Transport Proceedings of the Institution of Civil Engineers

Volume 167 Issue TR1 Transport 167 February 2014 Issue TR1
Pages 3–14 http://dx.doi.org/10.1680/tran.11.00023
Design for efficient drainage of railway Paper 1100023
track foundations Received 26/04/2011 Accepted 24/10/2011
Published online 19/09/2012
Rushton and Ghataora Keywords: groundwater/mathematical modelling/rail track design

ICE Publishing: All rights reserved

Design for efficient drainage

of railway track foundations
j
1 Ken R. Rushton PhD, DSc, CEng, MICE, MCIWEM j
2 Gurmel Ghataora PhD, MIMM, MILT
Emeritus Professor, School of Engineering, University of Birmingham, Senior Lecturer, School of Engineering, University of Birmingham,
Birmingham, UK Birmingham, UK

j
1 j
2

The efficient drainage of water from railway track support systems is critical in maintaining the stiffness of the
subgrade following substantial rainfall. In a preliminary investigation, the results from a full-scale laboratory
experiment of drainage through ballast and sub-ballast were examined; a numerical model was developed to
represent these experimental results. That initial study is extended to consider the effect of the geometry and the
hydraulic properties of the track foundation on the rate at which water drains from the system. A design
methodology is presented which shows that the time for almost complete drainage depends on the permeability and
specific yield (drainable porosity) of the underlying sand blanket. A number of additional conditions are examined
including the slope of the subgrade, the intensity and duration of the rainfall, and the occurrence of a surcharge in
the drainage ditch. The provision of a laterally and vertically permeable geocomposite directly above the subgrade
can reduce the time for drainage by a factor of ten.

Notation that the track support system (ballast to subgrade) provides a

h water table elevation uniform response to loading with a relatively consistent stiffness
K permeability along the length of the track, allowing for seasonal variations.
m saturated thickness of layer Since the subgrade has a significant impact on stiffness, any
q recharge effects of water on subgrade stiffness should be designed out. In
SY specific yield (drainable porosity) terms of railway track support system it is necessary to ensure
t time that any surface water, arising from precipitation or localised
x horizontal coordinate flooding, can be removed rapidly. The scale of the problem on the
s, b suffixes for sand blanket and ballast UK railway network has been described by Ghataora et al.
(2006), who found that in certain regions there are about five wet
1. Introduction spots per mile.
The need for efficient drainage of water from the support systems
of railway tracks is stressed in the texts of Hay (1982) and Selig In an important experimental study of the drainage of water from
and Waters (1994). A number of documents (Lichtberger, 2005; railway track support systems (Heyns, 2000), results were
Network Rail, 2005; Selig and Waters, 1994; UIC, 1994) describe obtained from a full-scale laboratory investigation in which
methodologies for railway track design including drainage. ballast and sub-ballast were located above an impermeable base;
Drainage of water from railway track foundations is important simulated rainfall entered the track support system and then
because an increase in moisture can result in a decrease in drained laterally to a monitored outlet. A series of experiments
strength and hence in the bearing capacity of the subgrade soil. were conducted with different rainfall intensities and durations;
This in turn leads to an increase in settlement due to shear/plastic the slope of the impermeable base was adjusted between
deformation of the soil, leading to a change in stiffness of the horizontal and 5% (1:20). Heyns measured the change of the
track support system. Both the magnitude and rate of deteriora- water table elevation with time and monitored the outflows from
tion of the subgrade depend on soil mineralogy, stress history, the system as the water table rose into the ballast. After the
void ratio and density. Since soils tend to be variable in nature, rainfall stopped, the water table fell from the ballast into the sub-
their response to changes in moisture also differ, resulting in ballast until it drained to the lowest point on the sloping base.
variations in stiffness. The aim of a good track design is to ensure Important insights gained from the experiments included faster

