You are on page 1of 19

This article was downloaded by: [USP University of Sao Paulo]

On: 02 April 2013, At: 07:40


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Vehicle System Dynamics: International


Journal of Vehicle Mechanics and
Mobility
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/nvsd20

Real-time catenary models for the


hardware-in-the-loop simulation of the
pantograph–catenary interaction
a a a
A. Facchinetti , L. Gasparetto & S. Bruni
a
Department of Mechanical Engineering, Politecnico di Milano, Via
La Masa 1, 20156, Milan, Italy
Version of record first published: 17 Dec 2012.

To cite this article: A. Facchinetti , L. Gasparetto & S. Bruni (2013): Real-time catenary models
for the hardware-in-the-loop simulation of the pantograph–catenary interaction, Vehicle System
Dynamics: International Journal of Vehicle Mechanics and Mobility, 51:4, 499-516

To link to this article: http://dx.doi.org/10.1080/00423114.2012.748920

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation
that the contents will be complete or accurate or up to date. The accuracy of any
instructions, formulae, and drug doses should be independently verified with primary
sources. The publisher shall not be liable for any loss, actions, claims, proceedings,
demand, or costs or damages whatsoever or howsoever caused arising directly or
indirectly in connection with or arising out of the use of this material.
Vehicle System Dynamics, 2013
Vol. 51, No. 4, 499–516, http://dx.doi.org/10.1080/00423114.2012.748920

Real-time catenary models for the hardware-in-the-loop


simulation of the pantograph–catenary interaction
A. Facchinetti, L. Gasparetto and S. Bruni*
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

Department of Mechanical Engineering, Politecnico di Milano, Via La Masa 1, 20156 Milan, Italy

(Received 16 July 2012; final version received 7 November 2012 )

Hardware-in-the-loop (HIL) simulation is a promising technique to study the pantograph–catenary


interaction problems by realising the interaction of a physical pantograph with a mathematical model
of the overhead equipment (catenary). However, the computing power presently available on real-time
CPUs only allows to run simplified models of the overhead equipment. Therefore, it is important to
define catenary models that are suitable for real-time simulation and at the same time capable of accu-
rately representing the dynamic behaviour of the catenary. In this paper, the use of a catenary model
based on modal superposition is considered, and the effect of changing the number of modelled spans
and the number of modal components allocated to the contact and messenger wires is investigated in
view of finding the best model compatible with real-time simulation. Comparisons between HIL sim-
ulation results and line measurements are presented, to quantify the accuracy of the hybrid simulation
method developed.

Keywords: pantograph–catenary interaction; hardware-in-the-loop; hybrid simulation; catenary


models

1. Introduction

The hybrid simulation approach, also known as ‘hardware-in-the-loop simulation’ (HILS),


is based on the integration between a hardware component and a virtual simulation model,
based on the use of a suitable testing apparatus and of a real-time computing facility, bridging
the gap between pure numerical simulation and the conventional physical testing of hardware
components.
With respect to the numerical simulation, HILS removes the modelling errors associated
with defining a numerical model for the hardware component, which may be highly relevant to
the accuracy of the overall simulation process on account of simplifying assumptions implied
by the modelling (e.g. neglecting nonlinear effects, flexibility of bodies, etc.) and of uncer-
tainty in the model parameters. Furthermore, HILS allows to reproduce highly realistic testing
conditions for the hardware, thereby representing an efficient approach for rapid prototyping
and validation of complex engineering systems.
Applications of the HILS method have been proposed in several fields of engineering.
Initially, the method was mainly applied to the testing of electronic drives and controllers.
However, in recent years the concept has found application in mechanical, aerospace and

*Corresponding author. Email: stefano.bruni@polimi.it

© 2013 Taylor & Francis


500 A. Facchinetti et al.

structural engineering, in which the hybrid simulation entails significant power flow between
the hardware tested and the interface devices in the HILS test-rig, thereby making the design
of the HILS experiment more challenging. Examples of the application of HILS in the
above-mentioned fields include various types of active suspension and safety systems for
road vehicles [1–4], power train components [5], actuators for flight control in aircrafts [6]
and the assessment of the behaviour of structures under seismic loads [7]. Applications
of HILS in the railway field have mainly addressed the study of the traction and braking
systems [8–10].
This paper deals with the HILS of the pantograph–catenary dynamic interaction. This is
obtained by testing a physical pantograph subjected to dynamic excitation defined by means
of a mathematical model of the overhead equipment. To this aim, two approaches have been
proposed, called ‘open-loop’ and ‘closed-loop’ HILS [11]. In the ‘open-loop’ HILS, a sim-
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

