You are on page 1of 11

Journal of Sound and Vibration 464 (2020) 114992

Contents lists available at ScienceDirect

Journal of Sound and Vibration


journal homepage: www.elsevier.com/locate/jsvi

Railway track degradation: The contribution of a spatially


variant support stiffness - Global variation
€l Steenbergen*
Mehran Sadri, Tao Lu, Michae
Delft University of Technology, Faculty of Civil Engineering and Geosciences, Railway Engineering Group, Stevinweg 1, 2628, CN Delft, the
Netherlands

a r t i c l e i n f o a b s t r a c t

Article history: The effect of large-scale variation in the support stiffness on railway track degradation is
Received 23 May 2019 studied using a frequency-domain approach. The model used can deal with parametric
Received in revised form 26 September 2019 excitation due to both the discrete sleeper spacing and arbitrary large-scale spatial track
Accepted 27 September 2019
non-uniformity. Adopted stiffness profiles are based on realistic datasets in the literature.
Available online 28 September 2019
Handling Editor: D.J. Thompson
The sensitivity to degradation is assessed by quantifying the energy dissipation in the
substructure over the influence zone. Results show that the effect of spatial stiffness
variation generally increases with the speed, for any subgrade condition; system resonance
Keywords:
Track degradation
however leads to increased degradation at resonance speeds and increases with the mean
Spatial variation value of the track stiffness. The speed is shown to have a larger influence in the presence of
Frequency-domain model non-uniformity than it has for uniform track with a mean value of the same non-uniform
Energy dissipation track stiffness, independent of this mean value. In general, support stiffness non-
Non-uniform track support uniformity and poor track conditions (in terms of a low overall stiffness) may have com-
Dynamic track stiffness parable effects; the combination of both is a worst-case scenario. Predictions are inde-
pendent from the randomness for measured datasets and have therefore general validity.
Further, an excellent correlation is found between the spatial variation of the dynamic
track stiffness, the differential energy dissipation in the substructure, and the work per-
formed by the moving contact load with respect to the track, independent of the train
speed. This confirms existing empirical evidence of the dynamic track stiffness for non-
uniform track as an indicator for degradation.
© 2019 Elsevier Ltd. All rights reserved.

1. Introduction

In a previous study [1] a frequency-domain model of ballasted railway track has been derived, able to deal with both
parametric excitation due to the discrete sleeper support and arbitrary forms of spatial system non-uniformity. Different
measures to describe the energy dissipation in the substructure over the non-uniform section and its influence length were
benchmarked with respect to their performance to describe the sensitivity to degradation in defined cases of parametric
variation. This work applied the derived model to specifically investigate the case of local non-uniformity in the support
stiffness, also known as ‘hanging sleeper/sleepers’. The present work will shift the scope to large-scale or global variation of
the track stiffness. In other terms: the main object of this study consists in tracks with a non-uniform stiffness profile, where

* Corresponding author.
E-mail address: m.j.m.m.steenbergen@tudelft.nl (M. Steenbergen).

https://doi.org/10.1016/j.jsv.2019.114992
0022-460X/© 2019 Elsevier Ltd. All rights reserved.
2 M. Sadri et al. / Journal of Sound and Vibration 464 (2020) 114992

‘stiffness’ basically denotes the static property of the support and with a minimum spatial resolution equal to the sleeper
spacing.
A literature review, both with respect to the definition, role and measurement of the track stiffness in the degradation
process under train loading and the frequency-domain modeling of train-track interaction in the presence of non-uniformity,
has been given in the previous work [1] and is therefore not repeated here. This included the state of the art with respect to
the effects of local non-uniformity in the form of hanging sleepers. Here, this literature review will be extended to large-scale
variability in the track support stiffness. Like in the case of local non-uniformity, the state of the art in this domain typically
concentrates on model development and effects in terms of train-track interaction at the level of the wheel-rail interface
[2e7], however with a few recent exceptions that merit a discussion in more detail at this place. Studies [8,9] explicitly
address the role of track stiffness variability in the context of track degradation and settlement, using a computational time-
domain model, measured stiffness records (obtained with the help of the Swedish Rolling Stiffness Measurement Vehicle
(RSMV)) and parametric updates to predict the development of the actual geometry. This work establishes a clear correlation
between the measured dynamic track stiffness along the track and the development of the track geometry in terms of set-
tlement. The effect of the probabilistic distribution of the track support stiffness, together with different vehicle and track
variables, on the track degradation rate, is investigated in Ref. [10], using a numerical time-domain model. Also this work finds
that stiffness variability contributes significantly to track deterioration. In Ref. [11], support stiffness variability is considered
in the framework of environmental vibration radiation; numerical analysis and its experimental validation show the actual
loaded track geometry to be a governing parameter for the analysed case. The present study mainly distinguishes itself from
the state of the art in the scientific literature in the following points:

