Anda di halaman 1dari 11

Transport Proceedings of the Institution of Civil Engineers

Volume 166 Issue TR2 Transport 166 April 2013 Issue TR2
Pages 113123 http://dx.doi.org/10.1680/tran.9.00048
Three-dimensional modelling of high speed Paper 900048
ballasted railway tracks Received 18/05/2009 Accepted 27/03/2012
Published online 09/11/2012
Banimahd, Woodward, Kennedy and Medero Keywords: maintenance & inspection/mathematical modelling/rail track
design
ICE Publishing: All rights reserved

Three-dimensional modelling
of high speed ballasted
railway tracks
j
1 Meysam Banimahd BSc, MSc, PhD j
3 Justin Kennedy, BEng, MSc
Senior Geotechnical Engineer, Technip UK (formerly of Heriot-Watt Geotechnical Engineer, Technip UK (formerly of Heriot-Watt
University) University)
j
2 Peter Woodward BEng, PhD j
4 Gabriela Medero, MEng, MSc, PhD
Professor of Railway Geotechnical Engineering, School of the Built Lecturer in Civil Engineering, School of the Built Environment,
Environment, Heriot-Watt University, Edinburgh, UK Heriot-Watt University, Edinburgh, UK

j
1 j
2 j
3 j
4

In the present paper, a three-dimensional dynamic railway finite-element model, is utilised to study the geotechnical
performance of high speed railway tracks. This model takes into account many of the railway track components such
as multi-layered ballast and subgrade, nonlinearity in resilient behaviour of the sub-soil, and tracktrain interaction.
Different features of the developed model are briefly presented. The effects of train speed, ballast and subgrade
stiffness, and nonlinearity of the soil behaviour are then investigated. The results of the numerical simulations
regarding the track response and required maintenance levels, in terms of the stress and vibration level experienced
by the ballast and subgrade are also discussed and compared to site measurements. Track design methods are also
reviewed, and a design method based on the dynamic finite-element analysis is proposed and discussed, which aims
at minimising railway track maintenance levels by protecting the subgrade from dynamic ground amplification for
speeds approaching the Rayleigh wave velocity, and also aims at minimising ballast maintenance by setting a
maximum ballast particle velocity.

Notation in rapid deterioration of the track structure, causing ground


Er resilient modulus of the soil failure in the worst cases. In 1997 the Swedish Rail Authority
k 13 material parameters (Banverket) started running X-2000 high speed trains between
patm atmospheric pressure Gotenberg and Malmo. At speeds over 184 km/h large ground
p mean stress vibrations were observed in and around the track. The train speed
q deviatoric stress was immediately reduced and track investigations revealed that
Vr Rayleigh wave velocity Rayleigh waves were developing in the track due to the soft soils
Vs shear wave velocity (Madshus and Kaynia, 1999, 2000). This is a clear example of
v volumetric strain Rayleigh wave development at conventional train speeds due to
q deviatoric strain soft subgrades. In addition to the ground wave development, the
Poissons ratio more frequent passage of trains will subject the track to
r density of the soil medium permanent settlements at a faster rate, which makes the rail lose
its track geometry and alignment. The irregularity generated on
1. Introduction the surface of the rail itself increases the dynamic train load and
High speed trains have brought a number of challenges to consequently accelerates the track deterioration even more.
traditional railway track engineering, specifically to track geo- Ground wave propagation with respect to the cut-off frequencies
technics. Experimental observations (see e.g. Madshus and has been discussed by Yang et al. (2003).
Kaynia, 2000) indicate that running high speed trains on railway
tracks constructed on soft soils induces high levels of vibration in Despite all of the geotechnical complexities associated with high
the track and the surrounding ground. These vibrations can result speed tracks, the design and maintenance of the track is often still

113
Downloaded by [ National University of Singapore] on [17/12/16]. Copyright ICE Publishing, all rights reserved.
Transport Three-dimensional modelling of high
Volume 166 Issue TR2 speed ballasted railway tracks
Banimahd, Woodward, Kennedy and Medero

strongly based on empirical relationships and simple models such nonlinearity and stiffness changes in sub-soils are simulated, and
as beam-spring type models. Although these types of model give the track maintenance level is related to the stress level in the
some insight into the dynamic behaviour of railway tracks, they subgrade, and also the stress and vibration level in the ballast
have several drawbacks. They do not consider the continuity of layer. It is finally discussed how such a three-dimensional
the soil medium and do not properly take into account the inertia dynamic model can be used to design railway track (and
of the sleeper, ballast and ground layers. In addition, the three- substructure improvement) through a design methodology which,
dimensional nature of wave propagation phenomena in the rail- in this paper, is aimed at protecting the subgrade from deteriora-
way track have not been taken into account in these types of tion and minimising the ballast maintenance as the speed
simple model. increases.