ulated or measured time history of collector motion is applied on the pantograph, and the
corresponding contact force is measured. Eventually, the measured force reacted by the pan-
tograph can be introduced (off-line) into a mathematical model of the catenary to provide
a better estimate of collector motion and the test repeated with the new collector motion
estimate.
In the ‘closed-loop’ HILS, the actuation system is driven by a real-time board where the
force applied by the collector head to the contact wire is processed according to a catenary
model, cf. Section 2. This second approach is more accurate than the ‘open-loop’ HILS, as it
accounts for the feedback effect associated with the change in the collector head displacements
produced by the force reacted by the pantograph, but requires the use of a simplified catenary
model, to achieve real-time computation time. Hence, the main benefit which is sought for in
this research is to have a simulation method for the pantograph–catenary interaction which is
free from modelling errors as far as the pantograph is concerned, though enabling a sufficiently
accurate mathematical modelling of the catenary dynamic behaviour. It should be recalled that
pantographs may be difficult to model due to nonlinearities, presence of servo-components,
effect of friction and backlashes.
Catenary models which can be used for ‘closed-loop’ HILS need to be highly efficient from
the point of view of CPU usage, in order to allow real-time simulation, but at the same time
should be accurate enough to account for the main flexibility and inertial effects affecting the
dynamic behaviour of the catenary when subjected to pantograph passage. The aim of this paper
is therefore to investigate the effect of the catenary model on the HILS of pantograph–catenary
interaction.
Considering a class of models based on modal superposition (called hereafter ‘real-time
model’), the effect of changing the number of modelled spans and the number of modal
components used to describe the vibration of the contact and messenger wires was investigated
in view of finding the best model compatible with real-time simulation. This process was
performed by comparing alternative configurations of the real-time catenary model using as
term of reference a finite-element (FE) model of the catenary. Then, the different versions of
the real-time catenary model were compared in terms of HILS performance, to confirm the
selection of the best performing model.
The paper is organised as follows: in Section 2, the principle of HILS as applied to pan-
tographs is briefly outlined, making reference to previously published work. In Section 3,
the FE and real-time models of the overhead equipment are introduced. In Section 4, the
numerical experiments performed to assess the performances of different real-time catenary
models compared with the reference FE model are described, while Section 5 presents the
results of HIL experiments for different real-time catenary models and comparisons with
line measurements of the pantograph–catenary interaction. Finally, conclusions are drawn in
Section 6.
Vehicle System Dynamics 501

2. HIL simulation of the pantograph–catenary interaction

A concept for the ‘closed-loop’ HIL simulation of the pantograph–catenary interaction was
initially proposed by Zhang et al. [12,13] based on a simple linear model of the catenary. This
was extended by Resta et al. [14] to include the nonlinear effects related with the slackening
of droppers and further by Facchinetti and Bruni [15] introducing the independent actuation
of the two collectors as well as the effect of the stagger and comparing results to line test
measurements.
The concept is schematically shown in Figure 1. The contact forces generated at the two
collector strips in the pantograph head are measured and fed in a real-time board which
computes the vertical displacements of the contact wire at the collector strips, based on a
mathematical model of the overhead equipment described as a flexible continuum. The contact
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

wire displacements produced by the real-time calculation are then fed back to the pantograph
head by means of suitable actuators, so that the dynamic interaction of the two systems is
represented. The frequency range addressed by this HIL simulation is 0–20 Hz, corresponding
to the range of frequency used in line testing of pantographs to evaluate the quality of the current
collection.
The hardware-in-the-loop (HIL) testing principle introduced above is implemented on a
dedicated test-rig as shown in Figure 2, which allows to apply on the pantograph head vertical
displacements reproducing the contact with the oscillating contact wire. The test-rig also
reproduces the periodic change of the excitation point position across the pantograph head
caused by the stagger in the contact wire.
The rig is composed of a vertical actuation system consisting of two hydraulic actuators, a
lateral actuation system consisting of a slider moved by an AC motor via a screw and spindle,
and the real-time board. The vertical actuators are operated in stroke control mode based on the
references elaborated by the real-time board as shown in Figure 1. The lateral actuation system
moves the vertical actuators across the pantograph head in the lateral direction according to
a triangular waveform whose amplitude and frequency parameters are defined based on the
stagger geometry and pantograph speed being simulated during the test. The vertical and lateral
actuation systems are both mounted on a reaction frame which is designed to have negligible
deformation up to 30 Hz.

Figure 1. Concept of hybrid simulation of the pantograph–catenary interaction.


502 A. Facchinetti et al.
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

Figure 2. The HIL test-rig. Left: photo; right: layout and components.

The real-time board driving the HIL test-rig is a dSpace DS1006 model, with a Quad-core
AMD Opteron processor with clock frequency 2.6 GHz and 256 MB ram. The communication
between the board and the test bench is performed via analog/digital and digital/analog
converter units embedded in the board working at a sampling frequency of 2 kHz.

3. Mathematical models of the overhead equipment

Railway catenaries are flexible structures that provide a sliding contact to the pantographs
mounted on railway vehicles and therefore enable the transfer of electrical energy from the
railway infrastructure to the trains. A typical construction includes:

• a messenger wire, suspended to masts and presenting a curvature in the vertical plane due
to gravitational forces (hence the name of ‘catenary’);
• a contact wire, making contact to the pantograph(s);
• droppers suspending the contact wire to the messenger wire;
• registration and steady arms which connect the contact wire to the masts and produce the
stagger.

The catenary considered in this paper is a modified version of the Italian High-Speed
catenary C270, with tension increased to 30 kN to raise the maximum service speed of the
line. The number of droppers in each span is nine and the total span length is 57 m. Table 1
presents the main mechanical data of the catenary.

Table 1. Mechanical properties of the modified C270 catenary with increased


contact wire tension.

