Vehicle System Dynamics

International Journal of Vehicle Mechanics and Mobility

ISSN: 0042-3114 (Print) 1744-5159 (Online) Journal homepage: https://www.tandfonline.com/loi/nvsd20

Modelling of wheels and rail discontinuities in

dynamic wheel–rail contact analysis

Michaël J. M. M. Steenbergen

To cite this article: Michaël J. M. M. Steenbergen (2006) Modelling of wheels and rail
discontinuities in dynamic wheel–rail contact analysis, Vehicle System Dynamics, 44:10, 763-787,
DOI: 10.1080/00423110600648535

To link to this article: https://doi.org/10.1080/00423110600648535

Published online: 17 Feb 2007.

Submit your article to this journal

Article views: 723

View related articles

Citing articles: 3 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=nvsd20
Vehicle System Dynamics
Vol. 44, No. 10, October 2006, 763–787

Modelling of wheels and rail discontinuities in dynamic

wheel–rail contact analysis
MICHAËL J. M. M. STEENBERGEN*
Faculty of Civil Engineering and Geosciences, Department of Road and Railway Engineering,
Delft University of Technology, P.O. Box 5048, NL 2600, GA Delft, The Netherlands

A classification of wheel–rail contact is given. Difference is made between modelling of a running

wheel with continuous single-point-contact, as is common practice in wheel–rail contact analysis,
and a wheel with transient double- or multi-point-contact, which may occur for rail irregularities
with curvatures larger than that of the wheel circumference. It is shown that application of the first
model for these irregularities will strongly underestimate the contact forces as it does not describe
occurring mechanisms correctly. Further, it is shown that in principle it is not possible to describe the
second type of contact fully correct with a lumped wheel model. Both wheel models are formulated
mathematically for some basic contact cases. Afterwards, results are applied to a linear track model.
Analytical closed-form solutions are found in the frequency domain for arbitrary type of contact and
numerically transformed to the time domain. Finally, the necessity is shown to avoid situations where
transient multiple-point-contact may occur (like rail joints) in practice.

Keywords: Rail joint; Dipped joint; Wheel–rail contact; Wheel model; Impact load

Nomenclature

A cross-sectional area of a rail

E steel Young’s modulus
F wheel–rail contact force
i imaginary unit
I moment of inertia of the rail cross-section
kH Hertzian wheel–rail contact stiffness
kf foundation stiffness of an embedded rail
ksusp stiffness of the primary suspension corresponding to one wheel
mw mass of a train wheel (half unsprung mass)
M train wheel centre
Q static wheel-load
R train wheel radius
Rrail railhead radius
s Laplace parameter, defined as s = σ + iω with σ some small positive value
and ω the frequency

*Email: [email protected]

Vehicle System Dynamics

ISSN 0042-3114 print/ISSN 1744-5159 online © 2006 Taylor & Francis
http://www.tandf.co.uk/journals
DOI: 10.1080/00423110600648535
764 M. J. M. M. Steenbergen

t time
u vertical degree of freedom of the lumped wheel-mass
u0 step height in the vertical rail-geometry
u1 total height difference in the vertical rail geometry for a multi-linear vertical
rail- geometry
v train velocity
v0 initial velocity of the train wheel-mass
w vertical degree of freedom of the rail
x longitudinal coordinate of the rail-geometry
z vertical height of the rail-geometry
γ relative angle in the vertical rail-geometry
ρ steel density
υ Poisson’s ratio of steel

1. Introduction

In the analysis of wheel–rail contact, several modelling methods are applied. In general, differ-
ence can be made between mixed discrete-continuous descriptions and finite element models.
The first type of models describes each discrete element (track and vehicle components) by
its equation of motion. The complete set of coupled equations is then solved analytically in
the frequency domain, or with a time-stepping routine in the time domain.
In this type of models as well as in several finite element models, the train wheel is com-
monly modelled as a lumped mass connected to a vertical Hertz spring, mostly non-linear,
representing the wheel–rail contact stiffness [1, 2].
A problem arises when the wheel–rail contact has a spatial discontinuity. Such discontinuity
may occur e.g. for dipped joints, insulated rail joints (IRJs) (figure 1); bolted rail joints, where
both so-called fishplates have a smaller vertical bending stiffness than the rail itself; crossings,
where apart from a gap, the transition from the wing rail to the crossing nose often has a
difference in height; and even in the case of badly welded connections in continuously welded
track (CWR).
In 1974, Jenkins, Lyon and others (British Rail) presumably were the first to study wheel–
rail contact for a dipped joint in their well-known article [1] where they present approximating