A three-dimensional time domain finite-element model can 2. Model configuration


provide an appropriate tool for simulation of railway tracks in A symmetrical finite-element representation of the railway track
different conditions. The finite-element method is applicable to is shown in Figure 1. The sleeper, ballast (and sub-ballast) and
almost any arbitrary geometric condition, allowing the inclusion ground are discretised using 20 noded brick elements. Three-
of embedded structures and multi-layering of the ground. It is dimensional beam-column elements are employed to represent
also possible to include material nonlinearity which affects the the rail. The train loading is introduced either as a sequence of
track response, and the wave radiation damping to infinity can be constant magnitude loads or through train track models. The train
taken into account through the implementation of a appropriate models have been coupled within the 3D track model (Dart3D) to
boundary condition which absorbs the travelling wave energy study the response of the train-track system in the vicinity of
from the source. Accurate modelling of critical velocity and structural/geometrical irregularity, for example stiffness transi-
ground wave propagation is therefore possible (Ekevid and tions near bridges and tunnels (see Banimahd, 2008; Banimahd
Wilberg, 2002). and Woodward et al., 2007). In this particular paper, track
geometry is assumed to be good and hence the effects of track
Limited application of three-dimensional dynamic modelling of irregularities have not been considered (i.e. the train suspension
railway track has been reported in the literature. Lei and Mao system). The train loading is therefore represented by a sequence
(2001) studied the track structure using a three-dimensional of constant magnitude moving loads as shown in Figure 1.
finite-element model. Using a beam on spring type model, the Loading frequencies are simulated by this constant magnitude
loading spectrum on the sleepers was first derived. This load moving load traversing the sleepers across the finite-element
spectrum was then inputted into the three-dimensional finite- mesh via the rail.
element model to study the dynamic response of the track. An
increase in vibration level in the ballast was observed with the In order to reduce stress waves from reflecting back into the
increase in train speed. Similar studies had also been conducted domain, a viscous boundary condition has been employed. The
by Luo et al. (1996), in which they extended the finite-element dynamic equilibrium equation is solved in the time domain using
boundaries to be far enough from the domain of interest instead
of employing any absorbing boundary. Hall (2000, 2003) con- Train locomotive
ducted a limited number of three-dimensional finite-element 26 m
simulations of ballasted railway tracks and simulated the displa- 77 m
26 m
cement of railway track at different track speeds. In this analysis,
Plane of symmetry
the train was modelled as a sequence of loads and dashpots were
placed on the boundaries to avoid any fictitious wave reflections.
Ju and Lin (2004) developed a three-dimensional finite-element
model and studied railways on a homogeneous substructure. They
reported that the vibration level in the area surrounding the track
increases with the train speed. Furthermore, the change in
displacement pattern under the train, when the train speed
approaches the so-called critical track velocity, was simulated. Ju
(2007) also used a similar model to study vibration in a building
near a railway bridge.
Viscous boundary
In this paper, a three-dimensional time domain dynamic nonlinear
finite-element model, called Dart3D (Dynamic Analysis of Rail-
way Track 3D), is presented to simulate the dynamic response of
a typical railway track substructure for speeds approaching the Figure 1. A schematic presentation of the railway track finite-
Rayleigh ground wave velocity. Using the dynamic model, the element model
effects on the performance of the railway track of train speed,

114
Downloaded by [ National University of Singapore] on [17/12/16]. Copyright ICE Publishing, all rights reserved.
Transport Three-dimensional modelling of high
Volume 166 Issue TR2 speed ballasted railway tracks
Banimahd, Woodward, Kennedy and Medero

a modified explicit time integration method as suggested by Zhai the displacement, and consequently the strain, are known at the
(1996). Using this explicit method, no equations need to be start of the time step in the explicit time integration, and
solved and integration may therefore be performed more effi- therefore if the constitutive relationship is expressed in terms of
ciently when compared to implicit methods. In addition, when the strain invariant, its implementation in the dynamic finite-
compared to the classical explicit method, the modified explicit element code will be straightforward; the stiffness is updated
scheme could lead to higher accuracy, particularly if the time step based on the strain level in the corresponding time step, and
chosen is relatively large (Banimahd, 2008). The reader should hence no iteration is required (Banimahd, 2008). By considering
refer to Woodward et al. (2005), Banimahd and Woodward the relationships between stress and strain variables in elasticity,
(2007) and Banimahd (2008) for more details on the model and the universal models can be written in terms of volumetric and
its validation against theoretical solutions and site measurements. deviatoric strains as

In this paper the finite-element mesh is relatively coarse,


k 
particularly in the directions perpendicular to the railway; this is 2: E r k 1 Patm v 2 qk 3
to reduce the running time. Although the current model config-
uration reflects the essential features of high speed tracks, it
should be noted that finer meshing and larger domains will be
where
required to study wave propagation under a very high speed
regime in more detail (particularly post critical responses). This
is the subject of future research, which will employ parallel   k 2   k 3
1 3
k 1 k 1
1=(1 k 2  k 3 )
computing technology to study track response at very high speeds
1  2v 2(1 v)
in a more detailed and efficient way, in particular the track
vibration interaction with adjacent structures. k2
k 2
1  k2  k3
3. Nonlinearity in response of substructure k3
Under typical traffic loading, after a number of load cycles have k 3
1  k2  k3
been applied, the plastic deformation accumulation rate decreases
and the sub-soil will start to show mostly resilient behaviour,
usually represented by the resilient modulus. The rotation of
principal stresses plays a role in this accumulation process. v and q are volumetric and deviatoric strains respectively. This
Nonlinearities in the resilient behaviour of ballast and clayey model was implemented in Dart3D for this work. For clayey
subgrade have been studied for many years and are mainly linked subgrade, the model was calibrated (k1 120.24, k2 0.0 and
to the stress-strain state in the soil medium. Uzan (1992) related k3 0.46) to the cyclic triaxial tests results on clay (Seed et
the resilient modulus of the soil to the mean and deviatoric stress al., 1962). For the ballast, the model parameters are adjusted to
in the soil though a so-called universal model represent the reasonable change of ballast stiffness with induced
stress level (k1 1000, k2 0.40 and k3 0.65). The range of
  k2   k3 ballast stiffness, presented by the calibrated model, compares well
3p q
E r k 1 patm with those considered by different researchers such as Shahu et
1: patm patm
al. (1999) and Hall (2000).