Mechanical tension (kN) Mass per unit length (kg/m)

Contact wire 30 1.35


Messenger wire 16 1.08
Vehicle System Dynamics 503

In this paper, two different levels of modelling are considered for the overhead equipment:
a FE model, which is too demanding in terms of the computational effort to be imple-
mented on present real-time processors, and a family of simplified real-time catenary models
based on the modal superposition principle, which allow a mathematical formulation sim-
ple enough to be run in real time at the price of introducing some simplifying assumptions.
The more accurate FE model, described in Section 3.1, is considered here as a reference
to assess the accuracy of the simpler real-time catenary models, which are introduced in
Section 3.2.

3.1. The FE model of the overhead equipment


Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

The use of FE catenary models has become the state-of-the-art in the modelling and numerical
simulation of the pantograph–catenary interaction [16]. Therefore, the reference catenary
model considered in this paper is a FE one which allows the detailed reproduction of the
structural behaviour for a wide range of overhead equipments in use in railways.
The contact wires, messenger wires, auxiliary wires, registration arms and suspensions are
schematised using 3D Euler–Bernoulli tensioned beam elements. Droppers are represented
as nonlinear lumped elements, considering the experimental force–displacement curve as
measured from laboratory tests [17]. The inertial effects associated with the end mountings of
the droppers are represented by means of lumped mass elements.
The catenary equations are treated in their complete nonlinear form to derive the static
configuration and static internal stress state in the catenary corresponding to the gravitational
loads applied and then linearised around the static equilibrium position. The linearised equa-
tions are used in the dynamic simulation of the pantograph–catenary interaction. Inertial effects
are accounted for in the linearised catenary equations according to a consistent mass-matrix
formulation. Catenary damping is introduced in the linearised catenary equations according
to the ‘proportional damping’ model.

3.2. The real-time models of the overhead equipment

The real-time catenary models considered in this paper are based on the modal superposition
principle, allowing the efficient implementation of the numerical integration procedure in
terms of the required CPU effort, and yet providing an accurate representation of the space-
varying flexibility of the overhead equipment. The real-time models also account for the
nonlinear effects related with dropper slackening. As shown in Figure 3, real-time models
consist of:

• the contact and messenger wires, modelled as tensioned beams, with pinned–pinned
boundary conditions at the extremities of the domain modelled;
• the droppers, modelled as bi-linear force elements connecting the two wires;
• the catenary suspensions at the masts, modelled as lumped visco-elastic elements having
stiffness ks and viscous damping cs ;
• the registration arms, represented as lumped masses mr .

The model only represents a limited length of the overhead equipment, with a minimum
of three and a maximum of five spans. To minimise the effect of boundary conditions, the
pantograph is kept within the central portion of the model: this requires that the catenary model
is periodically ‘forward shifted’ each time the pantograph gets too close to the boundaries of
the modelled section.
504 A. Facchinetti et al.

ks cs ks cs

rm, Am, EJm, Sm

x
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

mr fp2 fp1 mr rc, Ac, EJc, Sc

Figure 3. The real-time catenary model.

The vertical displacements of the contact wire zc and of the messenger wire zm are obtained
according to the following expressions:


Nc  
iπ x
zc (x,t) = c (x)T qc (t) = sin qci (t),
i=1
LT
  (1)

Nm
iπ x
zm (x,t) = m (x)T qm (t) = sin qmi (t),
i=1
LT

where LT is the total length of the model, Nc and Nm are the number of modal coordinates of
the contact and messenger wires, respectively, qci and qmi are their modal coordinates, qc and
qm are the vectors collecting them, the vectors c and m collect the sinusoidal modal shape
functions.
Using Lagrange equations, the catenary equations of motions are written in the following
form:

Mq̈ + Cq̇ + Kq = Qd (q, t) + Qp (q, t) (2)

in which: q(t) = {qTc qTm }T , Qd is the vector of the Lagrangian components of the forces
generated by the droppers, and vector Qp contains the Lagrangian components of the forces
applied by the pantograph.
The mass matrix M accounts for inertia effects in the contact and messenger wires and in
the registration arms, and takes the following form:
 
Mc + Mr 0
M= , (3)
0 Mm

where Mc is a Nc × Nc diagonal matrix containing the modal masses associated with the contact
wire, Mm is the diagonal modal mass matrix of the messenger wire having size Nm × Nm and
Mr is a Nc × Nc non-diagonal matrix accounting for the effect of the registration arms modelled
Vehicle System Dynamics 505

as concentrated masses. These matrices are defined as follows:

1 1
Mc = ρc Ac LT I, Mm = ρm Am LT I, (4)
2 2

Ns
Mr = c (iL) mr c (iL)T , (5)
i=0

with ρc and ρm being the density of the contact and messenger wire materials, Ac and Am the
area of the section for the same wires, LT the total length of the model and L the length of one
span.
The stiffness matrix K accounts for the elastic forces associated with the contact and
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

messenger wires and with the suspensions. The expression is


 
Kc 0
K= (6)
0 Km + Ks

with Kc being the diagonal matrix containing the modal stiffness terms associated with the
contact wire, Km the diagonal modal stiffness matrix of the messenger wire and Ks the stiffness
matrix accounting for the messenger wire suspension:
⎡     ⎤
π2 π2 LT i2 π 2 i2 π 2 LT
⎢ EJc + Sc ··· EJc + Sc · · ·⎥
⎢ L2 L2 2 LT2 LT2 2 ⎥
Kc = diag ⎢ 2 T2  T2 2  ⎥
⎣ Nc π Nc π LT ⎦
2 2
EJc + Sc
LT LT 2
⎡ 2  2    ⎤
π π LT i2 π 2 i2 π 2 LT
⎢ EJm + Sm ··· EJm + Sm · · ·⎥
⎢ L2 L2 2 LT2 LT2 2 ⎥
Km = diag ⎢ 2 T2  T2 2  ⎥ (7)
⎣ Nc π Nc π LT ⎦
2 2
EJm + Sm
LT LT 2

