SPECIAL ISSUE PAPER 1

A comparison of railway track

foundation design methods
M P N Burrow1
, D Bowness2, and G S Ghataora3
1
Department of Civil Engineering, The University of Birmingham, Birmingham, UK
2
School of Civil Engineering and the Environment, University of Southampton, Southampton, UK
3
Department of Civil Engineering, The University of Birmingham, Birmingham, UK

The manuscript was received on 25 January 2006 and was accepted after revision for publication on 30 May 2006.

DOI: 10.1243/09544097JRRT58

Abstract: One of the primary functions of the layers that make up the trackbed in a conventional
railway track system is to distribute wheel/rail contact forces so that the stresses in the subgrade
are of an acceptable level. To ensure that the trackbed layers perform this task adequately, there
are a number of design procedures that may be used to calculate an appropriate thickness of the
layers. This paper describes and compares five such design procedures from the USA, UK,
Europe, and Japan. The comparisons show that the design procedures give large differences
in the calculated layer thicknesses which are due to the design methodologies used in each pro-
cedure. Consequently, to enable an appropriate design procedure to be adopted for a given set
of conditions, it is necessary to have a thorough knowledge of the methodologies employed
together with their inherent assumptions.

Keywords: subgrade, ballast, sub-ballast, trackbed layers, granular material, design procedure,
analytical design, material performance, resilient modulus

1 INTRODUCTION primary objective of design is to protect the subgrade

and to this end, the trackbed layers (ballast and sub-
Conventional railway track combines a number of ballast) should be of an appropriate thickness for the
components, including the rail, fastening system, traffic and subgrade conditions at hand. If they are of
sleepers, ballast and sub-ballast, in a structural insufficient thickness, then the large train induced
system (Fig. 1). repetitive stresses in the subgrade are likely to cause
The structure should be designed to withstand the excessive subgrade deformation, significantly increas-
damaging effects of railway traffic and climate so that ing maintenance requirements, whereas unnecess-
the subgrade is adequately protected and that rail- arily thick trackbed layers will be inefficient in terms
way vehicle operating costs, safety, and passenger of construction and maintenance costs [3].
comfort are kept within acceptable limits during The trackbed generally comprises ballast and sub-
the design life [1, 2]. ballast layers. The ballast is crushed granular
The cumulative effect of repeated traffic loads material, of uniform size, placed as the top layer of
deteriorates the track substructure over time. How- the substructure in which the sleepers are embedded
ever, while the ballast lends itself to periodic mainten- (Fig. 1). The most important functions it performs
ance to adjust track line and level, subgrade related are resisting vertical, lateral, and longitudinal forces
problems are less easily rectified. Consequently, a applied to the sleepers to maintain track in its
desired position, provision of resiliency and energy
absorption for the track, provision of drainage,


Corresponding author: Railway Research Centre, Gisbert Kapp reduction of traffic induced stresses in the underlying
Building, School of Engineering, University of Birmingham, layers, and facilitating maintenance operations [4].
Edgbaston, Birmingham B15 2TT, UK. email: m.p.n.burrow@ The sub-ballast is a granular layer between the
bham.ac.uk ballast and the subgrade which also helps to reduce

JRRT58 # IMechE 2007 Proc. IMechE Vol. 221 Part F: J. Rail and Rapid Transit
Downloaded from pif.sagepub.com at University of Hawaii at Manoa Library on June 13, 2015
2 M P N Burrow, D Bowness, and G S Ghataora

Fig. 1 Simplified components of conventional ballasted railway track

the traffic induced stresses to acceptable levels in the shear) [4]. These modes are mostly associated with
subgrade and facilitates drainage. In addition, the fine-grained soils such as clays. Problems due to sub-
sub-ballast layer prevents interpenetration of the grade attrition, progressive shear failure, and an
subgrade and ballast, prevents upward migration of excessive rate of settlement through the accumu-
fine material emanating from the subgrade and lation of plastic strain are associated with the upper
helps prevent subgrade attrition by ballast [4]. most part of the subgrade where cyclic shear stresses
There are a number of design procedures for deter- are likely to be at their highest. Attrition arises when
mining appropriate trackbed layer thickness, includ- the ballast penetrates and wears away the subgrade
ing standards issued by infrastructure operators and surface, whereas progressive shear failure occurs
research published in the literature. As the structural when the cyclic stresses are sufficiently high to
properties of the ballast and sub-ballast layers are cause the soil to be sheared and remoulded (Fig. 2).
similar, such procedures usually recommend a Excessive settlement may cause pockets to form in
single thickness for the trackbed layers and the the subgrade (Fig. 3) as a result of ballast punching
proportion of ballast and sub-ballast is not usually into the underlying soil [4]. Water can collect in the
specified. As the ballast is more expensive than the pockets further weakening the subgrade and possibly
sub-ballast, it is assumed that a minimum thickness resulting in mud pumping (wet spots). Attrition
of ballast, usually between 0.2 and 0.3 m, will be used may be prevented using an appropriately thick sub-
to facilitate maintenance operations which are car- ballast layer, and progressive shear failure occurs at
ried out periodically to readjust the line and level of stress levels below that causing massive shear.
the track. Therefore, foundation design methods should expli-
This paper describes five such procedures and citly prevent progressive shear failure and excessive
compares the calculated thicknesses for a variety of plastic deformation. Several approaches may be
conditions. The methods analysed are from the adopted to help prevent these failure modes
USA (a method proposed by Li et al. [5]), Europe
(The International Union of Railways standard UIC
719R [6]), the UK (a method developed by British
Rail Research [7] and the current Network Rail code
of practice [8]), and Japan (the West Japan Railway
Company standards for high speed and commuter
lines [9, 10]).

