KSCE Journal of Civil Engineering (2020) 24(10):2939-2948 pISSN 1226-7988, eISSN 1976-3808

DOI 10.1007/s12205-020-5372-5 www.springer.com/12205

Railroad Engineering

A Priori Determination of Track Modulus Based on Elastic Solutions

a a
Tulika Bose and Eyal Levenberg
a
Dept. of Civil Engineering, Technical University of Denmark, Nordvej, Kgs. Lyngby 2800, Denmark

ARTICLE HISTORY ABSTRACT

Received 5 March 2019 The standard approach for modeling railway tracks idealizes the rails as two infinite beams,
Revised 11 February 2020 each supported over a separate continuous spring foundation. The foundation is characterized
Accepted 28 May 2020 by a track modulus that embodies all components and materials underlying each rail as well as
Published Online 17 August 2020 any cross-rail interaction. Track modulus is considered a basic parameter governing the field
performance of tracks. Therefore, a priori determination of track modulus is needed in design
KEYWORDS of traditional railways, as well as in evaluating the performance-potential of non-traditional
track solutions. In this study, a new method was suggested for a priori track modulus
Track modulus determination based on elastic solutions. Specifically sought were closed-form analytical
Railway formulations that could be representative and tractable. In this connection, a 3-D track model
Elasticity was developed, wherein: rail-pads were considered as linear springs, sleepers as finite beams,
Track stiffness
and all underlying soil-like materials as a homogenous half-space. Ultimately, track modulus
Rail track modeling
was determined by linking calculations in the 3-D model and the standard model. This was
done by requiring equal maximal displacement as well as identical load distribution along the
rail under the weight of a single railcar axle. The method was illustrated considering a wide set
of values for the different model parameters. The calculated results are comparable in
magnitudes and exhibit similar sensitivities to the input parameters as reported in field studies
or as derived from elaborate numerical schemes.

1. Introduction moment as basis for equivalency. Determination methods of

subgrade reaction for foundation design were proposed by
Winkler’s (1867) hypothesis of subgrade reaction is widely Terzaghi (1955) after combining full-scale field experience and
practiced for design and analysis of soil-coupled constructions theoretical considerations. Similarly, full-scale experiments and
such as: 1) pile and raft foundations (Hemsley, 2000), 2) concrete theoretical considerations were employed to guide practitioners
pavements (Westergaard, 1948; Ioannides, 2006), 3) buried pipes on determining subgrade reaction for slab-on-grade constructions
(Rajani et al., 1996; Klar et al., 2005), and 4) tunnel linings (Wood, (Vesić and Saxena, 1969; Daloglu and Vallabhan, 2000;
1975; Lee et al., 2001; Mair, 2008). Essentially, this hypothesis is Khazanovich et al., 2001; Setiadji and Fwa, 2009).
a radical mathematical simplification of actual soil behavior; it A mathematically equivalent concept to subgrade reaction -
does not directly represent any basic material property. Because of called ‘track modulus’ is commonly utilized within the field of
this, problem-specific methods or guidelines are needed to suitably railway engineering (Selig and Li, 1994). In words, track modulus
determine subgrade reaction values for subsequent structural expresses the supporting force per unit length of an infinite rail
evaluation. In an early contribution, Biot (1937) offered such per unit vertical rail displacement. The concept serves as a basic
guidelines for the case of an infinite beam resting on an elastic input parameter for: 1) calculating rail bending stresses and
half-space and loaded by a concentrated force. The development deflections under the weight of railcar axles (Hay, 1982; Selig and
was founded on theoretical arguments, and was based on Waters, 1994; AREMA Manual, 2006; Kerr, 2003; Sadeghi and
requiring equivalency of maximal beam moments. Vesić (1961) Barati, 2010), 2) analysis of vibrations caused by vehicle dynamics
extended Biot’s work to include beams of finite length. The (Newton and Clark, 1979), and 3) assessing overall in-service
proposed guidelines were based on theoretical considerations track quality (Ebersöhn et al., 1993; Read et al., 1994; Roghani and
reinforced by experimental evidence. They also utilized maximal Hendry, 2017). Unlike soil-coupled constructions, track modulus

CORRESPONDENCE Tulika Bose [email protected] Dept. of Civil Engineering, Technical University of Denmark, Nordvej, Kgs. Lyngby 2800, Denmark
ⓒ 2020 Korean Society of Civil Engineers
2940 T. Bose and E. Levenberg

