You are on page 1of 28

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/303595370

Static form-finding analysis of a railway catenary using a dynamic equilibrium


method based on flexible multibody system formulation with absolute nodal
coordinates and controls

Article  in  Multibody System Dynamics · March 2017


DOI: 10.1007/s11044-016-9522-y

CITATIONS READS

9 2,675

6 authors, including:

C. J. Yang Jing. Zhang


Southwest Jiaotong University Southwest Jiaotong University
28 PUBLICATIONS   171 CITATIONS    28 PUBLICATIONS   397 CITATIONS   

SEE PROFILE SEE PROFILE

Guiming MEI Ning Zhou


Southwest Jiaotong University Southwest Jiaotong University
80 PUBLICATIONS   843 CITATIONS    59 PUBLICATIONS   565 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

High-speed pantograph and catenary dynamics View project

pantogrpah and catenary dynamics View project

All content following this page was uploaded by Jing. Zhang on 24 April 2018.

The user has requested enhancement of the downloaded file.


Multibody Syst Dyn (2017) 39:221–247
DOI 10.1007/s11044-016-9522-y

Static form-finding analysis of a railway catenary using


a dynamic equilibrium method based on flexible
multibody system formulation with absolute nodal
coordinates and controls

C.J. Yang1,2 · W.H. Zhang1 · J. Zhang3 · G.M. Mei1 ·


N. Zhou1 · G.X. Ren2

Received: 7 March 2014 / Accepted: 6 May 2016 / Published online: 27 May 2016
© Springer Science+Business Media Dordrecht 2016

Abstract This paper proposes a dynamic equilibrium method for finding the initial equilib-
rium configuration of a railway catenary. In the proposed method, the catenary composed
of flexible wires is modeled using two-node cable elements with absolute nodal coordinates
based on a flexible multibody system formulation. Dynamic conditions that characterize
the initial equilibrium configuration of the catenary are given and addressed as control pro-
cesses in the form-finding procedure. The key feature of the proposed method is that the
catenary configuration is continually evolved by dynamic simulation until characterization
conditions are attained and an equivalent configuration of the centenary at static equilibrium
is thus computed. It is validated using two examples and applied to the form-finding anal-
ysis of a two dimensional simple railway catenary. The obtained results are analyzed and
discussed. It is general and can be applied to catenaries with complex configurations.

Keywords Railway catenary · Form-finding · Dynamic equilibrium method · Control ·


Absolute nodal coordinates · Flexible multibody system formulation

1 Introduction

Today, the high speed electric train has become a convenient and rapid transport vehicle. It is
usually powered by electricity, collected from the contact wire using a pantograph. Being the
transportation mode for the next generation, it has prominent advantages of high velocity,
high stability and high passenger capacity when compared to other traditional transportation
modes.
Higher speed on train operations is constantly desired. An accompanying problem for
this request is to ensure a stable current collection. The current collection of a high speed

B C.J. Yang
ycj78_2012@163.com
1 State Key Laboratory of Traction Power, Southwest Jiaotong University, Chengdu, 610031, China
2 Department of Engineering Mechanics, Tsinghua University, Beijing, 100084, China
3 School of Electrical Engineering, Southwest Jiaotong University, Chengdu, 610031, China
222 C.J. Yang et al.

train is largely dependent on the dynamic interaction [1, 2] between the catenary system
and the pantograph system. Incompatibility of the pantograph and catenary dynamics has an
influential effect on the quality of the current collection. For this reason, it is important to
study the dynamic pantograph–catenary interaction.
During the past several years, the dynamic interaction between the catenary and the pan-
tograph system has been studied extensively. A number of dynamic models [3–13] in this
field have been developed. In the use of these models, necessary conditions representing
the initial equilibrium configuration of the catenary are of importance to be provided for
dynamic simulation since they can strongly influence the obtained results. Thus, accurate
analysis of the initial equilibrium configuration of the catenary is required.
Determination of initial equilibrium configuration of flexible tension structures as the
railway catenary, also known as the form-finding problem, is not an easy task due to many
unknowns and high nonlinearities. Various form-finding methods currently exist. A review
of existing form-finding methods can be found in [14]. Among these methods, the finite
element methods [15, 16] are commonly used in practice. However, the application of finite
element methods is quite tiresome due to the need of adjusting lengths of droppers to fit into
the predefined equilibrium state of the system.
The nonlinear displacement method [17] and the force density method [18] are another
two common formulations for the from-finding problem. Being the oldest method for tension
cable structures, the nonlinear displacement method [17] is based on the geometry variation
of the structural configuration up to the point that the equilibrium with a prescribed force
distribution is attained. This method has an important shortcoming where the geometrical
initial configuration mapping with the guess of nodal coordinates is required. To relax this
requirement, Schek [18] proposed the force density method for the form-finding problem of
a cable net. The force density method is regarded as one of the most powerful form-finding
formulations. The feature of the force density method is that nonlinear equilibrium equa-
tions for nodal coordinates of the cable net are transformed to a set of linear equations by
prescribing the force density. Later on, many researchers resumed this method and presented
improved forms, e.g., the extended force density method [19] and the adaptive force density
method [20].
Recently, an accurate, robust and fast method based on analytical catenary equations
[12] for a two dimensional form-finding problem was previously proposed by Garcia [21].
Miguel [22] extended this method to analyze three-dimensional tension cable structures.
This method [21, 22] addressed the equilibrium of forces at nodes where the droppers
connect with the messenger and contact wires. Projecting these forces to every node and
considering the force–displacement law defined by the catenary equation yields a set of
transcendental equations of the unknowns for nodal tensions and nodal coordinates. The
Newton–Raphson method is used to solve nonlinear equations to obtain tensions at cable
nodes and nodal coordinates.
Besides the above-mentioned methods, other useful form-finding methods are available
in the literature, e.g., the geometry variation method [6, 23], the separated model method
[7], the negative sag method [11], the optimization methods on constraint problems [13,
24], the minimal error of the dropper tension [10], minimizing the distance [5] between the
static deformed geometry of the contact wire and its specified position and the minimization
[8] of a residual function constituted by the weighted differences between the actual and the
design values of the tensile load in the wires and of the lateral position of the steady arms,
and so on.
This paper proposes a dynamic equilibrium method to compute the initial equilibrium
configuration of the railway catenary. The proposed method relies on dynamic simulation of
Static form-finding analysis of a railway catenary using a dynamic. . . 223

the catenary model to change the structural configuration. The interim structural configura-
tion computed is identified and, if necessary, updated for the next form-finding computation
until dynamic conditions that characterize the initial equilibrium configuration of the cate-
nary are attained. An equivalent configuration of the catenary at static equilibrium is thus
computed. In the proposed method, the catenary that is composed of flexible wires is mod-
eled using two-node finite cable elements with absolute nodal coordinates in the flexible
multibody system formulation. The established catenary model allows for large displace-
ments and deformations of the wires in the working conditions.
The remainder of this paper is organized as follows. After a description of the dynamic
equilibrium method in Sect. 2, the implementation of the proposed method is discussed
in Sect. 3. Two examples are used to validate the proposed method in Sect. 4. In Sect. 5,
the proposed method is applied to the form-finding of a two dimensional simple railway
catenary. Conclusions are finally provided.