4. Track response
where patm is atmospheric pressure; p and q are mean and The track responses have different characteristic features, for
deviatoric stresses respectively; k 1 , k 2 and k 3 are the material example they depend on properties like the subgrade stiffness,
parameters. Assigning a negative or positive value to k 3 can model train speed, substratum depths and layering properties as well as
the softening or hardening effect of the shear stress on the resilient the train characteristics and track superstructure. At low speeds,
modulus. In clayey subgrades, the resilient modulus depends the displacement pattern is mainly symmetrical; every wheel has
mainly on shear (deviatoric) stress; as shear stress increases, the its own footprint and the displacement field moves with the train
resilient modulus decreases. For granular soil such as ballast, (Figure 2). When the train speed approaches and exceeds the
increases in resilient modulus with increases in mean stress are wave velocity in the ground, it is observed that the train moves
reported. The effect of shear stress is more complicated. Despite ahead of the displacement bowl, indicating that the train is
slight increases in resilient modulus with shear strain, at low moving faster than the subgrade stress waves (e.g. Figure 3(a)).
stress/strain levels, considerable decreases in the resilient modulus This confirms that the wave velocity in the substructure plays a
have been observed as shear stress approaches the failure level. very important role in the high speed track response. In this paper
(for ballasted tracks) high speed is defined by train speeds over
Implementation of Equation 1 should be in an iterative manner to 200 km/h and very high speed by train speeds over 300 km/h
find the stress and resilient modulus at each time step. However (ultra high speed is for train speeds over 400 km/h). The current

115
Downloaded by [ National University of Singapore] on [17/12/16]. Copyright ICE Publishing, all rights reserved.
Transport Three-dimensional modelling of high
Volume 166 Issue TR2 speed ballasted railway tracks
Banimahd, Woodward, Kennedy and Medero

Distance along track: m


Figure 4 shows the normalised speed (normalised to the Rayleigh
0 10 20 30 40 50 60
wave velocity of subgrade) versus the maximum normalised
displacement obtained numerically for a typical track on subgrade
05
Wheel with ES 25 MPa (ES: resilient modulus of subgrade). Experi-
Displacement: mm

mental (field) observations at different sites (Stilton Fen (UK),


15 Ledsgard (Sweden), Amesterdam-Utrecht (Netherlands)) available
in the literature (Madshus et al., 2004) have also been plotted in
Figure 4. A dramatic increase in the displacement is observed, as
25 the train speed approaches the subgrade Rayleigh wave velocity
(Vr ). The 3D numerical simulation reflects the experimental
35 trends, showing that the dynamic displacement may be more than
three times the maximum static displacement when the train
Figure 2. Typical simulated sleeper displacement under a train speed equals the subgrade Rayleigh wave velocity (Vr ).
locomotive at low to medium train speed (V 30 m/s)
The Rayleigh wave velocity itself is close to shear wave velocity
and can be estimated from Equation 3 below
0
5 5
0:87 1:12
Depth: m

10 10 Vr  Vs
3: 1
15 15
20 20
25 25 where and Vs are the Poissons ratio and the shear wave velocity
0 10 20 30 40 50
Distance along track: m respectively. Vs is found from Equation 4
(a) s
0
E
Vs
5 5 4: 2r(1 )
Depth: m

10 10
15 15
20 20 where E and r are Youngs modulus and density of the soil
25 25 medium respectively.
0 10 20 30 40 50
Distance along track: m
(b) An example of the calculation can be given for E 25 MPa and
a typical unit weight of 1800 kg/m3 with Poissons ratio of 0.5 for
Figure 3. At depth views of simulated track displacement the soft clay. The Rayleigh wave velocity for this clay is therefore
responses on soft soil at high speed (V 70 m/s) with different
ballast stiffness (a) Soft ballast, (b) Stiff ballast (contours in mm) 50
ES 25 MPa (simulation)
45
Max. dynamic disp./max. static disp.

Ledsgard (experiment)
Stilton Fen (experiment)
record for ballasted tracks is 574.8 km/h set by a French TGV 40
Amsterdam-Utrecht (experiment)
between Paris and Strasbourg on 3rd April 2007. 35
30
The effect of the stiffness of the ballast layer (BD 300 mm,
25
BD: ballast depth) is also studied in Figure 3. In comparison to a
railway on a stiff ballast (EB 300 MPa, EB: resilient modulus 20
of ballast) a railway system on soft ballast (EB 30 MPa,) 15
results in significantly more asymmetrical displacement patterns 10
with higher displacements at high speeds (V 70 m/s). This
05
indicates that a ductile, fatigue-resistant reinforcement technique
that fully stiffens the ballast will have a positive effect on the 0
0 02 04 06 08 10
track response at higher speeds (e.g. Woodward et al., 2007). Normalised train speed
This has been shown using in situ polyurethane reinforcement at
Newham Bog on the UK East Coast Main Line. Polyurethane Figure 4. Maximum simulated and experimental normalized
reinforcement of the ballast can significantly increase the ballast displacement versus normalized train speed
stiffness over poor ground.