embodies the continuous support offered to a rail by discrete achieve this the elastic properties of all model components were
track components, e.g., rail-pads, sleepers, under-sleeper pads, and assumed, and track modulus was determined solely based on a
ballast mats, as well as by soil-like materials, e.g., ballast, sub- criterion of maximal rail displacement, i.e., by equating maximal
ballast, and subgrade. This means that track modulus is governed rail displacement in GEOTRACK and in the infinite beam on a
by a large set of attributes, i.e., material properties, component continuous spring model. One drawback of this approach is the
dimensions, interconnectivity, etc. Perhaps because of this intricacy incompatibility in the distribution of vertical forces along the
the vast majority of existing methods for determining track modulus track in the sleeper locations - resulting in a mismatch of bending
rely on field measurements (Kerr, 2000; Narayanan et al., 2004; stress distribution in the rails. Regardless, the GEOTRACK
Norman et al., 2004; Lu et al., 2008; Zakeri and Abbasi, 2012; program is neither accessible nor handy at this time; it is incompatible
Nafari et al., 2017). with modern computing platforms and non-upgradable - available
The idea of track stiffness, taken to mean the point-load only in compiled form (Mishra et al., 2016).
required to produce a unit deflection of the rail (at the location A priori determination of track modulus was specifically
where the load is applied), is often employed for evaluating the targeted by Cai et al. (1994). Their work commenced by analyzing a
quality of existing tracks (Tzanakakis, 2013). In effect, whenever finite length beam representing a single sleeper, supported on a
assuming a beam on Winkler model for the rail, track stiffness Winkler foundation (representing the entire track structure below
and track modulus are directly linked. However, while track the sleeper), and loaded by two equal forces representing loading
stiffness encapsulates all track components to provide some at the rail seats. The sought track modulus was calculated based
global stiffness, track modulus is only focused on the support on the displacement of the finite beam under one of the forces,
offered to the rail. The advantage of so doing is related to the fact taking into account the added flexibility due to rail-pad and sleeper
that rail flexural properties are engineered and well defined compressibility (jointly represented as an additional discrete spring
whereas the support offered to the rail may vary widely; thus, in series). The subgrade reaction offered to the finite beam was
track modulus is closely linked to the sources governing the track obtained from an approximate formula (Galin, 1943). This formula
quality. expresses the interaction between a finite beam and a linear elastic
A priori determination of track modulus can be attained with (homogenous) half-space based on the assumption that the beam’s
realistic and representative track models that are able to accept width tends to zero. Steinbrenner’s (1936) formula was also
basic material properties as input. Such determination is needed suggested in this study for calculating a single ‘effective’ Young’s
in new-design and rehabilitation-design of railways, e.g., for modulus from a layered half-space (representing ballast, sub-
guiding the selection of components and materials; it is equally ballast, subgrade, and deeper soil layers). The main limitation of
important when evaluating the performance-potential of non- the work comes from the analysis of a single sleeper. Doing so
traditional track solutions for which limited field experience disregards the effect of adjacent sleepers, and therefore results in
exists (if any). Nonetheless, only a handful of studies were overestimation of the track modulus. Other limitations include
specifically dedicated to this task. representing the sleeper as a zero-width finite beam, and utilizing
The GEOTRACK program (Adegoke et al., 1979; Chang et Steinbrenner’s formula, which is essentially applicable for layered
al., 1980), successor of the MULTA code (Prause and Kennedy, systems wherein the modulus increases (rather than decreases)
1977), represents one of the first efforts to develop an elaborate with depth (Poulos and Davis, 1974).
and realistic track model. In the program, the rail and sleepers Consequently, it is the aim of the current study to contribute a
were considered as Euler-Bernoulli beams of finite length, rail- new method for a priori determination of track modulus based on
pads were represented by discrete linear springs, and all soil-like elastic solutions. Three development attributes are sought as
layers (ballast, sub-ballast, and subgrade) were represented by a means of enhancing any possible acceptance and long-term
stratified elastic half-space. The total model size was limited to usefulness of the work: 1) the formulation should be founded on
eleven sleepers supported on the surface of the layered half- closed-form analytic solutions, offering a step-by-step reproducible
space, jointly carrying two rails. As means of simulating the approach; 2) the method should accept as input all pertinent
weight action of a railcar axle, vertical point-loads were applied material properties, component dimensions, and connectivity to
to the rails, just above the central sleeper. Stress distributions at offer user flexibility in evaluating a wide range of scenarios, and
the sleeper bottoms were approximated according to Barden’s 3) to better strengthen the connection with the track modulus
(1962) formulation. In turn, the corresponding stress distributions at concept, the method should also consider the correct interaction
the surface of the half-space were approximated as circular loads across sleepers, and not rely solely on equivalency of maximal
(Burmister, 1945a, 1945b, 1945c). Ultimately, by means of a rail displacement. The paper commences with an overview of the
stiffness matrix formulation, the displacements at the interaction method followed by a detailed presentation of two elastic
points between all model elements were matched. The GEOTRACK solutions. It then explains how these solutions are to be linked for
program was not explicitly designed for performing a priori track modulus determination. Next, the method is illustrated with
determination of track modulus. Nonetheless, it was applied for a reference set of representative values, along with a limited
this purpose (Stewart, 1985), with reported reasonable agreement parametric investigation. The paper ends with a short summary
between calculated and field-measured track modulus values. To and recommendations for further development.
 KSCE Journal of Civil Engineering 2941

2. Method Overview cement concrete, stabilized ballast, etc. The exclusion of materials
from in-between and along the sides of the finite beams is
The standard approach for modeling railway tracks idealizes the justified given their marginal effect (if any) on vertical track
rails as two weightless separate infinite beams (IBs), each responses. Fig. 1(b) shows a part of this model consisting of
individually supported on a continuous spring foundation; this seven FBs, with the model sides and bottom truncated for visual
model is graphically shown in Fig. 1(a). The figure also includes clarity. The two dark shaded patches seen at the top of each FB
a Cartesian coordinate system where the X-axis is oriented along represent rail-pad locations; each modeled as a linear discrete
the IBs, the Y-axis is transverse to the IBs, and the Z-axis points spring (DS).
vertically downward. As can be seen, the two IBs are simultaneously Figure 1(b) shows that the DSs in the 3-D model are directly
loaded by point forces with intensity PZ representing (jointly) a loaded by forces with intensities S nZ ’s. These forces are spring
single railcar axle. The foundation is characterized by a track reactions calculated in the standard model with an assumed kZ,
modulus with spring constant kZ that embodies all components after converting the ‘continuous’ springs into individual ones
and materials underlying each rail as well as any cross-rail (also with spacing s). For this purpose, the two models are aligned
interaction. such that the position of PZ in Fig. 1(a) corresponds to the center
To facilitate a priori determination of kZ, a 3-D elastic quasi- of the 3-D model in Fig. 1(b). The relative size of the arrows
static track model is developed, comprising all components and schematically indicate the difference in force intensities applied
materials below the rail that govern the track modulus. It consists on top of individual DSs. Considering that all 3-D model
of an equidistant array of sleepers (spacing denoted by s) modeled parameters are known, the displacement at the top of the zeroth
as Euler-Bernoulli finite beams (FBs) with uniform cross- DS (on either side) is calculated quasi-statically due to the
section. Each FB is identified by an index n, going from positive simultaneous application of all S nZ ’s. This displacement is composed
to negative, with the zeroth FB located at the center of the 3-D of the shortening of the zeroth DS and the surface deflection of the
model. The FBs are supported by an underlying elastic half- HS directly below it. In this calculation, it is assumed that the FBs
space (HS) that is linear, homogenous and isotropic. The HS are incompressible, and that they remain in frictionless contact with
represents all soil-like materials, which for a traditional track the HS. Therefore, as part of the calculation, it is necessary to resolve
includes: ballast (excluding crib and shoulder ballast), sub- the contact stresses at the interface of the FBs and underlying HS.
ballast, and deeper soil layers; for a non-traditional track the soil- The frictionless contact assumption means that FBs are fully bonded
like materials may also include: asphalt concrete, Portland to the HS on which they are supported – allowing for full transfer of
stresses in the vertical direction, without allowing shear stresses to
develop at the bonding interface. This neglect of interface shear
stresses is justified given their second order effect, and given the
focus on track modulus.
Ultimately, a representative track modulus is determined by
manipulating the assumed kZ until a match is attained between
the calculated displacement at the top of the zeroth DS and the
maximal IB displacement from the standard track model. Thus,
an intimate and close connection is established between the 3-D
model and the standard model accounting not only for maximum
displacement equivalence but also for equivalence in the longitudinal
load spreadability, i.e., identical moment distribution along the
rail. More details on the method and underlying calculations are
provided in what follows.