2 Description of the proposed dynamic equilibrium form-finding method

Dynamic simulation provides a better tool to learn the evolution of dynamic behaviors of
a complex structure in time domain. A consecutive dynamic state of the structure at an
arbitrary time starting from a given initial state can be attained by dynamic simulation.
Generally, the structure at a specific state, e.g., the initial static equilibrium state, can be
uniquely characterized using available conditions. Consider that the structure is forcedly
evolved into a desired state where characterization conditions are attained. The configuration
of the structure at current state can be regarded as an equivalent configuration of the structure
required by designers. This is the feature of the dynamic equilibrium form-finding method
proposed in this paper.
To implement the proposed dynamic equilibrium method in the form-finding analysis of
the railway catenary system, two critical issues need to be addressed: (1) what conditions
can uniquely characterize the initial static equilibrium state of the catenary system and (2) in
what way the catenary system can reach the state where the prescribed characterization
conditions are attained. They are discussed in what follows.
Figure 1 depicts a typical railway catenary extensively used in China. It is mainly com-
posed of the contact wires, the messenger wires, the vertical wires known as the droppers,
and the supports. The contact wires provide the energy supply to the train by their contact
with the pantograph installed on the train. The messenger wires give a physical support to
hold the contact wires by droppers. The droppers are fixed to the messenger and contact
wires by clamps, providing the whole catenary with a sufficient stiffness distribution.
For such an extremely flexible tension cable structure with a large slenderness ratio
shown in Fig. 1, accurate characterization of the structural initial equilibrium configura-
tion is quite difficult due to many unknowns and high nonlinearities. As it is well known,
the initial equilibrium configuration of the cables largely depends on how the assembly of
the catenary system is performed. An actual installation often leads to differences of the as-
sembled catenary from the pre-design structure due to unavoidable reasons, e.g., installation
error. But, the assembled structure is able to be better responsible for the system in operation
despite of these differences. In this sense, several conditions are required to characterize the
catenary system at the initial static equilibrium.
After the assembly, the messenger and contact wires are highly tightened. To maintain a
sufficient mechanical stiffness distribution of the catenary, the messenger and contact wires
have constant prescribed tensions and are connected by a finite number of droppers regularly
224 C.J. Yang et al.

Fig. 1 Picture of a conventional


Chinese railway catenary

Fig. 2 Schematic of the form-finding process in the proposed method

installed in spans along the track. With the help of droppers, the contact wires are suspended
at connection points at prescribed heights. Without loss of generality, characterization con-
ditions for the form-finding analysis are thus summarized as (1) the messenger and contact
wires have constant prescribed tensions Tm and Tc , see Fig. 2; (2) finite droppers, say Nd ,
are regularly installed along the track at intervals of di in a number of spans; (3) connection
points between the contact wires and the droppers are positioned at prescribed heights hi ,
and so on.
These conditions that characterize the initial static equilibrium of the catenary are ad-
dressed as control processes in the form-finding analysis in this paper, as stated early. Con-
trol processes are complex and rely on the dynamic model of the catenary. Relevant details
are provided in the following section. Figure 2 briefly represents a schematic of the form-
finding processes in the dynamic equilibrium method. As can be observed in Fig. 2, the dy-
namic equilibrium method proposed in this paper is to compute an equivalent configuration
of the catenary at static equilibrium based on an iteration procedure, which is dependent
on the catenary dynamic model and control conditions that characterize the initial static
equilibrium of the catenary.
Static form-finding analysis of a railway catenary using a dynamic. . . 225

Fig. 3 Undeformed and


deformed configurations of the
ANCF cable element

3 Implementation of the proposed dynamic equilibrium form-finding


method

The implementation of the proposed dynamic equilibrium method in the form-finding analy-
sis of the railway catenary is discussed below in detail. As stated above, the proposed method
relies on dynamic simulation of the catenary model under control conditions to change the
structural configuration and compute an equilibrium configuration, which is equivalent to
the initial equilibrium configuration of the catenary. In this section, the dynamic model of
the catenary at a proper starting configuration is firstly established based on the flexible
multibody system formulation and the modeling of relevant control problems is followed.

3.1 Dynamic description of the catenary at a proper starting configuration

The railway catenary that is composed of wires is an extremely flexible structure with a large
slenderness ratio. It has highly nonlinear behaviors, undergoes large displacements and de-
formations and holds significant tensions in the working conditions. An accurate model for
cable dynamics is therefore required. In this paper, two-node cable elements [25] based on
the absolute nodal coordinate formulation (ANCF) [26] are adopted to model the messenger
and contact wires, as well as the droppers. The ANCF cable element can allow for large
axial and bending deformations of the wires. All cable elements that model the wires are
assembled using dynamic constraints and loads of the cable element.

3.1.1 ANCF cable element

The messenger and contact wires and droppers have smaller radii. In this case, the cable
element is treated as an elastic line with two nodes i and j , see Fig. 3.
In the absolute nodal coordinate formulation, the absolute displacement vector of an
arbitrary point P on the axis of the cable element at l in a global inertia coordinate frame
OXY Z can be expressed as
r = Sqe (1)
where q =e
[rTi , ri T , rTj , rjT ]T
is the element nodal generalized coordinates vector, in which
ri and rj are respectively the absolute displacement vectors of two nodes, and ri and rj are
respectively the element slopes at two nodes. The superscript “T ” denotes the transpose of
a vector or a matrix. The prime “ ” denotes the derivative of the variable with respect to
226 C.J. Yang et al.

the arc-length coordinate l defined in the initial undeformed configuration. S is the cable
element shape matrix, which is defined by
⎡ ⎤
N1 0 0 N3 0 0 N2 0 0 N4 0 0
S=⎣ 0 N1 0 0 N3 0 0 N2 0 0 N4 0⎦ (2)
0 0 N1 0 0 N3 0 0 N2 0 0 N4

where interpolation functions Ni for i = 1, 2, 3, 4 are defined as

1 1
N1 = (s − 1)2 (s + 2), N2 = (s + 1)2 (2 − s), (3)
4 4
L L
N3 = (s − 1)2 (s + 1), N4 = (s + 1)2 (s − 1), (4)
8 8
in which s is a dimensionless coordinate defined as
2l
s= −1 (5)
L
and L denotes the initial undeformed length of the cable element. It is remarkably noted that
the element displacement field, see Eq. (1), is expressed using a cubic Hermite interpolation
in terms of nodal coordinates. Thus, the cable element has C 2 continuity and can better
model nonlinearities of the wires.
Differentiating Eq. (1) with respect to time t yields

