Engineering Structures 167 (2018) 431–444

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

A novel cable element for nonlinear thermo-elastic analysis T

⁎
M. Rezaiee-Pajand , M. Mokhtari, Amir R. Masoodi
Department of Civil Engineering, Ferdowsi University of Mashhad, Iran

A R T I C LE I N FO A B S T R A C T

Keywords: The exact solution of inextensible catenaries in Cartesian coordinates is utilized to propose an efficient two-node
Three-dimensional cable element cable element for static analysis of three-dimensional cable structures. This element can consider out of plane
Elastic catenary inclination without using any transformation matrices. Since the element is formulated within the framework of
Large sag large curvature assumption, cables with large sag, as encountered in long-span cable-stayed bridges and sus-
Pretension
pension bridges, can be modeled accurately. The proposed element also accounts for the thermal effects. By
Thermal loading
defining the stiffness component as the ratio of infinitesimal load increment to infinitesimal increase in length,
explicit entries of the tangent stiffness matrix are derived through equating the total differentiation of the
strained length and the elastic elongation of the cable. The tangent stiffness matrix is available in a closed form
and the need of taking the inverse of the flexibility matrix, which is faced in the solution procedure of elastic
catenary, is eliminated. The robustness of the suggested technique is established through investigation of sig-
nificant case studies, including slack and pre-tensioned spatial cable networks. Excellent agreement between the
present results and those found in the literature indicates the versatility of the proposed scheme.

1. Introduction method provided satisfactory results in some cases, it was rather in-
efficient since a large number of truss elements was required to achieve
Over the past two centuries, analysis and design of cable-supported an acceptable level of accuracy. Later, Knudson embarked on the im-
structures have received huge attention as a crucial topic in the main- provement of this method in 1971 [2]. Various researchers have further
stream of scientific research. Owing to their unique mechanical and developed the truss element by introducing the nonlinear behavior and
aesthetic features, cables are widely applied as constituent parts of various loading conditions [3,4]. Besides, different types of two-node
many engineering structures, such as, suspension roofs, long-span sus- elements with rotational degrees of freedom have been proposed by
pension bridges, cable supported bridges and power transmission lines. several researchers [5–7]. The cable members have been also modeled
Cables are flexible members that exhibit highly nonlinear behavior based on the isogeometric approach with Lagrangian shape functions.
when subjected to external loads. This structure, within a cable-sup- In this method, the shape of an infinitesimal cable element is approxi-
ported body, undergoes large displacements and rotations and sustains mated using multi-node curved elements [8,9]. Wu and Su im-
significant portions of load. Pretension is proposed as a simple tech- plemented a Four-node isogeometric element for analysis of cable
nique to alleviate the deflection of cable structures. Numerous studies structures [10]. In 2013, a six-node isogeometric element was proposed
can be found in the literature addressing various schemes for in- by Wang et al. [11]. The main drawback of the isogeometric elements in
vestigation of the behavior of cable structures. In fact, the cable modeling of cable assemblies is their complexity and large number of
members have been widely modeled, based on two different ap- degrees of freedom. This makes the analysis laborious and significantly
proaches, namely the finite element method with interpolation func- time consuming. Further, since the explicit form of the tangent stiffness
tions and also the analytical approach which makes use of explicit ex- matrix is not available, numerical approaches must be iteratively
pressions of a catenary. adopted to derive the tangent stiffness matrix. In some cases, such
In the first scheme, a cable is represented by a number of two-node, analysis approaches lead to the numerical instabilities [12].
multi-node or generally curved elements. The displacement field within On the other hand, an element based on the analytical expressions
the element domain is approximated using the interpolation functions. of the elastic catenary was first utilized by O’brien and Francis [13].
In 1965, Ernst suggested that a cable member can be modeled by truss They showed that each cable member within a cable structure can be
elements for the first time. He also introduced a modified axial stiffness modeled using a single analytical element. In this method, the overall
to account for the sag effects of a hanging cable [1]. Although his equilibrium of a stretched cable element is satisfied in the Lagrangian

⁎
Corresponding author.
E-mail address: [email protected] (M. Rezaiee-Pajand).

https://doi.org/10.1016/j.engstruct.2018.04.022
Received 18 January 2018; Received in revised form 6 April 2018; Accepted 7 April 2018
0141-0296/ © 2018 Elsevier Ltd. All rights reserved.
M. Rezaiee-Pajand et al. Engineering Structures 167 (2018) 431–444

coordinates, and the exact profile is derived by imposing the boundary

conditions at the end of the cable. Many researchers have developed the
elastic catenary element by introducing thermal effects and different
loading types [14–20]. Salehi Ahmad Abad et al. proposed an extended
three-dimensional catenary element which takes the thermal effects and
distributed lateral loads in different directions into account [21]. Na-
ghavi Riabi and Shooshtari implemented the elastic catenary along with
the Ramberg-Osgood stress-strain relationship to investigate the effects
of material nonlinearity on the behavior of cable networks [22]. Re-
cently, Crusells-Girona et al. have employed a mixed variational ap-
proach in curvilinear coordinates based on the elastic catenary ex-
pressions to model cables with material and geometric nonlinearity
[23]. Moreover, a number of researchers have adopted the parabola
approach for analysis and design of practical cable structures. Since the
parabola approach disregards the large sag effects, it provides an ap-
proximate solution to the hanging cables. It is proved that the error of
the method increases by increasing the sag to span ratio. Therefore, this Fig. 1. Configuration of the hyperbolic cable element under self-weight.
approach is unsuitable for modeling deep cables [24–26].
In addition to the aforementioned finite element approaches, many
self-weight per unit unstained length are denoted by F, u, S and w, in a
researchers have developed innovative ways for nonlinear analysis of
respective manner.
cable structures. Lewis employed the principle of minimum total po-
The cross-sectional area, elastic modulus and thermal expansion
tential energy along with the dynamic relaxation method to assess the
coefficient of the cable are supposed to be constant, and the formulation
efficiency of pure numerical approaches in analysis of pre-tensioned
is developed within the framework of small strains. Note that L and H
cable nets [27]. A two-link structure was utilized by Kwan to develop a
correspond to the horizontal projected length and the horizontal force
simple technique for nonlinear analysis of pre-tensioned cable struc-
of the cable along the local axis, ζ , respectively. Herein, it is assumed
tures. In this approach, similar to a spatial truss, the nonlinear equili-
that the profile of the cable is sufficiently deep. In other words, no
brium equations were written for each node, and then, they were solved
limitations are imposed on the curvature of this structural element. It is
by using an iterative method [28]. Stefanou and Moossavi Nejad
worth mentioning that removal of this assumption leads to the simpli-
minimized the total potential energy of the entire structural assembly
fied elastic parabola approach. Based on the foregoing hypotheses, the
by the conjugate gradient method to obtain the equilibrium state of the
profile of the cable hanging under its self-weight with respect to the ζ
cable structures [29]. The efficiency of various dynamic relaxation
axis can be defined by the following hyperbolic function [17]:
methods in analysis of cable structures was studied by Hüttner et al.
[30]. To model single-span cables considering extensibility and thermal H⎡ w
z (ζ ) = cosh ⎛ ζ + ξ ⎞−cosh(ξ )⎤
strains, the finite difference approach was applied by Bouaanani et al. w⎣ ⎝H ⎠ ⎦ (1)
[31,32].
Although the elastic catenary provides highly accurate results, the It should be added that satisfaction of the boundary conditions, i.e.
tangent stiffness matrix is not explicitly available. Therefore, a com- z (0) = 0 and z (L) = l z , will lead to the subsequent constants:
plicated iterative procedure must be adopted to determine the nodal
forces and establish the flexibility matrix. To perform a very systematic ξ = sinh−1 ( λlz
Lsinh(λ ) )−λ
analysis, the inverse of the flexibility matrix must be also computed to λ=
wL
obtain the stiffness matrix. Thus, many difficulties arise during the 2H (2)
analysis procedure of the elastic catenary. On the other hand, simplified Another parameter is defined as L = l x2 + l y2 . It can be easily shown
cable approaches, such as elastic parabola or elastic straight shape, are that the cable tension at the Cartesian coordinates has the next ex-
problematic in addressing the large sag effects in the deep cables. pression:
Moreover, these elements must be transformed from a local axis to the
global one via transformation matrices to be able to consider inclina- 2
dz w
tion. This action further increases the computational complexities. To T (ζ ) = H 1 + ⎛⎜ ⎞⎟ = H cosh ⎛ ζ + ξ ⎞
dζ
⎝ ⎠ ⎝ H ⎠ (3)
improve these drawbacks, a three-dimensional cable element is pro-
posed in this study for static analysis of the general cable structures. where T is the tension of the cable. After deformation, the strained
The elemental shape considers inclination without using transforma- length of the cable element can be obtained as:
tion, and it takes both the large sag and thermal strain effects into ac-
2 2
count. To make the nonlinear analysis easier, the components of the L dz L
P= ∫0 1 + ⎛⎜ ⎞⎟ dζ = ⎛ sinh(λ ) ⎞ + lz2
tangent stiffness matrix are presented by relatively simple closed-form dζ
⎝ ⎠ ⎝λ ⎠ (4)
expressions. Since the profile of the hanging cable is given by hyper-
bolic functions, for convenience, the proposed element is referred to as where P stands for the strained length of the cable element. Since a real
the ‘elastic hyperbola’. The numerical outcomes of the studied problems cable has finite axial flexibility, the inextensibility condition must be
reveal the accuracy and efficiency of the present element in the non- relaxed to obtain its elastic elongation. For an extensible cable with
linear analysis of spatial cable structures. constant material properties, the Hooke’s law is held:
T (ζ )
2. Formulation of the hyperbolic cable ε (ζ ) = + α Δϑ
EA (5)