1.1 Subgrade failure caused by repeated loading

The primary modes of traffic-induced deterioration
in the subgrade are attrition by the ballast (subgrade
erosion), progressive shear failure (cess heave), an
excessive rate of settlement through the accumu-
lation of plastic strain, and massive shear failure
(i.e. where the weight of the train and track super- Fig. 2 Subgrade progressive shear failure (after
structure may cause the underlying subgrade to reference [4])

Proc. IMechE Vol. 221 Part F: J. Rail and Rapid Transit JRRT58 # IMechE 2007
Downloaded from pif.sagepub.com at University of Hawaii at Manoa Library on June 13, 2015
 Railway track foundation design methods 3

Fig. 3 Excessive subgrade plastic deformation (after reference [3])

including using non-ballasted track forms, introdu- where K is the ratio of dynamic to static wheel
cing an asphalt layer, increasing the flexural rigidity loads, V the train speed (km/h), and D the wheel
(EI) (where E is the Young’s modulus of the rail and diameter (m).
I is the moment of inertia) of the rail and using tech-
niques, such as soil stabilization, to permit higher 2.1.2 Subgrade performance
stresses [11]. Usually, however, the use of appropri-
ate thicknesses of the trackbed layers is likely to be To determine allowable plastic strains and defor-
effective and economical [3]. Procedures that adopt mations under repeated loading, cyclic load triaxial
the latter approach are the subject of this paper tests were conducted on various fine-grained soils
and are described subsequently. [12]. From these, it was found that the subgrade
cumulative plastic strain (1p) could be related to
soil deviator stress (sd) and the number of repeated
2 DESIGN PROCEDURES stress applications (N) as follows

m
2.1 Li et al. [5]’s method sd
1p (%) ¼ a Nb (2)
The method proposed by Li et al. [5] aims to prevent ss
both progressive shear failure and excessive plastic
deformation. This is achieved by limiting the stresses where ss is the compressive strength of the soil, and
in the subgrade such that plastic strain is of an accep- a, b, and m are parameters dependent on the soil
table level. Subgrade stresses are determined using type. Integrating over the depth of the deformable
an analytical model of the track system, wheres the part of the subgrade, the total cumulative defor-
allowable stresses are determined from an equation mation can be determined as
that relates plastic strain to the number of loading ðT
cycles. r¼ 1p dt (3)
0
2.1.1 Track model
where T is the subgrade layer depth in metres.
A three-dimensional, multi-layered elastic model For design, Li et al. [5] suggest that 1p and r should
known as GEOTRACK [4] was built to determine be limited to 2 per cent and 25 mm, respectively.
the subgrade stress distribution under various traffic These values are used for the comparisons described
loadings. For design purposes, the model simplifies subsequently.
the substructure as a single trackbed layer overlying
an homogenous subgrade. To account for the
2.1.3 Design charts
increase in track loading resulting from track and
vehicle irregularities, Li et al. [5] suggest that Equations 2 and 3, together with GEOTRACK, were
dynamic loads should be used. Where this infor- used to produce two sets of design charts [5]. The
mation is unavailable, they prescribe the use of the first set of charts give a minimum thickness of the
following empirical equation, developed by the trackbed layers to prevent progressive shear failure
American Railway Engineering Association (AREA), and are functions of trackbed layer and subgrade
to modify static wheel loads resilient moduli (defined as the repeated deviator
stress divided by the recoverable (resilient) axial
0:0052  V strain), soil type and traffic loading. The second set
K ¼1þ (1)
D of charts, which are additionally a function of

JRRT58 # IMechE 2007 Proc. IMechE Vol. 221 Part F: J. Rail and Rapid Transit
Downloaded from pif.sagepub.com at University of Hawaii at Manoa Library on June 13, 2015
4 M P N Burrow, D Bowness, and G S Ghataora

subgrade depth, give a thickness of the trackbed 2.2.3 Traffic

layers to prevent excessive plastic deformation.
To characterize the traffic, UIC 719 R refers to UIC
714 [13]. UIC 714 classifies a particular line as a
2.2 The International Union of Railways method function of the tonnage hauled, tonnage of tractive
units, line speed, traffic mix (i.e. freight and/or pas-
UIC 719 R [6] is a set of recommendations for the senger), and wear effects of vehicles. According to
design and maintenance of the track substructure. the classification determined using UIC 714, lines
Specifications are given for a single thickness of the that carry faster and heavier traffic are required to
trackbed layers and for the prepared subgrade have thicker trackbed layers.
(Fig. 4).