ṙ = Sq̇e (6)

where the overdot denotes time differentiation. Thus, the kinetic energy of the cable element
can be expressed by
 L
1  e T  e 
T e = ρA Sq̇ Sq̇ dl (7)
2 0

where ρ is the mass density of the cable element, and A is the cross-sectional area of the
cable element.
In the description of the strain energy of the cable element, a rigorous curvature defined
using deformed nodal coordinates [27] of the cable element based on the Seret–Frenet frame
theory [28], instead of the approximate second order derivate of lateral displacements of the
element nodal coordinates [29–31], is adopted. The current curvature allows for the effects
of geometry nonlinearities in the cable element. Moreover, a nonlinear axial strain based on
Green strain tensor is adopted. Therefore, the current cable element can better describe large
displacements and deformations of the wires in the working conditions. Without considering
the element cross-sectional effects, the strain energy of the cable element can be simply
written using two variables representing the bending curvature κ(l) and the axial strain ε(l)
of the cable element axis at l as
 
1 L  1 L
Ue = EAε 2 + EJ κ 2 dl = (N ε + Mκ) dl (8)
2 0 2 0

where E is the Young’s modulus of the cable element and J is the inertia moment of the
cross-section of the cable element. N is the axial tension in the cross-section of the cable
Static form-finding analysis of a railway catenary using a dynamic. . . 227

element at l, and M is the moment in the cross-section of the cable element at l. The axial
strain ε of the cable element axis at l is defined by

1   
ε(l) = r ·r −1 . (9)
2
The bending curvature κ of the cable element at l is defined by

r × r 
κ(l) = (10)
r 3

where the symbol  ·  denotes the modulus of a vector. In terms of the beam theory, the
axial tension N of the cable element at l is calculated as

N (l) = EAε(l). (11)

Using the Lagrange formulation of the first kind, equations of motion of a cable element
can be written as
⎧ e
⎪ d ∂T ∂U e  T
⎨ = − + Qef − Ceq λe
dt ∂ q̇e ∂qe (12)

⎩ e e e 
C q , q̇ , t = 0
or
⎧  
⎨Me q̈e = Qeq + Qef − Ceq T λe
(13)
⎩Ce qe , q̇e , t  = 0

where Me is the mass matrix of the cable element, which is defined by


 L
M = ρA
e
ST S dl. (14)
0

e
Qeq is the elastic force vector of the element, and Qeq = − ∂U
∂qe
. Qef is the generalized force
e
vector of the element including the gravity force. C denotes the constraint equations of
the element. Ceq is the Jacobian matrix of constraint equations Ce . λe denotes the unknown
Lagrange multipliers associated with the element constraint equations.

3.1.2 Constraints and loads of the cable element

To model the wires using the ANCF cable element in Eq. (12) or Eq. (13), constraints and
loads of the element are considered and modeled in what follows. In the assembly, the mes-
senger wire is connected to stays, and the contact wire is connected to registration arms.
The external structural components such as stays and registration arms are installed along
the track. Without considering flexible effects of external structural components, the mes-
senger wire, as well as the contact wire, is equivalently constrained at a given point in space.
On the other hand, the messenger and contact wires in spans are required to be tightened
or loosened in the form-finding so that prescribed tensions in them are attained. To fit these
conditions, sliding joints are used to model the connections between the messenger/contact
wire and the support, see Fig. 4. Similarly, the connection between the dropper and the mes-
senger/contact wire is modeled using a sliding joint so that droppers are immovably placed
228 C.J. Yang et al.

Fig. 4 Constraints and loads of the ANCF cable element

along the track in the form-finding. In the flexible multibody system formulation, constraint
conditions of the sliding joint can be expressed as [32–34]

Ces = rsi − rsj = 0 (15)

where rsi and rsj are the absolute displacement vectors of interacted points in two elements,
respectively. Constraint equations in Eq. (15) are linear in generalized coordinates of the
cable element, but are nonlinear in the parameter s defined in Eq. (5). Since the parameter
s is variable and three constraint equations are imposed, the sliding joint eliminates two de-
grees of freedom. Note that Eq. (15) models the connections between the messenger/contact
wire and the support when rsj indicates the displacement vector of a point in the support.
While rsi and rsj are the displacement vectors of two points in the messenger/contact wire
and the dropper, Eq. (15) models the connections between the messenger/contact wire and
the dropper.
Cable elements are required to model a long wire. Two neighboring cable elements are
constrained together using a fixed joint. In terms of dynamics of the cable element, constraint
conditions of the fixed joint between two cable elements ii and jj can be expressed as
[35, 36]
 f f

rii − rjj
Cf =
e
f f f f
=0 (16)
(rii )T rjj − rii  · rjj 
f f
where rii and rjj are respectively the absolute displacement vectors of two joint points in
f f
two elements ii and jj . rii and rjj are respectively the slopes of two elements ii and jj at
joint points.
The wires are subjected to the gravity forces in the form-finding. The generalized force
vector Qeg of the cable element associated with the gravity force can be calculated using the
principle of virtual work over the element volume as [35, 36]
  
L   L
δW = (ρAg dl)n · δ Sqe = ρAg ST n dl δqe = Qeg δqe (17)
0 0

where n = [0, −1, 0]T is a unit vector along the direction of the gravity force, and g is the
gravitational constant.
Static form-finding analysis of a railway catenary using a dynamic. . . 229

Fig. 5 Schematic of the catenary at an assembled configuration and a starting configuration

3.1.3 Dynamic model of the catenary at a proper starting configuration

The form-finding of a tension structure in design is an inverse problem with unknowns. In


this analysis, lengths of the messenger and contact wires are unknown. To model them using
finite length cable elements introduced above, their guessed lengths are given. The railway
catenary is a periodic structure and is symmetrical in finite spans; see Fig. 5(a). The distance
in spans can be a guessed length of the messenger/contact wire for the analysis. In this sense,
the messenger wire placed horizontally has a line-like shape at a starting configuration; see
Fig. 5(b). The messenger wire is divided into many segments. Each segment is modeled us-
ing a cable element. Two neighboring cable elements are connected using a fixed joint. Two
end points of the messenger wire are jointed to predefined positions in space (representing
the support) using sliding joints. The contact wire is modeled similarly.
Droppers vary in length and tension along the track. In this paper, the dropper is modeled
in a special manner. Two identical cables are responsible for dynamics of a dropper in the
form-finding. Each cable is attached to either the messenger wire or the contact wire by
sliding joints and has a length equal to the vertical height between the messenger wire and
the contact wire; see Fig. 5(b). Thus, a simple configuration of the catenary in Fig. 5(b) that
is employed as input for the next form-finding iteration is obtained. Considering geometric
properties of the catenary in Fig. 5(b), nodal coordinates of cable elements that model the
wires can be analytically determined. Assembling all cable elements using fixed and sliding
joints and considering the element gravity forces, one can establish the dynamic model of
the catenary in Fig. 5(b), whose motion equations are expressed by

⎨Mq̈ = Qq + CTq λ
(18)
⎩C(q, q̇, t) = 0

where q, q̇, and q̈ are the generalized coordinate, velocity, and acceleration vector of the
catenary system, respectively. M is the mass matrix of the system. C denotes the constraint
equations of the system. Cq is the Jacobian matrix of constraint equations. Qq is the gen-
eralized force vector of the system. λ is the unknown Lagrange multipliers’ vector of the
system.
230 C.J. Yang et al.