The configuration of a perfectly flexible and elastic cable element where ε , E, A, α and Δϑ refer to the cable strain, elastic modulus, cross-
stretched between two nodes, namely i and j , is depicted in Fig. 1. As it sectional area, thermal expansion coefficient and uniform variation in
can be seen, the projected lengths along the x, y and z directions are the temperature, respectively. Substituting for the cable tension from
designated lx, ly and lz, respectively. Further, the nodal forces and nodal Eq. (3) into Eq. (5) and performing some mathematical manipulations
displacements along the global axes, initial unstrained length and the yield:

432
M. Rezaiee-Pajand et al. Engineering Structures 167 (2018) 431–444

2 2
H ⎛ dz ⎞ ⎛ dz ⎞
d ΔS = 1 + ⎜⎛ ⎟⎞ ⎟ dζ + α Δϑ 1 + ⎜ dζ ⎟ dζ
EA ⎜ ⎝ dζ ⎠ ⎝ ⎠ (6)
⎝ ⎠
where d ΔS corresponds to an infinitesimal increment in the unstrained
length of the cable. The elastic elongation of the cable element can be
easily obtained by integrating both sides of Eq. (6). The solution pro-
ceeds as:
2
wL2sinh(2λ ) ⎡ 1 ⎛ lz ⎞ 1 ⎤
ΔS = ⎢ 2λ2 + Lsinh(λ ) + λ sinh(2λ ) ⎥
⎜ ⎟

4EA ⎣ ⎝ ⎠ ⎦
2
L
+ α Δϑ ⎛ sinh(λ ) ⎞ + lz2
⎝λ ⎠ (7)
In which ΔS refers to the elastic elongation of the cable element.
Because the strained length of the cable is equal to sum of the un-
strained length and the elastic elongation, the following equality holds Fig. 2. Infinitesimal increase in the nodal forces due to infinitesimal increment
of l x .
true:
P = S + ΔS (8)
Since the initial unstrained length of the cable is a constant value, its
It follows from Eq. (8) that for a given H, the unstrained length of derivative equals to zero. Hence, for any n, differentiating both sides of
the cable can be directly calculated. When the unstrained length of the Eq. (8) yields:
cable element is given instead, Eq. (8) represents a nonlinear equation
dPn = d ΔSn (17)
by which the value of the horizontal force must be determined. A
simple and efficient iterative procedure, based on the modified Newton- Clearly, Eq. (17) indicates that the total differentiation of the elastic
Raphson technique, will be discussed later to handle this issue. elongation, and the strained length of the cable are equal. Upon sub-
As it can be seen in Fig. 1, the nodal forces Fjx and Fjy represent the stitution of Eqs. (15) and (16) into Eq. (17), the following interesting
projected resultants of the horizontal force H along the global x and y expression will be achieved:
axes. In order to derive the tangent stiffness matrix, independent of
∂Pn ∂P ∂P ∂P ∂ΔSn ∂ΔSn ∂ΔSn
transformation matrices, the following relationships must be exploited: dFjn + n dl x + n dl y + n dl z = dFjn + dl x + dl y
∂Fjn ∂l x ∂l y ∂l z ∂Fjn ∂l x ∂l y
lx ∂ΔSn
Fjx = H + dl z
L (9)
∂l z (18)
ly Eq. (18) forms the basis for calculation of the stiffness components.
Fjy = H
L (10) As shown in Fig. 2, assuming that an infinitesimal increment equal
Moment equilibrium in the O−ζ plane will lead to the next re- to dl x is applied to node j, such that the projected length of the cable
lationship: along the x direction is increase to l x + dl x . This gives rise to in-
finitesimal load increments, dFjx , dFjy and dFjz in the nodal forces of Fjx ,
lz wS Fjy and Fjz . Note carefully that in such a case, dl y and dl z equal to zero
Fjz = H +
L 2 (11) and vanish.
Substitution for the horizontal force by Eqs. (9)–(11) in Eqs. (4)–(7) Next, substituting by x for n in Eq. (18) leads to:
gives rise to: ∂Px ∂P ∂ΔSx ∂ΔSx
dFjx + x dl x = dFjx + dl x
2 ∂Fjx ∂l x ∂Fjx ∂l x (19)
Pn = ⎛ L sinh(λn ) ⎞ + l 2
⎜ ⎟ for n = x ,y,z
z
⎝ λn ⎠ (12) The first entry of the tangent stiffness matrix for node j, i.e. k11, is
defined as the ratio of the infinitesimal increment in the nodal force Fjx
2
wL2sinh(2λn ) ⎡ 1 ⎛ lz ⎞ 1 ⎤ to the infinitesimal length increase dl x [33]. Dividing Eq. (19) by dl x
ΔSn = ⎢ 2λ 2 + Lsinh(λn ) + λnsinh(2λn) ⎥
⎜ ⎟

4EA ⎝ ⎠ and rearranging the resultant, the stiffness component can be obtained
⎣ n ⎦
as:
+ α ΔϑPn for n = x ,y,z (13)
∂ΔSx ∂Px
dFjx − ∂l B1−A1
In which n is a dummy variable and other parameters have the k11 = =
∂lx x
=
dl x ∂Px ∂ΔSx A 4 −B4
following form: −
∂Fjx ∂Fjx (20)
wl wS
λn = n ,τx = Fjx ,τy = Fjy and τz = Fjz− The tangent stiffness formulae k21 and k31 which denote the ratio of
2τn 2 (14)
infinitesimal increment in the nodal forces Fjy and Fjz to dl x , respec-
tively, can be evaluated in a similar manner:
3. Derivation of the tangent stiffness matrix ∂ΔSy ∂Py
dFjy ∂lx
− ∂l B5−A5
x
k21 = = ∂Py ∂ΔSy
=
In the beginning, to determine the components of the tangent dl x − ∂F A8 −B8
∂Fjy jy (21)
stiffness matrix in an explicit manner, the following expressions are
available based on Eqs. (12) and (13): ∂ΔSz ∂Pz
dFjz ∂lx
− ∂l B9−A9
x
∂Pn ∂P ∂P ∂P k31 = = =
dPn = dFjn + n dl x + n dl y + n dl z dl x ∂Pz ∂ΔS
− ∂F z A12 −B12
∂Fjn ∂l x ∂l y ∂l z (15) ∂Fjz jz (22)