116
Downloaded by [ National University of Singapore] on [17/12/16]. Copyright ICE Publishing, all rights reserved.
Transport Three-dimensional modelling of high
Volume 166 Issue TR2 speed ballasted railway tracks
Banimahd, Woodward, Kennedy and Medero

44
Vr 233 km/h. Many sites in the UK have lower values of Vr
than this, for example Newham Bog and Stilton Fen on the East

Subgrade resillient modulus: MPa


40 V 30 m/s
Coast Main Line. These sites have Rayleigh velocities generally
around 200 km/h (for example at Stilton Fen it was measured
36
around 185 km/h). The transmission of ground vibration from
these sites shows why three-dimensional analysis is necessary.
For example, in order to predict the induced vibration levels of 32
adjacent structures, it is necessary to be able to fully capture the
three-dimensional aspects of the Rayleigh wave development (or 28
V 70 m/s
any other ground vibration mechanism) and its propagation. It is
also important to be able to measure the ground natural frequency 24
and ballast particle velocity (under loading) and have a know-
ledge of the track layering and bedrock level. 20
0 02 04 06 08 10 12 14
Time: s
As discussed previously, the presence of stiffer and/or deeper
upper layers (e.g. three-dimensional ballast reinforcement (Wood- Figure 5. Resilient modulus change in subgrade under a sleeper
ward et al., 2007) on soft ground can reduce the displacement at from nonlinear simulations (nonlinear subgrade)
higher speeds and therefore may shift the track critical velocity
slightly higher. This can be attributed to higher bending stiffness
of the modelled ballast layer and hence a better distribution of stiffness which changes with the induced shear stress level
train loading, and a shift in the system natural frequency. (Equations 1 and 2).
However due to the discrete tensionless nature of unreinforced
ballast, the bending stiffness of the ballast layer is low and the 5. Track maintenance
ballast stiffness will be degraded excessively under a high speed The experience of operating high line speeds on conventional
loading regime and the corresponding modification of the track ballasted track has clearly shown that a considerable increase in
critical velocity cannot be achieved or sustained. Based on the maintenance level is required to ensure good track geometry.
observations, the Rayleigh wave velocity of the subgrade can Dahlberg (2001) has presented a review of several empirical
therefore be practically considered as the track critical velocity of models developed mainly on track loading levels and the number
the railway track system for conventional unreinforced ballasted of load cycles or train passages. The vibration level in the ballast
track systems. It is also seen, from Figure 4, that the displace- layer has also been considered as an important factor in the
ment response is essentially static for train speeds lower than ballast deterioration process. Substantial increases in ballast
50% of the Rayleigh wave velocity and dynamic effects start to acceleration and consequently in the attrition rate of ballast
significantly accelerate when the train speed passes around 70% particles have also been reported with increasing train speed (Pita
of the Rayleigh wave velocity. Therefore 70% of the Rayleigh et al., 2004). Eisenmann and Rump (1997) suggested that the
wave velocity can be treated as the practical speed limit for ballast particle velocity should not exceed 1518 mm/s in order
conventional unreinforced ballasted tracks. to prevent excessive deterioration of the ballast and subsequently
the track geometry.
Based on nonlinear simulations, the resilient modulus of the
subgrade, under the passage of a train locomotive, is depicted in Figures 6(a) and 6(b) present the maximum vertical ballast
Figure 5 for two different train speeds. In both cases, the resilient particle velocity and deviatoric (shear) stress in the subgrade
modulus degrades while the train is passing the point of interest layer respectively, obtained from the numerical model for a
(where the response is presented) due to the induced shear conventional track, plotted against the normalised train speed. A
stresses. While degradation of subgrade stiffness is limited for continuous rise in the ballast particle velocity with train speed is
low speed cases, a significant decrease in subgrade stiffness is observed. According to Eisenmann and Rump (1997), criteria
observed as the train speed increases to 70 m/s (as shown in based on ballast particle velocity can be considered as a guide to
Figure 5). The nonlinear simulation can therefore yield higher track maintenance classification. As seen in Figure 6, as long as
displacements when compared to the linear simulation for these the train speeds are lower than 50% of the Rayleigh wave
higher speeds, implying that both the Rayleigh wave velocity and velocity, the track maintenance levels are seen to remain low
track critical velocity are affected by nonlinearity of the subgrade (ballast particle velocity is lower than 18 mm/s). The advantage
during train loading. of using ballast particle velocity is that it is easily measured on
site.
In embankments, nonlinearity in the track behaviour with train
speed and its effect on the near-surface Rayleigh wave velocity is As the train speed increases beyond the 50% Rayleigh wave
of particular importance (Madshus and Kaynia, 1999). This is velocity value, more frequent levels of track maintenance will be
because the Rayleigh wave velocity is calculated from the soil required to retain the track geometry in unreinforced ballasted