2.2.1 Basis for design 2.3 British Rail method

UIC 719 R specifies that the substructure may con- British Rail research developed a method which
tain some or all of the following layers: ballast, a sought to protect against subgrade failure by exces-
granular sub-ballast, a geotexile, and prepared sub- sive plastic deformation [7]. To this end, a series of
grade (Fig. 4). design charts were produced to relate the required
The combined thickness of the trackbed layers is thickness of the trackbed layers to a measure of the
determined from the type of soil forming the subgrade, strength of the subgrade known as the threshold
traffic characteristics, track configuration and quality, stress. The charts were developed by combining traf-
and thickness of the prepared subgrade. No infor- fic induced subgrade stresses predicted from a linear
mation is given on how the individual thicknesses of elastic model of the track system, with soil threshold
the ballast and sub-ballast should be determined. stresses determined by laboratory testing.
The prepared subgrade is the layer below the sub-bal-
last that has been treated to improve its engineering 2.3.1 Track model
properties. Its inclusion in the design is optional,
unless the subgrade requires improvement (discussed A single-layer elastic model of the track (i.e. the
subsequently). A geotexile may also be used. trackbed layers and subgrade are treated as hom-
ogenous) was developed to predict the stress distri-
bution in the subgrade for various assumed sleeper
2.2.2 Soil quality
spacings and contact pressures. Measurements of
The type of soil forming the subgrade is classified stresses at a site on the UK’s East Coast Main Line
according to a simple system based primarily on were used to verify the model.
the percentage of fines in the soil. There are four
quality classes of soil: QS0, for soil that is deemed 2.3.2 Subgrade performance
to be unsuitable without improvement; QS1, for
‘poor’ soils which are considered acceptable in In order to determine a suitable material parameter for
their natural condition subject to adequate drainage use in design, a series of cyclic triaxial compression
and maintenance, although improvement should be tests were performed on London Clay. The results of
considered; QS2 for soils of ‘average’ quality and QS3 the tests indicated the existence of a threshold stress,
for soils which are considered to be ‘good’. Poorer above which repeated load applications cause large
quality soils require thicker trackbed layers. permanent deformations that increase exponentially
with the number of loading cycles. Below this
threshold stress, the plastic strain associated with
each load cycle reduces until a stable condition is
reached where the permanent deformations are small.

2.4 Network Rail code of practice

Recommendations for the thickness of the trackbed
layers on the UK network are incorporated in a
recent Network Rail code of practice, NR/SB/TRK/
9039 ‘Formation treatments’ [8]. The code recog-
nizes that the condition of the railway substructure
affects track geometry and maintenance require-
Fig. 4 Calculation of the minimum thickness of the ments. On the basis of this premise, and where
track bed (after UIC [6]) track geometry has been adequate in the past

Proc. IMechE Vol. 221 Part F: J. Rail and Rapid Transit JRRT58 # IMechE 2007
Downloaded from pif.sagepub.com at University of Hawaii at Manoa Library on June 13, 2015
 Railway track foundation design methods 5

without the need for excessive maintenance, the Table 1 Required depth of trackbed layers for the West
code suggests that the subgrade possesses adequate Japan Railway Company
strength and stiffness. Where this has not been the
Annual tonnage Required trackbed
case, the required thickness of the trackbed layers Line (MGTa/year) layer depth (mm)
can be determined from a chart given in the code.
Shinkansen NA 300
Commuter lines 10 4 MGT 250
2.4.1 Trackbed layers thickness chart 10 5 MGT 200
a
The chart relates the required thickness of the MGT stands for million gross tonnes.
trackbed layers to undrained subgrade modulus
(i.e. Young’s modulus) for three different values of
the dynamic sleeper support stiffness (30, 60, and passenger trains operating at average speed of
100 kN/mm/sleeper end). The values of the dynamic 200 km/h. In contrast, the commuter lines use a
sleeper support stiffness relate to minimum require- narrow gauge (1067 mm) and may carry mixed
ments for existing main lines both with and without traffic. For both types of line, the required depth of
geogrid reinforcement and new track, respectively. the trackbed layers is given in Table 1. The substructure
No technical details of how the chart was derived is assumed to have a bearing capacity (sb) of 288 kPa
are given, although the document states that it was and when it is less than this value, ground improve-
‘derived using a combination of empirical data and ment is required. (Note a bearing capacity of 288 kPa
multi-layer elastic theory’. equates to a compressive strength, ss of approximately
112 kPa assuming a cohesion model plastic solution to
a simple strip footing where sb ¼ 2.57 ss.)
2.5 West Japan Railways standards
The West Japan Railway Company (WJRC) have
issued construction and maintenance standards 3 METHODOLOGY
for Shinkansen and commuter lines [9, 10]. The
Shinkansen lines are of standard gauge (i.e. A comparison of the design methods was made by
1435 mm) and are dedicated for high-speed determining the combined thickness of the trackbed

Table 2 Factors accounted for in the design procedures reviewed

Li et al. [5] UIC 719 R British Rail Method NR Code 039 WJRC

Static axle load From GEOTRACK Yes From an elastic model No – but 25.4 tonne No
model used to – charts only go up axle load limit on UK
formulate their to an axle load of 24 network
design charts tonnes
Sleeper type, Via GEOTRACK Yes No difference in No No
length and stresses found for
spacing sleeper spacings of
630– 790 mm
Rail section Via GEOTRACK No No No No
Speed By using a dynamic Yes No – field results Via minimum Crude variation –
axle load (can use showed response requirements for the Shinkansen has
the AREA equation) was quasi-static up dynamic sleeper greater depth
to 100 km/h, but support stiffness. than commuter
could be Also 125 mile/h is lines
incorporated by fastest speed on UK
using a dynamic axle network
load
Annual Yes Yes No No For commuter lines
tonnage only
Cumulative From annual tonnage No No No No
tonnage multiplied by the
design life
Subgrade Charts are provided for Yes (using soil quality Using a threshold Undrained subgrade Bearing capacity of
condition different subgrade determined stress for the modulus or subgrade
types in terms of the primarily from the material in question undrained shear assumed to be
resilient modulus number of fines in strength 288 kPa
and soil strength the soil) otherwise
ground
improvement
must be carried
out

JRRT58 # IMechE 2007 Proc. IMechE Vol. 221 Part F: J. Rail and Rapid Transit
Downloaded from pif.sagepub.com at University of Hawaii at Manoa Library on June 13, 2015
6 M P N Burrow, D Bowness, and G S Ghataora

layers specified by each method under a variety of

conditions relating to: subgrade, axle load, speed,
cumulative tonnage.
A summary of the factors accounted for in these
comparisons is given in Table 2.