3. Elastic Solutions

Two closed-form independent elastic solutions are presented in

this section. The first analyzes a FB subjected to two line-loads
of equal intensity (i.e., a straight track) and a non-uniform support
reaction. The line-loads represent rail-seat forces distributed over
the length of rail-pads, and the non-uniform support reaction
represents an interaction with a HS. The second elastic solution
deals with the surface deflection of a HS that is loaded over a
rectangular stress patch. Then after, both solutions are combined
Fig. 1. A Sketch Depicting the Models Considered for Track Modulus to analyze the interaction at the FB-HS interface with the aim of
Determination: (a) Standard Model, (b) Elastic 3-D Model resolving the contact stress distribution between a sleeper and its
2942 T. Bose and E. Levenberg

wherein aT = 2(a1 + a2 + a3 + a4 + a5 + 1) .
Equations (2) and (3) are inserted into Eq. (1), and a solution
for uZ FB ( y ) is obtained after applying the boundary conditions
,

of zero shear force and zero bending moment at beam ends as

well as zero slope at the beam center. The result is
wZ ⎛D L D ⎞
uZ FB ( y ) = ⎜ +
1
⎟ − yD + D ,
2 2
(4)
EFB I X ⎝ 24 LaT ⎠
, 3 4

, FB

where
4
⎛ ( y – L + L 1 + L2 ⁄ 2 ) H ( y – L + L 1 + L2 ⁄ 2 ) – ⎞
Fig. 2. A Finite Beam Subjected to Two Line-Loads and a Non-Uniform ⎜ ⎟
4
Support Reaction ⎜ ( y – L + L 1 – L 2 ⁄ 2 ) H ( y – L + L1 – L 2 ⁄ 2 ) ⎟
D1 = ⎜ ⎟, (5)
⎜ + ( y – L1 + L2 ⁄ 2 ) 4 H ( y – L1 + L 2 ⁄ 2 ) – ⎟
supporting medium. Ultimately, these solutions are applied to ⎜ ⎟
⎝ 4
( y – L1 – L 2 ⁄ 2 ) H ( y – L 1 – L 2 ⁄ 2 ) ⎠
determine the surface deflection of the HS under the action of
several S nZ ’s acting simultaneously (refer to Fig. 1(b)).
⎛ 4 4 ⎞
⎜ ∑ ( ak – ak + 1 ) ( ( y – kL ⁄ 12) H ( y – kL ⁄ 12 ) – ⎟
3.1 Solution1: Analysis of a FB ⎜k = 1 ⎟
⎜ 4 ⎟
A FB of length L with free ends is shown in Fig. 2; a 3-D D2 = ⎜ ( y – L + kL ⁄ 12 ) H ( y – L + kL ⁄ 12 )) ⎟, (6)
Cartesian coordinate system is also included in the Figure with ⎜ ⎟
⎜ + ( a5 – 1 ) ( ( y – 5L ⁄ 12)4 H ( y – 5L ⁄ 12 ) – ⎟
similar orientation as in Fig. 1. Here, the origin is located at the ⎜ ⎟
⎝ 4
( y – 7L ⁄ 12 ) H ( y – 7L ⁄ 12 )) – a1 y
4
⎠
leftmost end of the FB, with the Z-axis pointing vertically
downward, the Y-axis pointing along the length of the beam, and
⎛ ⎞
the X-axis oriented perpendicular to the YZ plane. The beam’s ⎜ ( L − 2L + L ) H ( L − 2L + L ) + H (2L − L − L) ⎟
⎜ ( 3
) ⎟
elastic modulus is EFB, and moment of inertia about the X-axis is
1 2 1 2 1 2

9wZ ⎜ −( L − 2L − L ) ( H ( L − 2L − L ) + H (2L + L − L) ) ⎟ ,
D = 3

IX,FB. The beam is symmetrically loaded by two line-loads of 3

432EFB I X FB ⎜
1 2
⎟ 1 2 1 2

⎜ LL ⎟
,
2

equal magnitude wZ (units of force/length), acting over a length ⎜− (91a + 61a + 37a + 19a + 7a + 1)
2
⎟
9a
1 2 3 4 5