117
Downloaded by [ National University of Singapore] on [17/12/16]. Copyright ICE Publishing, all rights reserved.
Transport Three-dimensional modelling of high
Volume 166 Issue TR2 speed ballasted railway tracks
Banimahd, Woodward, Kennedy and Medero

100 L.M. H.M. S.D. cause high levels of settlement in the subgrade, contributing to the
frequent loss of track geometry (severe deterioration region).
Max. particle velocity: mm/s

80 ES 25 MPa

In terms of track stabilisation, if the operational speed exceeds


60 70% of the Rayleigh wave velocity, which is mainly affected by
subgrade stiffness, ballast layer improvements (for conventional
40 ballast structures) would probably not be able to fully stop ground
vibration but could reduce it however to a level that was consid-
20 ered acceptable depending on the particular Railway Standard
being applied. To fully stop ground vibration and amplification,
0 subgrade treatment is probably necessary using, for example,
0 02 04 06 08 10 lime/cement columns. Potentially the most economical solution
Normalised train speed would be a combination of both subgrade and appropriate ballast
(a) treatments.

40
It is well understood that the closer the induced stress state is to
35 the static failure surface in the soil, the higher the permanent
ES 25 MPa deformation generated and subsequently the higher the level of
30
qmax (dyn.)/max (stat.)

track maintenance required. Since the strength of unreinforced


25 ballast depends significantly on the mean (confining) pressure
level, the stress ratio (i.e. the ratio of the deviatoric stress to the
20
mean stress) can be used to investigate the susceptibility of the
15 ballast layer to permanent settlement by comparing it to the peak
Mohr-Coulomb failure criterion value (given as the stress ratio
10
for ballast). The ratio of induced stress ratio to the stress ratio at
05 plastic yielding, is hereby termed the plastic ratio.

0
Figure 7 shows the plastic ratio distribution in the ballast layer
0 02 04 06 08 10
Normalised train speed under a locomotive. As the plastic ratio approaches one, where
(b) the stress state approaches the static failure surface, more plastic
deformation is generated and hence more frequent maintenance is
Figure 6. (a) Maximum ballast vertical particle velocity versus required. As can be clearly seen, the plastic ratio generally
normalized train speed for a conventional ballasted track increases with train speed. Moreover, the softer the ground is, the
(H.M.: High Maintenance; L.M.: Low Maintenance; S.D.: Severe higher the level of induced plasticity in the ballast. In other
Deterioration). (b) Normalized maximum deviatoric stress applied words, an increase in train speed, along with traversing softer
to subgrade versus normalized train speed for a conventional ground, contributes significantly to the permanent settlement in
ballasted track the ballast layer and thus the loss of track geometry. Therefore
not only does the issue of formation failure arise due to soft
ground, but also soft ground causes the track stiffness to reduce
which means that the cyclic ballast shear strains increase leading
to more track settlement. Principal stress rotation of the ballast
tracks since the ballast particle velocity is now exceeding 18 mm/s. particles has a role in this settlement process. The track stiffness
When the train speed exceeds around 70% of the Rayleigh wave is therefore a key parameter when assessing track maintenance
velocity, high levels of vibration (Figure 6(a)) and displacement levels. It should be noted that simply increasing the unreinforced
(Figure 4) are observed, which would cause rapid deterioration of ballast depth may not yield good results since track stiffness may
the track geometry in unreinforced track. Therefore the track not significantly increase due to the tensionless nature of ballast.
maintenance for unreinforced ballasted tracks can be categorised In these instances what is required is the ability to form a
into three different classes with respect to train speed: low resilient geopavement over the soft ground (Woodward et al.,
maintenance (V , 0.50Vr ), high maintenance (0.50Vr , V , 2007).
0.70Vr ), and severe deterioration (V . 0.70Vr ) regions. The general
trend between the deviatoric stress in the subgrade and normalised The plastic ratio distribution from the nonlinear simulation is, to
train speed fits well with the concept of track maintenance (Figure a large degree, similar to the plastic ratio from the linear
6(b)). The dynamic effect starts to appear when the train speed simulation (ES 40 MPa) for train speeds of 30 m/s and 50 m/s.
exceeds 0.50Vr : A considerable rise in the stress occurs when the However, at a train speed of 70 m/s, the plastic ratio obtained
train speed passes 70% of the Rayleigh wave velocity, which may from the nonlinear simulations (nonlinear subgrade), is generally

118
Downloaded by [ National University of Singapore] on [17/12/16]. Copyright ICE Publishing, all rights reserved.
Transport Three-dimensional modelling of high
Volume 166 Issue TR2 speed ballasted railway tracks
Banimahd, Woodward, Kennedy and Medero

V 30 m/s V 50 m/s V 70 m/s

Nonlin.
BD

Lin. (ES 40 MPa)


