;w

X
v;w

T F i 0 x; y

T z F iz mpi g 0

22

23

where TA is the tension of the cable and its origin of the local frame is the projection of the node, pi, on the XY plane of the global
system, and the TB is not. T( = x, y) are the components of the tension from the six cables at the node, pi, in the global system.
mpig is the force of gravity at the node, pi. Fi is the force of the driving cable at the node, pi. The components of the tension at the
coordinate axis can be expressed as
X Mi X pi
F ix F i r


2

2

2
X Mi X pi Y Mi Y pi Z Mi Z pi

24

Y Mi Y pi
F iy F i r


2

2

2
X Mi X pi Y Mi Y pi Z Mi Z pi

25

Z Mi Z pi
F iz F i r


2

2

2
X Mi X pi Y Mi Y pi Z Mi Z pi

26

where X Mi ,Y Mi , and Z Mi are the coordinate values of the fixed point of the driving cable in the global frame, which are known. X pi , Y pi ,
and Z pi are the coordinate values of the node, pi, in the global frame which should satisfy the equation of the spherical dish. The
equation of the spherical dish in the global coordinate is given by
2

X Y ZOO0 R

27

where OO0 is the distance between the spherical center and the origin of the global frame, as shown in Fig. 6.
The components of the tension, TA and TB, can be obtained by transforming the tension of the six cables from the local frame,
okxkykzk, to the global frame, OXYZ. The components of the cable in the local coordinate system should be obtained before it is in
the global frame. The upper right-hand portion of Fig. 6 shows the relationship between the tangent direction of the tension of a
cable and the coordinate axis of the local system. The component of the tension of a cable in the local system can be written
as
2

3 2
3
A
A
T kx
T cos kA
6 A 7 4 k
5
T k 4 T ky 5
0
A
A
T k sin kA
T

28

kz

where kA is the angle between the tangent of the cable endpoint and the x-axis in the local system, shown in the upper
right-hand portion of Fig. 6. The component of the tension of the cable in the global system can be solved by the rotation
matrix
T Ri Tk

29

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

221

where T = [Tkx, Tky, Tkz] ' contains the components of the tension of the cable in the global system and the rotation matrix Rk is
defined as
2

Rk

cosk
4 sink
0

sink
cosk
0

3
0
0 5:
1

30

According to Eqs. (17)(23) and (27), five equations can be set up for one main cable, three equations can be set up for one
node, and an additional one equation can be set up for one driving cable. Therefore, the number of the unknowns and the
equations is equal and the solution of all of the equations is unique. The cable-net structure is very large and complex, so a large
number of nonlinear equations need to be solved. The key problem in solving the nonlinear equations is determining an
appropriate initial value.
4. Layer division of the cable-net structure
The cable-net structure of the FAST is equipped with more than 2000 nodes and driving cables and 6000 main cables, resulting
in more than 40,000 nonlinear equations of the cable-net structure. This large amount of equations cannot be solved because of
the unobtainable initial value. According to the characteristics of the cable-net structure, a layer-wise substitution method is
proposed. With this method, the initial value of the equations can be solved step by step.
The cable-net is divided into 31 layers from the center node, p11, to the boundary, where each layer includes main cables and
nodes, and the outer three layers belong to the boundary layer, as shown in Fig. 7. For example, the first layer is the pentagon,
p12p22p32p42p52, which includes the main cables, p11p12, p11p22, p11p32, p11p42, p11p52, and the node, p11, as shown in Fig. 8a. The second layer is
1 2
2 3
3 4
5 1
located in the curved pentagons, p12p22p32p42p52 and p13p23p33p43p53p63p73p83p93p10
3 , which include the main cables, p2p2, p2p2, p2p2, p2p2 and
1 10
1 1
1 2
2 2
2 3
2 4
3 4
3 5
3 6
4 6
4 7
4 8
5 8
5 9
5 10
1
2
3
4
5
6
p2p3 , p2p3, p2p3, p2p3, p2p3, p2p3, p2p3, p2p3, p2p3, p2p3, p2p3, p2p3, p2p3, p2p3, p2p3 , and the nodes, p2, p2, p2, p2, p2, p2, as shown in
Fig. 8b. The following layers are similar. Note that the subscript letters and superscript letters in Fig. 8 refer to the layer number
and the node number in the same layer, respectively.
The main cable of each layer is divided into two types: longitude main cables and latitude main cables. A main cable on the
concentric curved pentagon is called a latitude main cable, such as the main cables,p12p22, p22p32, p32p42, p52p12, in Fig. 8b. A main cable
1 1
1 2
2 2
2 3
connected to two adjacent curved pentagons is called a longitude main cable, such as the main cables, p12p10
3 , p2p3, p2p3, p2p3, p2p3,
p22p43, p32p43, p32p53, p32p63, p42p63, p42p73, p42p83, p52p83, p52p93, p52p10
,
in
Fig.
8b.
Each
layer
of
the
cable-net
structure
includes
the
longitude
main
3
cables and the latitude main cables as shown in Figs. 7 and 8. It should be noted that each node is connected to one driving cable.
5. The layer-wise substitution method
The layer-wise substitution method is based on the similarity of the main cables with a similar shape parameter, which means
that the initial values for the equations of latitude or longitude main cables are similar. The first layer of the cable-net structure
can be solved where the initial values come from the solution of one cable. Where the point, pi2(i = 1, 2, , 5), is the boundary
node and the position coordinate of the node, p11, is unknown, as shown in Fig. 8a. After the solution of the first layer is obtained,