L2 at a distance of L1 from the beam ends. It is also loaded at the ⎝ T ⎠

(7)
bottom by a segmented array of opposing line-loads representing
a support reaction. There are 12 such segments of equal length and D4 is a free constant accounting for rigid body displacement.
(L/12), numbered from left to right; each with uniform magnitude.
The overall shape of the support reaction is variable, governed 3.2 Solution 2: HS Loaded over a Rectangular Stress
by the values of five independent unitless coefficients denoted as Patch
ai’s. These unitless coefficients represent a reaction intensity A linear elastic half-space, characterized by Young’s modulus
relative to unity, i.e., if all five ai’s equal unity then the support EHS and Poisson’s ratio νHS, is loaded at the surface by a uniform
reaction becomes uniform across the entire length of the beam. vertical stress with intensity qZ acting over a rectangular area of
In the interest of clarity, conditions of symmetry have not been dimensions b × a . This situation is shown in Fig. 3, which also
invoked in the FB analysis. includes a 3-D Cartesian coordinate system placed at the center
The governing differential equation of the FB is of the patch; the Z-axis is pointing downward, and the X and Y
axes are oriented parallel to the patch sides. The surface deflection
d uZ FB ( y)
4
pZ ( y ) − rZ ( y )
,

4
= , (1) of any point along the Y-axis due to the stress patch is denoted by
dy EFB I X FB p
,
u Z HS ( y ) . By integration of Boussinesq’s point-load solution, the
,

wherein uZ FB ( y ) denotes the displacement of the FB along Z-

,
related expression is
axis direction, pZ(y) represents the loading that acts from the top,
1 −ν HS
2

rZ(y) is the segmented bottom support reaction, and y identifies uZp HS ( y ) = qZ f ( y ) , (8)
2π EHS
,

the calculation location along the FB, i.e., 0 ≤ y ≤ L . Utilizing the

Heaviside function H ( ⋅ ) , and enforcing equilibrium considerations, wherein
these are expressed as follows: f − + (a − 2 y )
( f + + b)( f − + b)
f ( y ) = ( a + 2 y ) ln + 2b ln + , (9)
⎛ H (2 y − 2 L1 + L2 ) − H (2 y − 2 L1 − L2 ) + ⎞ + −
( f − b)( f − b) f − (a + 2 y )
pZ ( y ) = wZ ⎜ ⎟, (2)
⎝ H (2 y − 2 L + 2 L1 + L2 ) − H (2 y − 2 L + 2 L1 − L )
2 ⎠
and
⎛ 4 ⎞
rZ ( y ) =
∑
24wZ L2 ⎜ (ak − ak +1 ) ( H ( y − kL / 12) − H ( y − L + kL / 12) ) ⎟
f±= (a ± 2 y )
2
+ b2 . (10)
LaT ⎜⎜ ⎟ ,(3)
k =1
⎟
⎝ + (a5 − 1) ( H ( y − 5L / 12) − H ( y − 7 L / 12) ) − a1 ⎠
 KSCE Journal of Civil Engineering 2943

Fig. 3. An Elastic Half-Space Loaded over a Rectangular Stress Patch: (a) Top View, (b) Cross-Sectional View

3.3 Interaction Analysis at the FB-HS Interface 4. Track Modulus Determination

The two separate elastic solutions are hereafter combined to
resolve the contact stress distribution below a single FB having All above-described formulations are ultimately interlinked
width b and supported by a HS. This means determination of the hereafter to comprise a method for a priori track modulus
five ai’s as well as the rigid body displacement D4. To achieve determination. The calculations require all parameters to be
this, the displacement of the FB and the surface deflection of the known and fixed, with only assumption being a trial value for
HS are matched in the least-squares sense. Such matching kZ. The first step is to solve Eq. (12) and obtain a*i ’s that
represents a bonded yet frictionless contact area of dimension represent the interaction between a single FB supported on a
L × b between the two elastic bodies. HS. Next, the load applied on top of the nth DS, i.e., S nZ (refer
Based on the coordinate system defined in Fig. 2, and for a to Fig. 1) is calculated from the standard model:
given FB loaded by an arbitrary wZ, Eq. (4) is utilized to calculate β Z ns
sPZ β Z e
-

uZ FB ( y ) at y = yj = jL ⁄ J (j = 0, ..., J), i.e., at J + 1 equidistant

, S Zn =
2
( cos β Z ns + sin β Z ns ) , (13)
points within L. Concurrently, by means of superposition, Eqs.
(8) − (10) are used to compute uZ HS ( y ) , i.e., the HS surface
,
where βZ = kZ / (4EIB IY IB ) in which EIB and IY,IB are the Young’s
4
,

deflection at the same yj’s: modulus and moment of inertia about the Y axis of the IB,
respectively. This expression is obtained by multiplying skZ with
(1 − ν HS ) 1 ⎛ mL L ⎞ ⎛
2 12
⎛ mL L ⎞ ⎞
u Z HS ( y ) =
, ∑ rZ ⎜ − ⎟ f y − ⎜⎝ 12 − 24 ⎟⎠ ⎟⎠ . (11)
2π EHS m = b ⎝ 12 24 ⎠ ⎜⎝
n
uZ IB , i.e., the displacement of the IB at location corresponding to the
,

1
nth FB (Bose et al., 2018). Here n varies from −N to +N with 2N + 1
←⎯⎯⎯⎯
Expression I ⎯
→ ←⎯⎯⎯⎯⎯⎯
Expression II →
denoting the total number of FBs considered in the analysis.
Then, the line-load applied on top of the nth FB is
This equation represents the contribution of the 12 adjacent
patches indexed m = 1, ..., 12 with m = 1 identifying the patch wZn = SZn / L2 , (14)
closest to the coordinate origin. With reference to Eq. (8), Expression
where L2 signifies length of rail-pads (refer to Fig. 2). Also, the
I represents qZ and Expression II represents f(y). Expression I
maximum IB displacement is given by
gives the stress intensity at center of the mth patch calculated
using Eq. (3); the division by b implies that the stress distribution uZ0 , IB = SZ0 / skZ . (15)
is taken as uniform across the width of the FB. Expression II is
Further, the shortening of the zeroth DS is
computed according to Eq. (9), with a = L/12 and b.
The five sought ai’s as well as D4 are determined by minimizing ΔuZ0 , DS = SZ0 / KZ , DS , (16)
the sum of the square of the differences between the HS and FB
where KZ DS is the stiffness of DSs.
displacements across all considered yj’s. The formulation is ,

