You are on page 1of 15

312 Int. J. Heavy Vehicle Systems, Vol. 20, No.

4, 2013

High frequency modelling and validation of railway


dampers

A. Sañudo* and J. Nieto


CEIT, Manuel de Lardizábal 15,
San Sebastián 20018, Spain
E-mail: asanudo@ceit.es
E-mail: jnieto@ceit.es
*Corresponding author

J. Pérez
CAF, S.A. J.M. Iturrioz 26,
Beasain 20200, Spain
E-mail: xperez@caf.net

A. Alonso
CEIT, Manuel de Lardizábal 15,
San Sebastián 20018, Spain
Tecnun, University of Navarra,
Manuel de Lardizábal 15,
San Sebastián 20018, Spain
E-mail: aalonso@ceit.es

Abstract: This paper presents and validates an appropriate physical damper


model to predict the dynamic behaviour of a railway damper for frequencies of
up to 200 Hz. The model is computationally efficient and, therefore, appropriate
for implementation in NVH-CAE models used to study the transmission of
vibration paths to the train body. The model parameters are related to the physical
characteristics of the damper (volume of the chambers, piston rod sections,
characteristics of the valves, etc.). However, it is shown that in the absence of
internal constructive information, which is most often the case, the model
parameters can be fitted from a few tests carried out at low frequency (0–20 Hz),
based on its theoretical background.

Keywords: railway dampers; high frequency modelling; validation; NVH;


vibrations; transmission paths.

Reference to this paper should be made as follows: Sañudo, A., Nieto, J., Pérez,
J. and Alonso, A. (2013) ‘High frequency modelling and validation of railway
dampers’, Int. J. Heavy Vehicle Systems, Vol. 20, No. 4, pp.312–326.

Biographical notes: Álvaro Sañudo is currently working in the railway laboratory


of the Applied Mechanics department of CEIT as a PhD student. His research
interests are noise, vibration and hardness, structure-borne noise, suspension
elements modelling and finite element analysis.

Copyright © 2013 Inderscience Enterprises Ltd.


High frequency modelling and validation of railway dampers 313

Javier Nieto received his PhD in Mechanical Engineering in 1996 from Tecnun,
the Technological Campus of the University of Navarra. He is a researcher at the
Experimental Analysis and Design Unit (Applied Mechanics Department) at CEIT.
His main research and technical interests are in the field of noise and vibrations;
characterisation of the dynamic behaviour of structures or machines; design,
construction and use of test benches; instrumentation and data acquisition systems;
training simulators and railway technology. He has participated in more than 40
research projects, has directly supervised one doctoral thesis, is author or co-author
of about 10 scientific papers in refereed journals and 20 articles in national and
international conferences and is author or co-author of four international patents.

Javier Perez is a Senior Technology Advisor at the rolling stock manufacturing


company Construcciones y Auxiliar de Ferrocarriles (CAF). He obtained his PhD
in Mechanical Engineering in 2002 from Tecnun, the Technological Campus of
the University of Navarra. In the past, he has been a researcher at the Applied
Mechanics Department at CEIT and a research fellow at the Rail Technology Unit
of the Manchester Metropolitan University. His area of expertise is the dynamic
behaviour of railway vehicles, the modelling of mechanical and mechatronics
components of vehicles and the study of wheel/rail interface-related phenomena.
He has co-authored around 10 papers in international peer reviewed journals
and has participated as a researcher or manager in several research projects and
consultancy projects with private and public funding.

Asier Alonso obtained his PhD in Mechanical Engineering in 2005 from TECNUN,
the Technological Campus of the University of Navarra. He has been a researcher in
the Applied Mechanics Department at CEIT since 2002. At present he is in charge
of the Railway Laboratory of CAF-CEIT. His main research lines are railway
dynamics involving different areas of knowledge, such as wheel rail contact,
dynamic stability of railway vehicles, modelling techniques, characterisation
of dynamic properties of visco-elastic materials, vibration transmissibility and
interior and exterior noise in vehicles. He has participated in 14 research projects,
is involved in the supervision of five PhD theses and has co-authored around 15
scientific papers in international peer reviewed journals and 14 articles in national
and international conferences.

