You are on page 1of 6

12th IFToMM World Congress, Besançon (France), June18-21, 2007 CK-xxx

Modelling the longitudinal dynamics of long freight trains


during the braking phase
Luca Pugi* Duccio Fioravanti† Andrea Rindi‡
Departments of Energetics, University of Florence
Florence Italy
† ‡
E-mail: [* luca, fioravanti, rindi] @mapp1.de.unifi.it

Abstract—The common use of UIC Pneumatic braking plant reservoir” (“SC”). More precisely, for safety reasons, the
in long freight trains may involve time delays in the application braking is activated by a reduction of the brake pipe
of the braking forces along the convoy. This variable braking pressure of at least 0.15 bar with a pressure gradient of at
action along the train may cause heavy longitudinal efforts that least 100-200mbar/s from a reference pressure of about
could lead to wagons damaging and even cause train
derailment. Simulations of these phenomena are very important
5bar (6barA). Depression signal in the brake pipe is
both for safety and economic reasons since longer freight trains generated by the driver using a brake valve that can be
imply higher transport costs reductions. In this paper authors roughly modelled by a programmable pressure regulator
present simplified models used to evaluate, in a convenient according to the desired manoeuvre. In order to partially
time-saving way how train composition, buffers mechanical compensate plant delays and to optimize both braking and
features, braking plant design and other key parameters may comfort performances, the distributor works as a highly
influence the longitudinal behaviour of the convoy. non linear/gain scheduled pressure regulator with variable
dynamical behaviour, according to the pressure and
Keywords: freight trains, pneumatic brake, gradient values of both brake pipe and cylinders.
longitudinal efforts
Since air depression signal propagation has finite speed,
at a certain time, different clamping forces are applied to
I. The longitudinal behaviour of a train during a
wagons, thus producing variable longitudinal forces along
braking: a typical multi-disciplinary problem
train length. Notice that the calculation of longitudinal
Wagons braking plant is usually a pneumatic one: air braking forces involves also the simulation of train
pressure inside braking cylinders gives rise to a clamping dynamics behaviour since friction factor on braking
force that is used to activate a disk or block brake acting elements is often variable with vehicle speed. Also there
on the wheelset. In figure 1 a simplified scheme of the are many situations like electric braking in locomotives or
plant from a previous publication [1] is shown. speed-dependent braking forces regulation that introduce
a dependence of the braking efforts from vehicle speed.
Wagons are linked each other by different kinds of
couplers: usually traditional couplers based on the
combined use of a drawgear and dampers are used in
Western Europe (see figure 2).

BUFFER

Fig. 1: pneumatic layout of a typical brake system for freight wagons.

Variable pressure inside braking cylinders is used to DRAW GEAR


modulate the braking effort. In trains equipped with the (mounted with compliance )

classical UIC pneumatic brake (referring to figure 1), Fig. 2: Layout of a traditional coupling system (draw gear + buffer).
every vehicle is provided with a complex pneumo-
mechanical device, called “distributor” (pointed out with Freight wagons are usually coupled without any pre-load
n° 4): this device modulates the pressure inside braking or with a small clearance so that coupling system works as
cylinder (n° 7) according a law that is proportional to the a non linear variable stiffness-damping spring with dead
difference between the pressure in the “brake pipe” zone and end runs. Since a train may be composed with
(“CG”) and the reference pressure stored in the “command several wagons (20-30 or more), the resulting model