Figure 1. Examples of IRJs. The wear pattern shows the occurrence of additional bending through impact at the
joint (left).
 Modelling of wheels and rail discontinuities 765

formulae for determining P 1 and P 2 forces. They use a vertical nonlinear Hertz spring
for modelling the wheel–rail contact. However, they do not consider the possibility of a
discontinuity in the contact at the dip.
In more recent literature, some attention has been paid to this problem or some specific
aspect of it.
Andersson and Dahlberg [3] analyse wheel–rail contact for crossings with a certain gap
between wing rail and nose (without height difference), applying a nonlinear Hertzian spring
in modelling the contact. They use a finite element model for the turnout. For the gap, a fictitious
beam element is used, trough-shaped between the two end-nodes of adjacent elements of the
rail. Thus, discontinuities in vertical wheel displacement and its time-derivatives are avoided.
Increases in contact force are found up to 100% at 70 km/h and 200% at 150 km/h when
compared to the static value. The dynamic amplification occurs due to two reasons: not only
the geometrical discontinuity is playing a role, but also the sudden change in track flexibility
when the wheel moves from the wing rail to the nose rail.
Dukkipati and Dong [4] study the problem of a dipped rail-joint. However, the wheel–rail
contact is modelled as a multi-spring-contact (partly continuous contact), so that a spatial
contact discontinuity cannot occur.
Chen and Kuang [5] study the contact stress variation near IRJs. They use a finite element
model for both wheel and rail. Results show that the Hertzian contact theory is no longer able
to predict the contact stress distributions around the rail joints. In a following study by Chen
[6], elastoplastic finite elements are used for the same analysis. It is found that serious plastic
deformations may occur at the rail ends.
Wu and Thompson [7] study impact noise due to a wheel passing over rail joints. A linear
model for wheel–rail impact at a joint in the time domain is presented and wheel centre
trajectories for dipped joints are used as model input. For a joint with a gap of 7 mm and a
step-up of 1 mm dynamic load amplifications are found of 400% and 600% for speeds of 80
and 160 km/h, respectively.
Wu and Thompson [8] study the effects of track nonlinearity on wheel–rail impact, which
may occur due to wheel and rail discontinuities. Only the case of a wheel flat is considered,
and the wheel–rail contact has no spatial discontinuity.
Schupp et al. [9] simulate the contact between wheel and rail for a railway vehicle running
through a switch with multi-body system simulation. The main focus is on the changing rail
profile and no specific attention is paid to the rail discontinuity.
Koro et al. [10] pay attention to the problem of multi-point-contact between wheel and
rail, occurring for discontinuous rail with bolted joints with certain gap. They develop a
finite element model for wheel–rail contact, applying the Hertzian spring, and use a modified
constitutive relation of the Hertzian contact model, introduced in ref. [11], with changing
parameters in the vicinity of rail edges. This way, they smoothly change the single-point-
contact state to a two-point-contact stage and avoid a discontinuity in the contact. With
the developed model, they investigate the influence of the train speed and the gap size on
the impulsive contact force excited by wheel passage on a rail joint. They find excited
loads that are much higher than those occurring for a wheel running on a continuous
railhead.
The software package ADAMS/Rail has different contact models [12]; however, none of
them accounts for a discontinuous contact with a shift in contact position along the wheel rim.
In all mentioned studies, except for [7], the spatial discontinuity in wheel–rail contact
is avoided where the Hertzian spring model is used. This article is mainly devoted to the
applicability of lumped wheel models for the different types of short-length rail irregulari-
ties, with emphasis on those irregularities where a spatial discontinuity in wheel–rail contact
occurs.
766 M. J. M. M. Steenbergen

Figure 2. Types of wheel-rail contact; continuous single-point-contact (left) and transient double-point-contact
(right).