 Arbitrary large-scale spatial variation in track properties is addressed with an approach in the frequency domain, allowing
for parametric analysis and fundamental understanding;
 The sensitivity to degradation in the presence of defined cases of support stiffness non-uniformity is assessed quantifying
the energy dissipation in the substructure; this approach avoids the need of predicting an actually developing and site-
specific geometry, allowing for only qualitative - but at the same time more general conclusions;
 The suitability of the dynamic track stiffness as an indicator for the expected degradation, used on an empirical basis in
railway engineering (see also [8,12]), is theoretically verified.

In Ref. [1] a systematic analysis was performed for all independent design variables of the entire track system, providing
insight in the role of each component and its properties in the degradation process. Since there is no reason to assume that
small-scale (or local) and large-scale (or global) non-uniformity would lead to different conclusions, this specific analysis is
not repeated in the present investigation. Like in Ref. [1], a viscous description is used for the energy loss in the track sub-
structure, which may be physically inexact but allows for a frequency-domain approach while sufficient for qualitative
analysis. Further, a constant damping-to-stiffness ratio is assumed again in the description of the non-uniformity of the
support stiffness.
The structure of this paper is as follows. Section 2 briefly reviews the theoretical model, derived in Ref. [1], and extends this
derivation with some formulations that are relevant in the scope of this work. Section 3 presents and discusses computational
results. Section 4 ends with conclusions.

2. Theoretical framework

This section discusses the theoretical framework to study spatially variant track; a detailed mathematical derivation is
given in Ref. [1]. The periodic model is shown in Fig. 1, along with important notations. The model incorporates a double
‘periodicity layer’: the first one accounting for the discrete sleeper spacing, and the second one for large-scale periodicity of
adjustable length, which may take into account arbitrary and global variation in system properties. As shown in Fig. 1, the
track is discretely supported by equidistantly positioned supports. Each support element consists of a railpad with stiffness kp
and viscosity cp, a sleeper with mass Ms and the ballast/subgrade is characterized by its stiffness kb and damping coefficient cb.
Timoshenko beam theory is employed to model the track with its elastic properties listed in Fig. 1. From the train vehicle, only
one single axle load is taken into account, by means of an unsprung wheel mass coupled to the track through a linearized
contact stiffness kH, and uniformly moving at speed v along the track. The track is divided into a number of identical sections
and each section consists of N1þN3 identical supports and N2 middle supports with different characteristics; N1 and N3 can be
chosen identical, however this is not necessary. The number of sections needed is designated as NOS.
In the framework of this study, it is meaningful to address the steady-state response of the system to subcritical train
loading; transient behaviour is not considered. Further, the model addresses large-scale non-uniformity of the support
stiffness, which disrupts the process of parametric excitation in the time domain for the ideal uniform case. The associated
issue of instability [13,14] of the moving oscillator due to unlimited growth of the response may therefore be ignored.
Moreover, in case the issue would play a role, potential instabilities would be suppressed by the necessary inclusion of
material damping in the model. Finally, in the case of non-uniformity, transition radiation under moving loading will occur at
track discontinuities [15,16]. The used model description is suitable to cope with this phenomenon.
M. Sadri et al. / Journal of Sound and Vibration 464 (2020) 114992 3

Fig. 1. Schematic model overview of the model for non-uniform track [1].