1 Introduction
Passenger comfort is closely related to the noise level inside the vehicle, which is mainly
composed of air-borne noise at middle to high frequencies and structure-borne noise at low
to middle frequencies, where the highest noise level is recorded (Eade and Hardy, 1977;
Thompson, 2009). The structure-borne noise is mainly caused by the roughness of the wheel-
rail contact (Thompson et al., 1996a; 1996b; Thompson, 2009), which generates vibrations
that propagate through the suspension to the car body (Thompson, 2009).
Comfort is an increasingly interesting topic for rolling stock manufacturers to increase
the competitiveness of their high speed vehicles. It must be considered as a design parameter,
as it is difficult to achieve any significant reduction in sound pressure level once the vehicle
has been built. Therefore, manufacturers must be able to accurately predict the vehicle’s
internal noise and vibration propagation at the design stage. Among the different techniques,
314 A. Sañudo et al.

the finite element method (FEM) is the most suitable to predict the low- to mid-frequency
structure-borne noise (Lalor and Priebsch, 2007; Osborne et al., 2010; Thompson, 2009).
Bogies and coach models can be developed to precisely study propagation through the
structure. To obtain accurate results, it is also necessary to develop accurate models of the
various elements that play an important role in vibration propagation, such as dampers and
rubber joints (Gil-Negrete, 2004; Gil-Negrete et al., 2006; Sjoberg, 2002; Stenti et al., 2010),
coil springs (Lee and Thompson, 2001) and the trim (Fan et al., 2009). In this paper, the
development of a model that reproduces the dynamic behaviour of a railway damper in a
frequency range up to 200 Hz will be addressed. Damper models can be classified into two
main groups: empirical models and physical models.
The empirical models are derived from the experimental measurements made in a specific
range of frequencies and amplitudes (Conde et al., 2006). The main disadvantage of this
approach is the large amount of experiments that must be carried out to obtain an accurate
model. These tests must represent the entire range of frequencies under which the damper
will operate. Also, a model based on experimental results requires a great deal of result
analysis, which is difficult to use again if the damper model is changed or any modification
is made to it.
The physical models are suitable for component design (Wang et al., 2011) and vehicle
dynamics analysis (Alonso et al., 2011; Van Kasteel et al., 2010) in a frequency range up
to 30 Hz. The main advantage of this method is that the model can be applied to different
dampers merely by modifying the physical parameters. As a result, the number of tests to
adjust the model can be reduced to a bare minimum. A possible disadvantage of this type of
model is that it may be too complex. If this is the case, the dynamic simulation of the whole
vehicle may be slowed down.
In a previous work (Alonso et al., 2011), a railway damper model based on physical
parameters was developed. The main objective of this work was to develop a model
that reproduces the behaviour of the damper to analyse the dynamic stability of railway
vehicles. The developed model takes into account the oil flow between chambers and the oil
compressibility. The results show that it is valid for accurate reproduction of the behaviour for
most railway dampers in the frequency range of interest. However, no tests were performed
to check the validity of the model for higher frequencies.
The aim of this paper is to determine if the physical model presented in Alonso et al.
(2011) could be used in vibration transmission analysis; that is, if the results provided by this
model are accurate enough up to 200 Hz.
Another objective of the paper is to show that the parameters of the model can be obtained
from few experiments performed at low frequencies. This is necessary because constructive
information regarding the dampers is not usually known by manufacturers. In addition, most
railway dampers are difficult to test on high stroke frequency machines due to their size. The
paper is arranged as follows. First, the differential equation which describes the behaviour of
the damper is set out to relate the model parameters to their physical meaning.
Then, the experimental results are analysed. The tests are performed using stepped sine
excitation with and without silent-blocks to evaluate the influence of the stiffness induced by
the oil compressibility in the system’s response.
A mathematical model of the damper is later obtained and the model response is compared
to the experimental data in frequency and time domain, resulting in an accurate match in both.
Finally, it is shown that adjustments can be made just with the data from the low-frequency
test. This will allow us to develop models for larger dampers that cannot normally be tested
High frequency modelling and validation of railway dampers 315

in a high stroke frequency machine due to their size, such as the yaw damper, which is the
main path for vibration transmission (Thompson, 2009).