1
12th IFToMM World Congress, Besançon (France), June18-21, 2007 CK-xxx

needs to be conveniently simplified, in order to reduce background of studies produced by different research
computational load, as presented in the iterative procedure groups that substantially agree on many aspects. In
described by the flowchart in figure 3. particular two different aspects are usually emphasized:
1) Modelling the braking signal along brake pipe as a
Preprocessor depression wave.
Definition of train compositions, vehicle-coupling
Train
Composition main features, line design, applied manoeuvres 2) Distributor response: The dynamic response of this
system it’s quite difficult to be modelled since it’s
dominated by non linear phenomena like sonic chocking
Pneumatic Simulation
Pneumatic
Plant Brake signal propagation along the train of calibrated orifices, elastic membranes etc. Also further
Simulation brake clamping forces according pressure
profiles in brake cylinders
components as pressure relay that works as intermediate
power amplifier and regulator introduce further non linear
Train
Lumped mechanical model of the full train behaviour of the system.
Model of wagons connected by non linear spring-damper
Longitudinal representing couplings. Braking forces calculated
Behaviour according pressure profiles and vehicles speeds in order
to reproduce variable friction factor of braking blocks A. Propagation of pressure signals along the brake pipe:
(pads) or speed dependent braking regulations
basic modelling considerations
Heuristic Evaluation of results Generally the brake pipe is modelled as a lumped system:
Loading condition on couplings are
Coupling Loads
are potentially evaluated with heuristic criteria; if discretization elements are defined by the linearization,
longitudinal loads are lower than a certain
dangerous?
safety value further analysis are not around a typical working point, of the equations
required describing flow dynamics, following the classical electro-
mechanical analogy [3].
YES NO
During the braking phase pressure variations and
Results associated flows are too large, so linearized lumped
DATABASE
Multibody
Simulation “Tested Train models are not well suited to fit the dynamical behaviour
Configuration it’s not
dangerous” of the system. In this case more complex lumped elements
functions, able to adjust equivalent parameters of
NO “Resistance”, “Capacitance” and “Inductance”, according
Coupling Loads to fluid calculated states, have to be used. Authors have
are really YES
dangerous? deeply studied the problem, first developing brake pipe
Results
DATABASE models in the Matlab-Simulink™ [4] environment and
“Tested Train
Configuration is
then using more specialized tools like Imagine
potentially dangerous”
AMESIM™ [1]. In particular in Amesim™, non linear
Fig. 3: General flow-chart of the proposed simulation procedure. lumped elements called “C” (capacitance) and “RI”
(Resistance and Inductance) are available to simulate the
Longitudinal forces exchanged by buffers along the train dynamical behaviour of pipes. “Bondgraph” or state space
are usually calculated through a lumped model where each approaches are also been used to simplify modelling and
wagon is modelled as a single lumped inertia, neglecting notation [5], [6]: in fact the choice of a right set of
additional degrees of freedom due to bogies, suspension variables allows to write system governing relations as
system, wheel-rail interaction, etc. This simplified first order differential equations, where derivatives of the
approach is thus used to predict wagons and train state variables, usually called “flows”, can be calculated
compositions with the worst loading condition during the as functions both of the actual state value and of a certain
braking phase. At this stage, a more accurate dynamical number of forcing elements, usually called “efforts”.
model can be used to study how the efforts transmitted by However the way and the size of the lumped elements
coupling systems can affect the stability of a single used to discretize the pipe deeply influence precision and
wagon: this can be performed using standard commercial numerical performances of the model. In order to
multibody software like Adams™, Simpack™ or understand how discretization affects the simulated brake
Vampire™. The model of a single wagon, with pipe response, a simple example can be done. Figure 4
appropriate boundary conditions on coupling system, may shows the propagation of the braking signal modelled in a
be used in such cases for accurate predictions of vertical simplified way: a pipe 300m long with a diameter of about
and lateral forces between rail and wheels, in order to 25mm and known relative roughness (0.01) is initially
quantify derailment risk. However it’s reasonable that in loaded with air at an absolute pressure of 6 bar. After the
short-medium period even complete train model may be simulation has started (simulation time = 0), the pipe is
implemented on standard commercial hardware platform. opened at one end, in order to reproduce the action of the
driver’s braking valve. The output pressure profile
II. Simulation of the UIC Braking Plant measured at the other pipe end is shown and compared in
Many researchers have worked to the simulation of the figure 4, for different numbers of discretization elements.
pneumatic plant behaviour (see i.e. [2]) : so there is a wide

2
12th IFToMM World Congress, Besançon (France), June18-21, 2007 CK-xxx

Brake pipe modelled as a long duct with thermal flow evaluation (AMESIM™) and
polytrophic isothermal transformations (corresponding to
the simplified libraries build by authors in Matlab™
Open end of the pipe (driver’s valve in
braking position) Closed end of the pipe (last coach)
Simulink™ environment [4]) are quite similar.
Thermal flow
D epression wave arrive according ideal (500Jm-2K-1s-1)
gas sound speed (about 340 m/s) Wave arrival according
ideal sound speed