The wheel-rail contact force Fc is solved in the mathematical derivation in Ref. [1]. After obtaining the contact force, other
quantities such as displacements can be expressed in terms of the contact force. The displacement of the track is given by:
 
1 X X
Nt Nt 2p
jm l*
x
wb ðx; tÞ ¼  qmþðNt þ1Þ;nþðNt þ1Þ ðt  x = vÞF n e ; (1)
v m¼N n¼Nt
t

in the time-domain where j is the imaginary unit. The energy dissipation in the ith support is obtained by:

Zþ∞ Zþ∞
Ediss;i ¼ fd;i ðtÞvs;i ðtÞ dt ¼ cb;i vs;i ðtÞ2 dt (2)
∞ ∞

where vs;i ðtÞ is the velocity response of the ith sleeper in the time domain obtained by taking the inverse Fourier transform of
its frequency response.
The work performed by the moving wheel-rail contact force with respect to the track can be described as:

Zt2 ¼x2 =v
t2Z

W¼ Fc vb dt ¼ Fc vb dt; (3)
t1 t1 ¼x1 =v

where Fc is the contact force and vb is the vertical velocity of the track at the contact point, equal to the time derivative of the
track displacement wb (Eq. (1)) at the contact point, namely the time derivative of wb ðvt; tÞ, which is given by:
2   3
X
Nt
6 2p X
Nt jm 2p
l*
vt
7
vb ðvt; tÞ ¼  4jm * qmþðNt þ1Þ;nþðNt þ1Þ ð0ÞF n e 5 ; (4)
m¼N
l n¼Nt
t

The integration range of Eq. (3) is determined by the time span, governing, together with the speed, the traveled distance
of the wheel and its contact position.

3. Parametric study of global non-uniformity in the track support

The model convergence, in terms of appropriate choices of minimum values for N1,N3 and NOS per simulated case, is
basically identical for cases of local (addressed in Ref. [1]) and global support stiffness non-uniformity, whereas the choice of
N2 is determined in the latter case by the length of the non-uniform section and the number of sleepers included.
With respect to the system parameters, the same approach is used as in both earlier works [1,17], i.e. nominal, lower and
upper values are assigned to each wheel/track parameter according to Table 1. A static wheel load of 100 kN and a shear
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
coefficient (k) of 0.34 are used, and the linearized contact stiffness (kH ) is calculated as 3 3  1022 ,Q , with Q the static wheel
4 M. Sadri et al. / Journal of Sound and Vibration 464 (2020) 114992

Table 1
Values of the input parameters of the model used in the parametric study [12].

Parameters Lower value Nominal value Upper value


Rail
Bending stiffness (EI), MNm2 e 4.25 (54E1 profile) 6.11 (60E1)
Mass per length (rA), kgm1 e 54.4 (54E1 profile) 60.34 (60E1)
Sleeper
Half sleeper mass (Ms), kg 50 142.5 202
Distance between sleepers (ls), m 0.5 0.6 0.7
Pads
Pad stiffness (kp), MNm1 30 200 1000a
Pad damping (cp), kNm1s 20 30 e
Ballast
Ballast stiffness (kb), MNm1 40 50 60
Ballast damping (cb), kNm1s e 55 e
Moving wheel
Speed (km/h) 0 e 300
Unsprung mass (kg) 600 900 1200
a
The pad stiffness of 1000 MN/m corresponds to the Corkelast pad used for typical Dutch track.

load. Finally, again the constant damping-to-stiffness ratio is set as default in the modeling of the non-uniform part of the
track support, taken as the ratio of the nominal values of both parameters, i.e. cb =kb ¼ 5:5  104 =5  107 .
Concerning the description of the expected degradation in terms of energy dissipation over the influence length of the
non-uniform zone, it was shown [1] that two measures are meaningful:

 The root mean square (denoted as RMS) of energy dissipation in the influenced area Ne, defined as:

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
u1 X Ne
Ediss;rms ¼ t E2 (5)
Ne i¼1 diss;i

with Ediss;i according to Eq. (2);

 The maximum differential dissipation (denoted as MDD) between adjacent sleepers over the entire influenced area Ne,
defined as:

MDD ¼ maxjεj; (6)

with elements of vector ε defined as DDi ¼ εi ¼ Ediss;iþ1  Ediss;i ; i ¼ 1; 2; :::; N1 þ N2 þ N3  1.