3.1 Subgrade
For the study, the subgrade was assumed to be a clay
soil of high plasticity and is typical of problematic
soils in the UK. The condition of the soil was rep-
resented by its resilient modulus which was assumed
could vary from 15 to 100 MPa. As some of the pro-
cedures use measures of soil condition other than
the resilient modulus, it was necessary to have a
means of determining the resilient modulus from
these measures in order to be able to compare the
procedures. For the clay considered herein, a sim-
plistic relationship between the resilient modulus, Fig. 5 Variation in design thickness with subgrade
Es, and the ultimate compressive strength (Li condition
et al. [5]’s method), ss, of the form Es  250  ss
was used [4]. The threshold strength (British Rail
method) was related to the resilient modulus by Li et al. [5]’s method and UIC 719 R account for
assuming that it was equal to the half of the compres- mixed traffic the other procedures do not. Conse-
sive strength [4]. For the European Standard UIC quently, for these procedures, a train with a 250 kN
719 R, guidelines relating the measure of soil quality axle load travelling at 200 km/h and one at 300 km/
used in the standard to any engineering measures of h were used to represent the traffic. A design life of
soil performance, such as the resilient modulus or 60 years was used with a design loading of
strength, are not given. Rather, as described earlier, 900 MGT (i.e. 15 MGT/year for 60 years) as this is
the guidelines relate the soil quality to the percentage similar to that of the CTRL [15]. The results of the
of fines in the soil, and these suggest that a soil with study are shown in Fig. 5. For the case of a passenger
more than 40 per cent of fines should be considered train travelling at 300 km/h only the results using Li
as a class QS1 type soil. Consequently, it was et al. [5]’s procedure are shown as the thickness of
assumed that the quality of the subgrade was class trackbed layers determined using the other four
QS1 and that the subgrade had not been prepared.
The Young’s Modulus, necessary for the Network
Rail code, was assumed to be equal to the resilient
modulus, which is a more common means of expres-
sing the modulus of materials subjected to repeated
applications of stress. For the WJRC standard, the
subgrade was assumed to have been improved to
the minimum required bearing capacity of 288 kPa
(discussed earlier).
Concerning the traffic loading, two different scen-
arios were considered. Both assumed a mixture of 50
per cent freight and 50 per cent passenger traffic, the
characteristics of which were representative of a
Class 66 locomotive travelling at 125 km/h pulling
wagons with axle loads of 250 kN, and a high-speed
locomotive-hauled passenger train with an axle
load of 170 kN [14]. For one scenario, the passenger
train was assumed to travel at 200 km/h, and for
the other at 300 km/h. These were considered to
be representative of conditions on a main line in
the UK and a high-speed line, such as the Channel
Tunnel Rail Link (CTRL), respectively. While Fig. 6 Variation in design thickness with axle load

Proc. IMechE Vol. 221 Part F: J. Rail and Rapid Transit JRRT58 # IMechE 2007
Downloaded from pif.sagepub.com at University of Hawaii at Manoa Library on June 13, 2015
 Railway track foundation design methods 7

methods are the same for a passenger train travelling trackbed layers was determined for speeds between
at 200 km/h when compared with one at 300 km/h. 80 and 350 km/h (Fig. 7).
For the West Japan Railways design, thicknesses
appropriate to Shinkansen trains were used for
3.2 Axle load speeds above .200 km/h and those for commuter
lines, for all other speeds. The subgrade conditions
The axle load study was carried out for a design on
and design life were the same as those for the axle
a clay subgrade with a resilient modulus of 40 MPa
load comparison described earlier.
and the results of the study are shown in Fig. 6.
The relationships described for the subgrade con-
dition study were used to determine the other
required measures of soil strength. To simulate 3.4 Cumulative tonnage
freight traffic, a train speed of 125 km/h was
The cumulative tonnage study used a Class 66 loco-
chosen with axle loads varying from 140 to
motive pulling fully laden wagons travelling at
350 kN, the latter figure is just above the current
125 km/h with axle loads of 250 kN. Subgrade con-
343 kN (35 tonnes) limit in the USA. As the current
ditions were the same as those for the axle load com-
axle load limit is 250 kN (25.4 tonnes) in the UK,
parison and the cumulative tonnage varied from 30
for the British Rail and Network Rail comparisons
to 900 MGT with an assumed annual tonnage of
the load was limited to this value. The design life
15 MGT/year (Fig. 8).
was chosen to be 60 years with a design loading
of 900 MGT.
For the Network Rail procedure, the desired
dynamic sleeper end stiffness was assumed to be 4 DISCUSSION
100 kN/mm/sleeper end, as this corresponds to a
line speed greater than 160 km/h [8]. The West From the comparisons shown in Fig 5 to 8, two
Japan standards for commuter traffic were used in general observations may be made.
the study as the chosen line speed was 125 km/h
1. For each comparison, there is a large variation in
(Table 1).
the specified thickness of the trackbed layers
among the procedures.
2. The design thickness specified by each procedure
3.3 Speed
is a function of at least one of the four variables
The study of design thickness with speed was made considered (subgrade resilient modulus, axle
for a high speed locomotive with a 170 kN axle load load, speed, and cumulative tonnage). However,
(this is similar to Eurostar high speed trains operat- only Li et al. [5]’s method gives a variation in
ing on the CTRL). The required thickness of the required thickness with all of the variables.