Lin. (ES 15 MPa)
01 02 03 04 05 06 07 08 09 10

Figure 7. Plastic ratio distribution in the ballast layer under the


locomotive wheels for simulations with linear and nonlinear
subgrades

higher, and the zones with high plastic ratio appear to extend 40
when compared with the linear simulation. According to this
35
observation, it is confirmed that the maintenance level in the
Deviatoric stress: kPa

ballast can be adversely affected by subgrade nonlinearity. 30


Nonlinearity of ballast behaviour will significantly affect track 25 Nonlinear ballast
response especially at higher speeds. As shown in Figure 8, a
20
subgrade under a nonlinear ballast layer could experience higher
deviatoric stresses and consequently higher permanent settle- 15
ments when compared to the subgrade under linear ballast. 10
Linear ballast
5 (EB 120 MPa)
6. Design implications
The design of railway tracks has been generally empirical/semi- 0
empirical and has followed different guidelines in different 0 02 04 06 08 10
Time: s
countries. The main objective of such design methods is to
protect the subgrade from deterioration (Burrow et al., 2007). Figure 8. Deviatoric stress on subgrade for simulations with linear
The primary modes of traffic-induced deterioration in the sub- and nonlinear ballast (V 70 m/s)
grade can be classified into five categories, namely subgrade

119
Downloaded by [ National University of Singapore] on [17/12/16]. Copyright ICE Publishing, all rights reserved.
Transport Three-dimensional modelling of high
Volume 166 Issue TR2 speed ballasted railway tracks
Banimahd, Woodward, Kennedy and Medero

surface attrition, mud pumping into the ballast, massive shear ments. A knowledge of the actual ground cut-off frequency is
failure, progressive shear failure and excessive plastic deforma- then required; this is best obtained through site measurement.
tion. Among these, subgrade erosion and mud pumping may be
prevented by using appropriately thick granular sub-ballast layers NR/SP/TRK/9039 (Network Rail, 2005) indicates that for train
and ensuring track drainage. Massive shear failure also occurs at speeds higher than 60 km/h, the minimum dynamic sleeper
stresses well above that causing progressive shear failure. Thus support stiffness (measured by a falling weight deflectometer)
the track foundation design methods have mainly been estab- needs to be 100 kN/mm per sleeper end. It also reveals that
lished on preventing excessive plastic deformation and progres- while it may be possible to make a slight increase in critical
sive shear failure in the subgrade (Burrow et al., 2007). Based on velocity by improving the track bed, a significant improvement
this concept, several design procedures have been proposed; see can be made only by deep ground treatment. It is another area
Burrow et al. (2007) and Banimahd (2008) for a detailed review in which a dynamic model can be of great use. As discussed in
and discussion of the design methods. the previous section, 70% of Rayleigh wave velocity can be
treated as the speed limit for conventional unreinforced ballasted
It is absolutely essential at critical velocity sites that any issues tracks; if the speed needs to be increased further, then the
with track drainage are rectified before the track becomes subgrade and/or ballast should be improved. Polyurethane rein-
operational (particularly if the track is saturated). The develop- forcement of the ballast has been applied at the Newham Bog
ment of excess pore water pressure under higher line speeds can site on the UK East Coast Mainline to improve track dynamics
result in a rapid loss of track geometry. Even if sand blankets are for Rayleigh wave problems. While ballast reinforcement will
installed the rate of excess pore pressure developing in the sand not fully stop the ground vibration at the Rayleigh wave velocity,
blanket and formation/subgrade under loading can exceed the it can be used to reduce its effect. Figure 9 proposes a procedure
dissipation rate leading to cyclic softening. The sand blanket for the design of railway tracks using three-dimensional dynamic
itself can liquefy if it is fully saturated. multi-layered models. The procedure fulfils two main objectives:
protecting the subgrade from deterioration and decreasing the
In conventional low-to-medium speed track, the ballast main- level of ballast maintenance at higher speeds (i.e. ground
tenance level is not typically high for passenger trains, as vibration).
discussed in the previous section. On this basis current track bed
design procedures concentrate mainly on reducing the stress level The first objective can be achieved by limiting the stress
and consequently plastic deformation in the subgrade. However, (deviatoric) in the subgrade to a threshold which is proportional
according to the results presented in this paper, as the train speed to the failure stress, or limiting the plastic strain or deformation
increases and approaches the Rayleigh wave velocity (in this as suggested by Li and Selig (1998a, 1998b).
paper taken to be close to the track critical velocity), the level of
required maintenance increases considerably, and frequent main- In order to achieve the second objective, the stress level in the
tenance of the ballast layer will be required. In this case, the ballast layer may be limited to a threshold to avoid generation of
maintenance level of the ballast should be taken into account in excessive plasticity and consequently rapid loss of track geometry
the design procedure or ballast reinforcement should be used at higher speeds which results in high levels of track mainte-
(Woodward et al., 2007). In addition, the speed effect has not nance. The threshold can be based on the deviatoric stress in the
been properly addressed in most design procedures; some ballast or more rigorously based on a plasticity ratio as discussed
empirical equations are used to estimate the dynamic load (e.g. previously. An approach can also be based on limiting the
Li and Selig, 1998a, 1998b) or some basic recommendations on granular layer deformation or strain, both of which are calculated
the minimum dynamic stiffness of railway tracks based on the using empirical relationships. A rigorous constitutive model,
expected train speed and track type (e.g. NR/SP/TRK/9039). By incorporating the effects of mean stress, density and stress-strain
employing a three-dimensional multi-layered track model, it is history, can also be used to estimate the plastic deformation in
possible to estimate the stress level and its change with train the ballast (and subgrade) under repeated dynamic loading caused
speed to a reasonable degree of accuracy, taking into account the by traffic (e.g. Banimahd, 2008; Woodward and Molenkamp,
nonlinearity of the subgrade and ballast response, and the train- 1999; Woodward et al., 2005). The analysis method automatically
track interaction in the presence of any track irregularity such as looks at reducing the wave propagation phenomenon (ground
a transition zone (Banimahd, 2008; Banimahd and Woodward, wave transmission) by setting a limiting value on the ballast
2007). particle velocity depending on the particular Railway Standard
used. The 3D dynamic finite-element approach is capable of
As far as the train speed effect is concerned, a quantitative analysing the dynamic response of the ground and can therefore
criterion is rarely given concerning whether the subgrade itself simulate frequency transmission.
needs to be treated or not. According to WJRC (2002a, 2002b), if
the bearing capacity of the subgrade is lower than 288 kPa, The design procedure presented in Figure 9 can be outlined
ground improvement is required. Wave propagation effects briefly as follows. The Rayleigh wave velocity is estimated for
become particularly important when looking at urban environ- the present configuration of the track in the first step. If