Whereas for the case of the local non-uniformity MDD was an indicator unrelated to the spatial configuration of the non-
uniformity, and therefore not providing spatial information, this can be expected to be essentially different in the case of
large-scale global non-uniformity. It can be expected that MDD will depend, much more than RMS, on the ‘bandwidth’ of the
stiffness variation, and therefore the standard deviation of the distribution of the non-uniform stiffness and its extreme
values. MDD is therefore a suitable parameter for probabilistic analysis, whereas for given or defined cases of non-uniformity,
localised information is obtained without inherent general validity. There are only few experimental records of stiffness non-
uniformity available in the literature (as will be discussed in paragraph 3.1), and the aim of this study is not to investigate in
depth the effect of the stiffness distribution itself on degradation, but to establish general effects of parametric system
variation in the presence of arbitrary non-uniformity. RMS of the energy dissipation over the affected zone is therefore the
most meaningful choice; for illustration purposes MDD will be computed and shown as well in some cases.
Further paragraphs of this section are set up as follows. Paragraph 3.1 discusses the model input in terms of non-uniform
support stiffness profiles; paragraph 3.2. Studies the effects of the degree of non-uniformity in relation to adjacent track, and
paragraph 3.3 investigates the suitability of the dynamic track stiffness as an indicator for expected degradation for tracks
with a varying support stiffness.

3.1. Model input: non-uniform stiffness profile

As input to the simulation, an amount of N2 non-identical adjacent sleeper support stiffness values must be defined.
Preferably, these values are obtained from experimental records in order to be realistic. In the published literature, two
M. Sadri et al. / Journal of Sound and Vibration 464 (2020) 114992 5

different stiffness data records [4,18] can be found with a sufficient, sleeper spacing resolution. They may represent both
relatively normal to good [4] and poor [18] track conditions respectively, and are therefore suitable to be implemented here,
allowing for comprehensive conclusions.
The first measurement is provided by Nielsen and Oscarsson [4], for track conditions that may be classified as normal to
relatively good. The authors applied five different levels of static preloading with a superimposed dynamic loading (at 2 Hz)
on 19 adjacent sleepers (unfastened from the rail) and measured the subgrade/ballast secant stiffness along a high-speed
track. For the analysis in this study, datasets for two levels of pre-loading are selected from the five datasets reported in
Ref. [4], and subgrade/ballast stiffness values in each dataset are then distributed randomly in space along the non-uniform
section of the model with N2 ¼ 19 sleepers, representing a 12 m track section. The first dataset (preloading at 15 kN per half-
sleeper, which may represent empty passenger or freight cars) and the second dataset (preloading at 45 kN per half-sleeper,
which may represent loaded cars) are used, with mean values respectively at 66 and 130 MN/m per half-sleeper. The stiffness
profiles corresponding to these two datasets are shown in Fig. 2a. The mean value of each dataset depicted in this figure is
used as the stiffness value for the sleeper supports (kb;1 ) within the adjacent and uniform sections of the track, and therefore
as a kind of ‘reference’.
The second measurement is reported in Ref. [18] and may represent relatively poor track conditions (with minimum
values of 25 MN/m per half-sleeper). In this case, sleeper support stiffness values in the non-uniform section of the model are
chosen randomly from the dataset, in order to obtain a non-uniform track section with equal length as in the previous case
(12 m). The resulting realization of the stiffness profile in this case is shown in Fig. 2b. It is worth noting that the mean value
given in this figure, as well as both extreme values used in the next paragraph, are calculated over the entire original track
section of 400 m long [18], and also this value is employed as reference (kb;1 ) for the uniform parts of the track.

3.2. Effects of the degree of non-uniformity in relation to adjacent track

As discussed, the sensitivity to degradation is assessed primarily on the basis of RMS of the mechanical energy dissipation
in the substructure in area Ne. Fig. 3 shows RMS as a function of the speed, for both datasets corresponding to normal/good
track conditions and the dataset corresponding to poor track conditions. For all three datasets taken into account, the per-
formance of the non-uniform track is benchmarked against that of a uniform track. In analogy to the results of earlier work
[1], it is obvious from the results that the choice of RMS as an evaluation parameter is appropriate, uniform and non-uniform
tracks showing qualitatively similar behaviour. For each given dataset, three cases of uniform track are considered; they are
assumed to have identical sleeper supports with a stiffness equal to respectively the mean, minimum and maximum values of
the dataset. The local peaks in the trends correspond to velocities where the first system resonance occurs. Comparison of the
trends for non-uniform track and for uniform track with a stiffness value corresponding to the mean of the corresponding
dataset shows that the speed has a larger influence on the expected degradation in the presence of non-uniformity than it has
for uniform track, independent of the track quality. Further, comparing the graphs for the three cases in Fig. 3 in absolute
sense, it can be concluded that the effect of non-uniformity (Fig. 3a) and a uniform but very low stiffness (28 MN/m) (Fig. 3c)
are worst-case scenarios with respect to degradation; improving the overall track stiffness is generally more favorable in this
sense than reducing the non-uniformity (Fig. 3b). On the other hand, for stiffer track the resonance peak becomes more
pronounced, limiting the validity of the previous conclusion to operational speeds outside the resonance zone.
As discussed at the beginning of this section, RMS as a description of the dissipated energy is not expected to be influenced
by the distribution of the stiffness data within the non-uniform section for each dataset, unlike MDD, which provides local
information and depends on extreme values of the stiffness variation within a record. The realisations in Fig. 2 are random for
all datasets. Fig. 4 shows, for each of the three datasets/track qualities, two other random realisations. Fig. 5 shows the
corresponding energy dissipation over the influence area Ne in terms of both RMS and MDD. As can be observed from this
figure, the trend of RMS is indeed almost identical for the different realisations, whereas the trends for MDD differ for the