3.2 Control problems

As can be known from the ANCF cable element theory, the catenary at the configuration
in Fig. 5(b) is free of strain. However, the assembled catenary is in tension. To achieve a
mechanical tension in the current catenary, one can tighten/loosen the messenger and contact
wires. Droppers can be tensioned by raising/lowering them. Tension adjusting continues
until the catenary reaches the desired state so that prescribed conditions defined in Sect. 2 are
attained. In this paper, tension adjusting is modeled as a control process. Note that droppers
statically stand along the track during tension adjusting since no frictions exist in sliding
joints.

3.2.1 Tension control

Tightening and loosening the messenger and contact wires in spans leads to the variation of
the wire tension. For example, the contact wire is tightened when its two ends Ec0 and Ec1 ,
see Fig. 6(a), move at a variable velocity of vc along two opposite directions n0c and n1c at
the same time. The tension Tcref of the contact wire at the joint is increasing. As the tension
Tcref is larger than the prescribed tension Tc , the two ends move in reverse directions. The
contact wire is then loosened. The tension Tcref is decreased. As the tension Tcref is within a
proper range of Tc , the tension of the contact wire is held constant by imposing vc = 0. The
tension control can be described using the following piecewise equations as


⎪ −v , Tc > T1
⎪ s


⎪ −T −T

⎪ −v ( T
) (3 − 2 T1 −T00 ),
0 2 T
T0 < T ≤ T1
⎨ s T1 −T0
vc = 0, −T0 < T ≤ T0 (19)



⎪ T +T1 2 T +T1

⎪ vs − vs ( T1 −T0 ) (3 − 2 T1 −T0 ), −T1 < T ≤ −T0



vs , T ≥ −T1

where vc is a reference velocity. vs is an allowable maximal velocity. Tc = Tcref − Tc de-


notes a tension difference between the current tension Tcref and the prescribed tension Tc
of the contact wire. T0 and T1 are two predefined control variables. Note that the reference
velocity vc is nonlinearly dependent on the tension Tcref for the sake of smooth controls, see
Fig. 7. The tension control model can be included in the system model of Eq. (18) using
nonholonomic constraints based on the flexible multibody system formulation [35, 36]
 
vc0 (qc0 , q̇c0 , t) − vc n0c
CcTension
(q, q̇, t) = =0 (20)
vc1 (qc1 , q̇c1 , t) − vc n1c

where vci , in which i = 0, 1, is the velocity of the node i at the end of the wire and which is
dependent on the nodal generalized coordinate and velocity vectors qci and q̇ci . The tension
control of the messenger wire can be modeled similarly.

3.2.2 Height control

In the assembly, the contact wire is suspended by the droppers at discrete connection points,
i
e.g., the point Pcd in Fig. 6. These connection points are in prescribed heights. As an exam-
ple, dropper i is considered. It can reach a proper height after being raised or lowered with
Static form-finding analysis of a railway catenary using a dynamic. . . 231

Fig. 6 Schematic of a three stepped process in form-finding

Fig. 7 Variations of the


reference velocity with the wire
tension

a state-dependent velocity vi . The velocity vi is defined as




⎪−vs , hi > h1



⎪ h −h h −h
⎪−vs ( h1i−h00 )2 (3 − 2 h1i−h00 ),
⎪ h0 < hi ≤ h1

vi = 0, −h0 < hi ≤ h0 (21)



⎪ i +h1 2 i +h1

⎪vs − vs ( h ) (3 − 2 h ), −h1 < hi ≤ −h0


h1 −h0 h1 −h0

vs , hi ≥ −h1

where hi is the difference between the vertical displacement component yi of the connec-
i
tion point Pcd in the ith dropper and the prescribed height hi . h0 and h1 are two control
232 C.J. Yang et al.

variables. Nonholonomic constraint equations of a height control are expressed as [35, 36]
Height
Ci (q, q̇, t) = ẏi − vi = 0. (22)

3.2.3 Control strategy and force equilibrium conditions of the dropper

The railway catenary is an extremely flexible tension structure with a large slenderness ra-
tio. A number of cable elements with many degrees of freedom are used to model large dis-
placements and highly nonlinear deformations of the cables. The computational efficiency
in dynamic simulation is considered. A three-step control strategy is adopted for the sav-
ings in computation. The control strategy, illustrated in Fig. 6, is described as follows. In
the first step, the contact wire has the prescribed tension Tc by tightening/loosening it to a
proper state. In the meantime, it is suspended at prescribed heights at connection points by
raising/lowering the droppers. The required forces Tcdi in droppers at the desired equilibrium
state are calculated using Eq. (11). In the second step, the messenger wire has a prescribed
tension Tm by tightening/loosening it to a proper state. Droppers move up and down so that
tensions in them at connection points attain proper values. Since a dropper is modeled using
two identical cables, one cable is compatible with another cable in terms of its tension. The
condition of force equilibrium for the ith dropper can be expressed using the tensions in two
cables as [35, 36]
Dropper
Ci (q, q̇, t) = Tmd
i
− Tcdi − Gi = 0 (23)
where Gi is a varying gravity force for the ith dropper associated with the vertical height
i i
of two points Pmd and Pcd . Equation (23) can be included in the system model of Eq. (18)
by considering a similar height control for the ith dropper with a variable velocity vi that
Dropper
is dependent on the quantity Ci . If the condition of Eq. (23) is satisfied, velocity vi
is zero. Otherwise, velocity vi varies according to a similar law as given by Eq. (19) or
Eq. (21). In the first two steps, the contact and messenger wires respectively reach prescribed
equilibrium states. In the third step, lengths of the droppers are determined. For example,
the length of the ith dropper is calculated as a height difference of two connection points
i i
Pmd and Pcd . The form-finding is fulfilled. In the first two steps, a reduced model containing
the droppers plus either the contact wire or the messenger wire is analyzed, instead of a full
model containing the droppers, the messenger and contact wires. Costs in computation are
therefore saved.

3.3 Numerical solving method

In the proposed method, the form-finding of the railway catenary reduces to solving Eq. (18),
together with Eqs. (20), (22) and (23), using proper numerical methods. These equations are
typically differential-algebraic equations (DAEs) with index 3. The backward difference
formulas (BDFs) [37] are adopted here to solve them. For this purpose, the equations are
written in a general implicit form as

f(t, q, q̇, λ) = 0 (24)

where q, q̇ and λ are the unknowns to be solved for. Discretization of Eq. (24) at the kth
time step yields
 
f t (k) , q(k) , q̇(k) , λ(k) = 0. (25)
Static form-finding analysis of a railway catenary using a dynamic. . . 233

In terms of BDF, the differential variables q̇ can be expressed by a polynomial of the vari-
ables q at current and previous time steps as


s
q̇(k) = di · q(k−s+i) (26)
i=0

where s is the order of the BDF integrator, and di are the coefficients of BDF. Substituting
Eq. (26) into Eq. (25), one can establish the following nonlinear algebraic equation:
 
s
(k)
f t ,q , (k)
di · q(k−s+i)
,λ (k)
= 0. (27)
i=0

Equation (27) is numerically solved using the Newton–Raphson method under given initial
conditions, and unknown variables q and λ at the kth time step are found.