∂ΔSn ∂ΔSn ∂ΔSn ∂ΔSn Likewise, the rest of the stiffness components for node j can be
d ΔSn = dFjn + dl x + dl y + dl z calculated by applying infinitesimal increments to the projected length
∂Fjn ∂l x ∂l y ∂l z (16)
of the cable along the y and z directions and repeating the

433
M. Rezaiee-Pajand et al. Engineering Structures 167 (2018) 431–444

next result will be obtained:

[− k k ]{ duix duiy duiz dujx dujy dujz }T = { dFjx dFjy dFjz }T (32)
The equilibrium of nodal forces within the element requires the
following relation:
{ Fix Fiy Fiz }T = {− Fjx − Fjy wS−Fjz }T (33)
Total differentiation of Eq. (33) results in the equilibrium equation
for node i as:
[ k − k ]{ duix duiy duiz dujx dujy dujz }T = { dFix dFiy dFiz }T (34)
Using Eqs. (32) and (34), the equilibrium equation of the cable
element will be achieved:
[K]{dU}T = {dF}T (35)
In which

Fig. 3. Final configuration of the cable element subjected to infinitesimal

[K] = ⎡ k − k ⎤
length increments. ⎣− k k ⎦ (36)

aforementioned procedure for n = y,z . The final configuration of the {dU} = { duix duiy duiz dujx dujy dujz } (37)
cable element is depicted in Fig. 3. The rest of the stiffness components
{dF} = { dFix dFiy dFiz dFjx dFjy dFjz } (38)
for node j are also given below:
∂ΔSx ∂Px where K , dU and dF correspond to the tangent stiffness matrix of the
dFjx − ∂l
∂ly y B −A element, incremental nodal displacement vector and incremental nodal
k12 = = ∂Px ∂ΔS
= 2 2
dl y − ∂F x A 4 −B4 force vector, respectively.
∂Fjx jx (23)
As a matter of fact, the tangent stiffness matrix of the proposed
∂ΔSy ∂Py element given by Eq. (36) is derived with respect to the global axes, and
dFjy − ∂l
∂ly y B6−A6 thus a transformation matrix is not required to consider inclination.
k22 = = ∂Py ∂ΔSy
=
dl y − ∂F A8 −B8 This significantly decreases the complexities and computational issues
∂Fjy jy (24)
encountered in analysis of three-dimensional cable structures. The
∂ΔSz ∂Pz flexibility matrix of an elastic catenary is a function of three unknown
dFjz − ∂l
∂ly y B10−A10 nodal forces. These unknown forces must be determined through a
k32 = = ∂Pz ∂ΔS
=
dl y − ∂F z A12 −B12 rather complicated iterative procedure. In contrast, the stiffness matrix
∂Fjz jz (25)
in Eq. (36) depends only on one unknown force, namely the horizontal
∂ΔSx ∂Px
− ∂l force of the cable element. Hence, only a single value, i.e. the horizontal
dFjx ∂lz z B3−A3
k13 = = = force, must be calculated at each iteration step. Contrary to the
dl z ∂Px ∂ΔSx A 4 −B4
∂Fjx
− ∂Fjx (26) common procedures, it is not required to take the inverse of the flex-
ibility matrix at each iteration step since the explicit tangent stiffness
∂ΔSy ∂Py
− ∂l matrix is available and can be directly utilized. For pre-tensioned
dFjy ∂lz z B7−A7
k23 = = ∂Py ∂ΔSy
= cables, the pretension force is known in place of the unstrained length.
dl z − ∂F A8 −B8
∂Fjy jy (27) Unlike the elastic catenary that makes use of a laborious iterative
procedure to determine the unstrained length, the unstrained length of
∂ΔSz ∂Pz
dFjz − ∂l the cable can be directly calculated in the present scheme by means of
∂lz z B11−A11
k33 = = ∂Pz ∂ΔS
= Eq. (8). The foregoing merits of the proposed element make it re-
dl z − ∂F z A12 −B12
∂Fjz jz (28) markably efficient in the analysis of a great variety of cable structures,
including slack or pre-tensioned cable networks.
where Ai and Bi constants are used for simplification purposes, and all
of them will be reported in Appendix A. At this stage, the stiffness
4. Nonlinear analysis process
formulae are available in a closed-form and the tangent stiffness matrix
for node j can be established, as follows:
As it was already discussed in Section 2, when the initial unstrained
⎡ k11 k12 k13 ⎤ length of the cable is known instead of the horizontal force, the non-
k = ⎢ k21 k22 k23 ⎥ linear equation given by Eq. (8) must be solved to obtain the value of H .
⎢k k k ⎥ The Newton-Raphson technique provides a simple and efficient nu-
⎣ 31 32 33 ⎦ (29)
merical procedure to handle this problem. However, convergence to the
The nodal force and length increments of node j are related as: correct solution cannot be guaranteed unless sufficient criteria are in-
k{ dl x dl y dl z }T = { dFjx dFjy dFjz }T troduced into the Newton-Raphson scheme.
(30)
A schematic illustration of f (λ ) = P−S−ΔS is depicted in Fig. 4. As
where k is the tangent stiffness matrix of node j and dl corresponds to it can be seen by dashed lines, the equation has three roots in terms of
the difference between the nodal incremental displacements of two λ . The positive root indeed corresponds to the correct solution by which
nodes. The nodal displacement increments and infinitesimal length the value of H can be determined, and the two negative roots are un-
increments are related as: acceptable. Contrary to a taut cable, the Newton-Raphson technique
{ dl x dl y dl z }T = { dujx −duix dujy−duiy dujz−duiz }T may converge to any of these three possible solutions in case of slack
(31)
cables. In order to avoid unwanted solutions, i.e. the negative roots, a
where du refers to the increment in the nodal displacements u. constraint is built into the Newton-Raphson scheme, as discussed by
Substituting Eq. (31) into Eq. (30) and carrying out some algebra, the Ahmadi-Kashani and Bell [16]. An initial value for λ is required to

434
M. Rezaiee-Pajand et al. Engineering Structures 167 (2018) 431–444

this approach [34–37]. In this study, Newton-Raphson scheme is em-

ployed to trace the equilibrium paths. In this way, the governing finite
element equation is obtained at time t+Δt as below:
R (Ut + Δt ,Γt + Δt ) = ΔΓt + Δt F ext −F int (Ut + Δt ) (42)
in which F int (Ut + Δt )
is the vector of internal forces. It is worth
mentioning that these quantities are obtained based on the displace-
ments, Ut + Δt . Moreover, the vector of external loads is defined by F ext
and ΔΓ is the load factor which remains constant at all iterations of
each increment. The linearized form of Eq. (42) for each iteration of
increments is given by:
Kit + Δt ΔUit++1Δt = R (Ut + Δt ,Γt + Δt ) (43)
where R refers to the residual force vector. In other words, external
loads are computed at the first iteration of each increment and remain
constant throughout the other iterations of this step. The load factor ΔΓ
is also defined at the beginning of the analysis. The system of equations
can be solved as:
Kij − 1δ U1i = ΔΓF ext for j = 1Kij − 1δ Uij = Rij − 1for j ⩾ 2 (44)
The flowchart of the incremental Newton-Raphson solution procedure
for nonlinear analysis of cable structures is provided in Appendix B.
Fig. 4. Schematic view of the possible solutions for the horizontal force.
5. Verification and numerical examples
begin the iterations. This initial value may either be underestimated or
overestimated. For an underestimated value, the convergence is always In this section, various popular benchmark problems are studied to
to the correct solution. In contrast, an overestimated initial value for λ establish the efficiency and applicability of the geometric nonlinear
is likely to yield a negative value for the updated λ , even in the first analysis program, which is developed by authors, for using the sug-
iteration cycle. This indeed leads to the unwanted solutions. As a re- gested element. In case of single span hanging cables subjected to self-
medy, wherever a negative value for λ is detected, its value is halved weight and a concentrated load at s = s1, where s denotes the
and used in place of the updated value obtained by the Newton- Lagrangian coordinate along the cable unstrained length, an analytical
Raphson method in that iteration. Hence, the modified iterative method solution exists based on the elastic catenary expressions [38]:
proceeds according to:

λi−f (λi )/ f ′ (λi ) for λi + 1 >0

x (s ) =
Hs
EA
+
H (1 + α Δϑ)
w
⎡sinh−1
⎣ ( )−sinh ( ) ⎤⎦
V
H
−1 V − ws
H
⎧
λi + 1 = V 2 V − ws 2 ⎤
⎨λ
⎩2
i
for λi + 1 < 0 (39)
z (s ) =
s
EA (V − ) +
ws
2
H (1 + α Δϑ)
w
⎡ 1+
⎢
⎣
( ) − 1+(
H H )⎦ ⎥
for 0 ⩽ s ⩽ s1 (45)
To accelerate the Newton-Raphson procedure and converge to the
required solution by fewer iterations, the initial value of λ must be
chosen as close to the actual solution as possible. This further improves
x (s ) =
Hs
EA
+
H (1 + α Δϑ)
w
⎡sinh−1
⎣ ( )−sinh (
V
H
−1 V − Fc − ws
H ) + sinh (−1 V − Fc − ws1
H )
the numerical stability of the method. An appropriate initial value for λ
can be chosen based on the following relations:
− sinh−1 ( ) ⎤⎦
V − ws1
H

V 2 V − Fc − ws 2
⎧ 120η−20 −10 for 1 < η ⩽ 3.67
z (s ) =
s
EA (V − ws
2 )+ H (1 + α Δϑ) ⎡
w ⎢
⎣
1+ ( )−
H
1+ ( H ) +
Fc w
HEA 1
(s −s )
⎪
λ ≈ 2.337 + 1.095ln(η) V − Fc − ws1 2 V − ws1 2 ⎤
⎨ for 3.67 < η ⩽ 4.5 × 105 + 1+ ( H ) − 1+ ( H )⎦⎥
⎪ −0.00473[7.909−ln(η)]2.46
⎩ (40) fors1 ⩽ s ⩽ S
where . If the cable is pre-tensioned, η falls outside the boundaries (46)
given by Eq. (40). In such a case, the elasticity effects cannot be ne- where Fc corresponds to the concentrated load at s = s1 and V stands for
glected and a good estimate for λ may be given by the vertical component of the cable tension at its left end. The unknown
λ = w l x2 + l y2 + lz2 /2T where the approximate tension can be obtained values can be achieved by substitution of the right end boundary con-
as: ditions in the second relations of Eqs. (45) and (46). These analytical
1/3 a expressions will be utilized for verification of the proposed element in
⎧b + for b1/3 > a
T≈
3 modeling of single span deep and shallow cables.
⎨a + b
for b1/3 ⩽ a
⎩ 2a2 (41)
5.1. Isolated cable subjected to concentrated load
l x2 + l y2 + lz2 − S ⎞ AE 2 2 lz 2
In which a = AE ⎛⎜ 2 2 2 ⎟ and b = ()
wS L .
⎝ lx + l y + lz ⎠
24 This example presents an isolated cable with the span of 304.8 m.
Since the cable structures inherently exhibit geometrical nonlinear This problem was first considered by Michalos and Birnstiel [39] and
behavior, it is necessary to implement an incremental-iterative solution later studied by several researchers [13–15,17–19,21,40]. The initial
method. As it is obvious, no snap-through or snap-backs are observed in geometry of the cable and the necessary data for analysis are taken from
analysis of cable structures and thus a load control scheme can be [19] and presented in Fig. 5 and Table 1. The cable is hanging under its
employed. Among the load control methods, Newton-Raphson tech- self-weight, and a concentrated force is applied at node 2. The cable
nique has been the most popular for nonlinear structural analysis. It was modeled using 2 elastic hyperbola elements. The displacements are
should be reminded that many researchers have developed and utilized compared with the previous studies and the analytical solution in

435
M. Rezaiee-Pajand et al. Engineering Structures 167 (2018) 431–444

Fig. 5. Initial geometry of the isolated cable.

Table 1
Characteristics of the isolated cable.
Item Data

Cross sectional area 548.4 mm2

Elastic modulus 131000 MPa
Self-weight per unit unstrained length 46.12 N/m
Cable sag under self-weight at node 2 29.276 m
Unstrained length of cable element 1–2 125.847 m Fig. 6. Load-displacement curves for node 2.
Unstrained length of cable element 2–3 186.855 m

span ratio is larger than 1: 8 and thus the cable is regarded as deep. The
Table 2 resulting profiles are illustrated in Fig. 7. As it can be seen, there is an
Comparison of present and previous results for the isolated cable. excellent agreement between the proposed results and the analytical
solution. This fact clearly demonstrates that authors' element is able to
Researcher(s) Element Type No. of Displacement at node
elements 2 (m) accurately model extensible cables with large sag. It is worth men-
tioning that Bouaanani and Marcuzzi [31] implemented a complicated
ux uz and rather cumbersome finite difference technique to obtain the same
profile whereas the present scheme readily bears the correct config-
Saafan [40] Truss – −0.845 −5.472
O’Brien & Francis [13] Elastic catenary 2 −0.860 −5.627
uration using a single element.
Michalos & Birnstiel [39] Truss – −0.845 −5.472
Jayaraman & Knudson [19] Elastic catenary 2 −0.859 −5.626
Jayaraman & Knudson [19] Truss 10 −0.845 −5.471 5.3. Thermo-elastic analysis of an isolated cable
Tibert [17] Elastic catenary 2 −0.859 −5.626
Tibert [17] Associate 2 −0.859 −5.655 The third example is included to validate the reliability of the
catenary proposed element in the thermal analysis of cables. The problem refers
Tibert [17] Elastic parabola 2 −0.866 −5.601
Andreu et al. [14] Elastic catenary 2 −0.860 −5.626
to a cable the left end of which is fixed at the coordinates (0,90) m while
Yang & Tsay [15] Elastic catenary 2 −0.859 −5.625 the elevation of its right end support is kept constant at 30 m and the
Thai & Kim [18] Elastic catenary 2 −0.859 −5.626 horizontal coordinate of the right end is varied between 0.02 and 100 m
Salehi Ahmad Abad Elastic catenary 2 −0.859 −5.626 (see Fig. 8). The cable is subjected to a temperature rise of Δϑ = 100 K .
et al. [21]
The necessary properties of the cable are outlined in Table 3. The
Crusells-Girona et al. [23] Discrete elastic 3 −0.861 −5.630
catenary problem was first studied by Pevrot and Goulois [20] and later analyzed
Present Analytical 1 −0.859 −5.626 by Yang and Tsay [11] and Salehi Ahmad Abad et al. [21]. It is aimed to
Present Elastic 2 −0.859 −5.626 obtain the horizontal and vertical reactions at the right end support. To
hyperbola

Table 2. As it can be seen, the displacements obtained by the proposed

element are in excellent agreement with those found in the literature.
Fig. 6 depicts the comparison of the load-displacement curves for node
2 where the excellent agreement between the proposed scheme, and the
elastic catenary is clearly demonstrated.

5.2. Deep cable subjected to self-weight and a temperature variation

This example serves to demonstrate the potential of the proposed

element in modeling deep cables. The problem consists of a cable
hanging under its self-weight which spans 100 m . Note that the supports
are at the same elevation. The axial stiffness AE, thermal expansion
coefficient and mass per unit length of the cable are equal to 5 × 10 4 kN ,
1.25 × 10−5 K−1 and 25 kg/m , respectively. This cable also undergoes a
temperature variation of Δϑ = 10 K . The cable profile is obtained for
three different values of horizontal force, namely
H = 10 kN, H = 15 kN, H = 20 kN , by using the proposed scheme and the
analytical solutions given via Eqs. (45) and (46). To utilize the analy- Fig. 7. Present and elastic catenary profiles of the deep cable subjected to self-
tical expressions, Fc and s1 must be set to zero. In all cases, the sag to weight.