120
Downloaded by [ National University of Singapore] on [17/12/16]. Copyright ICE Publishing, all rights reserved.
Transport Three-dimensional modelling of high
Volume 166 Issue TR2 speed ballasted railway tracks
Banimahd, Woodward, Kennedy and Medero

Track configuration

Subgrade improvement
Ballast improvement

Dynamic model

Subgrade vibration
Ballast vibration

Subgrade stress
Ballast stress

No Yes

Train speed higher than


Acceptable when No 70% of critical velocity?
compared to allowable or
stresses and vibration? subgrade too weak?

Yes

Yes Yes
Design too conservative

No

End

Figure 9. Proposed design procedure of a new high speed


railway track based on the three-dimensional finite-element
model when polyurethane reinforcement of the ballast is not
used.

polyurethane reinforcement of the ballast is not used then the (particle velocity) in the ballast, to acceptable levels. The
subgrade will need to be improved if either the train speed is dynamic model can be used to perform parametric studies on
higher than 70% of Rayleigh wave velocity and ground the depth of the subgrade improvement, ballast depth and
vibration must be prevented, or if the subgrade is too weak. stiffness improvement (and damping) to offer an optimised
The weakness of the subgrade can be judged on the basis of design in terms of lifelong performance and economic cost of
stress or deformation level under the static train load (or low track construction. Improvement of the formation soil can be
speed). Once the subgrade stiffness is considered to be achieved through techniques like lime/cement stabilisation and/
sufficient, improvement of the ballast layer is designed to limit or by partial soil removal and replacement (as determined by
the stress levels in the ballast and subgrade, and vibration level the finite-element model).

121
Downloaded by [ National University of Singapore] on [17/12/16]. Copyright ICE Publishing, all rights reserved.
Transport Three-dimensional modelling of high
Volume 166 Issue TR2 speed ballasted railway tracks
Banimahd, Woodward, Kennedy and Medero