Fig 7. Division of the layers of the cable-net structure.

222

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

Fig. 8. Each layer of cable-net structure.

the initial value of the second layer can be obtained based on the solution of the first layer. The two layer cable-net structure can
be solved together, where the point, pi3(i = 1, 2, , 10), is the boundary node and the position coordinates of the node, p11 and
pi2(i = 1, 2, , 5), are unknown, as shown in Fig. 8b. In the same way, the whole cable-net structure can be solved step by step.
The layer-wise substitution method can be expressed with the following formulas. The solution and equations of the main
cables and node of the first layer are given by



1
1i 2i
i
Ai
Bi
f 1 x f c1 ; c1 ; 1 ; T 1 ; T 1 ; F 1
1

g 1 x f x1 ; y1 ; z1

31
32

2i
i
Ai
Bi
1 1
1 2
1 3
1 4
1 5
where c1i
1 , c1 , 1, T1 , T1 are the solutions of the equations of the main cables, p1p2, p1p2, p1p2, p1p2, p1p2, F1 is the force of the
driving cable, p11M1, and x1, y1, z1 are the solutions of the equations of the node, p11. Here the superscript letter, i(i = 1, 2, 3, 4, 5),
refers to the cable number in the first layer.
After the first layer cable-net structure is solved, the initial values for the second layer are gained through the solutions of the
first layer. In Fig. 8b, the equations of the two-layer cable-net structure are listed as

F 2 x

f 1 x; f 2 x

n
o
1
2
G2 x g 1 x; g 2 x

33
34

where f22(x) and g22(x) are the equations of the main cables and nodes of the second layer and they can be written as



2
1j 2j
j
Aj
Bj
m
f 2 xf c1 ; c1 ; 1 ; T 1 ; T 1 ; F 1

35

 m m m
2
g 2 xf x2 ; y2 ; z2

36

2j
j
Aj
Bj
1 10 1 1 1 2 1 2
5 10 5 1
m
where c1j
1 , c1 , 1, T1 , T1 are the initial values of the equations of the main cables, p2p3 p2p3 p2p3 p2p2 p2p3 p2p2, and F1 is the
m
m
m
initial value of the driving cables, Mmpm
2 . All of the above initial values come from the solution of the first layer, and x2 , y2 , z2 are
1
2
3
4
5
the initial values of the equations of the nodes, p2, p2, p2, p2 and p2, which come from the boundary position coordinate of the first
layer. The initial values of f11(x) and g11(x) are the solutions of the equations in the one layer cable-net structure. Here the
superscript letters, j( j = 1, 2, , 9, 10), and the superscript letters, m(m = 1,2,3,4,5), refer to the number of nodes of the second
layer, respectively. Therefore, Eqs. (33) and (34) can be solved.