436
M. Rezaiee-Pajand et al. Engineering Structures 167 (2018) 431–444

framework of stiffness concept, to account for the pre-elongation of a

taut cable, the elasticity effects cannot be neglected and thus the cable
weight per unit length must be updated as the length of the cable is
increased throughout the solution process according to the following
relation:

w iP i = w 0S (47)

where w i and w 0 refer to the updated and initial value of the cable
weight per unit length, respectively. It is clear that this equation implies
the conservation of mass. Unless the proposed modification is in-
troduced into the formulations, the distributed load of the cable will not
be recomputed. This will result in larger vertical reactions and deflec-
tions than the reality. This modification has been simply built into the
solution program to analyze prestressed cables.
The fourth example is presented to assess the applicability of the
new formulation for analysis of a prestressed cable subjected to uniform
transverse loads. The problem is taken from Jayaraman and Knudson
[19]. The initial unstrained length of the cable has been 9990.00999 in .
Due to pretension, the length of the cable is increased to span 10000 in ,
as shown in Fig. 9. It is aimed to obtain the vertical displacement at
mid-span for different values of uniform transverse load. The horizontal
line joining the supports has been utilized as the starting geometry. The
cable was modeled by two elastic hyperbola elements. The obtained
results for five different values of w are reported in Table 5. It is ob-
Fig. 8. Various configurations of the cable.
served that the present results conform well to the benchmark out-
comes. Fig. 10 compares the variation of mid-span displacement of the
Table 3 cable under increasing uniform load obtained by the proposed element
Properties of the cable under thermo-elastic loading. and the analytical theory of elastic catenary.
Item Data

Cross sectional area

5.5. Longest cable of a cable-stayed bridge
1 m2
Elastic modulus 3 × 107 N/m2
Self-weight per unit unstrained length 1 N/m In this practical example, the proposed element is implemented to
Thermal expansion coefficient 6.5 × 10−6 K−1 analyze the longest cable of the Sutong Changjiang high-way bridge in
Unstrained length 100 m China. The cable is perfectly elastic with a chord length of 576.488 m
and an initial unstrained length equal to 574.805 m. The cross-sectional
area, elastic modulus and weight per unit unstrained length are equal to
solve this problem, a single elastic hyperbola element was implemented A = 0.012046 m2, E = 190 GPa and w = 988 N/m, respectively. The
in all cases. The results are compared to the literature in Table 4. As it coordinates of the left and right ends of the cable are given as (0,
can be seen, the proposed element has led to almost identical results 220.564 m) and (532.626 m, 0), respectively. Since the initial un-
with those predicted by previous researchers. strained length of the cable is shorter than its chord length, the strained
profile of the cable represents a taut behavior. This problem was
5.4. Prestressed cable under uniform load modeled only by one element. Table 6 compares the present results
against the benchmarks. A good agreement is achieved once again.
It was observed that the present element could accurately model
slack cables without difficulties. The analytical expressions of the 5.6. Prestressed plane cable net
elastic catenary are based on the flexibility concept and consistently
satisfy the equilibrium of internal forces with the elastic elongation of In this example, a prestressed plane cable net consisting of inclined
the cable. Since the present element is directly formulated within the and horizontal members is inspected. The geometry of the cable

Table 4
Comparison of reactions (N) for the cable under thermo-elastic loading.
Researcher(s) Reactions (N) Location (m)

0.02 20 40 60 80 100

Pevrot & Goulois [20] H 0 3.061 9.172 22.15 504 4,170,000

V 20.02 19.93 19.24 15.73 −328 −2,511,000
Yang & Tsay [15] H 0.01 3.061 9.172 22.15 504.1 4,255,700
V 20.02 19.93 19.24 15.73 −328.9 −2,555,340
Salehi Ahmad Abad et al. (DCC) [21] H 0.01 3.09 9.16 22.11 504.48 4,255,849
V 19.99 19.83 19.14 15.63 −329.4 −2,555,047
Salehi Ahmad Abad et al. (CCC) [21] H 0.011 3.06 9.172 22.145 504.104 4,258,491
V 20.02 19.93 19.24 15.73 −328.86 −2,555,044
Analytical H 0 3.060 9.172 22.146 504.103 4,258,491
V 20.02 19.93 19.24 15.73 328.87 −2,555,044
Present H 0 3.062 9.178 22.16 504.27 4,255,712
V 20.03 19.94 19.25 15.74 −328.96 −2,553,369

437
M. Rezaiee-Pajand et al. Engineering Structures 167 (2018) 431–444

Fig. 9. Prestressed cable under uniform load.

Table 5
Results and comparison for prestressed cable under uniform load. Fig. 11. Prestressed plane cable net.

Load w (lb/in) Mid-span displacement (in)

parameter Table 7
Elastic Elastic Analytical Present Comparison of displacements for node 4 in plane cable net.
catenary [19] parabola [41]
Researcher(s) Method Displacement at node 4 (mm)
1 0.02 131.63 131.60 131.49 131.49
3 0.06 234.19 234.49 234.20 234.22 x-direction y-direction z-direction
5 0.10 292.79 293.10 292.78 292.83
7 0.14 336.03 336.22 336.04 336.11 Saafan [40] Elastic straight −40.35 −40.35 −448.27
9 0.18 371.13 371.37 371.13 371.24 Jayaraman & Elastic catenary −39.62 −40.20 −446.32
Knudson [19]
Tibert [17] Elastic catenary −40.48 −40.48 −450.00
Tibert [17] Associate catenary −40.78 −40.78 −453.36
Tibert [17] Elastic parabola −40.78 −40.78 −453.36
Thai & Kim [18] Elastic catenary −40.13 −40.13 −446.50
Thai & Kim [18] Elastic catenary −40.28 −40.28 −448.88
(SAP2000)
West & Kar [43] Nonlinear −40.39 −40.39 −447.99
equilibrium
Present Elastic hyperbola −40.75 −40.75 −452.79

displacements for node 4 against those reported by other researchers.

As it can be seen, the results are very close to previous studies. This
implies that the proposed element provides satisfactory results in the
static analysis of prestressed cable nets.

5.7. Hyperbolic paraboloid net

A hyperbolic paraboloid cable network is selected for which the

experimental results are available in [44]. Various methods have been
implemented by several researchers to numerically investigate this
structure as well [14,18,28,44,45]. As shown in Fig. 12, the structure
Fig. 10. Variation of central deflection with increasing weight for prestressed contains 31 cable segments pretensioned by the force of 200 N prior to
cable under uniform load.
the application of external loads. Some internal joints are subjected to
concentrated loads of 15.7 N. All cables maintain an elastic modulus of
Table 6 128.3 GPa and a cross-sectional area of 0.785 mm2 . A uniform weight per
Comparison study for the longest cable of the Sutong Changjiang high-way unit length equal to 0.195 N/m is assumed to act. The vertical dis-
bridge. placements of this assembly are compared to the previous studies in
Researchers Left TA (KN ) Right TB (KN ) Cable strained Sag1/2/ P * Table 8. As it can be seen, the results predicted by the proposed scheme
length P (m) are in good agreement with those reported by the other researchers. It is

Ren & Gu [42] 7331.219 7114.005 576.616 1/101.725

Yang & Tsay [15] 7321.591 7104.359 576.616 1/101.236
Present 7327.549 7109.632 576.616 1/101.328

* Sag of the cable below the chord at x = L/2 .

structure is shown in Fig. 11. The cross-sectional area, elastic modulus

and self-weight per unit unstrained length of the cable are 146.45 mm2,
82737 MPa and 1.459 N/m, respectively. Furthermore, inclined and
horizontal cables maintain unstrained lengths of 30.419 m and
31.760 m, respectively. The problem was first studied by Saafan [40]
and subsequently investigated by Jayaraman and Knudosn [19], Tibert
[17] and Thai and Kim [18]. The internal nodes are subjected to a
concentrated force of Fc = 35.586 kN . Each cable was modeled using a
single elastic hyperbola element. Table 7 compares the present Fig. 12. Hyperbolic paraboloid net.