2 Theoretical background
In a previous work (Alonso et al., 2011), a damper model for railway vehicle dynamic analysis
was developed. This model was based on the physical characteristics of the damper (volume
of the chambers, piston rod sections, characteristics of the valves, etc.). The results obtained
for this model were in line with experimental results (up to 20 Hz). The theoretical basis of
the model will be briefly explained in this chapter. First, the typical working principle of a
railway damper is explained. Then, the mathematical model that describes the behaviour of
the damper is presented and analysed.
Railway dampers are hydraulic elements that usually consist of three chambers:
compression, rebound and auxiliary chambers. Most of these operate under the re-circulating
principle shown in Figure 1.

Figure 1 Damper operation principle (a) compression stroke, (b) rebound stroke

During the compression stroke, the compression chamber volume reduces, increasing the
pressure in this chamber. This increase closes the foot check valve and opens the valve on the
damper piston, resulting in oil flow from the compression chamber to the rebound chamber.
The oil volume that passes to the rebound chamber is greater than the volume generated
by the piston movement, resulting in oil flow from the rebound chamber to the auxiliary
chamber through the damping valve.
During the rebound movement, the piston comes out. The rebound chamber volume
reduces, increasing pressure. The oil volume passes from the rebound chamber to the
316 A. Sañudo et al.

auxiliary chamber through the damper valve. Meanwhile, there is a volume increase in
the compression chamber, resulting in a pressure drop. Therefore, the piston check valve
remains closed while the foot check valve opens.
It can be seen that the oil flows in the same direction, both in the compression and in the
rebound stroke. This fact prevents undesired effects such as cavitation and also situations in
which all the valves remain closed.
It can also be concluded that to develop an accurate model, special attention must be
given to the characterisation of the valves and the relationship between oil flows, volume
changes due to piston movement and oil compressibility.
The equation relating to the above-mentioned phenomena for each chamber is as follows
(Alonso et al., 2011):
dp
∑Q i , chamber + Achamber .υ = β ⋅ Vchamber ⋅
dt
(1)

where ∑ Qi ,chamber is the addition of the flow rates coming out from the chamber through
the different valves, Achamber the piston section corresponding to the considered chamber, υ
the velocity of the piston rod, β the bulk modulus of the oil and Vchamber the volume of the
chamber, which is calculated using the Bernouilli equation.
In Alonso et al. (2011) it was shown that for most railway dampers, the following hypotheses
can be used to obtain a simpler model without leading to a significant loss of accuracy:
• The check valve in the piston remains closed during the rebound stroke. Therefore,
it can be accepted that the flow rate from the compression chamber to the rebound
chamber is null.
• The pressure in the compression chamber during the rebound movement is zero. During
the compression stroke, the flow rate from the auxiliary chamber to the compression
chamber is null.
• The pressure in the rebound and compression chambers is approximately the same
during the compression stroke.
These hypotheses lead to a model which consists of a linear spring element in series with
a dashpot element (Figure 2).

Figure 2 Idealised damper

3
Arebound ρ
C= (2)
2 ( Cd Aout )
2

2
Arebound
K= (3)
β V ( x, p )
High frequency modelling and validation of railway dampers 317

F = Cxv2 = K ( x − xv ) (4)
C parameter is related with the resistance of the fluid passing through the damping valve
(Aout) and K is related to the compressibility of the fluid, the section and volumes of the
chambers. The dashpot element can be linear or quadratic depending on the constructive
characteristics of the valves. If it is linear, then equation (4) will become:
F = Cxv = K ( x − xv ) (5)
It is worth pointing out that in several cases, the constructive data of the damper is not
available and the parameters must be obtained from experimental data.
Figure 3 shows the behaviour of this model at different frequencies (this figure has
been obtained supposing that the damping element is linear). The curve represents the
stiffness magnitude versus the excitation frequency. The values are the typical ones for a
passive yaw damper (Kyd = 4⋅106, Cyd = 1⋅105 Ns/m). Also, two additional lines have been
included: one of them corresponds to the values obtained if the damper is modelled as a
single stiffness element (Kyd = 4⋅106 N/m) and the other one using a single viscous element
(Cyd = 1⋅105 Ns/m). Analysing this figure, it can be seen that for excitation frequencies lower
than ωc = Kyd/Cyd = 40 rad/s, the yaw damper tends to behave as a dashpot element; however,
for frequencies higher than 40 rad/s, the damper tends to behave as a stiffness element.