2. Classification of wheel–rail contact

A perfectly circular wheel running on a perfectly flat rail continuously has a single-point-
contact in time, modelling the contact patch as a point. The contact is a real point when the
elasticity of both elements is not taken into consideration, in a purely geometrical approach.
As soon as any irregularity occurs in the vertical rail profile, this may change into a double-
point-contact, or even a multi-point-contact or partly continuous contact. This second type of
contact can only occur for irregularities with a length-scale smaller than the wheel radius, or
when the positive curvature of the vertical rail geometry exceeds the curvature of the wheel
rim (figure 2).
If this type of contact is transient, as it is for a running vehicle, it will cause a development
of the wheel–rail contact force in time, which will be completely different from that occurring
in the case of a single-point-contact that is continuous in time.
Both types of contact as well as their dynamic modelling will be analysed separately in
the following. The main objective is to show the essential difference in the behaviour of the
wheel–rail contact force for situations where the contact is single-point and continuous in
time, and situations where the contact is multiple-point and transient. The analysis will be
carried out first for a wheel running on a completely rigid rail and foundation. Attention then
can be focused completely on the modelling of the wheel. Later on also the rail and track
modelling will be included (section 6).

3. Continuous single-point contact

In the case that wheel and rail have a continuous single-point-contact, in multi-body-dynamics
the wheel may be modelled as a lumped mass, with all mass concentrated in the gravity centre,
supported by a spring, representing the radial wheel stiffness or the Hertzian wheel–rail local
contact stiffness. This stiffness is nonlinear, but may be linearised in a simplistic approach
for small deviations round a certain load-level. As long as single-point-contact exists, any
irregularity in the vertical rail surface can be modelled as an excitation of the wheel–rail
contact-point, or the contact-point between the Hertzian spring and the rail. For a wheel on
a rigid rail with height z(x) along the rail the corresponding formulation of the wheel on the
Hertzian spring becomes that of a simple oscillator (figure 3, x = vt):

mw ü(t) + kH u(t) = kH z(t) (1)

Application of this wheel-model (with linear or nonlinear Hertzian spring) is common practice
in the analysis of wheel–rail contact, as stated in the introduction.
 Modelling of wheels and rail discontinuities 767

Figure 3. Wheel modelling for continuous single-point wheel–rail contact.

4. Transient double-point-contact

The double-point-contact will be considered first for a most basic case: the train wheel passes
an upward step in the vertical rail geometry. For this case, the basic mechanisms playing a
role are most easy to analyse. Afterwards, a more complex case will be treated. In this section
only kinematical or geometrical derivations are given. This means that inertia is not accounted
for, and both wheel and rail are considered to be infinitely rigid. Some implications of these
assumptions will be discussed in section 5. The expressions derived for kinematical behaviour
may serve as an input for dynamic calculations.

4.1 Step in the vertical rail geometry

The considered case is shown in figure 4, where some notations are also introduced. The wheel
radius is denoted by R and its centre by M; the step height is u0 and the velocity v.
The vertical displacement of the wheel centre is not described by the step function of the
contact surface, due to its non-zero radius. In figure 4 the trajectory of the wheel centre is
shown; it consists of two straight lines, connected by a circle segment. The circle segment has
centre P (the instantaneous centre of rotation when moving from position A to B) and radius
R. The rolling wheel has continuous single-point-contact with the rail, except for position
A, where a two-point-contact occurs, immediately followed by a shift of the contact-point
in circumferential direction of the wheel. This two-point-contact at position A causes a dis-
continuity in the first spatial derivative of the trajectory. In the time domain (transformation
is performed via the linear operator v, the train velocity), this implies a discontinuity in the