438
M. Rezaiee-Pajand et al. Engineering Structures 167 (2018) 431–444

Table 8
Comparison of vertical displacements (mm) of hyperbolic paraboloid net.
Node Vertical displacement (mm)

Experiment [44] Dynamic Relaxation [44] Minimum Energy [45] Elastic catenary [18] Approximation by series [28] Elastic catenary [14] Present

5 −19.50 −19.30 (1.03) *

−19.30 (1.03) −19.56 (0.31) −19.52 (0.10) −19.51 (0.05) −19.51 (0.05)
6 −25.30 −25.30 (0.00) −25.50 (0.79) −25.70 (1.58) −25.35 (0.19) −25.65 (1.38) −25.58 (1.10)
7 −22.80 −23.00 (0.88) −23.10 (1.32) −23.37 (2.50) −23.31 (2.24) −23.37 (2.50) −23.28 (2.10)
10 −25.40 −25.90 (1.97) −25.80 (1.57) −25.91 (2.01) −25.86 (1.81) −25.87 (1.85) −25.83 (1.69)
11 −33.60 −33.80 (0.60) −34.00 (1.19) −34.16 (1.67) −34.05 (1.34) −34.14 (1.60) −33.95 (1.04)
12 −28.80 −29.40 (2.08) −29.40 (2.08) −29.60 (2.78) −29.49 (2.39) −29.65 (2.95) −29.42 (2.15)
15 −25.20 −26.40 (4.76) −25.70 (1.98) −25.86 (2.62) −25.79 (2.34) −25.86 (2.62) −25.61 (1.62)
16 −30.60 −31.70 (3.59) −31.20 (1.96) −31.43 (2.71) −31.31 (2.32) −31.47 (2.84) −31.02 (1.37)
17 −21.00 −21.90 (4.29) −21.10 (0.48) −21.56 (2.67) −21.42 (2.00) −21.57 (2.71) −21.24 (1.12)
20 −21.00 −21.90 (4.29) −21.10 (0.48) −21.57 (2.71) −21.48 (2.28) −21.62 (2.95) −20.83 (0.81)
21 −19.80 −20.50 (3.54) −19.90 (0.51) −20.14 (1.72) −20.00 (1.01) −20.15 (1.76) −19.19 (3.08)
22 −14.20 −14.80 (4.23) −14.30 (0.70) −14.55 (2.46) −14.40 (1.40) −14.55 (2.46) −13.81 (2.74)
‖Error‖ 10.63 4.54 7.81 6.24 7.94 6.14

* Numbers in parentheses indicate the absolute error percentage with respect to experiment results.

Fig. 13. Spatial cable network.

Table 9
Comparison of displacements (mm) for spatial net.
Researcher Direction Displacements (mm)

Node

7 8 9 14 15 16

Lewis et al. [27] (Elastic straight) dx −5.14 −2.26 0.00 −4.98 −2.55 0.00
dy −0.42 −0.47 2.27 0.00 0.00 0.00
dz −30.41 −17.7 3.62 −43.49 −44.47 −41.65

Thai & Kim [18] (Elastic catenary) dx −5.03 −2.23 0 −4.92 −2.55 0.00
dy −0.41 −0.46 2.31 0.00 0.00 0.00
dz −29.86 −17.29 3.61 −42.85 −44.26 −42.08

Salehi Ahmad Abad et al. [21] (PCCC) dx −5.02 −2.24 0.00 −4.94 −2.56 0.00
dy −0.41 −0.43 2.23 0.00 0.00 0.00
dz −29.55 −17.55 3.19 −42.99 −44.30 −42.04

Salehi Ahmad Abad et al. [21] (PDCC) dx −5.05 −2.23 0.00 −4.93 −2.55 0.00
dy −0.40 −0.40 2.36 0.00 0.00 0.00
dz −29.55 −17.16 3.19 −42.94 −44.34 −42.14

Present (Elastic hyperbola) dx −5.03 −2.23 0.00 −4.93 −2.55 0.00

dy −0.40 −0.39 2.36 0.00 0.00 0.00
dz −29.50 −17.14 3.19 −42.94 −44.39 −42.20

439
M. Rezaiee-Pajand et al. Engineering Structures 167 (2018) 431–444

Fig. 14. Plan and side view of the pretensioned cable roof.

also drawn from this table that the present element leads to more sa- two different load cases. The first load case, indicated by LC.1, refers to
tisfactory results in terms of the error norm compared to other tech- a downward concentrated force of 1 kip applied to all joints. The second
niques. load case, denoted by LC.2, is similar to LC.1 plus an additional load of
14 kips applied at node 29. Table 10 compares the present vertical
5.8. Spatial cable network displacements of the sample joints against the benchmark results. As it
can be seen, the results predicted by the proposed element are in ex-
Another example examined here is a spatial cable network. The cellent agreement with previous studies.
configuration of the structure is depicted in Fig. 13, in which symmetry
about both centerlines is evident. The assembly is 24 m× 16 m in the 5.10. Suspended cable ring
plan, and 38 pretensioned cable segments have been used to divide its
grid in 4 m× 4 m quadrilaterals. All cables maintain an elastic modulus This example serves to illustrate the efficiency of the proposed
of 160 GPa and the pretension force along the x and y directions are scheme in analysis of slack cable networks. The initial configuration of
equal to 90 kN and 30 kN , respectively. All internal joints are subjected an axisymmetric suspended cable ring with inner radius of 35 m and
to a downward concentrated force of 6.8 kN . The cross-sectional areas of outer radius of 75 m is presented in Fig. 15. The assembly consists of 8
the cables along the x and y directions are also equal to 350 mm2 and radial and 8 tangential slack cable segments. The outermost joints are
120 mm2 , in a respective manner. A very small quantity is assumed for assumed to be fixed, while each of the inner joints maintains three
the cables’ weight per unit length to carry out the analysis. The ob-
tained displacements are compared to those reported by previous re- Table 10
searchers in Table 9. As it can be seen, the proposed scheme leads to Comparison of vertical displacements (ft) for example 9.
almost identical results.
Node Vertical displacement (ft)

5.9. Pretensioned cable roof LC. 1 LC. 2

Ahmadi- Thornton & Present Ahmadi- Thornton & Present

In this section, a pretensioned cable roof structure with symmetry Kashani Birnstiel Kashani Birnstiel
about both centerlines is analyzed. The initial geometry and z co- [47] [46] [47] [46]
ordinates for a quarter of the assembly are shown in Fig. 14. For the
1 0.00 0.00 0.00 0.00 0.00 0.00
first time, this problem was introduced by Thornton and Birnstiel [46]
3 −0.254 −0.254 −0.254 −0.259 −0.259 −0.259
and further studied by Ahmadi-Kashani [47]. The structure includes 64 7 −0.552 −0.552 −0.553 −0.605 −0.605 −0.606
pretensioned and weight-less cable segments. The side beams are as- 13 −0.772 −0.772 −0.774 −1.020 −1.020 −1.023
sumed to be rigid. The horizontal component of the pretension force is 21 −0.861 −0.861 −0.863 −1.722 −1.722 −1.727
50 kips for all cables. The cross-sectional area and elastic modulus of the 29 −0.772 −0.772 −0.774 −3.720 −3.718 −3.727
35 −0.552 −0.552 −0.553 −1.268 −1.268 −1.272
cables are given as 1 in2 and 2.4 × 107 psi , respectively. Since the cables
39 −0.254 −0.254 −0.254 −0.381 −0.381 −0.381
are assumed to be weight-less, a very small weight per unit volume of 41 0.00 0.00 0.00 0.00 0.00 0.00
10−6 kips/ft3 is assigned to each element. The structure was analyzed for

440
M. Rezaiee-Pajand et al. Engineering Structures 167 (2018) 431–444

Fig. 15. Plan and perspective view of the suspended cable ring.