3
Transport Design for efficient drainage of railway
Volume 167 Issue TR1 track foundations
Rushton and Ghataora

rates of drainage with higher subgrade slopes, while reduced uniform stress at the sub-ballast/subgrade interface. This layer
permeabilities of the sub-ballast resulted in slower drainage. also has the function of preventing the migration of fines from
the subgrade to the overlying ballast layer. Therefore, it should be
These experiments provide a benchmark for developing and designed both as a structural layer, with suitable strength and
testing analytical and numerical models (Youngs and Rushton, stiffness, and as a filter to contain the fines in or near to the
2009a, 2009b). Both two-dimensional and one-dimensional Du- subgrade surface. The sub-ballast and the ballast layers are
puit–Forchheimer solutions were explored; acceptable results for designed to protect the subgrade.
a sloping-bed two layer system were obtained using the Dupuit–
Forchheimer approximation. The importance of an appropriate 2.2 Effect of moisture on the subgrade
downstream boundary condition is discussed by Rushton and Subgrade soils can comprise a range of soil types that include
Youngs (2010). This Dupuit–Forchheimer approach was devel- both granular and cohesive materials. Studies undertaken by Hunt
oped further by Rushton and Ghataora (2009) by preparing a and Evans (1993) have shown that the subgrade has a significant
numerical model that successfully reproduced the time-variant effect on the behaviour of the track under loading. Changes in the
response of the experiments of Heyns (2000). A limited number properties of subgrade soils due to increases in moisture content
of alternative conditions were examined using the numerical were observed by Hornych et al. (1998). Figure 1 indicates that
model, including the effect of the inclination of the subgrade increases in moisture content under dynamic loading (minimum
surface and the ratio of the permeabilities of the sub-ballast and load p, maximum load q) result in a decrease in resilient modulus
the ballast. Mr (the ratio of difference in applied stress to strain under a
dynamic load) and an increase in the plastic strain. Thus, in order
In this paper the study of railway track drainage is extended to to ensure that the railway track performs in a predictable and
provide information to assist in design. The dominant properties consistent manner, it is essential that either movement of water in
in determining the rate of drainage are shown to be the per- the track layer is controlled or the layers affected by loading are
meability and the specific yield (drainable porosity) of the sand made insensitive to changes in moisture content by stabilising
blanket that is located between the ballast and the subgrade. A them. The latter can be expensive. Traditionally, water levels in
design graph is presented that indicates how the time for a water and around the track are controlled by the use of suitable
table to fall to the subgrade depends on the ratio of permeability drainage systems. These are designed to cater for storm events.
to specific yield of the sand blanket. Since the properties of the For example, in Britain drainage pipes in the cess area are
sand blanket are likely to change due to dynamic loading designed so that they will run full bore for a two-year rainfall
resulting from the passage of trains and rolling stock, laboratory event (I. Harper, personal communication, 2009). The worst-case
tests are used to estimate how these properties are modified. It is scenario is for a five-year event, where pipes are designed so that
common practice to place permeable geotextiles between the the water level will not rise above the base of the sleeper. Thus
ballast and the sand blanket. There have also been suggestions the ballast layer is used as a temporary reservoir. The track is still
(Alobaidi et al., 2005) that geocomposites with substantial lateral considered to be operational when water is at the base of the
and vertical permeabilities can improve the drainage. Alternative sleeper. Train operations are likely to be affected when water
locations for geocomposites are considered and the resultant rates appears at the surface of the ballast.
of drainage are estimated using the numerical model.
Cedergren (1974) investigated the effect of saturation of leaky
road pavements on their useful life. He introduced severity factors
2. Effect of moisture on subgrade
that depend of the cross-section thickness, the height of free
2.1 Components of railway track support systems
In general, railway track support systems consist of ballast, 250 1·4
overlying sub-ballast and the underlying subgrade soils. Each of
Resilient modulus Mr: MPa

Mr (p ⫽ 20 kPa, q ⫽ 40 kPa)
Axial plastic strain ε after

ε 1·2
these layers has functional requirements that relate to structural 200
80 000 cycles: %

1·0
stability and durability in the challenging environment of
150 0·8
dynamic loading and seasonal climatic change. In terms of
engineering the track support, one of the most critical issues is 0·6
100
track drainage. This is because each component (ballast, sub-
0·4
ballast and subgrade) is affected by water. In general the greater 50
0·2
the fines content, the larger the impact of water. Consequently
there is a larger impact when water is present in the sub-ballast 0 0
and on the subgrade layers, compared to water in the ballast, 5 7 9 11 13
Moisture content: %
which is normally a single-sized rock.
Figure 1. Effects of increases in moisture content on both the
Sub-ballast layers are designed to support the ballast layer by resilient modulus and the plastic strain (from Hornych et al., 1998)
spreading the concentrated stress from ballast particles to a more

4
Transport Design for efficient drainage of railway
Volume 167 Issue TR1 track foundations
Rushton and Ghataora

water in the structural section and the state of deterioration. essential to understand the rate at which water levels decline in
Figure 2, which is based on a simple calculation developed by the various track support layers.
Cedergren, indicates the useful life of a cross-section, expressed
as a percentage of the life of a perfectly drained section and how
3. Practical information and parameters
it varies with the time that the section is saturated. For average
relevant to the current study
severity conditions, Figure 2 indicates that, if the pavement is
saturated for about 10% of the time (5 weeks per annum), there 3.1 Practical aspects of track construction
is a reduction in pavement life of approximately 45% compared The railway track structure should be designed to withstand the
with a perfectly drained structural section. If the pavement is combined effects of traffic loading and climate so that the
saturated for 5% of the time (2.6 weeks per annum), there is a subgrade is adequately protected and railway vehicle operating
28% reduction in life. These findings are also applicable to costs, safety and comfort of passengers are kept within acceptable
railway tracks, hence the importance of maintaining low moisture limits. A layout used by Network Rail in the UK is shown in
in the track support layers. Figure 3. The track support system typically comprises the rail, a
fastening system, rail pads, sleepers, ballast, sub-ballast (often a
The above discussion highlights the importance of ensuring that sand blanket) and subgrade.
water is removed from above the subgrade layer with the
retention period kept to a minimum. In order to achieve this, it is 3.2 Rainfall and recharge
In this study, the main interest is drainage following periods of
of a perfectly drained structural section: %

heavy rainfall. It is assumed that all the rainfall becomes recharge

Useful life as a percentage of the life

100 to the ballast. Water draining from the ballast and sand blanket is
taken away by drains, but these drains also collect runoff from
80 locations such as the sides of a cutting. Consequently, surcharge
in the trackside drains will be considered.
60
Rainfall recharge is assumed to occur at a rate of 0.8 m/d
40 (33 mm/h) for 0.5 h; this is a typical English rainfall for a five-
year return period (Ward and Robinson, 1990). The effects of
20 different rainfall intensities or different durations are considered
in the sensitivity analysis of Section 4.4.
0
0 10 20 30 40 50
Percentage of time structural section is saturated: % 3.3 Hydraulic properties of sand blanket
Laboratory tests have been carried out to obtain provisional
Figure 2. Relationship between pavement life and period of estimates of the permeability and specific yield (drainable
saturation for conditions of average severity (after Cedergren, porosity) of a sand blanket. Tests were carried out on specimens
1974) of sand after compaction, described as ‘before dynamic loading’,
and also ‘after dynamic loading’ of 1 million cycles; this