Figure 3 Idealised behaviour of a typical railway damper (see online version for colours)

3 Test procedure
The mathematical model is validated against the experimental results of a primary suspension
vertical damper. The damper characteristics provided are:
• Damping force = 4800 N at 100 mm/s piston velocity. Tolerance ±20%
• Damping force = 10000 N at 300 mm/s piston velocity. Tolerance ±15%
• No internal information is given.
318 A. Sañudo et al.

The tests are carried out with an Instron MHF-25l universal high stroke frequency
hydraulic machine, whose principal characteristics are: test frequencies up to 400 Hz,
±25000 N dynamic force and maximum amplitude of 50 mm. The software uses FFT based
analysis techniques to calculate the dynamic stiffness and the phase.
Two different types of tests were performed. In the first one, the damper was tested
without silent-blocks to obtain the response of the damper without the influence of the
elastic elements. Then, the whole system (damper and elastic elements) was tested to
analyse the influence of the oil compressibility. The experimental setup for each case is
shown in Figure 4.

Figure 4 Experimental test set-up, Left: damper without silent-blocks, right: damper with
silent-blocks (see online version for colours)

The tests inputs were stepped sinus with constant zero-to-peak amplitudes of 100, 200 and
300 mm/s in the entire frequency range. The maximum velocity of the tests corresponds to
the nominal velocity of the damper.

4 Experimental results
In this section, the experimental results are shown in the frequency and time domains. The
objective is to analyse the performance of the damper in a frequency range up to 200 Hz.
First, the results of the tests carried out without silent-blocks are shown. These are compared
with the qualitative conclusions derived from the theory. Then, these results are compared
against the ones obtained in whole system tests to assess the influence of the rubber elements
and the oil compressibility.
Figure 5 represents the variation of damper dynamic stiffness with frequency. It
shows how the performance of the damper is quite similar to that described in Section 2
(Fig. 3).
• At low frequencies, the damper behaves as an ideal dashpot element: the relationship
between the modules of dynamic stiffness and frequency is almost linear. In addition,
the phase angle between the force and displacement is close to 90°.
High frequency modelling and validation of railway dampers 319

Figure 5 (a) Dynamic stiffness and (b) phase (see online version for colours)

As the frequency increases, the damper tends to behave as an ideal spring element: the
dynamic stiffness is constant with the frequency and the phase angle tends to zero.
A deeper insight on the damper behaviour can be obtained by means of the force–velocity
curves at different frequencies. Figure 6 shows the curves for 6, 40, 100 and 160 Hz and for
the three different test velocities (100, 200 and 300 mm/s).
The figure shows that at low frequencies, the damper behaves as an ideal dashpot element.
However, as the frequency increases, the elastic behaviour enlarges. Analysing the figures,
the following conclusions are made:
• The behaviour of the damper is symmetrical: no difference is observed between the
compression and the rebound stroke.
320 A. Sañudo et al.

Figure 6 Force velocity curves for different frequencies and velocities (see online version for colours)

• The relationship between forces and velocities is quite linear, independently from the
test speed and the frequency. This indicates that for this damper, the dashpot element
must be a linear element instead of a quadratic one (Eq. (4)).
Thus, the damper can be modelled by a linear C in series with a constant K in the entire
frequency range.
Figure 7 compares the results obtained by testing the damper with and without silent-
blocks to evaluate the influence of the oil compressibility in the system response.
Figure 7 Comparison of the results from tests carried out with and without silent-blocks (see online
version for colours)
High frequency modelling and validation of railway dampers 321

As expected, the influence of the silent-blocks in the damper behaviour at high frequencies
is very important. It can, however, be observed that the influence of the oil compressibility
is not negligible.