Fig. 8 Variation in design thickness with cumulative

Fig. 7 Variation in design thickness with train speed tonnage

JRRT58 # IMechE 2007 Proc. IMechE Vol. 221 Part F: J. Rail and Rapid Transit
Downloaded from pif.sagepub.com at University of Hawaii at Manoa Library on June 13, 2015
8 M P N Burrow, D Bowness, and G S Ghataora

The reasons for these differences, although complex, that the thickness of the trackbed layers specified is
may be explained in part with reference to the a function of the axle load. For the static component,
approach to design adopted. Each of the procedures it is evident from Fig. 6 that this is the case for all of
has varying amounts of empirical and analytical the procedures, expect for the WJRC and Network
elements. In an analytical approach to design, two Rail ones. However, it can be seen from Fig. 6 that
main processes are combined [16]. In the first process, even for the four procedures which consider the
stresses, strains, and deflections induced by traffic static load as a design parameter, the thickness of
loading in the subgrade are determined, using an the trackbed layers recommended are not in close
analytical model of the track system. In the second pro- agreement. The reasons for this concern the analyti-
cess, critical or allowable stresses, strains, and deflec- cal models used and how these simulate the distri-
tions are determined, often from experimentation on bution of the train load through the substructure, as
the subgrade soils. Induced stresses, strains, or deflec- discussed in section 4.2.1.
tions are compared with the allowable to formulate the Concerning the dynamic component, as it is a
design. The approach is summarized in Fig. 9. function of speed, the comparison of thickness of
the trackbed layers with design speed (Fig. 7) may
be used to help determine whether dynamic loading
4.1 The characterization of traffic has been considered.
4.1.1 Axle loads As described earlier, Li et al. [5]’s method consider
dynamic effects through the AREA formula (i.e.
Traffic characterization requires the magnitude, fre- equation (1)), and this is reflected in Fig. 7, which
quency, configuration, and duration of all loads to shows that the thickness of the trackbed layers for
be modelled. In terms of the magnitude of the their method is a continuous function of speed.
wheel loads, both static and dynamic components However, the AREA formula is based on empiricism
should be considered [4]. Research suggests that the and is believed to overestimate the dynamic incre-
dynamic component is a function of speed, vehicle ment (and therefore the required thickness of the
mass, and sources of irregularity in the wheel, run- trackbed layers) at high speeds [15].
ning surface, or vertical track geometry. Dynamic The recommended thickness of the trackbed layers
effects have been shown to increase significantly determined using UIC 719 R, the Network Rail stan-
track loading especially when vehicles which have dard, and the WJRC standards are irregular functions
out of round wheels are operating at high speeds of speed. For UIC 719 R, the increase in traffic loading
[17, 18]. As the stresses transmitted to the subgrade, with speed is taken into account using UIC 714 (dis-
and, therefore, the strains and deflections in the sub- cussed earlier) and for the Network Rail standard,
grade are a function of the load, it may be expected higher speed lines are required to have a greater
sleeper support stiffness and consequently, a greater
thickness of the trackbed layers. As mentioned ear-
lier, the WJRC standard for commuter lines is not a
function of speed; however, a thicker trackbed layer
is specified for the faster Shinkansen lines. Whether
the requirement to increase the thickness with
speed for both the Network Rail and WJRC pro-
cedures is due to setting higher standards of ride
quality for high-speed lines, or whether it is because
the procedures recognize that damage to the sub-
grade increases with speed, therefore, necessitating
thicker trackbed layers is unclear.
In the British Rail method, design thickness is not a
function of speed. However, it was recognized that
dynamic loads should be considered, and it was
reported that work was being undertaken to this
end, although the results of this work were unavail-
able for incorporation in the procedure when it was
published.

4.1.2 Traffic mix

The traffic using a particular line may be a mixture of

Fig. 9 Analytical design (after reference [1]) trains with different axle loads travelling at different

Proc. IMechE Vol. 221 Part F: J. Rail and Rapid Transit JRRT58 # IMechE 2007
Downloaded from pif.sagepub.com at University of Hawaii at Manoa Library on June 13, 2015
 Railway track foundation design methods 9