700 mm 700 mm
(min) (min) Clean ballast to
promote drainage

Separating Ballast
geotextile

Subgrade Blanketing sand

Minimum crossfall
1 in 30 nominal thickness 100 mm Uninterrupted path
Trackbed layer thickness to drain
0·4–0·8 m for speeds ⬍ 100 m/h
0·5–1·0 m for higher speeds

Figure 3. Typical cross-section of railway trackbed foundation

(drawn from Figure 8 of Network Rail (2005))

5
Transport Design for efficient drainage of railway
Volume 167 Issue TR1 track foundations
Rushton and Ghataora

Recharge q
represents an estimated 25 million cumulative tonnes of load (Li
and Selig, 1998). Water table
Ballast

The permeability of the sand was determined by first compaction

of the blanket sand, which complied with Network Rail line Subgrade
Sand blanket Geotextile
standard NR/SP/TRK/033 (Network Rail, 1998); compaction was
at the optimum dry density using the BS 1377:1990 method (a)
clause 3 (3-layer 27 blows) (BSI, 1990). Constant-head per-
Water table
meability tests were then conducted in accordance with the Ballast
procedure described by Roy and Sayer (1989). The specific yield
was estimated by saturating each specimen and ascertaining the
ratio of the volume of water drained under gravity to the total Sand blanket
volume of the specimen. Typical results for the permeability and (b)
specific yield are quoted in Table 1.
Water table
Ballast
3.4 Geotextiles and geocomposites
Various types and forms of geosynthetics, textiles and composites
are used in railways to prevent material from entering the ballast Sand blanket
from the subgrade or sand blanket layer. Separating geotextiles
(c)
(used in conjunction with a sand layer) may be either needle
punched or heat bonded with a permittivity (water flow normal to Ballast
the plane in term of quantity per unit area per unit time) in the Sand blanket Water table
range of 45 to 120 l/(m2 s). h ⫺ zb ⫺ ms
Seepage h Kb No
face ms
Ks flow
Geocomposites with both lateral and vertical permeabilities are zb
Datum
also available. The geocomposite considered in this study consists B x Top of A
of a highly permeable core with permeable geotextiles above and subgrade
below the core. The geocomposite is about 10 mm thick with an (d)
in-plane effective permeability of about 0.01 m/s for a slope of
Figure 4. Different stages of drainage following recharge due to
0.05 (1 in 20).
rainfall: (a) conditions towards the end of the recharge period;
(b) conditions early in the drainage stage; (c) conditions when
Geocomposites are also being used as a replacement for the sand
drainage is almost complete; and (d) mathematical formulation
blanket layer in railway track support systems.

4. Exploring drainage of railway tracks

drainage continuing from the sand blanket. When drainage is
using a numerical model
nearly complete, Figure 4(c), there are only small saturated
4.1 Conceptual model regions above the subgrade adjacent to the drains.
Various stages in railway track drainage are illustrated in the
conceptual model of Figure 4. The dual tracks are symmetrical A method of studying railway track drainage is described by
about the centreline. There is a sand blanket above the subgrade Rushton and Ghataora (2009). The technique is based on a time-
that slopes from the centreline, with ballast above the sand variant numerical model using the one-dimensional Dupuit–
blanket. Figure 4(a) indicates that when significant rainfall Forchheimer approximation for a two-layered system above a
recharge occurs, the water table rises into the ballast with water sloping subgrade. In the Dupuit–Forchheimer approach, the
discharged from the ballast and sand blanket. After recharge groundwater head is a function of the horizontal coordinate alone.
ceases, the water table falls below the ballast, Figure 4(b), with The model represents the unknown saturated thickness and the
permeability of the two layers; results from this model compare
well with two-dimensional analyses (Laplace solutions) that
Permeability: Specific represent the groundwater head distribution on vertical sections
105 m/s yield: % (Youngs and Rushton, 2009a, 2009b) This technique was initially
developed to represent the full-sized experiment in a specially
Before dynamic loading 13 2.5
designed flume (Heyns, 2000) in which water was sprayed onto
After dynamic loading 1.2 1.6
ballast and the discharges from the ballast and sub-ballast were
Table 1. Results (these are average values from three tests) monitored. Groundwater head information from a series of
manometers connected to the bottom of the flume was also