Ns
Ks = s (iL) ks s (iL)T (8)
i=0

with EJ c and EJ m being the bending stiffness of the contact and messenger wires, Sc and Sm
the tension in the same wires, ks the stiffness of the messenger wire suspension, LT the total
length of the model and L the length of one span.
Finally, the damping matrix C is defined as the sum of two terms: the first one, represent-
ing the structural damping in the contact and messenger wires, is defined according to the
proportional damping assumption; the second one, accounting for the lumped dashpot in the
messenger wire suspension, has a form similar to that of matrix Ks .
Equation (2) are numerically integrated in real time using the Euler integration scheme with
a time step of 0.5 ms. From this result, the vertical displacements of the contact wire at the
instantaneous position of the two collector strips are computed using Equations (1).
As mentioned above, a limited portion of the overhead line, consisting of three to five spans,
is modelled, and the catenary model is periodically shifted forward during the simulation, so
that the pantograph is always contained in the central spans of the model, to reduce the effect
of boundary conditions. Details about the ‘shift forward’ procedure are provided in [15].
506 A. Facchinetti et al.

4. Numerical experiments to optimise the real-time catenary model

The first step taken to choose the real-time catenary model consists of a series of numerical
experiments to compare different modelling options in terms of:
• number of modal coordinates for the contact and messenger wire;
• total number of catenary spans;
• position across the span where the ‘shift forward’ procedure is applied.
It was found from experiments that a maximum number of 185–190 modal coordinates
can be used in the model in order to ensure the correct integration in real time of the model
using the HIL test-rig CPU. Since the contact wire directly interacts with the pantograph,
the number of modal coordinates used to model the vibration of the contact wire shall be
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

greater than the number of coordinates allocated to the messenger wire. The adoption of
a lower number of coordinates for the messenger wire than for the contact wire is a rather
common option in catenary modelling, aimed at minimising the required computational effort.
For instance, Zhang et al. [18] models both wires with tensioned beam elements considering
different element lengths for the contact wire and for the messenger wire (0.25 and 0.5 m,
respectively), while both Lee [19] and Alberto et al. [20] represent the contact wire with
tensioned beam elements and the messenger wire with string elements, keeping the same
element lengths for both wires.
Four alternative configurations of the real-time model were compared, corresponding to
the different number of coordinates allocated to the contact and messenger wires and to the
different numbers of spans modelled, as detailed in Table 2. With respect to the number of
coordinates, two different options were considered, respectively, consisting in the allocation of
two-fifths of the total modal coordinates to the messenger wire and three-fifths to the contact
wire (75–115) and to the allocation of approximately one-fourth of the total coordinates to
the messenger wire and three-fourths to the contact wire (50–135). Note that the number of
total coordinates is slightly different for the two options, since the number of DOFs allocated
to the contact wire influences the computational effort required by other tasks that have to be
executed in real time than the pure numerical integration, such as the evaluation of the motion
of the contact point.
Concerning the number of spans, the two configurations modelling five spans of catenary
enable the pantograph to travel a longer distance before the ‘shift-forward’ process needs to be
applied, and therefore are less sensitive to the numerical perturbation entailed by this process.
Furthermore, for the five-span configuration the ‘shift-forward’ process can be applied both
when the pantograph is at the end of the third span or in the centre of the fourth span, whereas
for the three-span configurations the ‘shift-forward’ can only be taken when the pantograph
reaches the end of the second span.
Table 2 also reports for each configuration the minimum wavelength of deformation mod-
elled in the contact wire and in the messenger wire, λc,min and λm,min , respectively, computed

Table 2. Alternative configurations for the real-time catenary model considered in the sensitivity analysis.

Number of Minimum wavelength Minimum wavelength


Number of messenger wire modelled in the messenger Number of contact modelled in the contact
Name spans ns coordinates Nm wire λm,min (m) wire coordinates Nc wire λc,min (m)

3S-75-115 3 75 4.56 115 2.97


3S-50-135 3 50 6.84 135 2.53
5S-75-115 5 75 7.60 115 4.95
5S-50-135 5 50 11.40 135 4.22
Vehicle System Dynamics 507

according to the following formula:

2LT 2ns L 2LT 2ns L


λc,min = = ; λm,min = = (9)
Nc Nc Nm Nm

with Nc and Nm being the number of modal coordinates for the contact and messenger wires,
respectively, ns the number of spans in the model, L the length of one span (57 m) and LT the
total length of the model.
The alternative configurations of the real-time model are assessed with respect to their
capability to reproduce the catenary space-varying stiffness (Section 4.1) and to produce
results comparable to the FE model in terms of dynamic interaction with a lumped mass
model of a pantograph (Section 4.2).
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