F 2 x

f 2 x; f 2 x

n
o
1
2
G2 x g 2 x; g 2 x

37
38

The equations and solutions of a two-layer cable-net structure can be expressed as


1
1i 2i
i
Ai
Bi
f 2 x f c2 ; c2 ; 2 ; T 2 ; T 2 ; F 2

39

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

223


1
1 1 1
g 2 x f x2 ; y2 ; z2

40


2
1j 2j
j
Aj
Bj
m
f 2 x f c2 ; c2 ; 2 ; T 2 ; T 2 ; F 2

41

 m m m
2
g 2 x f x2 ; y2 ; z2

42

2i
i
Ai
Bi
1
1
1
where c1i
2 , c2 , 2, T2 , T2 , F2 and x2, y2, z2 are the solutions of the main cables, driving cable, and the node of the first layer, and
2j
j
Aj
Bj
m
m
m
m
c1j
2 , c2 , 2, T2 , T2 , F2 and x2 , y2 , z2 are the solutions of the main cables, driving cables, and nodes of the second layer.
It is worth noting that for the mesh generation of a cable-net by the geodesic method, the length of the main cable is the line
distance of the adjacent nodes. The coordinate value of each node in the global coordinates can be obtained and the initial value of
the nodes of each cable-net layer and the coordinate value of the boundary of the cable-net are given by reference to the
coordinate value.
The layer-wise substitution method is employed for solving the two-layer cable-net structure as described above. Then the
general formula can be used to describe the procedure for solving the entire cable-net structure. Because the initial value of n-th
layer of the cable-net comes from the solutions of (n 1)-th layer, the solutions of the equations of (n 1)-layer cable-net
structure must be completed prior to solving the equations of n-layer cable-net structure. The equations of (n 1)-layer
cable-net structure can be expressed as

F n1 x

n1

f n1 x; f n1 x f n1 x

43

n
o
1
2
n1
Gn1 x g n1 x; g n1 xg n1 x

44

where f rn 1(x) and grn 1(x) are the equations of the main cables and nodes of r-th layer cable-net structure and they can be
expressed as



r
1i
2i i
Ai
Bi
m
f n1 x f cn1 ; cn1 ; T n1 ; T n1 ; F n1

45

 m
r
m
m 
g n1 x f xn1 ; yn1 ; zn1

46

i i
2i
Ai
Bi
m
m
m
m
where c1i
n 1, cn 1, n 1, Tn 1, Tn 1, Fn 1 and xn 1, yn 1, zn 1 are the solutions for the main cables, driving cables, and
nodes of the r-th layer. Here the superscript letters and the superscript letters refer to the number of cables and nodes of the r-th
layer, respectively.
The initial value of the equations of the n-layer cable-net structure can be gained from the solutions of (n 1)-layer cable-net
structure. The equations of the n-layer cable-net structure are given by

o
1
2
n1
n
f n1 x; f n1 x; ; f n1 x; f n x

47

n
o
1
2
n1
n
Gn x g n1 x; g n1 x; ; g n1 x; g n x

48

F n x

Fig. 9. Symmetric cable-net.

224

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

where fnn(x) and gnn(x) are the equations of the main cables and nodes of the (n 1)-th layer cable-net and they can be written
as



n
2i
2i
i
Ai
Bi
m
f n xf cn1 ; cn1 ; n1 ; T n1 ; T n1 ; F n

49

 m m m
n
g n xf xn ; yn ; zn

50

2i
i
Ai
Bi
m
where c2i
n 1, cn 1, n 1, Tn 1, Tn 1, Fn are the initial values of the equations for the main cables and driving cables, and
m
m
m
xn , yn , zn are the initial values of the equations of the nodes. The initial values of fin 1(x) and gin 1(x) are the solutions of the
equations in the (n 1)-layer cable-net structure. With the initial values, Eqs. (49) and (50) can be solved.