Figure 4. Kinematical wheel displacement for a vertical step in the rail surface, u0 ≤ R.
768 M. J. M. M. Steenbergen

velocity (the time-derivative of the displacement) at position A. The result is an instantaneous

change in the upward wheel-mass velocity, or, an impact contact-force, as the duration of
the velocity change is zero. This also implies that the magnitude of the contact-force during
impact is undetermined. The instantaneous velocity change is also shown in figure 4; arriv-
ing in position A the velocity vector is horizontal; when leaving this position its direction is
parallel to the tangent line at the wheel in P, as its orientation is always normal to the wheel
radius towards the contact-point.
As has been done for the continuous single-point-contact, the double-point-contact now will
be formulated mathematically for the wheel-model of figure 3. Assuming the time moment of
reaching position A as t = 0, the initial vertical velocity of the wheel mass reads:


2u0
u̇(0) = v · sin θ = v · (2)
R
The momentaneous double-point-contact is followed by single-point-contact again, which
means that the subsequent vertical excitation of the wheel mass can be modelled as an excitation
at the base of the Hertz spring, as has been discussed before. This spring base displacement is
equal to the change in vertical distance between the wheel centre trajectory and the vertical rail
surface. Here, the displacement function is linearised (taken is the chord of the circle segment,
the dotted line in figure 4), to keep expressions simple. The excitation is shown in figure 5 as
a function of time.
The above now will be written in formula. For u0 ≤ R the following geometric relation
holds (see figure 4):
 
a= u0 (2R − u0 ), which is simplified as 2Ru0 for u0  R. (3)

Assuming again tA = 0, the vertical displacement of the spring base in space reads, for 0 <
x < a with x = 0 coinciding with position A:


u0 u0
z(t) = ·x = ·x (4)
a 2R

Assuming a constant train velocity and using x = v · t, in the time domain, this can be written
as follows:


u0
z(t) = ·v·t (5)
2R
The displacement function of the Hertz spring base (figure 5) may be written, in continuous
form, for the whole time domain t ≥ 0, as:
u0
z(t) = t · H (tB − t) + u0 H (t − tB ) (6)
tB

Figure 5. Linearized trajectory of the wheel centre vs. time for a step in the rail surface.
 Modelling of wheels and rail discontinuities 769

where H represents the Heaviside-function and tB is given by:

√
2Ru0
tB = (7)
v

Substituting equation (7) into (6), the final displacement function reads:

√   √ 
u0 2Ru0 2Ru0
z(t) = · vt · H − t + u0 · H t − (t ≥ 0) (8)
2R v v

4.2 Multi-linear vertical rail geometry

Now the somewhat more complex case of a piece-wise linear vertical rail surface (a ramp) will
be considered (figure 6). It can be observed immediately that a fictitious step with magnitude
u0 is present; the shape of the solid area has no influence. Before discussing this case, a
restriction on the applicability of the model of figure 6 is derived; compare figure 7. In this
figure, the situation is shown in which the positions A and C in figure 6 coincide; the linear
part in the displacement MM (figure 6) disappears and the situation is reduced to a simple
step (with u1 = u0 ), as in the model of figure 7 (again, the shape of the solid area has no
influence). This poses the following restriction on the combination of γ , R and u1 for which
the model of figure 6 is valid: u0 < u1 , or: u1 > R(1 − cos γ ).

Figure 6. Derivation of the wheel-model excitation for a multi-linear rail profile (ramp, dipped joint) for u1 ≤ R.
770 M. J. M. M. Steenbergen

Figure 7. Transition between step and multi-linear model.