Brake Activation pressure

D epression corresponding
to brake activation
1RC-I Polytrophic (isothermal)
Polytrophic (adiabatic) transformation
transformation Matlab Libraries [4]

3RC-I
5RC-I

D epression corresponding
to full braking
10RC-I
Figure 5: Comparison of simulated pressure profiles versus time at the
15RC-I pipe closed end according different hypotheses on flow thermodynamic
transformations.

Figure 4: Comparison of simulated pressure profiles at the closed pipe B. Model Calibration and Validation on experimental
end, for different pipe discretizations.
data
As visible in figure 4 a minimum number of 5 C-IR Experimental data for both model calibration and
elements sequences must be used to simulate with validation are essential for the successful development of
acceptable tolerance the time at which maximum service simulation tools, therefore in this paper authors have
braking is applied to brake cylinders (corresponding to a exploited a recent cooperation with Trenitalia DISQS
pipe pressure decrease of about 1,5 bar). Braking concerning studies about braking and longitudinal
activation instead, since is associated to small brake pipe behaviours of freight trains formed by SAADKMS low-
depression (about 0,15 bar) and to precise pressure floor wagons [5]. Train composition used for experimental
gradient (more than 100 mbar/s), needs a good pipe modal tests is briefly described in figure 6.
response reconstruction in order to be correctly simulated. Travel Direction Avaiable pressure
measurement on brake pipe
Figure 4 shows that pipe models with a sequence of 10 LOC 1 LOC 2 Strum. Mis. 1 1 2 3 4
and 15 C-IR elements give satisfactory predictions, even
5 6 7 8 9 10 11 12
showing a pressure wave propagation that is slightly faster
than ideal sound speed. In fact, starting from 10-15 C-IR 13 14 15 Mis. 2 16 17 18 19
elements, delays and pressure gradients involving brake
20 21 22 23 24 25 26 27
activation seems to be quite stable against any increase of
Avaiable pressure
discretization elements. In general we observe that further 28 29 30
measurement on brake pipe

increases in model elements number produce small SAADKMS


LOC Locomotive Mis. Measurement Coach
additional delays that can be easily compensated adjusting freight wagon

the distributor model dynamical behaviour in term of both Figure 6: Layout of the train used for experimental tests.
time response and sensitivity to pressure gradients. Pipe
described in figure 4 is in effect a continuous system with Experimental data concerning the braking behaviour of
infinite modal shapes and natural frequencies, that can be low-floor wagon SAADKMS have been used to validate
evaluated by the following relation (adiabatic small and calibrate our model. Experimental campaign was not
perturbations) : arranged for the aims of this work, so the authors
validated and customized the simulation model on
2i − 1 1 available experimental data including a very reduced set
fi = ⋅ KRT with i = 1, 2,...n (1)
2 2L of brake pipe pressure measurements. From these data
authors selected a rapid braking in G mode: in such test
To obtain a satisfying numerical simulation of the pipe the braking system of “measurement coach 2” was
response a sufficient number of low frequencies modes completely deactivated (see figure 6). Since data on
have to be modelled: as a natural consequence a sufficient locomotives and measurements coaches brake plants
number of C-RI elements, corresponding to the degrees of were not completely available, in this paper authors study
freedom of an equivalent mechanical flexible-dumped the braking signal propagation along SAADKMS wagons.
system, have to be used. The simulation described above The resulting plant, briefly described in figure 7, has been
has been also repeated using a lumped model with 5 C-RI modelled as follows:
elements series, but using different hypothesis for air - imposition of the pressure source (known from the
thermodynamic transformation. Figure 5 shows results experimental data) in the main brake pipe section

3
12th IFToMM World Congress, Besançon (France), June18-21, 2007 CK-xxx

corresponding to the first SAADKMS, as boundary as for example sonic transition of calibrated orifices, are
conditions; modelled with better accuracy. Distributor is a very
- model of the pipe of 15 SAADKMS freight wagons with complex system with different subsystems activated by
their respective pneumatic brake systems; several thresholds both of pressure and of pressure
- model of the pipe corresponding to the braking disabled gradients.
intermediate measurement coach; 3) Slow, smooth gradients on pressure profile associated
- model of another composition of 15 SAADKMS wagons to the braking of last wagon make easier to recognize
with their respective pneumatic braking systems. differences between the model developed by the authors
Imposed pressure profile Intermediate coach Calibration pressure with Amesim™ and E-train™ model.