6. The symmetry of the cable-net structure

The main characteristic of the cable-net structure of the FAST is rotational symmetry. As shown in Fig. 7, the cable-net is
divided into five parts: B1p11B2, B2p11B3, B3p11B4, B4p11B5 and B5p11B1, and any four parts can be obtained from the rest of the parts by
using a rotation transformation. In other words, the four parts of the cable-net: B2p11B3, B3p11B4, B4p11B5 and B5p11B1 can be obtained
from the B1p11B2 using a rotation transformation.
Another important characteristic of the cable-net structure is reflection symmetry, as shown in Fig. 9a. Part of the cable-net,
B1p11B2, is the asymmetric structure and the plane, Op11A, is the symmetry plane. Half of the cable-net, B1p11B2, can be expressed as
shown in Fig. 9b. For simplicity, part of the cable-net, B1p11A, is defined as the original symmetric cable-net. The nodes and the
main cables in the cable-net, B1p11A, are the original symmetric nodes and original symmetric main cables. The other cable-nets
such as B2p11B3, B3p11B4, B4p11B5, B5p11B1 and Ap11B2 are the symmetric cable-nets. The nodes and the main cables in the symmetric
cable-net are the symmetric nodes and symmetric main cables, which can be obtained by a rotation transformation or the
symmetry transformation of the original symmetric nodes and original symmetric main cables.
Due to the high coupling of the cable-net, the force analysis of the original symmetric cable-net involves other parts of the
cable-net. The boundary nodes of the original symmetric cable-net include some symmetric main cables and some originally
symmetric main cables. For example, the equilibrium equation of node, p11, not only includes the tension of the original symmetric
main cable, p11p12, but also the tension of the other symmetric main cables, p11p22, p11p32, p11p42 and p11p52. The equilibrium equations of
the boundary nodes on the symmetric planes, op11p1n and op11A, cannot only be solved by referring to the original symmetric
cable-net, B1p11A. These boundary nodes connect some other symmetric main cables. Fig. 10 shows the solution model including
the original symmetric cable-net and some symmetric main cables. The solid and dashed lines of the main cables are the originally
symmetric main cables and the symmetric main cables, respectively, and the parameters of the symmetric main cables come from
the corresponding originally symmetric main cables. The nodes with solid and virtual points are the original symmetric nodes and

Fig. 10. Solving model of the original symmetric cable-net.

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

225

Table 1
The solution of the node p11 and main cable p11p12 in different layers of the cable-net.
n

n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n
n

Parameters

=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

x/m

y/m

z/m

c1

c2

TA/N

TB/N

4.4e15
5.5e15
8.3e16
3.1e16
9.8e11
2.1e15
2.7e14
1.1e15
4.6e15
1.3e14
5.3e16
1.2e14
3.7e14
5.5e16
3.1e15
1.5e15
5.2e15
1.7e15
1.9e16
1.3e14
1.3e14
6.5e15
2.7e15
6.4e15
4.7e14
5.1e14
1.1e16
2.0e15
6.0e15
8.1e15
1.4e15

3.7e15
1.1e15
1.7e15
5.5e15
1.1e15
1.9e14
4.7e14
1.2e15
1.0e14
4.014
2.1e15
1.3e14
9.4e15
8.6e15
2.1e15
2.0e15
1.4e15
2.3e14
2.4e15
2.3e15
8. 8e15
5.4e15
1.9e15
5.3e15
1.7e14
9.2e15
3.7e14
6.3e14
1.2e15
1.7e14
1.0e14

134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699
134.1699

1,665,899.0
1,385,520.7
1,325,572.25
1,315,035.0
1,314,849.59
1,312,190.39
1,310,805.76
1,311,211.55
1,311,351.34
1,311,409.6
1,311,453.28
1,311,499.8
1,311,540.29
1,311,576.42
1,311,603.3
1,311,628.51
1,311,651.74
1,311,670.11
1,311,690.10
1,311,710.67
1,311,732.48
1,311,756.4
1,311,783.67
1,311,817.32
1,311,858.20
1,311,910.75
1,311,977.98
1,312,066.21
1,312,181.9
1,312,335.06
1,312,535.74

919,773.731
1,029,311.55
969,314.896
958,730.192
958,525.885
955,842.851
954,439.446
954,832.495
954,965.953
955,013.189
955,045.486
955,082.055
955,114.014
955,142.038
955,159.655
955,174.96
955,188.259
955,191.508
955,196.042
955,201.89
955,206.27
955,208.871
955,211.274
955,212.922
955,214.083
955,214.679
955,215.063
955,215.228
955,215.31
955,215.372
955,215.479