4.1. Analysis of the space-varying catenary stiffness

One critical issue with the modelling of the catenary is to reproduce the space-varying stiffness
which results from the geometry of the spans and from the different stiffness of the contact wire
at the droppers and in between two consecutive droppers. Therefore, a first set of analyses was
performed to evaluate the capability of the different real-time models in Table 2 to reproduce
this effect, using the FE model as a reference. The stiffness of the catenary was evaluated
considering two different contact force values, 100 and 200 N to investigate the importance
of nonlinear effects related with droppers slackening.
Figure 4 shows the spatial variation of the contact wire stiffness for two different contact
force values (100 and 200 N, respectively). The results of the real-time model configurations
listed in Table 2 are compared to the reference FE model. The overall shape of the diagrams
shows a higher stiffness near the catenary suspensions (x = 0 and x = 57 m), and local max-
ima at the droppers. The stiffness peaks corresponding to the first and last dropper in the
span are significantly affected by the contact force assumed in the calculation, because con-
sidering the higher force value of 200 N slackening occurs at these droppers. Furthermore, it
is observed that the largest deviations of the real-time model results with respect to the FE
model also occur at the first and last dropper in the span: this is due to the fact that in the FE
calculation the measured force–deformation characteristic curve is considered, whereas in the
real-time models a simpler bi-linear model is used to describe the slackening of droppers (cf.
Section 3.2).
It is observed that the 5s-50-135 configuration deviates significantly from the reference
FE model: this is because for this configuration, the minimum wavelength of deformation
represented in the messenger wire is 5.7 m, resulting in an artificial increase in the catenary
stiffness at the masts, which affects substantially the results of the static calculation presented
here. For this reason, this model configuration is not considered in the next steps of the analysis.
The other three configurations of the real-time model show minor differences, with the two
configurations modelling three spans showing a slightly better agreement with the FE model
than the 5s-75-115 configuration.

4.2. Simulation comparison

A second approach used to assess the performances of the real-time model configurations
consists in comparing the results of pantograph–catenary interaction simulations. To this aim,
the case of a single ATR95-25kV Italian High-Speed pantograph travelling at 300 km/h is
considered, and a three-mass-lumped model of the pantograph is used.
508 A. Facchinetti et al.

8000
F.E.M .
3s−75−115
3s−50−135
7000
5s−75−115
5s−50−135
Stiffness (N/m)

6000

5000
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

4000

3000

0 14.25 28.5 42.75 57


x (m)

8000
F.E.M
3s−75−115
3s−50−135
7000
5s−75−115
5s−50−135
Stiffness (N/m)

6000

5000

4000

3000

0 14.25 28.5 42.75 57


x (m)

Figure 4. Spatial variation of the contact wire stiffness. Above: contact load 100 N; below: contact load 200 N.

As mentioned above, the 5s-50-135 configuration of the real-time model is not considered
here, on account of the poor capability to reproduce the catenary stiffness. Instead, for the
5s-75-115 configuration, two alternatives are considered, one with the shifting process hap-
pening under the mast at the end of the third span, denoted as ‘5s-75-115-TUM’ and the other
with the shift forward happening in the middle of the fourth span, denoted as ‘5s-75-115-
TMS’. Results are compared in terms of time histories, spectral components and r.m.s. values
of the contact point displacement and of the contact force.
Vehicle System Dynamics 509

F.E.M.
3s−75−115
20 3s−50−135
5s−75−115−TUM
15 5s−75−115−TMS

10

5
zc (mm)

−5
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

−10

−15

−20
0 57 114 171 228
x (m)

Figure 5. Time history of the contact point vertical displacement along four spans.

Figure 5 shows the time history of the vertical contact point displacement: for this quantity
a periodic waveform is obtained with maxima occurring just after the centre of each span
and minima after the catenary suspension. Shorter wavelength fluctuations can be ascribed
to the effect of the droppers. All configurations of the real-time model show a general good
agreement with the reference FE model, but all real-time models tend to overestimate the
amplitude of the contact point displacement. This can be explained by noticing from Figure 4
that the real-time models overestimate the variation of the catenary stiffness between the
centre span and the positions at the first and last dropper in the span, thus causing a larger
excitation on the vertical oscillation of the pantograph head. In turn, the greater variation of
the catenary stiffness provided by the real-time model is due to limitations in the wavelengths
of deformation considered and to the simpler dropper model adopted.
The examination of the 1/3 octave band r.m.s. diagrams of the contact point displacement,
shown in Figure 6, allows to go into more depth with the comparison of models: considering the
frequencies affected by span-passing excitation (0–2 Hz), all four configurations of the real-
time model are in excellent agreement with the FE model. In the frequency range interested by
the dropper passing (7–18 Hz), the results are instead affected significantly by the configuration
of the real-time model, although there is no clear indication of one configuration behaving
better than the others. It is worth remarking that in this frequency range, the amplitude of
motion is very low (below 0.5 mm), so that even small differences in the dynamic stiffness of
the catenary play an important role.
The real-time models also appear to overestimate the vibration of the contact point
at frequencies 4–5 Hz: This frequency range is not related with specific sources of the
pantograph–catenary excitation and the relatively high level of vibration shown by the real-
time models is probably associated with the discontinuities introduced by the ‘shift-forward’
process, given that the frequencies interested are close to three times the span passing.
Figure 7 shows the 1/3 octave band r.m.s. diagrams of the contact force exchanged by the
pantograph with the catenary: the most significant harmonic contributions are related with
span-passing effects in the low frequency range (0–2 Hz) and with dropper-passing effects
510 A. Facchinetti et al.