……. 15 last wagon brake


1 16 ……. 30 pipe experimental
Travel Direction data
last wagon brake
Figure 7: Simplified plant used to study braking signal propagation along pipe sim. pressure
the convoy. profile

First wagon brake pipe


C. Results evaluation and comparison with experimental experimental data
data
First wagon brake pipe
Results of the model described in figure 7 have been sim. pressure profile
compared with experimental data given by Trenitalia
SPA. Some simulations have been also performed using
Figure 9: Pressure profiles in the main brake pipe for the first and the last
“E-train”, a special tool currently used by different wagon; comparison of experimental data and authors model results.
European railway companies, that has created and
customized for these particular applications. Figures 8 and III Simulation of train Dynamical response
9 shows the results concerning pressure behaviours inside
brake cylinders and main brake pipe at the first and at the A. Simplified Monodimensional Model of the train
last SAADKMS wagon. Lumped monodimensional models of the wagons, buffers
and other train vehicles have been created both in Matlab
First wagon cylinder pressure
(Amesim™ simulation) and Amesim environment. The aim of these simplified
First wagon cylinder pressure models is to create fast simulation tools useful for several
(experimental data) purposes:
First wagon cylinder 1) Statistical studies where thousands of different train
pressure
(E-train ™ simulation)
composition and manoeuvres have to be tested.
2) Calibration of more complex bi-tri-dimensional models
Typical last wagon cylinder pressure
of wagon and components like buffers, coupler, etc.
pressure drop (E-train™ simulation) 3) Real time simulation for HIL (hardware in the loop)
last wagon cylinder pressure
(experimental data)
applications like the MI-6 multipurpose bench of Firenze
Romito [9], developed by the authors for Trenitalia SPA .
last wagon cylinder pressure
(Amesim™ simulation) As visible in the simplified scheme of figure 10, wagons
are modelled as simple lumped masses subjected to
Figure 8: Pressure profiles in the brake cylinders for the first and the last
wagon. Comparison of experimental data and simulation results obtained
braking forces (computed by braking plant submodel) and
with E-train™ and Amesim™ (model developed by authors). to interaction forces, acting between vehicles, transmitted
by buffers; coupling elements are modelled in a separated
Good agreement between experimental data and subsystem. Braking forces are calculated according to
simulation results is clearly recognizable. In particular pressure values in brake cylinders and modelling variable
delays in pressure signal propagation and brake activation friction factor as a tabulated function of vehicle speed.
along the train are correctly simulated. In comparison with This assumption is very important especially for block
E-train, model developed by the authors has clearly brakes and for low speed braking simulation of long
obtained better results regarding the simulation of trains, where transition from kinematic to static friction
pressure behaviour of the last train wagon. This difference factor can produce considerable dynamical effects.
is mainly due to three factors:
1) Errors caused by bad brake pipe modelling-
discretization are more evident on last wagons where
pressure losses provoked by friction and delays introduced
by pipe length are more evident..
2) Distributor model developed by authors is more
precise: internal friction and other non linear phenomena, Figure 10: simplified mono-dimensional train model.