Fig. 2. Spatial distribution of sleeper support stiffness for the non-uniform track section; (a) relatively good track conditions, (b) relatively poor track conditions.
6 M. Sadri et al. / Journal of Sound and Vibration 464 (2020) 114992

Fig. 3. RMS of mechanical energy dissipation obtained for (a) normal track conditions e unloaded (b) normal track conditions e loaded, and (c) poor track
conditions.

three realisations, not only quantitatively but also qualitatively, and especially with increasing speed. Further, MDD is
significantly and systematically lower for a high ‘overall’ track stiffness level (or mean value); MDD is therefore not only
dependent on the ‘bandwidth’ of the variation but also on the mean value.

3.3. The dynamic track stiffness as an indicator for expected degradation: verification

In the previous study [17] the variation of the dynamic track stiffness, which is a measure to describe the ‘perceived’
stiffness by the moving train axles, and often used in practice on an empirical basis [5,8,12], was discussed as an engineering
parameter to assess sensitivity to degradation. The parameter was shown to be inappropriate for assessing the influence of
track design parameters in the case of an idealized uniform track, but at the same time it should be added that demonstrable
experimental evidence has only been reported precisely for tracks with a non-uniform support stiffness [5,8,12]. Therefore,
M. Sadri et al. / Journal of Sound and Vibration 464 (2020) 114992 7

Fig. 4. Two more random realisations of the stiffness profile for each of the three datasets in Fig. 2. (a) normal unloaded track; (b) normal loaded track; (c) poor
track conditions.

the concept and its suitability will be verified again in the framework of this study on spatially variant track. Two different
definitions of the dynamic track stiffness were proposed and elaborated in the previous study: (i) the ratio of the static load
(half the axle load or wheel load) to the dynamic displacement of the contact point defined in a moving reference system, and
(ii) the ratio of the contact force (including the static wheel load and dynamic terms) to the dynamic displacement of the
contact point defined in a moving reference system. In the case of non-uniform track, the second definition seems physically
less appropriate, as it is not uniquely related in space to local track properties; a local stiffness variation may lead to a dynamic
effect in the moving contact at a different position, whereas settlement can be expected to be governed by a locally perceived
stiffness variation. Further, from a measuring point of view the first definition is much more straightforward to apply.
Therefore, the first definition is adopted here, according to Eq. (54) in Ref. [17]. Fig. 6 shows both the variation of the dynamic
track stiffness and the variation of the energy dissipation, normalised by the damping value, in the substructure along the
track. The results are shown for poor track conditions according to the stiffness profile as shown in Fig. 2b, and for six different
train velocities. For both quantities, the differential value in absolute sense in space is taken; alternatively also the first spatial
derivative could be taken and would give similar results. Finally, both quantities are normalised with their respective
maximum in order to facilitate the comparison. The correlation level observed in Fig. 6 is impressive, providing indeed a
theoretical basis for the experimentally validated correlation between the intensity of dynamic stiffness variation and
degradation along a railway track [8]. Moreover, simulation results suggest that dynamic track stiffness measurements should
not necessarily be performed at relatively low speeds (according to experimental practice), but operational train velocities
can be used as well, which is a conclusion of great practical relevance. It is worth noting in this context that the recent
research [8] confirms a strong correlation between dynamic stiffness variation and the development of differential track
settlement especially for a low substructure stiffness; both aspects are perfectly in line with and confirmed by this study.
Fig. 7 shows the correlation of the dynamic stiffness variation with a second quantity that seems relevant in the framework
of degradation: the work performed by the moving total contact load with respect to the track. This work is calculated
8 M. Sadri et al. / Journal of Sound and Vibration 464 (2020) 114992