Thus, the following validity domains for application of the models of Figures 4 and 6 ensue:
• 0 < u1 < R(1 − cos γ ): step model, with u1 = u0 (9)
• R(1 − cos γ ) < u1 < R: multi-linear model
In figure 6 (top), the trajectory of the wheel centre is shown. It consists of two straight lines
(before position A and after position B), connected by the linear line segment MM and the
circle segment M M . A discontinuity in the first derivative (both space- and time-derivative)
or an instantaneous velocity change of the wheel mass occurs at position A, where a two-
point-contact occurs, immediately followed by a shift of the contact-point in circumferential
direction of the wheel, causing an impact. The impact is followed by continuous single-point-
contact, which can be described by a displacement at the base of the Hertz spring (figure 6,
bottom). This displacement consists of the straight line segment MM , where the base of the
Hertz spring moves from P to P  , followed by the circular segment M M , where the base
of the Hertz spring remains in P  , but the wheel centre moves from M to M . During this
movement, the vertical distance between the wheel centre and centre of rotation P  changes,
which can be considered as a displacement at the base of the Hertz spring. In order to keep
expressions simple, this displacement is linearised again, taking the chord of the circle segment
(figure 6, dotted line).
The above now will be written in formula. The time moment tA is chosen again as tA = 0.
For the initial velocity u̇(0) of the wheel mass can be written as follows:
v0 = v · sin γ (10)
where γ denotes the slope of the ramp. The following general expression holds for the vertical
displacement of the Hertz spring base (compare figure 6, bottom), for the whole time domain
t ≥ 0 (with 0 ≤ tC ≤ tB ):
z(t) = αt · H (tC − t) + (βt + d) · (H (t − tC ) − H (t − tB )) + u1 · H (t − tB ) (11)
The slope coefficients α and β are derived in the following. First, the time durations of passing
from position A to C (tAC ) and from C to B (tCB ) are determined.
MM
tAC = (12)
v · cos γ
 Modelling of wheels and rail discontinuities 771

The rail inclination γ is defined by tan γ = u1 / l. In the above expression for tAC , the denom-
inator may not be approximated by v as 0 < γ < 900 ⇒ 0 < cos γ < 1. In the same way, the
distance MM may not be replaced by its horizontal component.
(u1 − u0 )/ sin γ 2 (u1 − u0 )
tAC = = (13)
v · cos γ sin(2γ ) v
The time duration of passing from C to B is given by:
M M
tCB = (14)
v · cos ψ
In contrast to the expression for tAC , in this expression, the denominator may be approximated
by v, as shown in the following relations:
x
cos ψ =
M M
  
x = R 2 − (R − u0 )2 = 2Ru0 − u20 ≈ 2Ru0 (15)
  
M M = x 2 + u20 = 2Ru0 + u20 ≈ 2Ru0

With u0 = R, the term u20 is third-order and will be neglected, or cos ψ ≈ 1. Thus, for tCB
may be written:
√
2Ru0
tCB = (16)
v
which is the same result as obtained in equation (7) for the step, as expected. The coefficients
α and β now follow from:
u1 − u0 1
α= = sin(2γ ) · v (17)
tAC 2
and

u0 u0
β= = ·v (18)
tCB 2R
The value of d follows from:
u0
d = u1 − u0 − tAC tan β = u1 − u0 − tAC (19)
tCB
After some elaboration follows:


1 2u0
d = (u1 − u0 ) 1 − (20)
sin(2γ ) R

Substituting the values of equations (17) and (18) for the coefficients α and β, for the
displacement function z(t) is found:


1 u0
z(t) = sin(2γ ) · vt · H (tC − t) + · vt + d (H (t − tC ) − H (t − tB ))
2 2R


1 2u0
+ u1 · H (t − tB ); d = (u1 − u0 ) 1 − (21)
sin(2γ ) R

All parameters can be expressed in the wheel radius R, the rail ‘kink’ angle γ and the value
of u1 (where instead of u1 or γ also the length parameter l may be chosen; tan γ = u1 / l).
772 M. J. M. M. Steenbergen

This yields for, respectively, u0 , β and d (compare figure 6):

u0 = R − R cos γ = R(1 − cos γ ) (22)

 
β = (1/2)(1 − cos γ ) · v or β = sin2 (γ /2) · v = sin((1/2)γ ) · v (23)
 √ 
2(1 − cos γ )
d = (u1 − R(1 − cos γ )) 1 −
sin(2γ )
 
sin(γ /2)
= (u1 − R(1 − cos γ )) 1 − 2 (24)
sin(2γ )