6
Transport Design for efficient drainage of railway
Volume 167 Issue TR1 track foundations
Rushton and Ghataora

recorded. The Dupuit–Forchheimer model reproduced satisfact- specific yield are 1.16 3 102 m/s and 0.20 respectively (Rushton
orily the downstream discharges and the water table fluctuations and Ghataora, 2009). The rainfall recharge equals 0.8 m/d
in the ballast and sub-ballast. (33.3 mm/h) for a period of 0.5 h (0.0208 day), which is equiva-
lent to a total recharge of 16.7 mm. Due to symmetry, only one
Figure 4(d) illustrates the mathematical formulation; due to half of the cross-section is considered. In the numerical model
symmetry only the left-hand track is considered. In the Dupuit– there are 36 mesh subdivisions; the time step is 0.0001 d.
Forchheimer approximation, the water table elevation h above the
lowest point of the subgrade is a function of the horizontal Results from the numerical model for the change with time of the
coordinate x. However, the saturated thickness is also a function saturated thickness (water table elevation minus elevation of the
of this unknown water table elevation. Additionally, the sand subgrade) at the mid point of the right-hand track are included in
blanket has a hydraulic conductivity Ks ; for the ballast the Figure 5(a). There is a very rapid increase in saturated thickness
hydraulic conductivity is Kb : The impermeable base is at a height when recharge commences; at the end of the recharge period, the
of zb above datum. At location A the sand blanket is fully water table is in the ballast at 0.134 m above the sand blanket (or
saturated, hence the hydraulic conductivity of the sand is multi- 34 mm above interface between the sand blanket and the ballast).
plied by the sand thickness, ms : The saturated thickness of the Once the recharge ceases the water table falls quickly below the
ballast at location A is h  zb  ms ; this is multiplied by Kb . interface. Note that the duration of recharge is short compared
with the total time taken for the water table to fall through the
The differential Equation 1 describing the flow through the ballast lower-permeability sand blanket.
and sand blanket to the downstream drain has three components.
The left-hand expression describes the flow through the sand Figure 5(b) indicates water table elevations on the cross-section
blanket and the ballast; the first term on the right-hand side at different times. At the end of the recharge period of 0.5 h, the
defines water taken into or released from storage as the water water table is above the interface over the entire cross-section;
table rises or falls. The final term describes any recharge entering consequently, water is shed from both the sand blanket and the
the system ballast, as illustrated in Figure 4(a). At 1.0 h (0.5 h after the end
of recharge) the water table is just above the interface apart from
@ @h @h at the upstream end of the section where the water table is just
1:
[K s ms þ K b (h  zb  ms )] ¼ SY q
@x @x @t below the interface. This is followed by a slow fall in the water
table. At 1.0 day the water table is just above the subgrade at the
upstream end; by 3.0 days there is only a small region at the
For location B, the water table is within the sand; hence bottom of the slope where water is still present; see Figure 4(c).
modifications are required to the left-hand side of Equation 1.
The left-hand boundary is a seepage face. Rushton and Youngs To indicate when drainage is almost complete, a reference time is
(2010) describe a method of representing the height of the defined as the time when the water table falls to an arbitrary
seepage face. When the water table at the right-hand boundary is height of 1 cm above the subgrade at the mid point of the sloping
above the base, Figures 4(a) and 4(b), there is no lateral flow. side. For the example of Figure 5, the reference time is 2.6 days
However, as the water table falls, part of the base becomes dry as after recharge ceases; it is indicated by the vertical arrow on
indicated in Figure 4(c); consequently, the extent of the flow Figure 5(a).
region is gradually reduced. A finite-difference iterative technique
with small time steps is used to solve this equation (Rushton and 4.3 Design methodology
Youngs, 2010). When a geocomposite is present, an extra layer is From a series of simulations using the numerical model with
included in the mathematical formulation. different parameter values for the sand blanket, the permeability
divided by the specific yield was identified as the ratio that
4.2 Representative example governs the reference time. The higher the permeability of the
A representative example, Figure 5, is selected to illustrate how sand blanket, the more rapid the fall in the water table, whereas a
the numerical model provides information about both the rise in larger specific yield leads to a slower fall. In the design chart of
the water table in the sand blanket and ballast due to rainfall, and Figure 6, the solid line indicates the time, after the cessation of
the fall when recharge ceases. Parameter values for the represen- recharge, for the water table at the mid point to fall to 1 cm
tative example are as follows. For each of the twin tracks the above the base for different values of the permeability divided by
distance from the centreline to the ditch is 3.8 m, the subgrade the specific yield. This design curve is based on a slope of the
slope is 1:20 (5%). The sand blanket, which is 100 mm thick, is formation of 5%, a thickness of the sand blanket of 0.10 m and a
separated from the high-permeability ballast by a geotextile recharge of 0.8 m/d for 0.5 h.
through which water passes freely. The permeability of the sand
blanket is set at K ¼ 2.0 3 105 m/s; the specific yield (drainable As an example of the use of the design curve, consider a sand
porosity) of the sand blanket is SY ¼ 0.05; these values are blanket with a permeability of 3 3 105 m/s and a specific yield
consistent with Table 1. For the ballast, the permeability and of 0.03; hence K/SY ¼ 100 3 105 m/s. From Figure 6, the

7
Transport Design for efficient drainage of railway
Volume 167 Issue TR1 track foundations
Rushton and Ghataora

Recharge q

0·15 End of recharge period

Ballast
Water table falls
below geotextile
Water table above base: m

1·10
Sand blanket Geotextile

Water table above subgrade

at mid point Time to fall to
0·05 1 cm above base
at mid point

1 cm limit
0
0 1 2 3
Time: d
(a)
Time
0·30 0·5, 1, 3, 6, 12 h
Water table above datum: m