speeds and for design purposes, it is important to 4.2.2 Layer characterization

consider the effect of this spectrum of loads. How-
ever, only the Li et al. [5] and UIC 719 R methods As weaker subgrades can withstand lower stress
account for the variation in loads which may occur, levels, it would be expected that the thickness of
whereas the design of other procedures is based on the trackbed layers recommended by a procedure
the consideration of a single axle load. is a function of the engineering properties of the
To account for the variation in traffic, Li et al. [5] subgrade (Fig. 5). This is the case for the Li et al.
convert the estimated spectra of wheel loads over [5], British Rail, and Network Rail procedures.
the design life to the number of repetitions of a However, it is not the case for the UIC 719 R and
single design load that causes an equivalent WRJC standards. As mentioned earlier, the UIC
amount of damage [4, 19]. UIC 719 R adopts the 719 R design thickness is a function of the soil
approach developed by the UIC [13] which enables class which is related to the percentage of fines in
daily traffic to be represented in terms of a theoretical the soil. Using this system, two soils of different
traffic load. stiffness or strength are classified as the same if
the percentage of fines in each is similar. For the
example shown in Fig. 5, even though the soil resi-
4.2 Analytical model and layer characterization lient modulus varies from 15 to100 MPa, according
to UIC 719 R the soil is always classified as poor.
4.2.1 Analytical model Clearly, this may lead to the over design of the
In an analytical design procedure, it is assumed that trackbed layers in some cases. As described pre-
the railway substructure can be modelled as a system viously, the WJRC recommendations are based on
of elastic layers characterized by two properties, the the requirement for the subgrade bearing capacity
elastic modulus and Poisson’s ratio. The elastic mod- to be .288 KPa. However, this may result in
ulus is usually taken to be the resilient modulus, and an overly conservative design when the bearing
it may be determined from laboratory tests on the capacity is greater than this value.
materials, either directly or from an analysis of the
response measured in situ [20]. Poisson’s ratio is
4.3 Material performance
usually estimated.
Three of the procedures state that an elastic model In order to formulate the design using an analytical
was used to formulate their designs. Li et al. [5]’s approach, it is necessary to determine a measure of
model consisted of separate trackbed and subgrade material behaviour, which may be compared to the
layers with stress state dependent resilient moduli. subgrade stresses, strains, or deflections predicted
The model used to develop the British Rail procedure by the analytical model. For the procedures
represented the substructure by a single layer with a described here, Li et al. [5] use plastic strain and
single value of resilient modulus, whereas the cumulative deformation (equations (2) and (3)),
Network Rail code used multi-layer elastic theory. UIC 719 R a measure of soil ‘quality’, the British
The effect of these models on the thickness of the Rail method uses threshold stress, the Network Rail
trackbed layers may be seen with reference to code, the Young’s modulus, and the WJRC, a nominal
Fig. 5. This figure shows that although all three subgrade bearing capacity.
procedures give similar functions of design Ideally, these measures of performance should be
thickness with resilient modulus, the thicknesses determined by testing subgrade material under con-
recommended are not the same. The thickness of ditions which match the field situation as closely
the trackbed layers given by British Rail is greater as possible. The methods proposed by Li et al. [5],
than for the other two, partly because their single British Rail, and Network Rail used repeated load
layer model neglects both the effect of the much triaxial tests. Such tests, however, are unable to repli-
higher stiffness of the trackbed layers and also the cate the rotation of the principal stresses which
changes in resilient modulus which occur with load- occur in the soil under traffic loading [20, 21]. As a
ing. Given the limited information available about consequence, research suggests that the perform-
Network Rail’s model it is difficult to determine ance of many materials may be overestimated,
whether the differences in design thickness in resulting in the under design of the trackbed layers
relation to other procedures given by the code are [21]. The effect of the replication of principal stresses
due to the model used or to other factors. on material performance is the subject of on-going
It is not known whether models were used to for- research in this project.
mulate the specifications given by the UIC and The contrasting measures of material performance
WJRC; however, as mentioned previously, it is used by Li et al. [5] and the British Rail procedure
believed that these standards are largely based on are of particular interest. Li et al. [5]’s procedure
empiricism. incorporates a model of material fatigue under

JRRT58 # IMechE 2007 Proc. IMechE Vol. 221 Part F: J. Rail and Rapid Transit
Downloaded from pif.sagepub.com at University of Hawaii at Manoa Library on June 13, 2015
10 M P N Burrow, D Bowness, and G S Ghataora