4
12th IFToMM World Congress, Besançon (France), June18-21, 2007 CK-xxx

efforts along a convoy of known composition during a


The behaviour of coupling system (buffer and drawgear) known braking manoeuvre (typically a low speed
is quite complex for the presence of many non linear emergency braking) :
phenomena like variable stiffness-damping, hysteresis, MLD(i ) = max Bi (t) (2)
lower and upper end runs, preload on buffers, drawgear
compliance, etc. In order to reduce the computational cost Bi (t ) = load time hystory on i th couple of buffers
and to simplify model calibration, non linear behaviour of From a physical/scientific point of view MLD is not
coupling system has been modelled through a tabulated particularly significant, however this index is often used
function of run and buffers deformation speed (see figure in practice since it’s a useful plain parameter both to
11/a). In order to prevent chattering on end run, a define unsafe train compositions and to estimate the
position-tabulated viscous damping is introduced. In probability of critical events like wagon damage or train
figure 11/c is reported an example of calculated load derailment. Railway freight transport, in fact, includes lot
history, corresponding to an impact between two wagons of possible train composition, differing each other in
during a braking manoeuvre (low damping value adopted vehicles number and typology, goods loaded, train length
emphasizes load cycles and non linear behaviour). and even in vehicles disposition along the convoy (this
High stiffness last element is very important for long trains with
viscous damping(Ns/m)

for end run


High damping
different kind of wagons). MLD evaluation on huge
on end-run populations of diverse train compositions and
manoeuvres is often performed using specialized software
like E-train™[8], following a monodimesional lumped
approach and even simpler assumptions to model buffers
Dead Band response.
(compliance)
Fig.12/a shows the time history of longitudinal efforts
versus wagons number resulting obtained from simulation
Speed(m/s)
result: both the manoeuvre (emergency braking in G
Figure 11/a: example of run-speed Figure 11/b: example of
tabulated buffer response . run-speed tabulated viscous mode) and the train are the same used for the benchmark
damping factor. of braking plant model reported in Fig. 8 and 9.
Buffers load cycles “Third peak”
(kinematic to static “First peak”
friction factor) (braking delays)
Constant speed curve
(compliance) Stop
“Second peak”
Drawgear load (modal response)
cycles

Start

First wagon
Figure 11/c: example of run-speed tabulated buffer response. Last wagon
Figure 11 shows that the high non linear behaviour of the Figure 12/a: Time history of longitudinal loads (N) on coupling along
system can be easily reproduced and calibrated with a the train versus wagon order.
very simple and computational “light” tabulated model
1° Peak Max.loads
with straightforward implementations even on fast real- 3° Peak Max.loads 2° Peak Max.loads

time applications. Experimental


Experimental data concerning loads on buffers during a Data

braking manoeuvre were available for the same convoy


used in braking plant model calibration (see Section II B), Simulation Results Last buffer
coupling
so the authors have decided to compare real measurements
with simulation results. Since excessive compressive
efforts on buffer may cause wagons damaging and even First buffer
coupling
train derailment, railway companies typically adopt
performance-safety indexes in order to evaluate the
longitudinal behaviour of the convoy: in this paper the
“Maximum Load Distribution” (MLD) is used. MLD is Figure 12/b: Comparison of MLD distribution calculated (blue line)
defined as the distribution of the maximum compressive with experimental one (magenta line).