Ballast Geotex
tile at in
terface
0·20

Sand b Day 1
lanket
0·10 Subgra
de
Day 3

Slope 5%
0
0 1·0 2·0 3·0
Distance from centreline: m
(b)

Figure 5. Representative example: (a) height of the water table

above the subgrade at the mid point of one track; and
(b) cross-section water table elevations at different times. Slope
of subgrade 5%; recharge 0.8 m/d for 0.5 h (0.0208 d); sand
blanket, 0.1 m thick, K ¼ 2 3 105 m/s (1.73 m/d), SY ¼ 0.05;
ballast K ¼ 1.16 3 102 m/s (1000 m/d), SY ¼ 0.20

reference time for the water table to fall to within 1 cm of the For example, if there is substantial reduction in the duration of
base at the mid-point is just over 1.0 day. the recharge from 0.0208 d to 0.005 d (change 1), the design
curve would be of the form shown by the chain dotted line in
4.4 Sensitivity analysis Figure 6; note that the chain dotted line is parallel to the solid
A sensitivity analysis is used to explore the consequences of line. Consequently all reference times equal 0.83 of the values for
alternative parameter values on reference times. For each of the representative example; this factor is quoted in the fourth
the examples discussed below, a design curve was obtained column of Table 2. Doubling the duration of the recharge
The sensitivity of the design curves to alternative parameter increases the reference times by 1.01. If the recharge intensity is
values is summarised in Table 2; two alternative values of reduced to 0.30 m/d compared with the value for the representa-
individual parameters are usually considered. Modifications are tive example of 0.8 m/d (change 2), the reference times are
made to the duration and magnitude of the recharge, the reduced to 0.93 of the original values, whereas a threefold
properties of the sand blanket and ballast, and the slope of increase in the intensity of recharge to 2.4 m/d has only a small
the subgrade. effect on the reference times.

8
Transport Design for efficient drainage of railway
Volume 167 Issue TR1 track foundations
Rushton and Ghataora

100 Recharge q
Reference time after recharge ceases for saturated thickness at

Sand blanket Ballast

Subgrade slope 1:20
0·1 m thick
10
Geotextile Subgrade
3·8 m
mid point to fall to 1 cm: d

1
Recharge duration 0·0208 d

Recharge duration 0·005 d

0·1

0·01

1 10 100 1000 10 000

K /SY for drainage blanket: 10⫺5 m/s

Figure 6. Design graph; reference time when the water table

falls to 1 cm above the subgrade at the mid point of one of the
tracks as a function of K/SY for the sand blanket (solid line). Slope
of subgrade 5%, recharge 0.8 m/d for 0.5 h (0.0208 d); sand
blanket, 0.1 m thick, K and SY vary; ballast K ¼ 0.116 m/s
(10 000 m/d), SY ¼ 0.20. Chain dotted line refers to a recharge
duration of 0.005 d

Change Representative New Fraction of New Fraction of

representative representative

1. Duration of recharge 0.0208 d 0.005 d 0.83 0.0417 1.01

2. Recharge 0.8 m/d 0.3 m/d 0.93 2.4 m/d 1.01
3. Thickness of sand blanket 0.10 m 0.05 m 0.74 0.14 m 1.13
4. Permeability of ballast .
0 116 m/s .
0 001 16 m/s 1.22 .
0 000 116 m/s 1.43
10 000 m/d 100 m/d 10 m/d
5. Specific yield of ballast 0.20 0.10 1.00 0.30 1.00
6. Slope of subgrade 1:20; 5.0% 1:40; 2.5% 2.10 1:200; 0.5% 6.60
7. Surcharge in ditch 0.0 m 0.33 m 1.11

Table 2. Sensitivity analysis for the time taken, as a fraction of

that for the representative problem, for the water table to fall to
1 cm above the subgrade at mid point

Further results in Table 2 indicate that halving the thickness of (change 4) to one-hundredth of its original value to simulate
the sand blanket (change 3) causes a shortening of the reference fouling of the ballast, increases the reference times by a factor of
times, whereas increasing the sand blanket thickness lengthens 1.22; a further tenfold decrease in the ballast permeability leads
the reference times. Reducing the permeability of the ballast to reference times 1.43 times the original. These changes in