3.4 Discussions

The proposed method computes an equivalent structural equilibrium configuration based on


the evolution of the catenary configuration by dynamic simulation of the catenary model
under control conditions that characterize the structural initial equilibrium configuration. In
the proposed method, three-dimensional ANCF cable elements and the flexible multibody
formulation are adopted to model the catenary. In this sense, an advanced dynamic model of
catenaries with complex configurations can be established. Some factors, e.g., the supports
and registration arms, the clamps, Z wires, stitch wires, a prescribed sag and even installa-
tion error, can be considered in the catenary model. As an example, the support is modeled
using beam elements, and the flexible effect of the support can be addressed in the form-
finding analysis. Using the advanced model of the catenary and considering proper charac-
terization conditions, one can compute the initial equilibrium configuration of a prescribed
catenary. The proposed method is general. It can be applied to catenaries with complex
configurations.

4 Validation examples

The proposed method is validated below using two examples, including a horizontal rod
supported by four bars and a three-cable tension structure.

4.1 Example 1: Horizontal rod supported by four bars

Figure 8 depicts a horizontal rod supported by four elastic circular bars. The rod has length
L = 10 m, mass density ρ = 7800 kg/m3 , cross-sectional diameter d = 0.03 m. The Young’s
modulus E of the rod is assumed to be large enough in this example, e.g., 1018 Pa, so that it
can be treated as a rigid body. The bars are equally spaced at a span of 2 m.
The structure is statically indeterminate. The equilibrium and compatibility equations of
the structure are used to determine the axial forces in bars at static equilibrium. Equilibrium
of moments at joint O yields

Mo = 2F1 + 4F2 + 6F3 + 8F4 − 5G = 0 (28)
234 C.J. Yang et al.

Fig. 8 Schematic of a horizontal


rod supported by four bars

in which F1 , F2 , F3 and F4 are axial forces in bars 1, 2, 3 and 4, respectively. G is the gravity
force of the rod. The compatibility equation is based on the geometry of the structure. Under
the gravity force, the rod rotates about the joint O, and the vertical displacement δ at an
arbitrary point along the rod is therefore proportional to its distance from joint O. Thus, the
compatibility equation can be written as
δ1 δ2 δ3 δ4
= = = . (29)
1 2 3 4
Using Eq. (29) in Eq. (28), one obtains
1 1 1 1
F1 = G, F2 = G, F3 = G, F4 = G. (30)
12 6 4 3
Theoretical axial forces in bars 1, 2, 3 and 4 are thus calculated using Eq. (30) as 45.03,
90.05, 135.08, and 180.11 N, respectively. Note that the gravitational constant g is 9.8 m/s2
in the calculation.
Next, the axial forces in bars are numerically calculated based on height controls and
the structure dynamic model. The rod is modeled using 20 ANCF cable elements. Each
bar is treated as a cable of length 0.5 m and is modeled using two ANCF cable elements.
Other parameters of the bar are: Young’s modulus E = 1011 Pa, density ρ = 7800 kg/m3
and cross-sectional diameter d = 0.02 m. The rod is initially placed horizontally. Without
height controls exerted on four connection points in the rod, the rod rotates about joint O
under the gravity force. However, results shown in Figs. 9 and 10 imply that the rod is placed
horizontally since four connection points are still in their previous positions.
Figure 9 shows time histories of moving velocities of the top ends in bars during control
processes. Figure 10 shows time histories of vertical positions of four connection points in
the rod. Note that connection points are in the central axis of the rod. It is clearly shown
in Figs. 9 and 10 that the structure reaches an equilibrium state after a transient process of
about 5 s. Figure 11 shows the change of axial forces in bars against time. As shown in
Fig. 11, numerical solutions of axial forces in bars 1, 2, 3 and 4 are 45.18, 90.15, 135.03 and
180.06 N, respectively. Comparisons of analytical solutions with numerical ones of axial
forces in bars are given in Table 1. It can be observed in Table 1 that numerical solutions of
axial forces in bars are in good agreement with analytical ones. The relative errors between
numerical and analytical solutions for axial forces in bars are all less than 5 h. Example 1
shows good performance of ANCF cable elements. On the other hand, it shows the validity
of the height control model.
Static form-finding analysis of a railway catenary using a dynamic. . . 235

Fig. 9 Time histories of vertical positions of connection points in the rod during the controls

Table 1 Comparisons of
analytical and numerical No. Analytical (N) Numerical (N) Rel. error (h)
solutions for axial forces in bars
Bar 1 45.03 45.18 3.3
Bar 2 90.05 90.15 1.1
Bar 3 135.08 135.03 0.37
Bar 4 180.11 180.06 0.28

4.2 Example 2: Three-cable tension structure

A tension structure that consists of three cables is considered, as shown in Fig. 12. Three
cables are hereafter called cable 1, cable 2, and cable 3, respectively. Cable 1 is hinged at
two points O and B. Hinge points O and B are fixed on the ground. Cable 2 is hinged
at two points C and E. Hinge points C and E are rigidly positioned at a height of H .
Cable 3 connects with the cable 1 and the cable 2 at points A and D, respectively. In static
equilibrium, the tension of cable 1 at joint B is T1 and the tension of cable 2 at joint C is T2 .
Connection point A in cable 3 is at a height of h over the X-axis. The length of cable 3 is L.
In a pre-design, tension T1 is 3 kN and tension T2 is 1.5 kN. Other parameters are listed in
Tables 2 and 3.
In the modeling, cable 1 has a guessed length of 25 m and is modeled using 30 ANCF
cable elements. Cable 1 is hinged to point O using a sphere joint [35] and to point B using
a sliding joint. Cable 2 has a guessed length of 10 m and is modeled using 20 ANCF cable
236 C.J. Yang et al.