5
12th IFToMM World Congress, Besançon (France), June18-21, 2007 CK-xxx

The model is quite realistic since it’s able to reproduce a cases in fact, vehicle lateral dynamic and buffer contact
“three wave surface” that is typical of this kind of test. In geometry may produce relevant transversal forces that
figure 12/b calculated “MLD” is compared with data from may lead to train derailment. Authors, after several
experimental campaign. The first compression wave, optimization, are studying a pre-tabulated solutions
clearly pointed out in figure 12/a, is mainly associated to inspired to the simplified model proposed by researchers
the time delay in braking efforts application along the of Politecnico di Milano for their software TSdyn [10].
convoy as a consequence of pneumatic braking plant Further results will be presented when future extensions of
response; the second peak is more sensitive to the our work will be completed.
longitudinal modal response of the train and thus to model
buffer calibration (stiffness, end run, damping). The third Acknowledgements
peak is usually the strongest in particular for rapid braking Authors wish to thank all people that have shared data and
in G mode: it’s deeply influenced by the rapid increase of knowledge and that have been precious for the
the friction factor on brake blocks or pads, due to the development of this work. In particular authors would
transition from cinematic to static friction at the end of thanks: Paolo Presciani, Alessandro Palazzolo, Giacomo
braking manoeuvre. Cerfeda, Vincenzo Castellani from Trenitalia SPA and
Experimental and calculated “MLD” reported in Fig. 12/b Gianni Turchi from Frensistemi SRL. An additional
are clearly similar: small differences are compatible with thanks goes to the painstaking work of the students of
simplifications introduced in the simulation and normal Florence engineering faculty that have contributed to
tolerances of real mechanical and pneumatic components. collect data and to optimize the response of various model
subcomponents.
IV Conclusions and future developments
In this paper authors have presented the current status of References
their research concerning the simulation of long train [1] Pugi L., Palazzolo A., Presciani P. and Fioravanti D. "Simulation
longitudinal dynamics, pointing the attention to the and optimization of railway pneumatic braking system" World
braking phase. Libraries for an accurate simulation of the Congress for Railway Research WCRR2006, Montreal 5-7 June
2006.
braking plant behaviour and a simple but effective
monodimensional model of a freight train have been [2] Murtaza M.A. and Garg S.B.L. Brake modelling in train simulation
successfully developed and validated on a experimental studies. Proc. Instn. Mech. Engrs, Part F: J. Rail and Rapid Transit,
1989, 203(F2), 87–95.
data from Trenitalia SPA .
Actually authors are working to the developments of 2D [3] Belforte G. Manuale di pneumatica Tecniche Nuove. "ISBN
88-481-0541-6.
and 3D mechanical models in order to better investigate
train stability on curved tracks . [4] Pugi L., Malvezzi M., Allotta B., Banchi L. and Presciani P. A
parametric library for the simulation of a Union Internationale des
In particular authors are working with an approach based Chemins de Fer (UIC) pneumatic braking system Proc. Instn.
co-simulation concept. Multibody-model of wagons, Mech. Engrs Vol. 218 Part F: J. Rail and Rapid Transit F01403 #
complete of suspension system, and an accurate model of IMechE 2004 pag 117-132.
wheel-rail contact is implemented on standard commercial [5] Wargo J. D. (1991) “Vector Bond Graph Analysis of Mechanical
software (MSC Adams-Rail™). Additional Systems” Trans. of the ASME, J. of Dynamic Systems Measurement
train components like buffers and pneumatic braking plant and Control, 113(3), pp.344-353.
are instead implemented in a more suitable environment [6] Nagai S., Kukuminato T., Sekine T. (1986) Simulation Program
like Amesim™ and Matlab-Simulink™ since both tools For Pneumatic Systems Fluid Control and Measurement, 1, pp.89-
94.
support co-simulation with MSC Adams™. Many
vehicles models (locomotives, coaches, wagons) have [7] Rindi A., Fioravanti D., Pugi L., Rinchi M., Auciello J.
"Optimization of special freight wagons with small wheel
already been developed and validated in past research diameter" COMPRAIL 2006 Tenth International Conference on
activities of the authors. In particular a full validated Computer System Design and Operation in the Railway and Other
model of SAADKMS freight wagon [7] has been recently Transit Systems, Prague 10-12 July 2006 Czech Republic.
assembled and validated with data based on experimental [8] I.V.E E-train Technical documentation.
campaign in which stability and riding quality of the
[9] Pugi L., Rinchi M., Malvezzi M., Cocci G., “A multipurpose
vehicle have been tested in different condition with platform for HIL testing of safe relevant railway subsystem” AIM
various wheel-rail profiles. At present time authors are 2005 IEEE-ASME Advanced Intelligent Mechatronics
working to the optimization of different three-dimensional 2005 25-30 July 2005 Monterey California USA.
buffer models in Matlab Simulink™ with a particular [10] Belforte P., Cheli F., Diana G., Facchinetti A., Favo F., Melzi S.
attention to the computation of forces and torque and Sgroi F., “Effetto della dinamica longitudinale sulle
introduced by multiple contact points with friction. condizioni di svio dei rotabili” Convegno C.I.F.I. “Evoluzione del
materiale rotabile tecniche di progettazione ed attività di testing”
Authors think that three-dimensional wagons behaviour Pistoia 18 Ottobre 2005 Italy.
cannot be neglected especially when high compressive
longitudinal forces are associated to curved track: in these

You might also like