9
Transport Design for efficient drainage of railway
Volume 167 Issue TR1 track foundations
Rushton and Ghataora

ballast permeability are consistent with the effect of ballast table elevation at various times. At 0.5 h, the end of the recharge
fouling described by Selig and Waters (1994). Changes to the period, the water table is similar to that of Figure 5 although it is
specific yield of the ballast (change 5) have a negligible effect. higher towards the ditch due to the rising water level in the ditch.
By 4.8 h, when the ditch water level is at a maximum, the entire
Reducing the slope of the subgrade (change 6) has a substantial water table is in the ballast. As the ditch water level drops, the
effect on the times taken for the water table to fall. If the slope of water table falls; at 8.4 h it is in the ballast for about one-third of
the subgrade is halved from 1:20 to 1:40, reference times for the cross-section. The reference time for the water table to fall to
drainage increase to 2.1 times the original values. If the slope of 1 cm above the subgrade at the mid-point of the section is 2.9 d;
the subgrade is nearly horizontal at 1:200, drainage of water from this is about 10% longer than for the situation when there is no
the sand blanket takes six times longer. Consequently, failing to surcharge in the ditch.
provide a sufficient slope on the subgrade radically affects the
drainage of water from the sand blanket. The results of Table 2 indicate that the rate at which water drains
from the railway track and its foundations depends primarily on
For the final example of the sensitivity analysis (change 7), the the permeability and specific yield of the sand blanket. The slope
effect of a surcharge in the drainage ditch is examined; a specific of the subgrade is also a critical feature; if the slope is flatter than
example is presented in Figure 7. The build-up of water in the 1:20, the time for the water to drain out is increased substantially.
drainage ditch is assumed to continue for some time after the Fouling of the ballast also results in longer times for drainage.
rainfall event due to water entering the drainage ditch from the
sides of a cutting or from upstream. A uniform rate of rise is 5. Incorporating a geocomposite in the
assumed, with a maximum water level in the ditch of 0.33 m track foundation
occurring after 4.8 h (0.2 d). The water level then falls; this fall is In the examples of the preceding section, a geotextile is located
effectively complete by 0.7 d (16.8 h). Figure 7 shows the water between the sand blanket and the ballast; this geotextile is

Recharge q

Max. water Surcharge

Ballast
table in ditch

Sand Subgrade
Geotextile
blanket
(a)
Changing
surcharge
4·8 h
4·8 h
0·30
Water table above datum: m

0·5 h

0·20 Ballast
1d 8·4 h 8·4 h
1h
Sand b
Geotex lanket
0·10 tile at in
terface 3d 1·0 h
Subgra
de 0·5 h
Slope 5%
0
0 1·0 2·0 3·0
Distance from centreline: m
(b)

Figure 7. Effect of surcharge in drainage ditch: (a) sketch of

cross-section with surcharge; and (b) change of water table
elevation with time. Maximum water level in ditch is 0.33 m
above subgrade at 0.2 d (4.8 h) after recharge commences; slope
of subgrade 5%; recharge 0.8 m/d for 0.5 h (0.0208 d); sand
blanket, 0.1 m thick, K ¼ 2 3 105 m/s (1.73 m/d), SY ¼ 0.05;
ballast K ¼ 1.16 3 102 m/s (1000 m/d), SY ¼ 0.20

10
Transport Design for efficient drainage of railway
Volume 167 Issue TR1 track foundations
Rushton and Ghataora

sufficiently permeable to have no effect on the flow of water The results of Figure 8(c) demonstrate that locating a geocompo-
between the ballast and the sand blanket. However, Section 3.4 site immediately above the subgrade results in very rapid
introduces a laterally permeable geocomposite that has a substan- drainage. An alternative design curve has been prepared for this
tial permeability both vertically and laterally. The potential situation. In Figure 9 the solid line refers to a laterally and
advantages of incorporating such a geocomposite in railway track vertically permeable geocomposite 1 cm thick with
foundations are explored using the numerical model. The rate at K ¼ 300 3 105 m/s (259 m/d) located immediately above the
which water drains from railway track foundations following subgrade. Most reference times are substantially less than when
rainfall is determined for three alternative arrangements: a con- there is no geocomposite (the broken line is the design curve
ventional permeable geotextile between the sand blanket and the from Figure 6). As the ratio of the permeability divided by the
ballast; a laterally permeable geocomposite between the sand specific yield of the sand blanket decreases, the reference time
blanket and the ballast; and finally a laterally permeable geocom- tends to a constant value of 0.12 d. However, this curve only
posite between the subgrade and the sand blanket with a conven- applies when the vertical permeability of the sand blanket is
tional permeable geotextile between the sand blanket and the sufficiently high for the recharge and drainage water to move
ballast. For each of these examples the permeability and specific downwards to the geocomposite. Consequently, for the parameters
yield of the sand blanket are respectively 2.0 3 105 m/s and used in this example, the vertical permeability of the sand blanket
0.05; for the ballast the permeability and specific yield are should equal or exceed the recharge rate, which is 0.8 m/d
1.16 3 102 m/s and 0.20. (0.9 3 105 m/s).