27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997
27.997

12,619.3
9917.510
9339.468
9237.489
9235.521
9209.671
9196.150
9199.937
9201.222
9201.677
9201.989
9202.341
9202.649
9202.919
9203.089
9203.236
9203.364
9203.395
9203.439
9203.496
9203.538
9203.563
9203.586
9203.601
9203.613
9203.619
9203.622
9203.624
9203.625
9203.625
9203.626

12,623.6
9922.977
9345.274
9243.359
9241.391
9215.558
9202.046
9205.830
9207.110
9207.570
9207.881
9208.233
9208.541
9208.811
9208.980
9209.128
9209.256
9209.287
9209.331
9209.387
9209.429
9209.454
9209.477
9209.493
9209.504
9209.510
9209.514
9209.515
9209.516
9209.517
9209.518

symmetric nodes, respectively. The position coordinates of the boundary nodes come from the corresponding originally symmetric nodes by the coordinate transformation.
In the cable-net structure, the numbers of the nodes, driving cables, and main cables are 2326, 2326, and 7140, respectively.
According to these numbers, the number of unknown parameters is 45,004 (3 2326 + 5 7140 + 2326), and the number of
the equations is also 45,004. After the symmetry of the cable-net structure is adopted, the numbers of the nodes, driving cables,
and main cables are 272,272, and 748, respectively. Therefore, the number of unknown parameters is then 4828
(3 272 + 5 748 + 272). After the simplification of the symmetry of the cable-net, the number of the equations is greatly
reduced.
7. Example
The average initial length of the main cables is about 11 m and the average initial length of the driving cables is about 3 m. Other
properties of the cables are the linear density of 2.74 kg/m, the diameter of 28 mm, and Young's modulus Es = 132 kN/mm2. The
quality of the tension of the nodes is 7.8 kg, the cables of the structure have a coefficient of thermal expansion of = 7e106 N/m,
and the ambient temperature is = 25 C [28]. The whole cable-net structure can be solved with the layer-wise substitution method.
Due to limitations, the only solution of the main cable, p11p12, in different layers is given. As shown in Table 1, increasing the
number of layers results in similarity in the parameters of the main cable, p11p12. The results of the whole cable-net at the tension
state are shown in Fig. 11.
8. Conclusion
As a multiple-input multiple-output (MIMO) flexible parallel mechanism, the cable-net structure of the active reflector for the
FAST has been explored in this paper. The mathematical model of the cable-net structure is established by the catenary equations.
In the cable-net structure the catenary equations of the main cables are expressed in the local coordinates and the equilibrium
equations of the nodes are established in the global frame. The tensions of the main cables are transformed from the local frame to
the global frame with a transformation matrix of coordinates when the static force equilibrium analysis of the nodes is conducted.
Based on the characteristics of the cable-net structure, the layer-wise substitution method has been proposed to deal with the

226

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

initial value of the catenary equations. This method solves the cable-net problem from the first layer to the last layer and solves all
of the cable-net layers one by one. In other words, this method uses the solution of the (n 1)-th layer cable-net equations as the
initial value of the n-th layer, allowing all of the n-th layer cable-net equations to be solved simultaneously. In order to reduce the

n=1

n=5

n=10
Fig. 11. The cable-net in different layers.

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

n=15

n=20

n=25
Fig. 11 (continued).

227

228

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

n=30

n=31
Fig. 11 (continued).

number of equations, the symmetry of the cable-net structure is introduced. With the layer-wise substitution method and the
symmetry of the cable-net structure, the initial value of the mathematical model can be achieved. The results provide an
important reference for the initial value for the control of the FAST while it is in its working state. Finally, the solution process of
the cable-net structure is shown. The solution of one main cable and one node has been given, which verifies that the method is
correct and efficient.

Acknowledgments
This work is partially supported by the Research Fund of State Key Lab of MSV, China (Grant No. MSV-ZD-2010-02), which is
gratefully acknowledged.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.mechmachtheory.2014.05.003.