cyclic loading to determine allowable stresses and 300 km/h subgrade study (Fig. 5) with a compressive
strains. Such models relate subgrade deformation strength of 100 kPa [22]. Using these values and
to the expected number of applications of load and assuming a resilient modulus value of 25 MPa (sec-
are widely used in the design of roads [1]. In contrast, tion 3.1), the thickness of the trackbed layers deter-
the British Rail procedure uses the threshold stress mined from each procedure are given in Table 3.
concept, which suggests that provided the stresses From this table, it may be seen that the trackbed
in the subgrade are always less than the threshold layer thicknesses recommended by the British Rail
stress, the subgrade may in theory undergo an infi- procedure is similar to UIC 719 R, although the
nite number of load cycles before failure. Evidently, former was not produced with high-speed lines in
using the former approach, it is necessary to specify mind. In contrast, Li et al. [5]’s procedure gives a
a design life, whereas for the latter it is not. This is value greater then the UIC 719 R one by 30 per
demonstrated by Fig. 8 which shows that the cent, which may be attributed in part to the use of
design thickness recommended by Li et al. [5]’s the AREA formula (see earlier). However, as
method only is a function of cumulative traffic (and mentioned previously, the overconsolidated clay
therefore design life). The concept of threshold subgrade on the CTRL was replaced with sand (to a
stress is not well understood. However, its use in depth of 0.65 m) which is likely to have somewhat
both railway and highway engineering is currently larger compressive strength and resilient modulus
gaining credence as more research is undertaken to values than the clay. Therefore, it may be argued
better understand the concept [20]. that a truer thickness of the trackbed layers for the
CTRL built on the overconsolidated clays is some-
where between 1 and 1.65 m (i.e. 1 þ 0.65 m),
4.4 Case studies depending on the engineering properties of the
Folkestone sands. The thickness given by the Li
In the UK, lines exist with a thickness of trackbed et al. [5]’s procedure is within these limits. Network
layers from ,0.3 to 1 m or more. For example the Rail’s procedure gives a smaller value than UIC 719 R
CTRL, whose trackbed layer design was based on by 20 per cent and may suggest that the code may
UIC 719 together with French TGV best practice, not be suitable for designing high-speed lines. A very
has a thickness of the trackbed layers (ballast þ sub- low thickness, in comparison with UIC 719 R, is
ballast þ prepared subgrade) on ballasted sections given by the WJRC procedure, but the subgrade
approximately between 0.85 and 1 m [22]. The would require improvement to achieve the required
lower thickness is for sections of the track where the 288 kPa bearing capacity (section 2.5).
subgrade comprised Folkestone sand and the greater A further example is a mixed traffic line near
thickness was used for track on overconsolidated Leominster in Herefordshire that has been investi-
clays. In addition, for the sections constructed on gated for some time prior to and during the current
overconsolidated clays, 0.65 m of the clay below the project [24]. The trackbed layer thickness varies
base of the prepared subgrade was dug out and along the site from 0.9 –1.3 m and has increased
replaced with Folkestone sand sandwiched between from its original thickness over time due to contin-
geotextiles. Since opening in September 2003, there ued ballast replacement. Sections of the site show
have been no reported problems with the substruc- large amounts of deterioration and there is a need
ture, and it has been suggested that the interval for frequent maintenance. The deterioration is
between planned tamping and realignment mainten- believed to be due to poor drainage causing localized
ance of every 3 years may be increased [23]. This softening, fines migration into the ballast, hetero-
would, therefore, indicate that the thickness of the geneous dynamic sleeper support stiffness, and con-
trackbed layers on the CTRL is adequate. sequent non-uniform track settlement (Fig. 10).
The CTRL sections constructed on overconsoli- Lower bound estimates of the subgrade strength
dated clays correspond approximately, in terms of and resilient modulus values found at the site are
traffic loading and subgrade conditions, to the 100 kPa and 25 MPa, respectively [22]. The design
line speed for the section concerned is 128 km/h
(although there are speed restrictions in place), and
Table 3 CTRL trackbed layer thickness the annual tonnage at the site is 6 MGT/year.
Using these values and assuming 50 per cent of the
Depth of trackbed layers (m) traffic is freight, the trackbed layer thicknesses rec-
Li et al. [5] UIC 719 R British Rail Network Rail WJRC ommended by the five procedures are given in
1.31 1.00a 0.97 0.79 0.30b Table 4. It may be seen from this table that the rec-
a ommended thicknesses given by Li et al. [5] and
Including prepared subgrade.
b
Subgrade improvement required as bearing capacity of soil is UIC 719 R are similar. However, the British Rail rec-
likely to be less than the minimum of 288 kPa (section 2.5). ommended thickness is 15 per cent greater than

Proc. IMechE Vol. 221 Part F: J. Rail and Rapid Transit JRRT58 # IMechE 2007
Downloaded from pif.sagepub.com at University of Hawaii at Manoa Library on June 13, 2015
 Railway track foundation design methods 11

Fig. 10 Track stiffness at Leominster (after reference [24])

these thicknesses, and the Network Rail and WJRC being monitored in this project, including the
recommendations are significantly lower. The CTRL. The sites monitored and the techniques used
greater thickness given by the British Rail method are described in a companion paper [25].
may be attributed to their model of the track sub-
structure that neglects the effect of the much
higher stiffness of the trackbed layer (section 4.2.1). 5 CONCLUSION
All of the recommendations are less than, or at the
lower end in the case of the British Rail procedure, This paper presented five different procedures that
the 0.9 to 1.3 m trackbed layer thickness found at may be used to determine the thickness of the
the site. This would suggest that if the thickness trackbed layers for railway foundation design. For
of the trackbed layers is the sole consideration in each procedure, comparisons were made of the
trackbed design, then there should be no need for thickness as a function of subgrade condition,
the excessive maintenance which has occurred. axle load, speed, and cumulative tonnage. The
However, the fact that there has been a large comparisons were made to reflect existing UK con-
amount of maintenance illustrates that other factors ditions and to enable extrapolation to future traffic
should be taken into account in any design process. loading conditions. For all comparisons, the pro-
These include appropriate drainage, the prevention cedures show a large variation in recommended
of subgrade attrition (through the use of a suitable thickness of the trackbed layers and that for four
sub-ballast layer), and the necessity to ensure a uni- out of five of the methods, the thickness is not a
form track stiffness (which as Fig. 10 illustrates is not function of all of the factors examined. Reasons
the case at Leominster). for this were discussed with reference to the
To better understand these factors as well as the approach to design adopted by the methods. This
relationship between the thickness of the trackbed showed that Li et al. [5]’s procedure follows most
layers and track performance, a number of sites are closely an analytical methodology. The research
presented has also shown that there are a number
of areas which require additional research. These
Table 4 Leominster trackbed layer thickness include traffic characterization, methods used to
determine material properties to reflect accurately
Depth of trackbed layers (m)
field conditions, models used to specify material
Li et al. [5] UIC 719 R British Rail Network Rail WJRC behaviour under repeated loading, and the use of
0.86 0.82 0.97 0.49 0.20a the methods for heavier axle loads and faster
a
Subgrade improvement required as bearing capacity of soil is speeds than those which currently occur on the
likely to be less than the minimum of 288 kPa (section 2.5). UK network.