Fig. 5. Effect of randomness on energy dissipation measures. (aeb) normal unloaded track; (ced) normal loaded track; (eef) poor track conditions.

according to Eq. (3); each calculation value corresponds to integration with boundaries at the position of this value and the
position of the next sleeper (in analogy to the calculation of the differential dynamic stiffness), and therefore comprises the
work performed within one sleeper bay. Also this correlation can be observed to be excellent. Together with the results in
terms of energy dissipation, this analysis provides therefore a strong theoretical basis for the use of the dynamic track
stiffness as an indicator for track degradation in terms of differential settlement.

4. Conclusions

This study used a frequency-domain model, able to deal with parametric excitation due to both the discrete sleeper
spacing and arbitrary large-scale spatial variation in track properties, to study the effect of such variation in the support
stiffness on the long-term railway track behaviour. The sensitivity to degradation in the presence of defined cases of large-
scale non-uniformity, obtained from realistic datasets in the literature, was assessed quantifying the energy dissipation in
the substructure over the influence zone. The following conclusions applied:

i. The energy dissipation in the non-uniform section of the substructure is well described by its root mean square (RMS)
value over the influence area; the extreme value of the differential dissipation (MDD) provides localised information
related to extreme stiffness variations within a record;
ii. The effect of spatial variation in track support stiffness on the expected degradation generally increases with the speed,
for any subgrade condition;
M. Sadri et al. / Journal of Sound and Vibration 464 (2020) 114992 9

Fig. 6. Spatial distribution of the normalised differential values (in absolute sense) of the dynamic stiffness and the energy dissipation in the substructure for
poor track conditions (stiffness profile as depicted in Fig. 2(b); (a) v ¼ 25 km/h, (b) v ¼ 40 km/h, (c) v ¼ 80 km/h, (d) v ¼ 140 km/h, (e) v ¼ 200 km/h and (f)
v ¼ 300 km/h.

iii. Similar to uniform track, system resonance remains in the case of non-uniform track and leads to increased degradation
at resonance speeds; the resonance peak disappears with decreasing track stiffness;
iv. The speed has a larger influence on the expected degradation in the presence of non-uniformity than it has for uniform
track with a mean value of the track stiffness, independent of the magnitude of this track stiffness itself;
v. Effects of realistic support stiffness non-uniformity and a uniform but low stiffness (poor track conditions) are com-
parable in order of magnitude; improving the overall track stiffness is generally more favorable than reducing the non-
uniformity, unless the speed is within the resonance area, a stiffer track increasing the effect of resonance;
vi. Trends in terms of RMS, as a function of speed and overall track stiffness, are rather independent on the randomness for
given datasets and therefore - qualitatively - generally valid; this is not true for MDD.

Finally, the suitability of the dynamic track stiffness as a practical indicator for the expected degradation was verified
theoretically for non-uniform track. Simulations showed under these conditions an excellent correlation in space between:

 The variation of the dynamic track stiffness (defined as the ratio of the static load and the dynamic displacement of the
contact point in a moving reference system),
 The differential energy dissipation in the substructure,
 The work performed by the moving contact load with respect to the track.

Therefore, the practical experience reported in the literature concerning the performance of the dynamic track stiffness is
confirmed theoretically in this study. Moreover, the correlation was established independent of the train speed, whereas no
10 M. Sadri et al. / Journal of Sound and Vibration 464 (2020) 114992

Fig. 7. Spatial distribution of the normalised differential values (in absolute sense) of the dynamic stiffness and work performed by the contact force to the track
for poor track conditions (stiffness profile as depicted in Fig. 2(b); (a) v ¼ 25 km/h, (b) v ¼ 40 km/h, (c) v ¼ 80 km/h, (d) v ¼ 140 km/h, (e) v ¼ 200 km/h and (f)
v ¼ 300 km/h.

practical experience has been gained yet for regular operational train speeds. The conclusion is therefore of significant
practical relevance.