Table 11
Comparison of joint coordinates (m) at the final equilibrium state.
Researcher(s) Direction (m) Joint No.

1 2 3 4 5 6 7 8

Hüttner et al. [49] x 41.469 29.451 0.000 −29.451 −41.649 −29.451 0.000 29.451
y 0.000 29.451 41.469 29.451 0.000 −29.540 −41.649 −29.451
z −21.713 −21.713 −21.713 −21.713 −21.713 −21.713 −21.713 −21.713
Deng et al. [48] x 41.469 29.451 0.000 −29.451 −41.649 −29.451 0.000 29.451
y 0.000 29.451 41.469 29.451 0.000 −29.451 −41.649 −29.451
z −21.713 −21.713 −21.713 −21.713 −21.713 −21.713 −21.713 −21.713
Present x 41.469 29.451 0.000 −29.451 −41.649 −29.451 0.000 29.451
y 0.000 29.451 41.469 29.451 0.000 −29.451 −41.649 −29.451
z −21.717 −21.717 −21.717 −21.717 −21.717 −21.717 −21.717 −21.717

translational degrees of freedom. The cross-sectional area, elastic corresponding equilibrium paths of cable structures, the Newton-
modulus and weight-per unit unstrained length of all cables are equal to Raphson iterative method was also employed.
A = 1963.44 mm2 , E = 170 GPa and w = 151.047 N/m , respectively. Unlike the elastic catenary approach, which requires a complicated
Further, the radial and tangential cables have unstrained lengths of iterative procedure for inverting the flexibility matrix and determina-
40 m and 32 m, respectively. The entire cable network is released from tion of three unknown nodal forces at each step, the proposed for-
the initial state to deform under its self-weight. It is aimed to obtain the mulation leads to an explicit stiffness matrix. It has only one unknown
nodal coordinates of the final equilibrium state. Deng et al. used a ra- force, namely the horizontal force. This makes the new element more
ther arduous shape finding technique to solve this problem [48]. In this efficient in terms of the analysis time. Significant case studies, ranging
study, the analysis was simply carried out by modeling each cable from single span cables to slack and pretensioned cable networks, were
segment with a single element. The coordinates of the inner joints at the performed to illustrate the robustness of the present technique in ana-
final equilibrium state are compared to those reported by previous re- lysis of various types of cable assemblies.
searchers in Table 11. As it can be seen, the proposed scheme has In addition to the simplicity and explicit nature of the equations and
predicted identical results. This clearly indicates the reliability and relations, the numerical outcomes clearly demonstrate that the pro-
applicability of the new formulation in shape finding and analysis of the posed element furnishes accurate results and can be conveniently uti-
slack cable structures. lized in research, analysis and design of practical deep and shallow
cable-supported structures such as pretensioned cable roofs, long-span
6. Conclusions cable stayed bridges and suspension bridges.

Based on the concept of inextensible catenary, a novel formulation Declaration of conflicting interests
for three-dimensional cables was developed in this study. Authors’
element is able to consider inclination without using any transforma- No potential conflicts of interest are declared by the authors with
tion matrices, takes the large sag effects into account, and it is applic- respect to the authorship, research and/or publication of this paper.
able to the cables undergoing general load cases, such as, concentrated,
uniformly distributed and thermal loads. The inextensibility condition Funding
was relaxed and equality of the total differentiation of the elastic
elongation and strained length of the cable was employed to derive the The authors received no financial supports either for the research or
explicit components of the tangent stiffness matrix. To trace the the authorship and publication of this article.

441
M. Rezaiee-Pajand et al. Engineering Structures 167 (2018) 431–444

Appendix A

As it is given by Eqs. (20)–(28), the components of the tangent stiffness matrix are simplified using Ai and Bi constants. These constants are
outlined below:
sinh(λ )(λL2cosh(λ )−l y2sinh(λ ))
A1 = (1−α Δϑ)
λ2l x P (I.1)

ly sinh2 (λ )
A2 = (1−α Δϑ)
λ2P (I.2)

lz
A3 = (1−α Δϑ)
P (I.3)

2L2
A 4 = −(1−α Δϑ) sinh(λ )(λ cosh(λ )−sinh(λ ))
l x λwP (I.4)

lx
A5 = A2
ly (I.5)

sinh(λ )(λL2cosh(λ )−l x2sinh(λ ))

A6 = (1−α Δϑ)
λ2l y P (I.6)
A7 = A3 (I.7)

lx
A8 = A 4
ly (I.8)
A9 = A5 (I.9)
A10 = A2 (I.10)

2H 2 lz
A11 = (1−α Δϑ) (λ sinh(λ )−2sinh2 (λ )) +
l z w 2P P (I.11)

lx
A12 = A 4
lz (I.12)
2
β= ⎛ wl z
⎜
⎞ +4 ⎟

⎝ H sinh(λ ) ⎠ (I.13)
1
B1 = 4βEALlx λ sinh(λ )
{Hβ sinh(λ )[λL2 (cosh(2ξ + 4λ ) + cosh(2ξ ))
− l y2 (sinh(2ξ + 4λ )−sinh(2ξ )) + 2λ (l x2−l y2)]−wl z (cosh(2ξ + 4λ )−cosh(2ξ ))(λL2coth(λ )−l y2)} (I.14)

ly ⎧ w Hl z
B2 = 2
(sinh(2ξ + 4λ )−sinh(2ξ ) + 4λ )− 2 (cosh(2ξ + 4λ )−cosh(2ξ ))⎫
2EA ⎨
⎩ 4λ βL sinh(λ) ⎬
⎭ (I.15)
H
B3 = (cosh(2ξ + 4λ )−cosh(2ξ ))
2EAβ sinh(λ ) (I.16)

λH
B4 = {−βH sinh(λ )[λ (cosh(2ξ + 4λ ) + cosh(2ξ )) + sinh(2ξ )−sinh(2ξ + 4λ )−2λ]−wl z (cosh(2ξ )−cosh(2ξ + 4λ ))(λ coth(λ )−1)}
βw 2EAl x sinh(λ ) (I.17)

lx
B5 = B2
ly (I.18)
1
B6 = {Hβ sinh(λ )[λL2 (cosh(2ξ + 4λ ) + cosh(2ξ ))−l x2 (sinh(2ξ + 4λ )−sinh(2ξ )) + 2λ (l y2−l x2)]
4βEALl y λ sinh(λ )
−wl z (cosh(2ξ + 4λ )−cosh(2ξ ))(λL2coth(λ )−l x2)} (I.19)
B 7 = B3 (I.20)

lx
B8 = B4
ly (I.21)

lx
B9 = {βH sinh(λ )(sinh(2ξ + 4λ )−sinh(2ξ ) + 4λ )−wl z (cosh(2ξ + 4λ )−cosh(2ξ ))}
4βEALλ sinh(λ ) (I.22)

ly
B10 = B9
lx (I.23)

442
M. Rezaiee-Pajand et al. Engineering Structures 167 (2018) 431–444

H
B11 = {Hβ sinh(λ )[λ (cosh(2ξ + 4λ ) + cosh(2ξ ))−sinh(2ξ + 4λ ) + sinh(2ξ )−2λ]−wl z (cosh(2ξ + 4λ )−cosh(2ξ ))(λ coth(λ )−2)}
2βEAwl z sinh(λ ) (I.24)

λH
B12 = − {Hβ sinh(λ )[λ (cosh(2ξ + 4λ ) + cosh(2ξ ))−sinh(2ξ + 4λ ) + sinh(2ξ )−2λ]−wl z (cosh(2ξ + 4λ )−cosh(2ξ ))(λ coth(λ )−1)}
βEAw 2l z sinh(λ ) (I.25)

Appendix B

The flowchart of the incremental Newton-Raphson solution procedure is given below:

443
M. Rezaiee-Pajand et al. Engineering Structures 167 (2018) 431–444

Appendix C. Supplementary material

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.engstruct.2018.04.022.