JRRT58 # IMechE 2007 Proc. IMechE Vol. 221 Part F: J. Rail and Rapid Transit
Downloaded from pif.sagepub.com at University of Hawaii at Manoa Library on June 13, 2015
12 M P N Burrow, D Bowness, and G S Ghataora

Accordingly, it is hoped that this paper will help 11 Stirling, A. B., Konstantelias, S., Ghataora, G. S.,
the reader to understand the scientific basis and Brough, M., and Madelin, K. B. Improving existing
inherent assumptions of the five design procedures railway subgrade stiffness. A case study of ground
described and will assist in selecting and modifying, improvement techniques. World Congress on Railway
research, Edinburgh, UK, November 2003, pp. 909 – 921.
where appropriate, any procedure to give an appro-
12 Li, D. and Selig, E. T. Cumulative plastic deformation
priate design for a given set of conditions.
for find grained soils. J. Geotech. Eng. ASCE, 1996,
122(12), 1006–1013.
13 UIC. Classification of lines for the purpose of track
ACKNOWLEDGEMENT maintenance. UIC 714 (A.A.72), International Union of
Railways, Paris, France, 1989.
The financial support of the Engineering and Phys- 14 Fox, P., Pritchard, R., and Hall, P. British Railways.
ical Sciences Research Council is noted with grati- Locomotives and Coaching Stock 2004, 2004 (Platform
tude. The authors are grateful to Mr Alan Stirling 5 Publishing Ltd, Sheffield, U.K).
and Dr Amanda Burrow for their advice and 15 Lord, J. A., O’Riordan, N. J., and Phear, A. G. Design
support. and Analysis of Railway Track Formation Subgrade for
High Speed Railways, Rail Technology for the Future,
ICE; London, 1999.
16 Ullidtz, P. (2002) Analytical tools for pavement design.
REFERENCES Key note address, 8th International Conference on
Asphalt pavements, Copenhagen, Denmark, August
1 McElvaney, J. and Snaith, M. S. Analytical design of 2002.
flexible pavements. Highways, the location, design, 17 Esveld, C. Modern railway track, Second edition, 2001
construction and maintenance of pavements (Ed. C. A. (MRT-Productions, Zaltbommel, The Netherlands).
O’Flaherty), 2002, Chapter 15 (Butterworth – Heine- 18 Shenton, M. J. Ballast deformation and track deterio-
mann, Oxford, U.K). ration. Track Technology, Nottingham, UK, 1984, pp.
2 Burrow, M. P. N., Ghataora, G. S., and Bowness, D. 253 – 265 (Thomas Telford Ltd, London, UK).
Analytical track substructure design. ISSMGE TC3 19 Li, D. and Selig, E. T. Method for railroad track foun-
International Seminar on Geotechnics in pavement dation design. II: applications. J. Geotech. Geoenviron.
and railway design and construction, NTUA – Athens, Eng. ASCE, 1998, 124(4), 323 – 332.
16 – 17 December 2004, pp. 209 – 216. 20 Brown, S. F. Soil mechanics for pavement engineers.
3 Li, D. and Selig, E. T. Method for railroad track foun- In Transportation Geotechnics, proceedings of a
dation design. I: development. J. Geotech. Geoenviron. symposium held at Nottingham Trent University,
Eng. ASCE, 1998a, 124(4), 316 – 322. London, 2003, pp. 1 – 24 (Thomas Telford Ltd).
4 Selig, E. T. and Waters, J. W. Track geotechnology and 21 Gräbe, P. J. and Clayton, C. R. I. Permanent
substructure management, 1994 (Thomas Telford Ltd, deformation of railway foundations under heavy axle
London, UK). loading. International Heavy Haul Association, Special-
5 Li, D., Sussman, T. R., and Selig, E. T. Procedure for ist technical session, Dallas, Texas, May 2003, pp.
railway track granular layer thickness determination. 3.25–3.32.
Report no. R-898, Association of American Railroads, 22 O’Riordan, N. J. and Phear, A. G. Design and construc-
Transportation Technology Center, Pueblo, Colorado, tion control of ballasted track formation and subgrade
USA, 1996. for high speed lines. Railway Engineering 2001,
6 UIC. Earthworks and track-bed layers for railway lines. London, 2001.
UIC Code 719 R, International Union of Railways, 23 Schofield, R. and Franklin, A. Maintaining track geo-
Paris, France, 1994. metry for 300 km/h (186 mph) operation on CTRL sec-
7 Heath, D. L., Shenton, M. J., Sparrow, R. W., and tion 1. Client and contractor in perfect harmony.
Waters, J. M. Design of conventional rail track foun- Permanent Way Institution Winter Track Engineering
dations. Proc. Inst. Civ. Eng., 1972, 51, 251 – 267. Conference, Leeds, UK, January 2005.
8 Network Rail. Company code of practice. Formation 24 Brough, M. J., Ghataora, G. S., Stirling, A. B., Madelin,
treatments NR/SB/TRK/9039, Network Rail, 40 Melton K. B., Rogers, C. D. F., and Chapman, D. N. Investi-
Street, London W1 2EE, December 2005. gation of Railway Track Subgrade. Part 2: case
9 WJRC. Construction and maintenance standards for study. Proc. Inst. Civ. Eng. Transp., 2006, 159, (TR2),
Shinkansen track. West Japan Railway Company, 83 – 92.
Osaka, Japan, 2002. 25 Bowness, D., Lock, A. C., Powrie, W., Priest, J. A., and
10 WJRC. Construction and maintenance standards for Richards, D. J. Monitoring the dynamic displacements
commuter and local railway track. West Japan Railway of railway track. Proc. IMechE, Part F: J. Rail and Rapid
Company, Osaka, Japan, 2002. Transit, 2007, 221(F1), 13– 22.

Proc. IMechE Vol. 221 Part F: J. Rail and Rapid Transit JRRT58 # IMechE 2007
Downloaded from pif.sagepub.com at University of Hawaii at Manoa Library on June 13, 2015