Y. Liu, F. Gao / Mechanism and Machine Theory 80 (2014) 214229

229

References
[1] R.D. Nan, G. Ren, W.B. Zhu, Y.J. Lu, Adaptive cable-mesh reflector for the FAST, Acta Astron. Sin. 44 (2) (2003) 1318.
[2] B.Y. Duan Duan, A new design project of the line feed structure for large spherical radio telescope and its nonlinear dynamic analysis, Mechatronics (1999)
5364.
[3] Y. Qiu, A novel design for giant radio telescopes with an active spherical main reflector, China Astrophys. 22 (3) (1998) 361368.
[4] B. Peng, R. Nan, Modeling FAST, the world's largest single dish, Radio Sci. Bull. 300 (2002) 1220.
[5] T. Hiroaki, Design optimization studies for large-scale contoured beam deployable satellite antennas, Acta Astronaut. 58 (2006) 443451.
[6] H. Schek, The force density method for form finding and computation of general networks, Comput. Methods Appl. Mech. Eng. (1974) 115134.
[7] J.Y. Zhang, M. Ohsaki, Adaptive force density method for form-finding problem of tensegrity structures, Solids Struct. 43 (1819) (2006) 56585673.
[8] A.S. Day, J.H. Bunce, Analysis of cable networks by dynamic relaxation, Civ. Eng. Pub. Work. Rev. (1970) 383386.
[9] H. Sang-Eul, L. Kyoung-Su, A study of the stabilizing process of unstable structures by dynamic relaxation method, Comput. Struct. 81 (17) (2003)
16771688.
[10] D. Bemd, F. Etienne, L. Vikram, Combining dynamic relaxation method with artificial neural networks to enhance simulation of tensegrity structures, Struct.
Eng. 129 (5) (2003) 672681.
[11] H.B. Jayaraman, W.C. Knudson, A curved element for the analysis of cable structures, Comput. Struct. 14 (34) (1981) 325333.
[12] W.J. Lewis, M.S. Jones, Dynamic relaxation analysis of the non-linear static response of pretensioned cable roofs, Comput. Struct. 18 (6) (1984) 989997.
[13] R. Karoumi, Some modeling aspects in the nonlinear finite element analysis of cable supported bridges, Comput. Struct. 71 (4) (1999) 397412.
[14] K. Imaia, D.M. Frangopol, Geometrically nonlinear finite element reliability analysis of structural systems. I: theory, Comput. Struct. 77 (6) (2000) 677691.
[15] Q.H. Liang, Theoretical and experimental research on supporting structure of FAST reflector, 2010. ((PhD thesis) China).
[16] X.F. Jin, Study of key issues of the structural safety and accuracy control of the Five-hundred-meter aperture radio telescope, 2007. ((PhD thesis) China).
[17] F. Fan, X.F. Jin, S.Z. Shen, Effect of non-uniform solar temperature field on cable-net structure of reflector of large radio telescope-FAST, Adv. Struct. Eng.
(2009) 503512.
[18] S. Ming, R. Jesus, An approach based on the catenary equation to deal with static analysis of three dimensional cable structures, Eng. Struct. 31 (9) (2009)
21622170.
[19] A.C. Pipkin, Catenary deformations of inextensible networks, J. Eng. Math. 28 (1993) 401406.
[20] Y. Suzuki, E. Miyata, S. Iverson, Static analyses of the triangular running skyline system: a three-dimensionally movable logging cable system, Comput.
Struct. 60 (4) (1996) 579592.
[21] D.A. Kiureghian, J. Sackman, Tangent geometric stiffness of inclined cables under self-weight, J. Struct. Eng. 131 (6) (2005) 941945.
[22] A. Andreu, L. Gil, P. Roca, A new deformable catenary element for the analysis of cable net structures, Comput. Struct. 84 (2930) (2006) 18821890.
[23] O. Lopez-Garciaa, A. Carniceroa, V. Torresb, Computation of the initial equilibrium of railway overheads based on the catenary equation, Comput. Struct. 28
(10) (2006) 13871394.
[24] Y. Huang, W. Lan, Static analysis of cable structure, Springer Appl. Math. Mech. 27 (10) (2006) 14251430.
[25] H. West, A. Kar, Discredited initial values analysis of cable nets, Solids Struct. 9 (1973) 14031420.
[26] A. Igo, A. Griv, Case studies optimization: catenary problem, Optim. Eng. 6 (2007) 463482.
[27] Y. Liu, F. Gao, The study of the initial state of the cable-net structure reflector for the FAST, ASME Paper No. DETC2012-70310, 2012, (pp.).
[28] A. Peyrot, A. Coulois, Analysis of cable structures, Comput. Struct. 10 (5) (1979) 805813.