7. Conclusion Ju SH and Lin HT (2004) Analysis of train-induced vibrations and


In this paper, the response of conventional ballasted track was vibration reduction schemes above and below critical
studied employing a three-dimensional dynamic finite-element Rayleigh speeds by finite element method. Soil Dynamics and
model. The increase in track displacement, ballast vibration and Earthquake Engineering 24(12): 9931002.
stress level in subgrade, and ballast with train speed were Lei X and Mao L (2001) Dynamic analysis of the track structure
simulated. It was seen that the track response was essentially of a high speed railway using finite elements. Proceedings of
static when the train speed is lower than half of the track critical the Institution of Mechanical Engineers, Part F: Journal of
speed for unreinforced ballast, being practically equal to Rayleigh Rail and Rapid Transit 215(4): 301309.
wave velocity in the subgrade for conventional unreinforced Li D and Selig ET (1998a) Method for railroad track foundation
ballast track. The track maintenance was also studied in terms of design. I: Development. Journal of Geotechnical and
vibration and stress level in the ballast and the stress level in the Geoenvironmental Engineering, ASCE 124(4): 316322.
subgrade; based on this the maintenance level was classified into Li D and Selig ET (1998b) Method for railroad track foundation
low maintenance, high maintenance and severe deterioration design. II: Applications. Journal of Geotechnical and
zones to fully prevent ground vibration. In low to high main- Geoenvironmental Engineering, ASCE 124(4): 323332.
tenance zones, improvement of the ballast layer will be sufficient Luo Y, Yin H and Hua C (1996) The dynamic response of
whereas the subgrade may need to be improved for the severe railway ballast to the action of trains moving at different
deterioration zone. Based on the outcome of the dynamic finite- speeds. Proceedings of the Institution of Mechanical
element analysis, a design method was finally proposed which Engineers, Part F: Journal of Rail and Rapid Transit
aims at minimising railway track maintenance levels by protect- 210(2): 95101.
ing the subgrade from deterioration and decreasing the level of Madshus C and Kaynia AM (1999) Dynamic ground interaction:
ballast maintenance. A critical issue for high speed train lines on soft soil. In
Geotechnical Engineering for Transportation Infrastructure
REFERENCES (Barends et al. (eds)). Balkema, Rotterdam, the
Banimahd M (2008) Advanced finite element modelling of Netherlands.
coupled train-track systems: a geotechnical perspective. PhD Madshus C and Kaynia AM (2000) Highspeed trains on soft
thesis, Heriot-Watt University, Edinburgh, UK. ground: Dynamic behaviour at critical speed. Journal of
Banimahd M and Woodward PK (2007) Numerical study of train Sound and Vibration 231(3): 689701.
speed effect on railway track response. In Proceedings of 9th Madshus C, Lacasse S, Kaynia A and Harvik L (2004) Geodynamic
International Conference on Railway Engineering, London, challenges in high speed railway projects. In International
UK. Conference on Geotechnical Engineering for Transportation
Burrow MPN, Bowness D and Ghataora GS (2007) A comparison Projects. ASCE, CA, WA, pp. 192215.
of railway track foundation design methods. Proceedings of Network Rail (2005) NR/SP/TRK/9039 Formation treatments.
the Institution of Mechanical Engineers Part F: Journal of Network Rail Track Standards, UK, Issue 1.
Rail and Rapid Transit 221(1): 112. Pita AL, Teixeira PF and Robuste F (2004) High speed and track
Dahlberg T (2001) Some railroad settlement models a critical deterioration: The role of vertical stiffness of the track.
review. Proceedings of the Institution of Mechanical Proceedings of the Institution of Mechanical Engineers, Part
Engineers, Part F, Journal of Rail and Rapid Transit 215(4): F: Journal of Rail and Rapid Transit 218(1): 3140.
289300. Seed HB, Chan CK and Lee CE (1962) Resilience characteristics
Eisenmann J and Rump R (1997) Ein Schotteroberban fur hohe of subgrade soils and their relation to fatigue failures in
Geschwindigkeiten. Eisenbahntechnische Rundschan 3: 99 asphalt pavement. In Proceedings of International Conference
107 (in German). on the Structural Design of Asphalt Pavements, University of
Ekevid T and Wilberg NE (2002) Wave propagation related to Michigan, Ann Arbor, MI, USA, pp. 611636.
high-speed train, a scaled boundary FE-approach for Shahu JT, Kameswara Rao NSV and Yudhbir (1999) Parametric
unbounded domains. Computer Methods in Applied study of resilient response of tracks with a subballast layer.
Mechanics and Engineering 191(36): 39473964. Canadian Geotechnical Journal 36(6): 11371150.
Hall L (2000) Simulations and analyses of train-induced ground Uzan J (1992) Resilient characterization of pavement materials.
vibrations: A comparative study of two and three dimensional International Journal for Numerical and Analytical Methods
calculations with actual measurements. PhD thesis, Royal in Geomechanics, 16(6): 453459.
Institute of Technology, Stockholm, Sweden. WJRC (2002a) Construction and Maintenance Standards for
Hall L (2003) Simulations and analyses of train-induced ground Commuter and Local Railway Track. West Japan Railway
vibrations in finite element models. Soil Dynamics and Company, Osaka, Japan.
Earthquake Engineering 23(5): 403413. WJRC (2002b) Construction and Maintenance Standards for
Ju SH (2007) Finite element analysis of structure-borne vibration Shinkansen Track. West Japan Railway Company, Osaka,
from high-speed train. Soil Dynamics and Earthquake Japan.
Engineering 27(3): 259273. Woodward PK and Molenkamp F (1999) Application of an

122
Downloaded by [ National University of Singapore] on [17/12/16]. Copyright ICE Publishing, all rights reserved.
Transport Three-dimensional modelling of high
Volume 166 Issue TR2 speed ballasted railway tracks
Banimahd, Woodward, Kennedy and Medero

advanced multi-surface kinematic constitutive soil model. Proceedings of the Institution of Civil Engineers Transport
International Journal for Numerical and Analytical Methods 160(3): 109115.
in Geomechanics 23(15): 19952043. Yang YB, Hung HH and Chang DW (2003) Train-induced wave
Woodward PK, Zettor B, Kaddouri A and Banimahd M (2005) propagation in layered soils using finite/infinite element
Advanced nonlinear dynamic finite element modelling of simulation. Soil Dynamics and Earthquake Engineering
railway track behaviour. In Proceedings of 8th International 23(4): 263278.
Conference on Railway Engineering, London, UK. Zhai WM (1996) Two simple integration methods for large-scale
Woodward PK, Thompson D and Banimahd M (2007) dynamic problems in engineering. International Journal for
Geocomposite technology: reducing railway maintenance. Numerical Methods in Engineering 39(24): 41994214.

WHAT DO YOU THINK?


To discuss this paper, please email up to 500 words to the
editor at journals@ice.org.uk. Your contribution will be
forwarded to the author(s) for a reply and, if considered
appropriate by the editorial panel, will be published as a
discussion in a future issue of the journal.
Proceedings journals rely entirely on contributions sent in
by civil engineering professionals, academics and students.
Papers should be 20005000 words long (briefing papers
should be 10002000 words long), with adequate illustra-
tions and references. You can submit your paper online via
www.icevirtuallibrary.com/content/journals, where you
will also find detailed author guidelines.

123
Downloaded by [ National University of Singapore] on [17/12/16]. Copyright ICE Publishing, all rights reserved.

Anda mungkin juga menyukai