ξ

The behaviour of the non-dimensional factor ξ is shown in figure 8, for 0 < γ < 90◦ (0 <
γ < (1/2)π ): − ∞ < ξ < (1/2). For small γ , the value 0.5 can be adopted.
Summarising, the total displacement function now can be written as (with tA = 0, tC = tAC
and tB = tAC + tCB ):
   
1 1
z(t) = sin(2γ ) · vt · H (tC − t) + sin γ · vt + d
2 2
· (H (t − tC ) − H (t − tB )) + u1 · H (t − tB ) (25)

where
 
sin(γ /2)
d = (u1 − R(1 − cos γ )) 1 − 2
sin(2γ )
2 (u1 − R(1 − cos γ ))
tC =
sin(2γ ) v
 
2 (u1 − R(1 − cos γ )) 1 R
tB = + 2 sin γ
sin(2γ ) v 2 v

Introducing the simplification γ  (1/2)π (which is a realistic assumption for practice), the
expressions can be simplified significantly; the complete formulation of the excitation of the

Figure 8. Behaviour of ξ vs. γ for 0 < γ < (1/2)π .

 Modelling of wheels and rail discontinuities 773

Figure 9. Example of a piecewise linear displacement function according to equation (27).

wheel model takes the following form:

u̇(0) = γ · v (26)

and
 
1 1 2
z(t) = γ · vt · H (tC − t) + γ · vt − γ R + u1 (H (t − tC ) − H (t − tB ))
2 2
+ u1 H (t − tB ) (t ≥ 0) (27)

where    
R u1 1 R u1 1
tC = − γ (>0) and tB = + γ (>0)
v γR 2 v γR 2
with model restrictions: 0 < γ  90◦ and (1/2)γ 2 < u1 /R < 1.
In figure 9, an example of the displacement according to equation (27) is shown, for the
arbitrary parameter values: γ = 18◦ , R = 0.4 m, u1 = 0.04 m (which is a theoretical value,
used for clearness of the graph) and v = 80 km/h.

5. Discussion

In both cases treated in the previous section, a linearization of the trajectory was used, for
simplicity and transparency of the analysis. It is obvious that taking the original, partly circular,
displacement will only increase dynamic effects.
The rail profiles as have been considered in the sections 4.1 and 4.2 only have theoretical
meaning, as these profiles will never occur in practice. However, they clearly demonstrate the
impact problem for two-point-contact. Further, the multi-linear case very clearly shows the
principle of what will occur e.g. in case of a dipped joint: impact followed by a continuous
system excitation (figure 10).
In general, the impact will be the dominating factor, and the vertical shift in the contact
point can be found, for given wheel radius, from the considered geometry. This vertical shift
determines the impact magnitude, which in reality is finite, as will be explained in the following.
In the previous section, all derivations were geometrical or kinematical. Both inertia and
elasticity were not accounted for. Also damping was not accounted for, but it does not play a
role of primary importance at the wheel–rail interface.
774 M. J. M. M. Steenbergen

Figure 10. Wheel centre trajectory with discontinuous derivative at a dipped joint.

The effect of elasticity is smoothing. In reality, the velocity change and the resulting impact
will not occur in a zero time-interval, but in a finite time-interval. Also the magnitude will not
be infinite, but reach only a finite value. Further, with high impact loads, plasticity will play
an important role. Its role, next to contact stress redistribution, is physical smoothing due to
plastic deformation (at least for the rail), which will decrease dynamic effects too [6].
Finally, the rail system (the foundation) was considered as completely rigid during impact.
As long as the time duration of impact approaches zero, this is a correct assumption also for
a non-rigid foundation; both wheel and rail inertia prevent them from displacing during this
time interval. However, when the local elasticity of the wheel and the track in the contact
patch is taken into account, the duration of impact will no longer be zero. This means that for
a real track, the above is only an approximation for the derivation of the relative wheel–rail
displacement, which is the best possible way with a discrete wheel model.
The above discussed smoothing effects of elasticity and plasticity are illustrated in figure 11,
where the momentum of the wheel mass in vertical direction p(t) is shown as a function of
time, as well as its time-derivative, the vertical contact-force F (t), corresponding to the step
transition in figure 4.
From the previous sections, it follows that in general it is not necessarily correct to model
the wheel as a lumped mass on a Hertz spring with excitation possibility only at the base
of the spring. In the case that a shift of the contact-point occurs, impact forces may result,
with dynamic effects according to equations (2) and (26) linearly increasing with the train
velocity and the dip angle (or the fictitious dip angle (2)), and with magnitude which may
reach much higher values than would occur in the case of continuous single-point-contact,
due to the impact character. The quasi-linear relation between P1 -forces and both train speed
and dip angle was also found by Jenkins et al. [1] for a non-rigid ballasted track.