5 Development and validation of a theoretical model


In this section, the parameters of the model are obtained. In Section 2, it was demonstrated that
the K and C parameters of the physical model are related to the constructive characteristics of
the damper. However, in most cases, this information is not available and, therefore, their value
must be obtained from experimental data. This is achieved by comparing the experimental
dynamic response of the damper with the theoretical one. The experimental results prove that
the damper can be modelled by a linear dashpot element in series with a linear spring element.
Thus, the model parameters can be fitted from frequency domain results.
If the damper is subjected to sinusoidal excitation:
x = Xeiωt
(6)
x = iω Xeiωt

As they work in series, each element bears the same force:


f = fdashpot = fspring (7)
The reaction force applied by the damper can be calculated with the following expression:
iωCK
f = x (8)
K + iωC

Therefore, the dynamic stiffness of model KM and the phase angle ϕM are:
ω 2C 2 K ωCK 2
K M (ω ) = +i 2
K +ω C
2 2 2
K + ω 2C 2
⎛ Im ( K (ω )* ) ⎞ (9)
ϕ M (ω ) = tan ⎜ −1

⎜ Re ( K (ω ) ) ⎟
*
⎝ ⎠

The parameters are tuned fitting the dynamic stiffness of model KM to the dynamic stiffness
obtained in the test conducted under nominal velocity EXP K minimising the sum S:
N

(
S = ∑ abs ( K EXP (ωn ) − K M (ωn ) ) )
2
(10)
n =1

The values of the adjusted parameters are: K = 30680 N/mm, C = 34.02 Ns/mm.
Figure 8 shows that both the dynamic stiffness and the phase angle of the model are
identical to the ones obtained experimentally.
To validate the proposed model completely, its response is compared to the experimental
one in the time domain using Simulink. The model is subject to the experimental input
displacement and the response obtained is compared to the experimental force output. In
this respect, it is worth pointing out that the test machine is not able to execute pure sinus.
322 A. Sañudo et al.

Figure 9 proves that the model is able to predict the response of the damper even in this
situation, confirming the validity of the model.

Figure 8 (a) Stiffness (b) Phase (see online version for colours)

The results show that the model is able to predict the damper performance in a low-
to mid-frequency range. The force–velocity curves validate the model and values of the
parameters.
It is interesting to point out that given their size, most railway dampers cannot be tested
on a conventional high stroke frequency machine. However, it is possible to achieve an
acceptable fitting in the low- to mid-frequency range from the data collected in low-frequency
tests, thanks to the theoretical background of the proposed model. To illustrate this, in the
following example, only the results of the test conducted below 20 Hz have been used to fit
the model. The new values are: K = 27681 N/mm, C = 34.5 Ns/mm.
High frequency modelling and validation of railway dampers 323

Figure 9 Comparison between experimental and model response (see online version for colours)

The value of the damping parameter is almost the same, while there is a slight difference
in the stiffness. This is because the influence of the stiffness increases with the excitation
frequency. Nevertheless, Figure 10 shows that the predicted results are within an error of
10%, which is less than the usual tolerance in the theoretical characteristics provided by the
suppliers.
Finally, to predict the response of the whole system (damper and silent-blocks), an
accurate model that represents the dynamic behaviour of the silent-blocks is absolutely
necessary. However, this falls outside the scope of this paper.

6 Conclusions and final remarks


A primary suspension vertical damper has been tested in a frequency range up to 200 Hz.
The theoretical equations and the experimental results show that as frequency increases,
the influence of the stiffness due to the compressibility of the oil is greater in the damper
response. Thus, the model presented, based on the physical one developed in Alonso et al.
(2011), is suited to be used to predict the level of noise vibration and harness (NVH) in a
low- to mid-frequency range.
The experimental data has been analysed to evaluate the behaviour of the damper. The
force velocity curves have shown that, for this damper, the relationship between force and
velocity is linear and the influence of the stiffness is almost symmetrical. In this case, the
324 A. Sañudo et al.

Figure 10 (a) Dynamic stiffness (b) Phase with 10% error bars (see online version for colours)

dry friction has been assumed to be negligible. These assumptions have lead to a simplified
model with linear constant parameters regardless of the movement of the piston in the entire
frequency range, which has been fixed from experimental data by the least square method in
the frequency domain. The results show that the model can accurately predict the response
of the damper in the frequency and time domains. It is also computationally cheap, making
it appropriate for implementation in NVH-CAE models.
The theoretical background of the methodology provides for the tuning of the damper’s
inner characteristics in the design stage, but also fitting the model based on experimental
data. The latter is interesting from the clients’ point of view, given the lack of constructive
High frequency modelling and validation of railway dampers 325

information provided by the suppliers. Furthermore, it has been shown that the parameters
can be fitted from tests performed at low frequencies, given their physical meaning, which is
a remarkable contribution providing for the development of NVH models for large railway
dampers that, given their size, cannot be tested in a common high stroke frequency test
machine.