Fig. 10 Time histories of moving velocities of top ends in bars during the controls

Table 2 Structural parameters of


the cable structure in Example 2 Parameter Value (m)

L1 3
L2 7
L3 15
L4 10
H 1.5
h 0.01

Table 3 Material parameters of


the cable system in Example 2 Parameter Cable 1 Cable 2 Cable 3

Young’s modulus E (GPa) 130 81.7 119


Cross-sectional area A (mm2 ) 120 153 10
Mass density ρ (Kg m−3 ) 7710 4020 8900
Bending stiffness EJ (Nm2 ) 400 311 4

elements. Similarly, cable 2 is hinged to point E using a sphere joint and to point C using a
sliding joint. Cable 3 is has a guessed length of 1.5 m and is modeled using 4 ANCF cable
Static form-finding analysis of a railway catenary using a dynamic. . . 237

Fig. 11 Time histories of axial forces in four bars during the controls

Fig. 12 Schematic of a three-cable tension structure

elements. Cable 3 is connected to cable 1 at point A using a sliding joint and to cable 2 at
point D using a sliding joint.
The structure in Fig. 12 is analyzed using the proposed method. The obtained results are
provided below. In the first step, cable 1 is put up at the height of 0.01 m at the connection
point A and has a prescribed tension of 3 kN at joint B under height and tension controls.
Figure 13 shows the variations of the tension of the cable 1 at joint B in the time domain
during the controls. Clearly, in Fig. 13, the tension of cable 1 at joint B attains a prescribed
value after about t = 20 s. Figure 14 shows the change of the height of connection point A
with time during the controls. The curve in Fig. 14 implies that the connection point A is
238 C.J. Yang et al.

Fig. 13 Time history of the


tension T1 of the cable 1 at joint
B in the first step

Fig. 14 Time history of the


height of connection point A in
the first step

Fig. 15 Time history of the


tension of the cable 3 in the first
step

tightly positioned at the height of 0.01 m by controls. Figure 15 shows the variations of the
tension of cable 3 at connection point A in the time domain. It is importantly noted that the
required tension of cable 3 at static equilibrium is about 121 N.
In the second step, cable 2 is tightened with a tension of 1.5 kN at joint C. Cable 3 is in
force equilibrium and holds a tension of 121 N at connection point A. The variations of the
Static form-finding analysis of a railway catenary using a dynamic. . . 239

Fig. 16 Time history of the


tension T2 of cable 2 at joint C in
the second step

Fig. 17 Time history of the


tension of cable 3 at connection
point D in the second step

tension of cable 2 at joint C in the time domain during the controls are depicted in Fig. 16. As
shown in Fig. 16, cable 2 holds a constant tension of 1.5 kN at joint C after about t = 70 s.
Dynamic simulation shows that cable 3 has a tension of about 122 N at connection point D
after the controls, see Fig. 17. Figure 18 shows the variations of the height of connection
point D in the time domain. From Fig. 18, one can conclude that connection point D is at a
height of about 1.317 m at static equilibrium.
In the third step, the length of cable 3 is determined. The length of the deformed cable 3 is
about 1.31 m, obtained by calculating the height difference of two connection points D and
A in cable 3. To validate the obtained results, the dynamic relaxation method is applied to
a reconstructed model of the structure. Figure 19 shows the variations of dynamic property
of the reconstructed structure in time domain. As shown in Fig. 19, the tension of cable 1
at joint B is a little smaller than its prescribed value. The tension of cable 1 at joint B is
2980 N. The tension of cable 2 at joint C is 1540 N. The connection point A is at the height
of 0.01 m. It is concluded from Fig. 19 that the structure is better at the desired equilibrium
state. The proposed dynamic equilibrium method is valid in the form-finding of the cable
structure in Example 2.
240 C.J. Yang et al.

Fig. 18 Time history of the


height of connection point D in
the second step

Fig. 19 Variations of dynamic property of the reconstructed structure in dynamic relaxation

5 Engineering application

To show the application of the proposed method in railway dynamics and for the sake of
simplicity, a two dimensional simple catenary as shown in Fig. 20 is considered in this
Static form-finding analysis of a railway catenary using a dynamic. . . 241

Fig. 20 Schematic of a two dimensional simple catenary system in three spans

Table 4 Parameter values of intervals di (measured in m) in the analysis

d1 d2 d3 d4 d5 d6 d7 d8 d9 d10 d11

4 9.5 9.5 9.5 9.5 9.5 9.5 8 9.5 9.5 9.5

section. Without considering flexible effects of external structure elements, e.g., supports,
the contact wire is hinged to two points A and B at the same height using sliding joints. The
messenger wire is hinged to four points C, D, E and F at the same height using sliding
joints. The contact wire is suspended from the messenger wire by 21 droppers. Droppers are
placed along the track at intervals di , in which an index i is 1, 2, and up to 21. Structural
parameters are listed in Table 4. Due to symmetry, only parameters of the left part of the
structure are given in Table 4. The contact wire, the messenger wire and the dropper are
respectively characterized using the same parameters of cables 1, 2 and 3 in Example 2. In
the form-finding analysis, the messenger and contact wires have constant tensions 15 kN.
Results obtained using the proposed method are provided and analyzed below.

5.1 Results on control

Figure 21 shows the variations of tensions of the messenger and contact wires at joints in
the time domain, respectively. As clearly shown in Fig. 21, the messenger and contact wires
hold constant tensions of 15 kN at joints after they reach prescribed equilibrium states. It is
also interesting to note that the tensions of the messenger at joints C and D have a growth
rate, while the tensions of the contact wire at joints A and B have another growth rate. In
current research, the contact and messenger wires respectively attain prescribed equilibrium
states in the first two steps. In the first step, the tension of the contact wire is adjusted under
the control parameters of vs = 0.2 mm/s, T0 = 5 N and T1 = 100 N. Due to symmetry, the
tensions of the contact wire at joints A and B have a growth rate. In the second step, the
tension of the messenger wire is adjusted under the control parameters of vs = 0.4 mm/s,
T0 = 20 N and T1 = 500 N. Therefore, two distinct growth rates exist in tension curves of
Fig. 21.
Figure 22 shows the equilibrium configuration of the contact wire. As can be observed in
Fig. 22, the contact wire has the maximum sag of about 6 mm. It can be also found in Fig. 22
that all connection points between the contact wire and the droppers are at the same relative
height. As pointed out above, no frictions exist in sliding joints. Therefore, the droppers
are immovably placed at previous locations along the track after the controls, as shown in
Fig. 22.
242 C.J. Yang et al.

Fig. 21 Variations of tensions of


the messenger and contact wires
at joints during the controls

Fig. 22 The shape of contact wire at the equilibrium state

From Figs. 21 and 22, one can conclude that the catenary at computed configuration is
responsible for all requests in the pre-design. It is remarkably noted that a time-dependent
gravitational coefficient g is adopted for an efficient numerical integration, which is given
as

9.8( tt1 )2 (3 − 2 tt1 ), 0 < t ≤ t1
g(t) = (31)
9.8, t > t1
where t1 is a time constant. The catenary structure in this problem has a three-span length
of 195 m. As it is well known, a long flexible structure can strongly vibrate under its own
gravity force in dynamic simulation. High frequency components of the structural vibration
harmfully influence numerical integration owing to stiffness effects. The use of a variable
coefficient g leads to the structure at a quasi-equilibrium state in simulation, and integration
Static form-finding analysis of a railway catenary using a dynamic. . . 243

Fig. 23 Two configurations of the catenary in simulation

conditions are improved. Note that the selection of control parameters does not influence
final results.