Next, Eq. (3) is employed with a*i ’s and wnZ replacing ai’s and
J wZ (respectively) to express r nZ ( y ) , i.e., the support reaction
( ai , D ) = argmin ∑ (uZ HS ( y j ) − uZ FB ( y j ) )
* * 2
( i = 1, …, 5 ) , (12)
4
a Di
, 4 j =0
, ,
distribution below the nth FB:

⎛ 4 * ⎞
where the optimal/final argument values are denoted with an
rZn ( y ) =
∑
24wZn L2 ⎜ (ak − ak +1 ) ( H ( y − kL / 12) − H ( y − L + kL / 12) ) ⎟
*

asterisk as superscript, i.e., a*i ’s and D*4 . Note that due to ⎜ k =1 ⎟,

LaT* ⎜ ⎟
⎝ +(a5 − 1) ( H ( y − 5L / 12) − H ( y − 7 L / 12) ) − a1
* *
assumption of linearity the ai’s are essentially unaffected by ⎠
choice of wZ. (17)
2944 T. Bose and E. Levenberg

Fig. 4. Top View of an Elastic Half-Space Loaded by Stress Patches and Point Forces Representing the Support-Interaction of Several Finite Beams

wherein aT* = 2(a1* + a2* + a3* + a4* + a5* + 1) . Note that in doing so intensity at the center of the mth patch for the nth FB, and Rnm is
any cross-interaction between the adjacent FBs is neglected the distance between the calculation point and the center of the
(with respect to the foundation support stress distribution), i.e., mth patch of the nth FB. Thus, the sought HS deflection is given
one interaction analysis between a FB and HS is utilized for all by the summation:
FBs.
uZT , HS = uZC, HS + uZ0 , HS (L1 ) . (21)
Subsequently, the deflection of the HS for a point located
directly underneath the zeroth DS is sought - denoted by uTZ HS . ,
Ultimately, by matching T
uZ, HS
plus the shortening of the zeroth
0
For clarifying the calculation rational, Fig. 4 offers a plan view of DS ΔuZ DS against the maximal displacement in the standard
,

the HS loaded by several FBs. The point of deflection calculation model u0Z IB the representative track modulus k*Z (associated with all
,

is depicted in the figure at y = L1 for n = 0 (as a triangular 3-D model attributes) is determined:
marker). Obtaining uTZ HS requires calculating u0Z HS ( L1 ) , i.e.,
, ,
uZT , HS + ΔuZ0 ,DS = uZ0 , IB for kZ = kZ* . (22)
evaluating Eq. (11) with rZ = r0Z ( y ) at y = L1 . It additionally
requires computing the deflection contributed by all adjacent Figure 5 presents a flowchart that summarizes the entire
FBs, i.e., considering the effect from all rnZ ( y ) ’s excluding n = 0. method of track modulus determination. As can be seen the
To quantify this latter contribution, it is sufficient to consider process commences by fixing all model parameters and an
Boussinesq’s point-load solution when superposing the effects assumed/trial value for the track modulus. Then after, calculations
from all stress patches: and exchange of information take place in the standard track
model and in the 3-D track model. The specific equations needed
2(1 −ν HS ) N
2 12
Pn
uZC HS =
,
π EHS n=
∑∑ m
n ,
m= Rm
(18) for each step are indicated in the flowchart. The final step (after
1 1
which the process ends) requires solving Eq. (22), i.e., finding
*
where kZ . This is done by improving the initial assumption for kZ and
iterating until satisfactory convergence. Thus, a track modulus is
L n ⎛ mL L ⎞
Pmn = ⋅ rZ ⎜ − ⎟, (19) determined, associated with the entire parameter set, and satisfying
12 ⎝ 12 24 ⎠
both maximal rail displacement and moment distribution along
⎛
2 the rail.
⎛ mL L ⎞ ⎞
Rmn = (ns)2 + ⎜ L1 − ⎜ − ⎟⎟ . (20)
⎝ ⎝ 12 24 ⎠ ⎠
5. Method Illustration
Equation (18) represents the deflection contribution of 12
adjacent patches indexed m = 1, ..., 12 from all considered FBs The purpose here is to illustrate, by means of application, the
excluding n = 0. Symmetry about the model center permits proposed track modulus determination method. Some intermediate
considering twice the contribution of the FBs indexed n = 1, ..., N results will also be shown to further facilitate clarity of the
(hence the multiplication by 2 in Eq. (18). Both P nm and Rnm are underlying calculations. A reference set of input parameters was
graphically presented in Fig. 4. As can be seen, P nm is the force chosen for this purpose – shown in Table 1. As can be seen, the
 KSCE Journal of Civil Engineering 2945

Fig. 5. Flowchart Describing the Proposed Track Modulus Determination Method

Table 1. Input Parameters for Method Illustration

Component Parameter/Role Symbol Units Reference value Range Comments
4 −5
Infinite beam (IB) Moment of inertia about Y axis IY,IB m 3.04 × 10 - UIC 60 rail section
Young’s modulus E IB GPa 208 -
Discrete spring (DS) Stiffness K Z,DS MN/m 100 25 to 250 Rail-pad stiffness
Finite beam (FB) Moment of inertia about X axis IX,FB m4 1.22 × 10−4 - -
Young’s modulus E FB GPa 30 - Concrete sleeper
Length L m 2.50 - Sleeper length
Base width b m 0.25 - Sleeper base width
Total number on one side N - 18 - See Eq. (18)
Spacing s m 0.60 0.50 or 0.60 See Fig. 1
Position of wZ L1 m 0.50 - See Fig. 2
Length over which w operates
Z L2 m 0.17 - Rail-pad length
No. of matching FB-HS points J - 50 - Eq. (12)
Half-space (HS) Young’s modulus E HS MPa 200 20 to 300 All soil-like materials
Poisson’s ratio νHS - 0.30 -