References [25] Gründig L, Bahndorf J. The design of wide-span roof structures using micro-com-
puters. Comput Struct 1988;30(3):495–501.
[26] Haber RB, Abel JF. Initial equilibrium solution methods for cable reinforced
[1] Ernst JH. Der E-modul von seilen unter berucksichtigung des durchhanges. Der membranes part I—formulations. Comput Methods Appl Mech Eng
Bauingenieur 1965;40(2):52–5. 1982;30(3):263–84.
[2] Knudson WC. Static and Dynamic Analysis of Cable-net Structures. Berkeley: [27] Lewis WJ. The efficiency of numerical methods for the analysis of prestressed nets
University of California; 1971. and pin-jointed frame structures. Comput Struct 1989;33(3):791–800.
[3] Liew JYR, Punniyakotty NM, Shanmugam NE. Limit-state analysis and design of [28] Kwan ASK. A new approach to geometric nonlinearity of cable structures. Comput
cable-tensioned structures. Int J Space Struct 2001;16(2):95–110. Struct 1998;67(4):243–52.
[4] Chen ZH, et al. Formulation and application of multi-node sliding cable element for [29] Stefanou GD, Nejad SEM. A general method for the analysis of cable assemblies with
the analysis of Suspen-Dome structures. Finite Elem Anal Des 2010;46(9):743–50. fixed and flexible elastic boundaries. Comput Struct 1995;55(5):897–905.
[5] Gambhir ML, Batchelor Bde V. A finite element for 3-D prestressed cablenets. Int J [30] Hüttner M, Máca J, Fajman P. The efficiency of dynamic relaxation methods in
Numer Meth Eng 1977;11(11):1699–718. static analysis of cable structures. Adv Eng Softw 2015;89(Supplement C):28–35.
[6] Gambhir ML, Batchelor BD. Finite element study of the free vibration of 3-D cable [31] Bouaanani N, Marcuzzi P. Finite difference thermoelastic analysis of suspended
networks. Int J Solids Struct 1979;15(2):127–36. cables including extensibility and large sag effects. J Therm Stresses
[7] Ozdemir H. A finite element approach for cable problems. Int J Solids Struct 2011;34(1):18–50.
1979;15(5):427–37. [32] Bouaanani N, Ighouba M. A novel scheme for large deflection analysis of suspended
[8] Coyette JP, Guisset P. Cable network analysis by a nonlinear programming tech- cables made of linear or nonlinear elastic materials. Adv Eng Softw
nique. Eng Struct 1988;10(1):41–6. 2011;42(12):1009–19.
[9] Ali HM, Abdel-Ghaffar AM. Modeling the nonlinear seismic behavior of cable-stayed [33] Choi D-H, Gwon S-G, Na H-S. Simplified analysis for preliminary design of towers in
bridges with passive control bearings. Comput Struct 1995;54(3):461–92. suspension bridges. J Bridge Eng 2014;19(3):04013007.
[10] Wu JH, Su WZ. The non-linear finite element Analysis of cable structures based on [34] Riks E. The application of newton’s method to the problem of elastic stability. J
four-node isoparametric curved element. J. Chongqing Jianzhu Univ. Appl Mech 1972;39(4):1060–5.
2005;27(6):55–8. [35] Mondkar DP, Powell GH. Evaluation of solution schemes for nonlinear structures.
[11] Wang Yang, Zuo SR, Wu C. A finite element method with six-node isoparametric Comput Struct 1978;9(3):223–36.
element for nonlinear analysis of cable structures. Appl Mech Mater [36] Powell G, Simons J. Improved iteration strategy for nonlinear structures. Int J
2013;275–277:1132–5. Numer Meth Eng 1981;17(10):1455–67.
[12] Tang JM, Zhao Y, Wu LH. A Eeulerian geometrically non-linear finite element [37] Padovan J, Tovichakchaikul S. Self-adaptive predictor-corrector algorithms for
method with two node cable element for the analysis of cable structures. Shanghai J static nonlinear structural analysis. Comput Struct 1982;15(4):365–77.
Mech 2005;21(1):89–94. [38] Irvine HM. Cable Structures. 1992: Dover Publications.
[13] Terence O'Brien W, Francis AS. Cable movements under two dimensional loads. J [39] Michalos J, Birnstiel C. Movements of a cable due to changes in loading. J Struct Div
Structural Div ASCE 1964;90(3):89–124. 1960;86(12):23–38.
[14] Andreu A, Gil L, Roca P. A new deformable catenary element for the analysis of [40] Saafan SA. Theoretical Analysis of Suspension Roofs. Journal of the Structural
cable net structures. Comput Struct 2006;84(29):1882–90. Division, 1970. 96(2): p. 393.405.
[15] Yang YB, Tsay J-Y. Geometric nonlinear analysis of cable structures with a two- [41] Wei-Xin R, Meng-Gang H, Wei-Hua H. A parabolic cable element for static analysis
node cable element by generalized displacement control method. Int J Struct Stab of cable structures. Eng Comput 2008;25(4):366–84.
Dyn 2007;07(04):571–88. [42] Ren SY, and Gu MJ. Static analysis of cables configuration in cable stayed bridges.
[16] Ahmadi-Kashani K, Bell AJ. The analysis of cables subject to uniformly distributed Tongji Uni. (Nat. Sci.), 2005. 33(5): p. 595–599.
loads. Eng Struct 1988;10(3):174–84. [43] West HH, Kar AK. Discretized initial-value analysis of cable nets. Int J Solids Struct
[17] Tibert G. Numerical analyses of cable roof structures. in Trita-BKN. Bulletin. 1999, 1973;9(11):1403–20.
KTH: Stockholm. p. x, 180. [44] Lewis WJ, Jones MS, Rushton KR. Dynamic relaxation analysis of the non-linear
[18] Thai H-T, Kim S-E. Nonlinear static and dynamic analysis of cable structures. Finite static response of pretensioned cable roofs. Comput Struct 1984;18(6):989–97.
Elem Anal Des 2011;47(3):237–46. [45] Sufian FMA, Templeman A. On the non-linear analysis of pre-tensioned cable net
[19] Jayaraman HB, Knudson WC. A curved element for the analysis of cable structures. structures. Struct Eng 1992;4:147–58.
Comput Struct 1981;14(3):325–33. [46] Charles H. Thornton and C. Birnstiel, Three-Dimensional Suspension Structures.
[20] Pevrot AH, Goulois AM. Analysis of cable structures. Comput Struct Journal of the Structural Division, 1967. 93(2): p. 247–270.
1979;10(5):805–13. [47] Ahmadi-Kashani K. U.o.M.I.o. Science, and Technology, Development of Cable
[21] Salehi Ahmad Abad M, et al. Nonlinear analysis of cable structures under general Elements and Their Applications in the Analysis of Cable Structures. 1983:
loadings. Finite Elem Anal Des 2013;73(1):11–9. University of Manchester Institute of Science and Technology (UMIST).
[22] Naghavi Riabi AR, Shooshtari A. A numerical method to material and geometric [48] Deng H, Jiang QF, Kwan ASK. Shape finding of incomplete cable-strut assemblies
nonlinear analysis of cable structures. Mech Based Des Struct Mach containing slack and prestressed elements. Comput Struct 2005;83(21):1767–79.
2015;43(4):407–23. [49] Hüttner M, Máca J, and Fajman P. Numerical Analysis of Cable Structures.
[23] Crusells-Girona M, Filippou FC, Taylor RL. A mixed formulation for nonlinear Proceedings of the Eleventh International Conference on Computational Structures
analysis of cable structures. Comput Struct 2017;186(Supplement C):50–61. Technology, Civil-Comp Press, Stirlingshire, Scotland, 2012.
[24] Gimsing NJ, Georgakis CT. Cable Supported Bridges: Concept and Design. Wiley;
2011.

444