5.2 Equilibrium configuration of the catenary

Figure 23 shows a starting configuration and a computed equilibrium configuration of the


catenary, respectively. In the current research, the configuration of the catenary in Fig. 23(a)
is employed as an input for the next form-finding iteration. By dynamic simulation under
control conditions, the configuration of the catenary in Fig. 23(b), which is evolved from the
starting configuration, is computed. The equilibrium configuration in Fig. 23(b) is equiv-
alent to the initial static equilibrium configuration of the catenary since characterization
conditions are attained. Note that high difference H in Fig. 20 is 1.5 m.
244 C.J. Yang et al.

Fig. 24 Tension distribution


along the messenger wire

Fig. 25 Tension distribution


along the contact wire

5.3 Tension distribution in the messenger and contact wires

Tensions in the messenger and contact wires are calculated using Eq. (11). Figure 24 shows
the tension distribution in the messenger wire. As shown in Fig. 24, the average tension in
the messenger wire is slightly more than 15 kN. The wire tension is largely dependent on
the axial strain. Both the axial stretching owing to the wire tensile force and the bending
of the wire contribute to the axial strain. The messenger wire subjected to the gravity force
and the forces at joints obviously bends downward with large sags, as shown in Fig. 23(b).
Thus, the messenger wire has a higher tension at the location where large bending exists. It
is the reason that the tension curve in Fig. 24 has sharp changes near the joints E and F .
Due to similar reasons, it can be observed in Fig. 24 that the tensions near connection points
between the droppers and the messenger wire sharply vary. As shown in Fig. 22, the contact
wire is suspended with small sags by the droppers. As a result, the contact wire has a com-
paratively uniform tension distribution. The tensions of the contact wire near connection
points slightly fluctuate due to the dropper effects; see Fig. 25.

5.4 Lengths and tensions of the dropper

By the calculation, lengths of the dropper at the equilibrium state are listed in Table 5. Ten-
sions of the dropper are also given, see Table 5. As shown in Table 5, lengths and tensions
of the dropper are symmetric within the structure. The second and 20th droppers have max-
imum tensions of about 87.72 N. The 11th dropper has the shortest length of 0.98 m.
Static form-finding analysis of a railway catenary using a dynamic. . . 245

Table 5 Lengths and tensions of the dropper at the equilibrium state

No. Length Tension No. Length Tension No. Length Tension


(m) (N) (m) (N) (m) (N)

1 1.38 63.35 8 1.38 80.51 15 1.38 80.51


2 1.16 87.72 9 1.16 87.51 16 1.16 87.51
3 1.03 87.42 10 1.03 87.42 17 1.03 87.42
4 0.99 87.43 11 0.98 87.43 18 0.99 87.43
5 1.03 87.42 12 1.03 87.42 19 1.03 87.42
6 1.16 87.51 13 1.16 87.51 20 1.16 87.72
7 1.38 80.51 14 1.38 80.51 21 1.38 63.35

6 Conclusions

In this paper, a dynamic equilibrium method is proposed to determine the initial static equi-
librium configuration of the railway catenary. The proposed method is based on the evo-
lution of the catenary configuration by dynamic simulation of the catenary model under
control conditions. In the proposed method, an arbitrary starting configuration of the cate-
nary that is easily described is considered as an input for the form-finding iteration. The
catenary at the starting configuration is modeled using ANCF cable elements with absolute
three-dimensional nodal coordinates in the flexible multibody system formulation. The cate-
nary model established allows for large displacements, deformations and high nonlinearities
in the wires. Control conditions that characterize the catenary initial equilibrium configura-
tion are addressed in terms of nodal coordinates of cable elements and implemented in the
numerical procedure by nonholonomic constraint equations. By dynamic simulation of the
catenary model under control conditions, the catenary reaches a prescribed dynamic equi-
librium state, which is dynamically equivalent to the initial equilibrium state. Though the
developed method is dependent on controls, the selection of control parameters has no influ-
ence on the obtained results at equilibrium state. The developed method is general and can
be applied to catenaries with complex configurations.
The developed method is validated using two examples. In Example 1, the comparison
between numerical and analytical solutions of the tensions in bars shows good performance
of the ANCF cable element and the validity of the height control model. In Example 2,
the initial equilibrium configuration of a three-cable tension structure is analyzed using the
developed method. Results show the structure at the configuration computed is responsible
for requests in the pre-design. Finally, the proposed method is applied to a two dimensional
simple railway catenary. The obtained results are properly interpreted.

Acknowledgements The authors would firstly like to thank the editor Prof. Jorge Ambrosio and anony-
mous Reviewers for their constructive and detailed comments on improving this paper.
The first author thanks the supports from the China Postdoctoral Science Foundation (No. 2013M542290)
and the Project of State Key Laboratory of Traction Power, Southwest Jiaotong University (No. 2015TPL_
T03). The second author thanks the support from the High Speed Railway Basic Research Fund Key Project
of China (No. U1234208). The third author thanks the support from the National Natural Science Foundation
of China (Grant No. 51405401). The fourth author thanks the support from Applied Basic Research Programs
of Sichuan Province (No. 2014JY0078). The fifth author thanks the support from the National Natural Science
Foundation of China (Grant No. 51475391).
246 C.J. Yang et al.