IB (rail) properties are taken as per UIC60, the stiffness of the FB plus 18 FBs on each side (see Fig. 1). This latter choice was
DS (rail-pads) is 100 MN/m, the FBs (sleepers) are of concrete, taken to guarantee that the load distribution between the sleepers
each 2.5 m long with a rectangular cross-section (base width was fully captured across the entire parameter range. With
0.25 m), spaced apart by 0.6 m; the elastic modulus of the respect to the flowchart in Fig. 5, the FB-HS interaction was
supporting HS is 200 MPa. solved first (Eq. (12)), giving: a*1 = 1.94 , a*2 = 1.69, a*3 = 1.87, a*4
All calculations were done considering PZ = 80 kN (a choice = 1.56, and a*5 = 1.21. Subsequently, the shortening of the zeroth DS
that does not influence the final result) and 37 FBs, i.e., a central (Eq. (16)) was: Δu0Z DS = 0.308 mm, corresponding to S 0Z =
,
2946 T. Bose and E. Levenberg

30.8 KN. The HS deflection directly underneath the zeroth DS (Eq. depicts track modulus vs. half-space modulus for five different
(21)) was: uTZ HS = 0.440 mm of which about 35% was from uCZ HS
, ,
rail-pad stiffnesses and two different sleeper spacings. Fig. 6(a)
(Eq. (18)). Finally, the sought track modulus was obtained by and Fig. 6(b) refer to s = 0.6 m and s = 0.5 m, respectively. In
matching the maximal IB displacement u0Z IB (Eq. (15)) with
,
both figures each line represents the variation of k*Z with change
T 0 *
uZ HS + ΔuZ DS = 0.748 mm. The result was: kZ = 68.7 MPa. It is
, ,
in EHS for a given value of KZ,DS. From these figures it can be
*
noted that the ai ’s above indicate a maximum contact stress to be observed that for any KZ,DS, k*Z increases when the EHS increases.
occurring near the edges of the FB; this is in contrast to the accepted The increase is not indefinite as each curve is bounded by a
viewpoint that the peak stresses occur below the rails (e.g., Zakeri horizontal asymptote defined by k*Z = KZ DS ⁄ s . This can be
,

and Sadeghi, 2007). This is the consequence of the assumption of a explained by the fact that at the limit, as EHS approaches infinity,
FB in contact with an elastic continuum, which mandates high the term uTZ HS becomes zero and the track flexibility is solely
,

stresses at the edges to fulfill deformation compatibility. governed by the rail-pads ( Δu0Z DS remain finite). By comparing
,

Additionally, a parametric investigation was conducted to Fig. 6(a) against Fig. 6(b) it can be observed that track modulus
study the behavior of k*Z . For this purpose the following parameter increases with a decrease in sleeper spacing.
values were varied: FB spacings s, half-space modulus EHS and As a final note, the parametric study was repeated with all
*
discrete spring stiffness KZ DS . The chosen range for these
,
ai ’s presumed unity (see Fig. 2). This is a simplification wherein
parameters is also shown in Table 1. All other parameters (within the stress distribution below the FBs is assumed uniform. Thus,
a realistic range) were noticed to have minor influence on track the calculation step in Eq. (12), which solves the FB-HS interaction,
modulus and therefore remained fixed at their reference value. is evaded. After doing so it was found that k*Z was only slightly
The results of this investigation are presented in Fig. 6 which overestimated; the error compared with the curves in Fig. 6 was
mostly perceptible for higher values of EHS and KZ,DS, and
reached a maximum of only 8% (for EHS = 300 MPa and KZ,DS =
250 MN/m). This finding also means that for the purpose of
track modulus evaluation, it is not necessary to conduct detailed
analysis (e.g., further refined segmentation) of the interaction
between FBs and underlying HS.

6. Conclusions

This paper offered a new method for a priori determination of

track modulus based on elastic solutions. Within an analytical
framework, a 3-D model was developed (Fig. 1(b)), capable of
representing a wide range of track-types, in terms of components
and materials that underlay the rails. Closed form equations were
presented for calculating displacements of a finite beam subjected to
two line loads and non-uniform support reaction, and calculating
surface deflection of a half-space loaded over a rectangular stress
patch. These equations were subsequently utilized to determine
track modulus by linking the 3-D model and the standard track
model of infinite beam on springs (Fig. 1(a)). The linkage was
based on simultaneous equivalency of both maximal displacement
as well as force distribution along the rail. A flowchart was
introduced to summarize the main calculation steps (Fig. 5).
Overall, the method is deemed flexible and robust, while at the
same time computationally ‘lightweight’ and straightforward to
reproduce.
Track modulus determination was demonstrated via application
of the method to a realistic set of input values. To better clarify
the approach, intermediate calculation steps as well as final
results were shown. The demonstration included a parametric
investigation, in which the sensitivity of track modulus to the
elastic modulus of the half-space, stiffness of rail-pads, and
sleeper spacing was studied (Fig. 6). Results are deemed realistic
Fig. 6. Track Modulus Determined as a Function of Half-Space Modulus
for Different Rail-Pad Stiffnesses and Sleeper Spacings Based and valid overall, given that they are similar in magnitude and
on Table 1 Parameters: (a) s = 0.6 m, (b) s = 0.5 m exhibit similar sensitivity to varying input parameters when
 KSCE Journal of Civil Engineering 2947