Figure 11. The smoothing effect of elasticity and plasticity in the wheel–rail contact patch on the vertical momentum
of the wheel mass as a function of time and its time-derivative, the contact-force, corresponding to the step transition
in figure 4.
 Modelling of wheels and rail discontinuities 775

The above mentioned will be illustrated with the help of the theoretical wheel-model of
figure 3, passing a step u0 in a rail on rigid foundation at a velocity v. Linear behaviour
is assumed, and the response will be determined for two different cases: assuming tran-
sient double-point-contact with corresponding impact on the wheel, and assuming continuous
single-point-contact. This basic problem is used to illustrate the differences in the main fea-
tures of the response for both cases. In figure 12, the response (both the displacement and the
velocity of the wheel-mass, and the contact-force) is shown for both cases. At the left, the
system response is shown for the case of an impact load on the wheel-mass. The equation of
motion of the system is given by:

mw ü(t) + kH u(t) = mw v0 δ(t) (28)

√
where δ designates the Dirac-delta-function; v0 = v · 2u0 /R according to equation (2). For
the amplitude of the solution for the displacement, it is found as follows:

2mw u0
û = v (29)
kH R

Figure 12. Qualitative model comparison for a wheel passing a step, assuming double-point-contact with corre-
sponding impact (left) and continuous single-point-contact with corresponding Hertz-spring excitation (right). Shown
are the displacement of the mass, its velocity, and the contact-force in the Hertzian spring.
776 M. J. M. M. Steenbergen

and the contact-force amplitude equals

2u0 kH mw
F̂ = v . (30)
R
In figure 12 (left), it is observed that the wheel-mass has a zero initial displacement and
a non-zero initial velocity; the wheel–rail-contact force starts with an infinite peak of zero
duration (corresponding to the initial condition), after which it is exactly in phase with the
wheel displacement on the Hertz spring.
At the right, the results are shown for a similar model, but assuming continuous single-point-
contact, with a corresponding step-excitation at the base of the Hertz-spring. The equation of
motion of the system is given by:

mw ü(t) + kH u(t) = kH u0 H (t) (31)

where H designates the Heaviside-function. The solution amplitude for the displacement is
given by:
û = 2u0 (32)
and for the contact-force amplitude is found:

F̂ = 2kH u0 . (33)

Comparison of equations (30) and (33) shows that in equation (33) the influence of wheel
mass and radius, as well as that of the velocity have vanished. This is obviously wrong when
applied to a running train wheel. Results are different, also with respect to the behaviour in
time. No impact occurs in the second case; the displacement of the wheel starts at zero, with
zero slope, though the contact-force starts at its amplitude, as a result of the instantaneous
compression of the Hertz-spring (figure 12, right).

6. Application to a continuously embedded rail–wheel system

The results of the previous sections will be applied now to a vertical wheel–rail interaction
model. A simple, linear model is chosen, so that analytical solutions can be obtained. The train
vehicle is modelled up to the level of the primary suspension. Physical as well as geometrical
linearity is assumed. This strongly reduces the practical applicability of the model, but enables
an approach in the frequency domain. The moving oscillator problem may be simplified to a
standing oscillator problem, where the rail irregularity moves with speed v between rail and
spring base, because of the fact that the wave propagation and reflection at the irregularity in
the system due to the moving oscillator itself is not of interest here and will have negligible
effects on the contact-forces at the discontinuity. The model is shown in figure 13, where also
some notations are introduced.
The following mathematical problem statement is valid:
• Beam (Euler–Bernoulli) on elastic foundation:
∂ 4 w(x, t) ∂ 2 w(x, t)
EI + ρA + kf w(x, t) = −F (t)δ(x) (34)
∂x 4 ∂t 2
• Oscillator:
d 2 u(t)
mw + ksusp u(t) = F (t) (35)
dt 2
 Modelling of wheels and rail discontinuities 777

Figure 13. Wheel–rail interaction model with excitation z(t).