2
10
FEM.
3s−75−115
3s−50−135

1
10
zc (mm)

0
10
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

−1
10

−2
10 0 1
10 10
f (Hz)
2
10
FEM.
5s–75–115–TUM
5s–75–115–TMS

1
10
zc (mm)

0
10

1
10

2
10 0 1
10 10
f (Hz)
Figure 6. 1/3 octave band r.m.s. diagrams of the contact point vertical displacement. Above: real-time catenary
models 3s-75-115 and 3s-50-135 versus FEM; below: real-time catenary models 5s-75-115-TUM and 5s-75-115-TMS
versus FEM.

taking place in the 7–18 Hz frequency range. As for the contact point displacement, the contri-
butions related to span passing obtained with the reference FE model are very well reproduced
by the real-time models, for all the considered parameter combinations. In the dropper-passing
frequencies and in particular for the frequencies above 10 Hz, the 3s-50-135 real-time model
and the two models considering five spans tend to overestimate the amplitude of the contact
force contributions, whereas the results of the 3s-75-115 configuration appear to be in better
agreement with the FE model ones.
Vehicle System Dynamics 511

2
10
F.E.M.
3s−75−115
3s−50−135
Fc (N)

1
10
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

0
10 0 1
10 10
f (Hz)
2
10
F.E.M.
5s−75−115−TUM
5s−75−115−TMS
Fc (N)

1
10

0
10 0 1
10 10
f (Hz)
Figure 7. 1/3 octave band r.m.s. diagrams of the contact force. Above: real-time catenary models 3s-75-115 and
3s-50-135 versus FEM; below: real-time catenary models 5s-75-115-TUM and 5s-75-115-TMS versus FEM.

This conclusion is further supported by the banded r.m.s. values (i.e. r.m.s. of the pass-band
filtered signal) of the contact force, reported in Table 3. It is apparent that for the 0–2 Hz
frequency range, all configurations of the real-time model provide similar results, but in
the 7–18 Hz frequency range, the results of the 3s-75-115 are in much better agreement with
the FE model than for the other configurations. As a result, the overall r.m.s. of the contact force
512 A. Facchinetti et al.

Table 3. Banded force r.m.s.

Frequency band (0–2) Hz % to FEM (7–18) Hz % to FEM (0–20) Hz % to FEM

FEM 15.35 19.11 26.73


3s-75-115 16.73 9.02 17.20 −9.96 26.84 0.41
3s-50-135 17.52 14.15 27.16 42.16 34.35 28.52
5s-75-115-TUM 17.76 15.67 24.24 26.83 31.61 18.29
5s-75-115-TMS 18.04 17.51 22.63 18.42 30.10 12.64

in the 0–20 Hz frequency range (which is the parameter used to assess the current collection
in real practice [21]) is also in much better agreement with the FE model for the 3s-75-115
configuration of the real-time model than for the other configurations.
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

Based on the numerical experiments examined in this section, it is therefore possible to


conclude that the 3s-75-115 configuration of the real-time catenary model provides the best
performances among the alternative configurations listed in Table 2, on account of the better
capability to reproduce short wavelengths of deformation in both the contact and messenger
wires. On the other hand, the location where the shift forward process is applied has a limited
influence of the results obtained for the 5s-75-115 model. To further support these conclusions,
HIL experiments and comparison with line measurements were performed, as described in
the next section.

5. HIL experiments and comparison with line tests

In order to further assess the performances of the different versions of the real-time catenary
model and to verify the capability of the HIL simulation to correctly reproduce the actual
pantograph–catenary interaction, HIL simulation results are compared with the experimental
data obtained from line tests.
Line measurements were obtained by an ETR500-class test train equipped with an ATR95
25 kV pantograph, instrumented according to the EN50317 standard [21] for measuring the
contact forces and the contact point displacements, and refer to the pantograph running at
300 km/h under an experimental portion of a modified C270 catenary with contact wire tension
30 kN, installed on the Italian High-Speed Line from Torino to Milano, consisting of three
consecutive sections, for a total length of approximately 3 km.
Figure 8 compares the HIL line results obtained with the different versions of the real-time
catenary model with the line measurement results, in terms of 1/3 octave band r.m.s. values of
contact point displacements. A very good agreement is found between all HIL results and line
measurements in the span-passing frequency range (0–2 Hz), but the results in the dropper-
passing frequency range are significantly affected by the real-time catenary model used, and
all real-time models show an overestimation of vibration at frequencies 4–5 Hz, as already
observed with reference to the results in Figure 6. Overall, no conclusive indication can be
obtained from the results in Figure 8 about which configuration of the catenary model behaves
better, in terms of reproducing the actual pantograph–catenary interaction measured on track.
Figure 9 shows the comparison of HIL results and line measurement results, in terms of 1/3
octave band r.m.s. values of the pantograph–catenary contact force. Again, all HIL results are
in good agreement with line measurements in the frequency range interested by span passing
(0–2 Hz), but in the frequency range corresponding to dropper passing (7–18 Hz) remarkable
differences are observed. In this frequency range, the results obtained using the 3s-75-115
configuration of the real-time model are in very good agreement with the experimental data
coming from line measurements (light blue line in the upper diagram), whereas the 3s-50-135
Vehicle System Dynamics 513