∫M
2
( y )dy
0
I X , FB = L
, (23)
M 2 ( y)
∫I
0 X , SL
( y)
dy

where M(y) is the bending moment at a distance y for a FB as

shown in Fig. 2, considering an arbitrary wZ and all ai’s taken as
unity; and IX,SL(y) denotes the actual moment of inertia, which
varies along the sleeper length.
In future studies, the method proposed herein can be extended
to determine track modulus in the longitudinal and lateral directions.
Moreover, by means of the elastic-viscoelastic correspondence
principle (e.g., Schapery, 1965) the modeling approach can be
enhanced to include time-dependence and thermal sensitivity of
Fig. 7. A Priori Determined Track Modulus Values Compared with components and materials. This will facilitate, for example, a
Track Modulus Values from Other Sources for Different Model priori determination of track modulus for asphalt overlayment
Parameters: Numbers in the Figure Represent the Lower Bound, solutions across different train speeds and environmental conditions.
the Nominal Value, and the Upper Bound of the Parameters as
per Selig and Li (1994)
Acknowledgments

compared to GEOTRACK (Selig and Li, 1994). Specifically, it The support from Innovation Fund Denmark is gratefully
was possible to closely reproduce the track modulus results acknowledged. This study is part of ‘Roads2Rails: Innovative
shown in Fig. 4 in Selig and Li (1994) for the entire range of and cost-effective asphalt based railway construction system’
subgrade moduli, fastener stiffnesses, and tie spacings - when (Grand Solutions 5156-00006B).
assuming ballast and subballast moduli were same as subgrade to
accommodate a homogenous half-space assumption. Also, the ORCID
results are within the range of track modulus values observed in
field studies (Narayanan et al., 2004; Norman et al., 2004). A Tulika Bose https://orcid.org/0000-0003-1483-8800
graphical depiction of these comparisons is provided in Fig. 7. Eyal Levenberg https://orcid.org/0000-0003-1188-8458
The development included simplifying assumptions, some
of which can be partially relieved to further enhance the applicability References
of the method. The treatment of all soil-like materials and track
components as linear is not realistic, e.g., loading magnitude is Adegoke CW, Chang CS, Selig ET (1979) Study of analytical models
known to affect track modulus determined in the field. Nonetheless, for track support systems. Transportation Research Record 733:2-20
the linear treatment can be utilized to approximate nonlinear AREMA Manual (2006) AREMA manual. American Railway Engineering
and Maintenance-of-Way Association, Lanham, MD, USA
behavior by choosing elastic constants such that they reflect the
Barden L (1962) Distribution of contact pressure under foundations.
anticipated load level. The treatment of the sleepers as beams Géotechnique 12(3):181-198, DOI: 10.1680/geot.1962.12.3.181
with zero compressibility and constant geometry can also be Biot MA (1937) Bending of infinite beam on elastic foundation. Journal
made more realistic. Whenever the vertical compression of the of Applied Mechanics 4(1):1-7
sleepers (FBs) is not negligible, e.g., in the case of wooden Bose T, Levenberg E, Zania V (2018) Analyzing track responses to train
sleepers, an additional linear spring can be assumed connected braking. Proceedings of the Institution of Mechanical Engineers,
(in series) to the discrete springs representing the rail-pads. Thus, Part F: Journal of Rail and Rapid Transit 232(7):1984-1993, DOI:
the entire calculation procedure remains unchanged except that 10.1177/0954409718761242
Burmister DM (1945a) The general theory of stresses and displacements
the new shortening/flexibility is added to the left-hand-side of
in layered systems I. Journal of Applied Physics 16(2):89-94, DOI:
Eq. (22). Finally, the issue of sleepers with variable cross-section 10.1063/1.1707558
can be addressed by converting the actual dimensions to equivalent Burmister DM (1945b) The general theory of stresses and displacements
constant dimensions. In this connection, the FB length (L) can be in layered soil systems II. Journal of Applied Physics 16(3):126-127,
taken as the actual length of the sleeper, and the FB width (b) DOI: 10.1063/1.1707562
can be obtained from dividing the actual sleeper base-area by Burmister DM (1945c) The general theory of stresses and displacements
L. Regarding the inertia moment, the following formula - based in layered soil systems III. Journal of Applied Physics 16(5):296-
on equivalency of bending strain energy - is suggested for 302, DOI: 10.1063/1.1707590
Cai Z, Raymond GP, Bathurst RJ (1994) Estimate of static track
calculating IX,FB:
modulus using elastic foundation models. Transportation Research
2948 T. Bose and E. Levenberg