Acknowledgements

The study carried out in this paper is part of a long-term research program aiming at an improved understanding of the
physics of track degradation and environmental vibration radiation under train loading. The program is executed by Delft
University of Technology with financial support from the Dutch rail infra manager ProRail. The authors would like to thank
especially Arjen Zoeteman from ProRail for his commitment to make scientific work in this domain possible.

References

[1] M. Sadri, T. Lu, M. Steenbergen, Railway track degradation: the contribution of a spatially variant support stiffness - local variation, J. Sound Vib. 455
(2019) 203e220.
[2] J. Oscarsson, Dynamic train-track interaction: variability attributable to scatter in the track properties, Veh. Syst. Dyn. 37 (1) (2002) 59e79.
[3] J. Oscarsson, Simulation of train-track interaction with stochastic track properties, Veh. Syst. Dyn. 37 (6) (2002) 449e469.
[4] J.C. Nielsen, J. Oscarsson, Simulation of dynamic trainetrack interaction with state-dependent track properties, J. Sound Vib. 275 (3e5) (2004)
515e532.
[5] M.X.D. Li, E.G. Berggren, A study of the effect of global track stiffness and its variations on track performance: simulation and measurement, Proc. Inst.
Mech. Eng. - Part F J. Rail Rapid Transit 224 (5) (2010) 375e382.
[6] V.A. Fernandes, F. Lopez-Caballero, S.C. d'Aguiar, Probabilistic analysis of numerical simulated railway track global stiffness, Comput. Geotech. 55
(2014) 267e276.
[7] L. Xu, W. Zhai, A new model for temporalespatial stochastic analysis of vehicleetrack coupled systems, Veh. Syst. Dyn. 55 (3) (2017) 427e448.
M. Sadri et al. / Journal of Sound and Vibration 464 (2020) 114992 11

[8] J.C. O Nielsen, E.G. Berggren, A. Hammar, F. Jansson, R. Bolmsvik, Degradation of railway track geometryeCorrelation between track stiffness gradient
and differential settlement, Proc. Inst. Mech. Eng. - Part F J. Rail Rapid Transit (2018), https://doi.org/10.1177/0954409719879780.
[9] J.C. Nielsen, X. Li, Railway track geometry degradation due to differential settlement of ballast/subgradeeNumerical prediction by an iterative pro-
cedure, J. Sound Vib. 412 (2018) 441e456.
[10] I. Grossoni, A.R. Andrade, Y. Bezin, The role of track stiffness and its spatial variability on long-term track quality deterioration, Proc. Inst. Mech. Eng. -
Part F J. Rail Rapid Transit 233 (1) (2019) 16e32.
[11] M. Germonpre , J.C.O. Nielsen, G. Degrande, G. Lombaert, Contributions of longitudinal track unevenness and track stiffness variation to railway
induced vibration, J. Sound Vib. 437 (2018) 292e307.
[12] E.G. Berggren, Railway Track Stiffness. Dynamic Measurements and Evaluation for Efficient Maintenance, PhD. thesis, Royal Institute of Technology
(KTH), 2009.
[13] S.N. Verichev, A.V. Metrikine, Instability of vibrations of a mass that moves uniformly along a beam on a periodically inhomogeneous foundation, J.
Sound Vib. 260 (5) (2003) 901e925.
[14] K. Abe, Y. Chida, P.E.B. Quinay, K. Koro, Dynamic instability of a wheel moving on a discretely supported infinite rail, J. Sound Vib. 333 (15) (2014)
3413e3427.
[15] M.J.M.M. Steenbergen, Physics of railroad degradation: the role of a varying dynamic stiffness and transition radiation processes, Comput. Struct. 124
(2013) 102e111.
[16] A.B. Fara
ga
u, A.V. Metrikine, K.N. van Dalen, Transition radiation in a piecewise-linear and infinite one-dimensional structureea Laplace transform
method, Nonlinear Dyn. (2019) 1e27.
[17] M. Sadri, M. Steenbergen, Effects of railway track design on the expected degradation: parametric study on energy dissipation, J. Sound Vib. 419 (2018)
281e301.
[18] M.J. Brough, G. Ghataora, A.B. Stirling, K.B. Madelin, C.D. Rogers, D.N. Chapman, Investigation of railway track subgrade. Part 2: case study, Proc. Inst.
Civ. Eng. Transp. 159 (2) (2006) 83e92.

You might also like