102
line test
3s−75−115
3s−50−135

101
zc (mm)

100
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

10−1

10−2
100 101
f (Hz)

102
line test
5s−75−115–TUM
5s−75−115–TMS

101
zc (mm)

100

10−1

10−2
100 101
f (Hz)
Figure 8. Comparison of HIL simulation results and line measurements for the ATR95 pantograph and C270
catenary (TCW = 30 kN), speed 300 km/h. 1/3 octave band r.m.s. of the contact point displacement. Above:
real-time catenary models 3s-75-115 and 3s-50-135 versus line measurements; below: real-time catenary models
5s-75-115-TUM and 5s-75-115-TMS versus line measurements.

real-time model significantly overestimates the contact force contributions in this frequency
range.
As far as HIL experiments performed using real-time catenary model configurations 5s-
75-115 and 5s-50-135 are concerned (lower diagram), the results appear to underestimate the
spectral contents of the contact force in the 7–13 Hz frequency range and to overestimate the
same in the 14–20 Hz range, when compared with line measurements.
514 A. Facchinetti et al.

102
line test
3s−75−115
Fc (N) 3s−50−135

101
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

100
100 101
f (Hz)
2
10
line test
5s−75−115–TUM
5s−75−115–TMS
Fc (N)

1
10

0
10
100 101
f (Hz)
Figure 9. Comparison of HIL simulation results and line measurements for the ATR95 pantograph and C270
catenary (TCW = 30 kN), speed 300 km/h. 1/3 octave band r.m.s. of the contact force. Above: real-time catenary
models 3s-75-115 and 3s-50-135 versus line measurements; below: real-time catenary models 5s-75-115-TUM and
5s-75-115-TMS versus line measurements.

Table 4 compares the banded r.m.s. of the contact force measured in line tests with the results
of HIL experiments. In the 0–2 Hz frequency range, all HIL results show a deviation from
line measurements in the range of 11–12%, regardless of the real-time model configuration
used. In the 7–18 and 0–20 Hz frequency range however, the HIL results obtained using the
3s-75-115 model configuration show a better agreement with line results than the results of
other HIL experiments, confirming the results of the numerical experiments in Section 4.
Vehicle System Dynamics 515

Table 4. Comparison of line measurements and HIL simulation results in terms of standard deviation and banded
r.m.s. of the contact force.

Frequency band (0–2) Hz % to line test (7–18) Hz % to line test (0–20) Hz % to line test

Line test 13.5 20.2 27.9


3s-75-115 15.1 12.0 20.6 1.6 29.4 5.3
3s-50-135 15.0 11.3 27.0 33.5 33.9 21.2
5s-75-115-TUM 15.0 11.3 21.5 6.3 29.0 3.8
5s-75-115-TMS 15.0 11.3 18.8 −7.0 26.4 −5.4

Notes: ATR95 pantograph and C270 catenary (TCW = 30 kN), speed 300 km/h.

In conclusion, the 3s-75-115 real-time model configuration appears to be best suited for HIL
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

simulation of the pantograph–catenary interaction, at least for the high-speed pantograph–


catenary couple considered in this work. Using this configuration, a deviation of 5.3% is
obtained for HIL experiments with respect to line tests in terms of the 0–20 Hz banded r.m.s.
This is well below the maximum deviation accepted in the purely numerical simulation of
the pantograph–catenary interaction, for which, e.g. the European Standard EN50318 [22]
prescribes a maximum deviation of 20%.

6. Conclusions

In this paper, the problem of defining a suitable real-time catenary model for the HILS of
the pantograph–catenary interaction was addressed. Considering a class of real-time catenary
models based on modal superposition, the effect of changing the number of modelled spans and
the number of modal components used to describe the vibration of the contact and messenger
wires was investigated by means of:
(i) numerical experiments consisting of static and dynamic calculations, using as reference
a catenary model based on the FE method;
(ii) comparison of HIL experiment results with line measurements.
Even if a specific catenary was considered in the investigation, the main outcomes, in terms of
best combination of modelling options, can be extended at least to the analysis of catenaries
of the same type, i.e. simple catenaries in which the contact wire is connected by droppers
directly to the messenger wire.
The obtained results show that, despite the limitation in the maximum number of modal
coordinates that can be managed in the real-time simulation, the catenary model shall be able
to describe as much as possible the short wavelengths of deformation in both the contact wire
and the messenger wire. On the other hand, having the possibility to reduce the number of
shift forward steps (by increasing the number of spans modelled from three to five) does not
improve noticeably the results of the catenary model. Hence, models considering three spans
are preferable to those including five spans, because for the same number of modal coordinates
used, shorter wavelengths of deformation (in the 3/5) ratio are considered.
Comparing the two models with three spans considered in this study, the 3s-75-115 model
(i.e. the one allocating more coordinates to describe the vibration of the messenger wire)
provides better results. For this model, the deviation from line measurements of the contact
force r.m.s. in the frequency range 0–20 Hz is 5% approximately. This is a very satisfactory
result, demonstrating that HIL can be considered as a mature approach to the simulation of
the pantograph–catenary interaction, with possible applications in the design of new overhead
lines and pantographs.
516 A. Facchinetti et al.