Record 1470:65-72 Technics Press, London, UK

Chang CS, Adegoke CW, Selig ET (1980) GEOTRACK model for Poulos HG, Davis EH (1974) Elastic solutions for soil and rock mechanics.
railroad track performance. Journal of the Geotechnical Engineering John Wiley & Sons, Hoboken, NJ, USA
Division 106(11):1201-1218 Prause RH, Kennedy JC (1977) Parametric study of track response.
Daloglu AT, Vallabhan CVG (2000) Values of k for slab on winkler Report DOT-TSC-FRA-77-75, US Department of Transportation,
foundation. Journal of Geotechnical and Geoenvironmental Engineering Washington DC, USA
126(5):463-471, DOI: 10.1061/(ASCE)1090-0241(2000)126:5(463) Rajani B, Zhan C, Kuraoka S (1996) Pipe-soil interaction analysis of
Ebersöhn W, Trevizo MC, Selig ET (1993) Effect of low track modulus on jointed water mains. Canadian Geotechnical Journal 33(3):393-
track performance. In: Proceedings of the Fifth International Heavy 404, DOI: 10.1139/t96-061
Haul Conference, International Heavy Haul Association, Beijing, China Read D, Chrismer S, Ebersöhn W, Selig ET (1994) Track modulus
Galin LA (1943) On the winkler-zimmermann hypothesis for beams. measurements at the pueblo soft subgrade site. Transportation
Journal of Applied Mathematics and Mechanics 7(4):293-300 Research Record 1470:55-64
Hay WW (1982) Railroad engineering. John Wiley & Sons, Inc., Hoboken, Roghani A, Hendry MT (2017) Quantifying the impact of subgrade
NJ, USA stiffness on track quality and the development of geometry defects.
Hemsley JA (2000) Design applications of raft foundations. Thomas Journal of Transportation Engineering Part A: Systems 143(7):1-10,
Telford Ltd., London, UK DOI: 10.1061/JTEPBS.0000043
Ioannides AM (2006) Concrete pavement analysis: The first eighty Sadeghi J, Barati P (2010) Improvements of conventional methods in
years. International Journal of Pavement Engineering 7(4):233-249, railway track analysis and design. Canadian Journal of Civil
DOI: 10.1080/10298430600798481 Engineering 37(5):675-683, DOI: 10.1139/L10-010
Kerr AD (2000) On the determination of the rail support modulus k. Schapery RA (1965) A method of viscoelastic stress analysis using
International Journal of Solids and Structures 37:4335-4351, DOI: elastic solutions. Journal of the Franklin Institute 279(4):268-289
10.1016/S0020-7683(99)00151-1 Selig ET, Li D (1994) Track modulus: Its meaning and factors influencing
Kerr AD (2003) Fundamentals of railway track engineering. Simmons- it. Transportation Research Record 1470:47-54
Boardman Books Inc., Omaha, NE, USA Selig ET, Waters JM (1994) Track geotechnology and substructure
Khazanovich L, Tayabji SD, Darter MI (2001) Backcalculation of layer management. Thomas Telford Ltd., London, UK
parameters for LTPP test sections, volume I: Slab on elastic solid and Setiadji BH, Fwa TF (2009) Examining k-E relationship of pavement
slab on dense-liquid foundation analysis of rigid pavements. Technical subgrade based on load-deflection consideration. Journal of
Report No. FHWA-RD-00-086, FHWA, McLean, VA, USA Transportation Engineering 135(3):140-148, DOI: 10.1061/(ASCE)
Klar A, Vorster TEB, Soga K, Mair RJ (2005) Soil-pipe interaction due 0733-947X(2009)135:3(140)
to tunnelling: Comparison between winkler and elastic continuum Steinbrenner WA (1936) A rational method for determination of the
solutions. Géotechnique 55(6):461-466, DOI: 10.1680/geot.2005.55. vertical normal stresses under foundations. Proceedings of the first
6.461 international conference on soil mechanics and foundation engineering,
Lee KM, Hou XY, Ge XW, Tang Y (2001) An analytical solution for a June 22-26, Cambridge, MA, USA
jointed shield-driven tunnel lining. International Journal of Numerical Stewart HE (1985) Measurement and prediction of vertical track
and Analytical Methods in Geomechanics 25:365-390, DOI: 10.1002/ modulus. Transportation Research Record 1022:65-71
nag.134 Terzaghi K (1955) Evaluation of coefficients of subgrade reaction.
Lu S, Arnold R, Farritor S, Fateh M, Carr G (2008) On the relationship Geotechnique 5(4):41-50
between load and deflection in railroad track structure. Proceedings of Tzanakakis K (2013) The effect of track stiffness on track performance.
the AREMA 2008 annual conference, September 21-24, Salt Lake City, In: The railway track and its long term behaviour, STTT 2, Springer-
UT, USA Verlag, Berlin, Germany, 79-87
Mair RJ (2008) Tunnelling and geotechnics: New horizons. Geotechnique Vesić AB (1961) Beams on elastic subgrade and the winkler’s hypothesis.
58(9):695-736, DOI: 10.1680/geot.2008.58.9.695 Proceedings of the 5th international conference on soil mechanics
Mishra D, Sharma S, Shrestha A, Li D, Basye C (2016) GEOTRACK- and foundation engineering, July 17-22, Paris, France
2015: An upgraded software tool for railroad track analysis. Joint Vesić AS, Saxena SK (1969) Analysis of structural behavior of road test
rail conference american society of mechanical engineers, March rigid pavements. Highway Research Record 291:156-158
23-26, San Jose, CA, USA Westergaard HM (1948) New formulas for stresses in concrete pavements
Nafari SF, Gül M, Cheng JR (2017) Quantifying live bending moments of airfields. Transactions of the American Society of Civil Engineers
in rail using train-mounted vertical track deflection measurements 113:425-439
and track modulus estimations. Journal of Civil Structural Health Winkler E (1867) Die Lehre von der Elastizitätund Festigkeit, mit
Monitoring 7(5):637-643, DOI: 10.1007/s13349-017-0248-1 Besonderer Rücksicht auf ihre Anwendung in der Technik. H.
Narayanan RM, Jakub JW, Li D, Elias SE (2004) Railroad track modulus Dominicus, Prague, Czech Republic
estimation using ground penetrating radar measurements. NDT & E Wood AMM (1975) The circular tunnel in elastic ground. Géotechnique
International 37(2):141-151, DOI: 10.1016/j.ndteint.2003.05.003 25(1):115-127, DOI: 10.1680/geot.1975.25.1.115
Newton SG, Clark RA (1979) An investigation into the dynamic effects Zakeri JA, Abbasi R (2012) Field investigation on variation of rail
on the track of wheelflats on railway vehicles. Journal of Mechanical support modulus in ballasted railway tracks. Latin American Journal
Engineering Science 21(4):287-297, DOI: 10.1243/JMES_JOUR_ of Solids and Structures 9(6):643-656, DOI: 10.1590/S1679-
1979_021_046_02 78252012000600002
Norman C, Farritor S, Arnold R, Elias SEG, Fateh M (2004) Design of a Zakeri JA, Sadeghi J (2007) Field Investigation on load distribution and
system to measure track modulus from a moving railcar. In: Proceedings deflections of railway track sleepers. Journal of Mechanical Science
of international conference of railway engineering 2004. Engineering and Technology 21:1948-1956, DOI: 10.1007/BF03177452