References
1. Ambrósio, J., Pombo, J., Pereira, M., Antunes, P., Mósca, A.: Recent developments in pantograph-
catenary interaction modelling and analysis. Int. J. Railw. Technol. 1(1), 249–278 (2012)
2. Pombo, J., Antunes, P.: A comparative study between two pantographs in multiple pantograph high-
speed operations. Int. J. Railw. Technol. 2(1), 83–108 (2013)
3. Pombo, J., Antunes, P., Ambrósio, J.: A study on multiple pantograph operations for high-speed catenary
contact. In: Proceedings of the Eleventh International Conference on Computational Structures Technol-
ogy, Stirlingshire, United Kingdom (2012)
4. Ambrósio, J., Pombo, J., Pereira, M., Antunes, P., Mósca, A.: A computational procedure for the dynamic
analysis of the catenary–pantograph interaction in high-speed trains. J. Theor. Appl. Mech. 50(3), 681–
699 (2012)
5. Ambrósio, J., Pombo, J., Antunes, P., Pereira, M.: PantoCat statement of method. Veh. Syst. Dyn. 53(3),
314–328 (2015). doi:10.1080/00423114.2014.969283
6. Cho, Y.H.: SPOPS statement of methods. Veh. Syst. Dyn. 53(3), 329–340 (2015). doi:10.1080/00423114.
2014.953182
7. Jönsson, P.-A., Stichel, S., Nilsson, C.: CaPaSIM statement of methods. Veh. Syst. Dyn. 53(3), 341–346
(2015). doi:10.1080/00423114.2014.999799
8. Collina, A., Bruni, S., Facchinetti, A., Zuin, A.: PCaDA statement of methods. Veh. Syst. Dyn. 53(3),
347–356 (2015). doi:10.1080/00423114.2014.959027
9. Ikeda, M.: ‘Gasen-do FE’ statement of methods. Veh. Syst. Dyn. 53(3), 357–369 (2015). doi:10.1080/
00423114.2014.968174
10. Massat, J.-P., Balmes, E., Bianchi, J.-P., Van Kalsbeek, G.: OSCAR statement of methods. Veh. Syst.
Dyn. 53(3), 370–379 (2015). doi:10.1080/00423114.2015.1005016
11. Zhou, N., Lv, Q., Yang, Y., Zhang, W.: <TPL-PCRUN> statement of methods. Veh. Syst. Dyn. 53(3),
380–391 (2015). doi:10.1080/00423114.2014.982136
12. Sánchez-Rebollo, C., Carnicero, A., Jiménez-Octavio, J.R.: CANDY statement of methods. Veh. Syst.
Dyn. 53(3), 392–401 (2015). doi:10.1080/00423114.2014.982135
13. Tur, M., Baeza, L., Fuenmayor, F.J., García, E.: PACDIN statement of methods. Veh. Syst. Dyn. 53(3),
402–411 (2015). doi:10.1080/00423114.2014.963126
14. Veenendaal, D., Block, P.: An overview and comparison of structural form finding methods for general
networks. Int. J. Solids Struct. 49(26), 3741–3753 (2012). doi:10.1016/j.ijsolstr.2012.08.008
15. Shan, Q., Zhai, W.M.: A macroelement method for catenary mode analysis. Comput. Struct. 69(6), 767–
772 (1998). doi:10.1016/S0045-7949(98)00130-8
16. Park, T.-J., Han, C.-S., Jang, J.-H.: Dynamic sensitivity analysis for the pantograph of a high-speed rail
vehicle. J. Sound Vib. 266(2), 235–260 (2003). doi:10.1016/S0022-460X(02)01280-4
17. Argyris, J.H., Angelopoulos, T., Bichat, B.: A general method for the shape finding of lightweight
tension structures. Comput. Methods Appl. Mech. Eng. 3(1), 135–149 (1974). doi:10.1016/0045-
7825(74)90046-2
18. Schek, H.J.: The force density method for form finding and computation of general networks. Comput.
Methods Appl. Mech. Eng. 3(1), 115–134 (1974). doi:10.1016/0045-7825(74)90045-0
19. Aboul-Nasr, G., Mourad, S.A.: An extended force density method for form finding of constrained cable
nets. Case Stud. Struct. Eng. 3, 19–32 (2015). doi:10.1016/j.csse.2015.02.001
20. Zhang, J.Y., Ohsaki, M.: Adaptive force density method for form-finding problem of tensegrity struc-
tures. Int. J. Solids Struct. 43(18–19), 5658–5673 (2006). doi:10.1016/j.ijsolstr.2005.10.011
21. Lopez-Garcia, O., Carnicero, A., Torres, V.: Computation of the initial equilibrium of railway
overheads based on the catenary equation. Eng. Struct. 28(10), 1387–1394 (2006). doi:10.1016/j.
engstruct.2006.01.007
22. Such, M., Jimenez-Octavio, J.R., Carnicero, A., Lopez-Garcia, O.: An approach based on the catenary
equation to deal with static analysis of three dimensional cable structures. Eng. Struct. 31(9), 2162–2170
(2009). doi:10.1016/j.engstruct.2009.03.018
23. Cho, Y.H., Lee, K., Park, Y., Kang, B., Kim, K.-N.: Influence of contact wire pre-sag on the dy-
namics of pantograph–railway catenary. Int. J. Mech. Sci. 52(11), 1471–1490 (2010). doi:10.1016/j.
ijmecsci.2010.04.002
24. Tur, M., García, E., Baeza, L., Fuenmayor, F.J.: A 3D absolute nodal coordinate finite element
model to compute the initial configuration of a railway catenary. Eng. Struct. 71, 234–243 (2014).
doi:10.1016/j.engstruct.2014.04.015
25. Yang, C.J., Hong, D.F., Ren, G.X., Zhao, Z.H.: Cable installation simulation by using a multibody dy-
namic model. Multibody Syst. Dyn. 30(4), 433–447 (2013). doi:10.1007/s11044-013-9364-9
26. Schwab, A.L., Meijaard, J.P.: Comparison of three-dimensional flexible beam elements for dynamic
analysis: classical finite element formulation and absolute nodal coordinate formulation. J. Comput.
Nonlinear Dyn. 5(1), 011010 (2009). doi:10.1115/1.4000320
Static form-finding analysis of a railway catenary using a dynamic. . . 247

27. Gerstmayr, J., Shabana, A.A.: Analysis of thin beams and cables using the absolute nodal co-ordinate
formulation. Nonlinear Dyn. 45(1), 109–130 (2006). doi:10.1007/s11071-006-1856-1
28. Kobayashi, S., Nomizu, K.: Foundations of Differential Geometry. Wiley-Interscience, New York (1996)
29. Abbas, L.K., Rui, X., Hammoudi, Z.S.: Plate/shell element of variable thickness based on the absolute
nodal coordinate formulation. Proc. Inst. Mech. Eng., Proc., Part K, J. Multi-Body Dyn. 224(2), 127–141
(2010)
30. Schwab, A.L., Gerstmayr, J., Meijaard, J.P.: Asme: comparison of three-dimensional flexible thin plate
elements for multibody dynamic analysis: finite element formulation and absolute nodal coordinate for-
mulation. In: Proceedings of the Asme International Design Engineering Technical Conferences and
Computers and Information in Engineering Conference 2007, Vol. 5, Parts A–C (2008)
31. Gerstmayr, J., Matikainen, M.K., Mikkola, A.M.: A geometrically exact beam element based on the
absolute nodal coordinate formulation. Multibody Syst. Dyn. 20(4), 359–384 (2008). doi:10.1007/
s11044-008-9125-3
32. Lee, S.-H., Park, T.-W., Seo, J.-H., Yoon, J.-W., Jun, K.-J.: The development of a sliding joint for very
flexible multibody dynamics using absolute nodal coordinate formulation. Multibody Syst. Dyn. 20(3),
223–237 (2008). doi:10.1007/s11044-008-9109-3
33. Sugiyama, H., Escalona, J.L., Shabana, A.A.: Formulation of three-dimensional joint constraints using
the absolute nodal coordinates. Nonlinear Dyn. 31(2), 167–195 (2003). doi:10.1023/a:1022082826627
34. García-Vallejo, D., Escalona, J.L., Mayo, J., Domínguez, J.: Describing rigid-flexible multibody systems
using absolute coordinates. Nonlinear Dyn. 34(1), 75–94 (2003). doi:10.1023/B:NODY.0000014553.
98731.8d
35. Haug, E.J.: Computer Aided Kinematics and Dynamics of Mechanical Systems: Basic Methods, Vol. I.
Allyn, Bacon (1989)
36. Shabana, A.A.: Dynamics of Multibody Systems. Cambridge University Press, Cambridge (2013)
37. Hairer, E., Nørsett, S.P., Wanner, G.: Solving Ordinary Differential Equations II: Stiff and Differential-
Algebraic Problems. Springer, Berlin (1996)

View publication stats

You might also like