Acknowledgements
This paper describes work undertaken in the context of the PantoTRAIN project, PANTOgraph and catenary interac-
tion: Total Regulatory Acceptance for the Interoperable Network (www.triotrain.eu). PantoTRAIN is a collaborative
project – medium-scale-focused research project supported by the European 7th Framework Programme, contract
number: 234015.

References

[1] S.-J. Lee, Y.-J. Kim, and K. Park, Development of hardware-in-the-loop simulation system for testing multiple
ABS and TCS modules, Int. J. Veh. Des. 36(1) (2004), pp. 13–23.
[2] C.-K. Huang and M.-C. Shih, Design of a hydraulic anti-lock braking system (ABS) for a motorcycle, J. Mech.
Sci. Technol. 24(5) (2010), pp. 1141–1149.
[3] H.-J. Kim and Y.-P. Park, Investigation of robust roll motion control considering varying speed and actuator
Downloaded by [USP University of Sao Paulo] at 07:40 02 April 2013

dynamics, Mechatronics 14(1) (2004), pp. 35–54.


[4] H. Metered, P. Bonello, and S.O. Oyadiji, An investigation into the use of neural networks for the semi-active
control of a magnetorheologically damped vehicle suspension, Proc. Inst. Mech. Eng. D, J. Automob. Eng.
224(7) (2010), pp. 829–848.
[5] M.D. Petersheim and S.N. Brennan, Scaling of hybrid-electric vehicle powertrain components for hardware-
in-the-loop simulation, Mechatronics 19(7) (2009), pp. 1078–1090.
[6] M. Karpenko and N. Sepehri, Hardware-in-the-loop simulator for research on fault tolerant control of
electrohydraulic actuators in a flight control application, Mechatronics 19(7) (2009), pp. 1067–1077.
[7] J.E. Carrion, B.F. Spencer Jr., and B.M. Phillips, Real-time hybrid simulation for structural control performance
assessment, Earthq. Eng. Eng. Vib. 8(4) (2009), pp. 481–492.
[8] J.N. Verhille, A. Bouscayrol, P.J. Barre, and J.P. Hautier, Validation of anti-slip control for traction system using
hardware-in-the-loop simulation, VPPC 2007 – Proceedings of the 2007 IEEE Vehicle Power and Propulsion
Conference, Arlington, Texas, 2007, pp. 440–447.
[9] C.-G. Kang, Analysis of the braking system of the Korean high-speed train using real-time simulations, J. Mech.
Sci. Technol. 21(7) (2007), pp. 1048–1057.
[10] B. Allotta, L. Pugi, M. Malvezzi, F. Bartolini, and F. Cangioli, A scaled roller test rig for high-speed vehicles,
Veh. Syst. Dyn. 48(S1) (2010), pp. 3–18.
[11] S. Bruni, A. Facchinetti, M. Kolbe, and J.-P. Massat, Hardware-in-the-loop testing of pantograph for
homologation, 9th World Congress on Railway Research WCRR 2011, Lille, France, May 2011.
[12] W. Zhang, G. Mei, X. Wu, and Z. Shen, Hybrid simulation of dynamics for the pantograph–catenary system,
Veh. Syst. Dyn. 38(6) (2002), pp. 393–414.
[13] W.H. Zhang, G.M. Mei, and X.J. Wu, and L.Q. Chen, A study on dynamic behaviour of pantographs by using
hybrid simulation method, Proc. Inst. Mech. Eng. F, J. Rail Rapid Transit 219(3) (2005), pp. 189–199.
[14] F. Resta, A. Facchinetti, A. Collina, and G. Bucca, On the use of a hardware in the loop set-up for pantograph
dynamics evaluation, Veh. Syst. Dyn. 46(S1) (2008), pp. 1039–1052.
[15] A. Facchinetti and S. Bruni, Hardware-in-the-loop hybrid simulation of pantograph–catenary interaction,
J. Sound Vib. 331(12) (2012), pp. 2783–2797.
[16] J. Ambrósio, J. Pombo, M. Pereira, P. Antunes, and A. Mósca, Recent developments in pantograph–catenary
interaction modelling and analysis, Int. J. Railw. Technol. 1 (2012), pp. 249–278.
[17] A. Collina and S. Bruni, Numerical simulation of pantograph-overhead equipment interaction, Veh. Syst. Dyn.
38(4) (2002), pp. 261–291.
[18] W. Zhang, Y. Liu, and G. Mei, Evaluation of the coupled dynamical response of a pantograph–catenary system:
Contact force and stresses, Veh. Syst. Dyn. 44(6) (2006), pp. 645–658.
[19] K. Lee, Analysis of dynamic contact between overhead wire and pantograph of a high-speed electric train, Proc.
Inst. Mech. Eng. F, J. Rail Rapid Transit 221(2) (2007), pp. 157–166.
[20] A. Alberto, J. Benet, E. Arias, D. Cebrian, T. Rojo, and F. Cuartero, A high performance tool for the simulation
of the dynamic pantograph–catenary interaction, Math. Comput. Simul. 79(3) (2008), pp. 652–667.
[21] European Committee for Electrotechnical Standardization EN 50317, Railway applications – current collection
systems – requirements for and validation of measurements of the dynamic interaction between pantograph and
overhead contact line, Brussels, 2002.
[22] European Committee for Electrotechnical Standardization EN 50318, Railway applications – current collection
systems – validation of simulation of the dynamic interaction between pantograph and overhead contact line,
Brussels, 2002.

You might also like