Acknowledgements
The Basque Government is gratefully acknowledged for the financial support through the
Research Personnel Training Programme of the Department of Education, Universities and
Research.

References
Alonso, A., Gimenez, J.G. and Gómez, E. (2011) ‘Yaw damper modelling and its influence on railway
dynamic stability’, Vehicle System Dynamics, Vol. 49, No. 9, pp.1367–1387.
Conde, A., Gómez, E. and Viñolas, J. (2006) ‘Advances on railway yaw damper characterisation
exposed to small displacements’, International Journal of Heavy Vehicle Systems, Vol. 13,
No. 4, pp.263–280.
Eade, P.W. and Hardy, A.E.J. (1977) ‘Railway vehicle internal noise’, Journal of Sound and Vibration,
Vol. 51, No. 3, pp.403–415.
Fan, R., Meng, G., Yang, J. and He, G. (2009) ‘Experimental study of the effect of viscoelastic
damping materials on noise and vibration reduction within railway vehicles’, Journal of Sound
and Vibration, Vol. 319, Nos. 1–2. pp.58–76.
Gil-Negrete, N. (2004) On the Modelling and Dynamic Stiffness Prediction of Rubber Isolators, PhD
Thesis, University of Navarra, San Sebastián, Spain.
Gil-Negrete, N., Viñolas J. and Kari, L. (2006) ‘A simplified methodology to predict the dynamic
stiffness of carbon-black filled rubber isolators using a finite element code’, Journal of Sound and
Vibration, Vol. 296, Nos.4–5, pp.757–776.
Lalor, N. and Priebsch, H-H. (2007) ‘The prediction of low- and mid-frequency internal road vehicle
noise: a literature survey’, Proceedings of the Institution of uechanical Engineering, Part D: J.
Automobile Engineering, Vol. 221, No. 3, pp.245–269.
Lee, J and Thompson, D.J. (2001) ‘Dynamic stiffness formulation, free vibration and wave motion of
helical springs’, Journal of Sound and Vibration, Vol. 239, No. 2, pp.297–320.
Osborne, G.M. Prater, G. and Shahhosseini, A.M. (2010) ‘Finite element concept modelling
methodologies for pickup truck boxes’, International Journal of Heavy Vehicle Systems, Vol. 17,
No. 1, pp.1–17.
Sjoberg, M. (2002) On Dynamic Properties of Rubber Isolators, PhD Thesis, KTH, Stockholm,
Sweden.
Stenti, A., Moens, D., Sas, P. and Desmet, W. (2010) ‘A three-level non-deterministic modelling
methodology for the NVH behaviour of rubber connections’, Journal of Sound and Vibration,
Vol. 329, No. 7, pp.912–930.
Thompson, D.J. Hemsworth, B. and Vincent, N. (1996a) ‘Experimental validation of the TWINS
prediction program for rolling noise, Part 1: Description of the model and method’. Journal of
Sound and Vibration, Vol. 193, No. 1, pp.123–135.
Thompson, D.J., Fodiman, P. and Mahé, H. (1996b) ‘Experimental Validation of the TWINS prediction
program for rolling noise, Part 2: Results’, Journal of Sound and Vibration, Vol. 193, No. 1,
pp.137–147.
326 A. Sañudo et al.

Thompson, D. (2009) Railway noise and vibration: mechanisms, modelling and means, Amsterdam,
Elsevier.
Van Kasteel, R., Wang, C.G., Lixin, Q., Jin-Zhao, L. and Guo-Hong, Y. (2010) ‘A new shock absorber
model for use in vehicle dynamics studies’, Vehicle System Dynamics, Vol. 43, No. 9, pp.613–631.
Wang, W.L., Huang, Y., Yang, X.J. and Xu, G.X. (2011) ‘Non-linear parametric modelling of a high-
speed rail hydraulic yaw damper with series clearance and stiffness’, Nonlinear Dyn, Vol. 65,
No. 13, pp.13–34.

You might also like