1. Figure 8(a) contains numerical model values for water table 6. Discussion and conclusion
heights for a conventional geotextile between the sand This investigation into the drainage of railway track foundations
blanket and the ballast. At the end of the recharge period of uses a numerical model that successfully reproduced results from
0.5 h (0.0208 d), the water table is within the ballast for the full-scale laboratory experiments. Insights from the numerical
entire cross-section. During the next 0.5 h the water table model indicate how railway track foundations can be designed to
falls quickly through the ballast. Thereafter there is a steady ensure that water drains rapidly from the system.
fall in the water table so that after three days most of the
sand blanket is dewatered, with the remaining small saturated 6.1 Design methodology
region extending for less than half of the cross-section. Design curves have been prepared to estimate the reference time
2. Figure 8(b) refers to the situation of a laterally permeable (the time after the cessation of recharge for the water table to fall
geocomposite between the sand blanket and the ballast. The to within 1 cm of the subgrade at its mid-point) for various
sand blanket thickness is 0.09 m and the geocomposite has a parameter values of the sand blanket and the ballast. The
thickness of 0.01 m and a lateral permeability more than 700 reference times depend primarily on the permeability divided by
times that of the sand blanket. The laterally permeable the specific yield of the sand blanket. The design curve of Figure
geocomposite prevents the water table from rising far into the 6 relates to a track and foundation geometry recommended by
ballast since it has a higher permeability than the ballast. Network Rail (2005). Sensitivity analyses, summarised in Table
However, once the recession starts, water table elevations are 2, show that modifications to the duration and magnitude of the
only slightly lower than those in Figure 8(a) with the normal recharge alter the reference times by no more than 17% and
geotextile. After 0.25 d the water tables are almost identical. +1%. Alternative values for the dimensions and parameters for
Consequently a laterally permeable geocomposite between the sand blanket and ballast have only a small effect on the
the sand blanket and the ballast has little effect on the long- reference times apart from when fouling leads to reductions in
term drainage. the ballast permeability.
3. An alternative location for the laterally permeable
geocomposite is immediately above the subgrade with a 6.2 Slope of subgrade
conventional geotextile between the sand blanket and ballast. The slope of the subgrade has a major effect on the rate of
The geocomposite permeability is set to 250 times the sand drainage. The reference time for water to drain out of the sand
permeability (if a higher permeability is used, the fall in the blanket for a subgrade slope of 1:40 is more than double that for
water table is so rapid that computational instabilities occur). a subgrade slope of 1:20. For a flat subgrade slope of 1:200,
At the end of the recharge period the water table is within the water drains out very slowly, taking 6.6 times the value for a
ballast. Simulations using the numerical model, Figure 8(c), subgrade slope of 1:20. Consequently, a subgrade slope of 1:20,
indicate that the presence of the laterally permeable or 5%, should be used whenever possible.
geocomposite beneath the sand blanket leads to a dramatic
reduction in the time taken for the water table to fall. One 6.3 Modifications to permeability and specific yield
hour after the end of the recharge period (total time due to dynamic loading
0.0625 d), the water table has fallen to intersect the subgrade Dynamic loading causes changes to the permeability and
towards the top of the slope. By 0.15 d (almost 4 h), most of specific yield of the sand blanket. Provisional values have been
the sand blanket is dewatered. obtained from laboratory experiments carried out following

11
Transport Design for efficient drainage of railway
Volume 167 Issue TR1 track foundations
Rushton and Ghataora

Geotextile at interface
0·3
Sand b
Water table above datum: m

lanket
Ballast
0·2
Subgra
de

0·1

Times 0·0208, 0·0417, 0·25, 0·5, 1·0, 2·0, 3·0: d Slope 5%

0
0 1 2 3
Distance from centreline: m
(a)
0·3
Geocomposite
Water table above datum: m

Sand b
lanket
Ballast
0·2
Subgra
de

0·1

Times 0·0208, 0·0417, 0·25, 0·5, 1·0, 2·0, 3·0: d Slope 5%

0
0 1 2 3
Distance from centreline: m
(b)
0·3 Geotextile at interface
Water table above datum: m

Sand b
lanket
Ballast
0·2
Subgra
de
Geocomposite
0·1

Times 0·0208, 0·0417, 0·0625, 0·100, 0·125, 0·14, 0·15: d Slope 5%

0
0 1 2 3
Distance from centreline: m
(c)

Figure 8. Water table elevations at different times: (a) geotextile

between sand blanket and ballast; (b) laterally permeable
geocomposite 1 cm thick with K ¼ 1.5 3 102 m/s (1300 m/d)
between sand blanket and ballast; and (c) laterally permeable
geocomposite 1 cm thick with K ¼ 0.5 3 102 m/s (430 m/d)
between subgrade and sand blanket and conventional geotextile
between sand blanket and ballast. For all examples, slope of
subgrade 5%; recharge 0.8 m/d for 0.5 h (0.0208 d); sand
blanket, K ¼ 2 3 105 m/s (1.73 m/d), SY ¼ 0.05; ballast
K ¼ 1.16 3 102 m/s (1000 m/d), SY ¼ 0.20

12
Transport Design for efficient drainage of railway
Volume 167 Issue TR1 track foundations
Rushton and Ghataora

100 Recharge q
Reference time after recharge ceases for saturated thickness at

Sand blanket Ballast

Subgrade slope 1:20
0·09 m thick
10
Geocomposite Subgrade
0·01 m thick Geotextile
3·8 m
mid point to fall to 1 cm: d

1 Design line without

geocomposite

Design line for geocomposite

directly above subgrade
0·1

0·01

1 10 100 1000 10 000

K /SY for drainage blanket: 10⫺5 m/s

Figure 9. Design graph for laterally permeable geocomposite

directly above subgrade; solid line indicating the reference time,
when the water table falls to 1 cm above the subgrade (at the
mid point of one of the tracks) as a function of K/SY for sand
blanket. Geocomposite is 1 cm thick, K ¼ 300 3 105 m/s
(259 m/d), SY ¼ 0.05; slope of subgrade 5%, recharge 0.8 m/d
for 0.5 h (0.0208 d); sand blanket, 0.1 m thick, K and SY vary;
ballast K ¼ 10 000 m/d (0.116 m/s), SY ¼ 0.20. Broken line is
from Figure 6 when no geocomposite is present

initial compaction but before applying the dynamic load and heavy rainfall, water levels can approach the bottom of the
also carried out after dynamic loading; the experimental values sleepers. This situation has been investigated using the numerical
are reported in Table 1. The application of a dynamic load leads model. Provided that water levels in the drainage ditches fall
to reductions in the magnitude of the permeability by approxi- within a few hours after the rainfall stops, the time taken for
mately one order of magnitude with the specific yield about water to drain from the sand blanket is only slightly increased.
two-thirds of the original value. For the example of Figure 6,
before applying the dynamic loading the ratio of the permeabil- 6.5 Advantages of a permeable geocomposite above
ity divided by the specific yield was approximately the subgrade
520 3 105 m/s; after applying the dynamic loading the ratio If a geocomposite with high lateral and vertical permeabilities is
approached 75 3 105 m/s. The corresponding reference times located between the sand blanket and the ballast, drainage times
are 0.22 d and 1.4 d; consequently, the time taken for water to are little different from those when no geocomposite is present.
drain from the sand blanket increases with the continuing use of However, locating a geocomposite between the subgrade and the
the track. These are preliminary results; more laboratory tests sand blanket leads to rapid drainage, with the reference times
are being conducted to identify the range of hydraulic parameter about one-twentieth of those without a geocomposite. The use of
values for sand blankets both before and after dynamic loading. such a geocomposite avoids the risk of the subgrade becoming
very wet.
6.4 Surcharge in drainage ditch
Following heavy rainfall it is possible that high water levels will REFERENCES
occur in drainage ditches that collect water both from the track Alobaidi I, Ghataora GS and Gilbert P (2005) The influence of
and from the sides of cuttings and surrounding land. Due to fines migration on trackbed performance: development and

13
Transport Design for efficient drainage of railway
Volume 167 Issue TR1 track foundations
Rushton and Ghataora

practical application. Proceedings of the 8th International Network Rail (1998) Rail Track Line Specification: NR/SP/TRK/
Conference Railway Engineering 2005, London, UK, CD- 033 Track Blanketing Sand. Network Rail, London, UK.
ROM. Network Rail (2005) Specification: RT/CE/C/9039 ‘Formation
BSI (1990) BS 1377-4:1990 Part 4. Methods of test for soils for Treatments’, Issue 1, December 2005. Network Rail, London,
civil engineering purposes. Compaction-related tests. BSI, UK.
London, UK. Roy M and Sayer SK (1989) In-situ permeability testing of sub-
Cedergren HR (1974) Drainage of Highway and Airfield bases on the M5 motorway. Proceedings of the 3rd
Pavements. Wiley, New York, NY, USA. Symposium on Unbound Aggregates in Roads (UNBAR3),
Ghataora GS, Burns B, Burrow M and Evdorides H (2006) University of Nottingham, Nottingham, UK, pp. 86–93.
Development of an index test for assessing anti-pumping Rushton KR and Ghataora G (2009) Understanding and
materials in railway track foundations First International modelling drainage of railway ballast. Proceedings of the
Conference on Railway Foundations, University of Institution of Civil Engineers – Transport 162(4): 227–
Birmingham, Birmingham, UK, pp. 355–366. 236.
Hay WW (1982) Railroad Engineering. Wiley, New York, NY, Rushton KR and Youngs EG (2010) Drainage of recharge to
USA. symmetrically located downstream boundaries with special
Heyns FJ (2000) Railway Track Drainage Design Techniques. reference to seepage faces. Journal of Hydrology 380(1–2):
PhD thesis, Department of Civil and Environmental 94–103.
Engineering, University of Massachusetts Amherst, MA, Selig ET and Waters JM (1994) Track Geotechnology and
USA. Substructure Management. Thomas Telford, London, UK.
Hornych P, Hameury O and Paute JL (1998) Influence de l’eau UIC (International Union of Railways) (1994) UIC Code 719 R.
sur le comportement mécanique des graves non traitées et Earthworks and Track-bed Layers for Railway Lines. UIC,
sols supports de chaussées. Symposium International AIPCR Paris, France.
sur le Drainage des Chaussées, Grenada, Spain, PIARC/ Ward RC and Robinson M (1990) Principles of Hydrology, 3rd
AIPCR, pp. 249–257. edn. McGraw-Hill, Maidenhead, UK.
Hunt GA and Evans E (1993) Optimisation of Track Formation Youngs EG and Rushton KR (2009a) Dupuit–Forchheimer
Stiffness Report 1. British Rail Research Track Mechanics analyses of steady state water table heights due to accretion
and Systems, Derby, UK. in drained lands overlying undulating sloping impermeable
Li D and Selig ET (1998) Method for railroad track foundation beds. Journal of Irrigation and Drainage Engineering, ASCE
design, II: Applications, Journal of Geotechnical and 135(4): 467–473.
Geoenvironmental Engineering 124(4): 323–329. Youngs EG and Rushton KR (2009b) Steady state drainage of
Lichtberger B (2005) Track Compendium: Formation, Permanent two-layered soil regions overlying an undulating sloping bed
Way, Maintenance, Economics. Eurailpress Tetzlaff-Hestra, with examples of the drainage of ballast beneath railway
GmbH & Co., Hamburg. tracks. Journal of Hydrology 377(3–4): 367–376.

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 2000–5000 words long (briefing papers
should be 1000–2000 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.

14