• Contact-force:
F (t) = kH (−u(t) + w(0, t) + z(t)) (36)

• Boundary conditions: no perturbations at −∞ and +∞

• Initial conditions are assumed zero, except for the initial velocity of the wheel mass and the
rail contact-point, which is assumed u̇(0) = v0 and ẇ(0, 0) = −v0 , respectively.

As has been established, the wheel gets an initial upward velocity at the moment of impact.
The rail at this moment gets the same initial velocity, but in downward direction, or ẇ(0, 0) =
−v0 . This can be explained as follows. At the moment of impact, when the wheel rim hits
the rail, the rail cross-section in question, which in theory has a zero mass as its length is
infinitely small, gets a certain initial velocity. Assuming a purely elastic impact, the value
of this velocity should be the same as the value of the initial vertical velocity of the wheel,
as displacements are zero. After impact, or for t > 0, wave propagation from the impact
coordinate occurs, introducing non-zero deflections and velocities (and thus momentum) in
adjacent cross-sections of the beam.
The functions z(t) for different rail irregularities have been derived in the previous sections.

7. Solution in the frequency domain

The problem (34)–(36) will be solved applying integral transform methods. The governing
equation (34) of the beam in the t, x domain is written as:


∂˜ 4 w(x, t) 1 ∂˜ 2 w(x, t) 1 EI kf
4
+ 2 + b2 w(x, t) = − F (t)δ(x) with a = , b=
∂x a ∂t 2 EI ρA EI
(37)

where the tilde indicates a generalised derivative, which is applied due to the discontinuity
at x = 0. Laplace-transforming this equation with respect to time, with initial displacement
assumed zero and arbitrary initial velocity ẇ(x, 0), yields:
 2 
∂˜ 4 w̃(x, s) s − ẇ(x, 0) 1
4
+ 2
+ b w̃(x, s) = −
2
F̃ (s)δ(x) (38)
∂x a EI
778 M. J. M. M. Steenbergen

where the transformed variable is indicated by a tilde. The general solution to the homogeneous
problem is of the form:

w̃ = C1 epx + C2 e−px + C3 eipx + C4 e−ipx (39)

where


4 s − ẇ(x, 0)
1 2
p = √ (1 + i) + b2
2 a2
Normally, the constants C can be determined applying the boundary conditions. However,
as the beam is undisturbed at both −∞ and +∞, it can be concluded immediately that all
constants are equal to zero, and thus that the homogeneous solution vanishes.
For the inhomogeneous problem, the particular solution to the right-hand-side must be
added, designated by w̃part . It will be determined applying a Fourier transform on the governing
equation with respect to x. The Fourier transformed governing equation is given as:
 
¯ s 2 − ẇ(x, 0) 1
w̃(k, s) k 4 + 2
+ b 2
= − F̃ (s) (40)
a EI

where the Fourier-transformed variable has been indicated by a bar. This equation yields for
the particular solution as follows:

1 F̃ (s)
w̃¯ = − (41)
EI k 4 + (s 2 − ẇ(x, 0))/a 2 + b2

The inverse Fourier transform of this solution yields the particular solution in the x, s domain:

∞
F̃ (s) e−ikx
w̃part (x, s) = − dk (42)
2πEI −∞ k 4 + (s 2 − ẇ(x, 0))/a 2 + b2

The integral can be calculated by contour integration, adding a semicircle joining −∞ to ∞

and applying, respectively, Jordan’s lemma and Cauchy’s residue theorem. This derivation is
carried out in Appendix A. The result of the integration reads for all x as follows:

∞
e−ikx π
dk = ((1 − i)e(−1+i)k0 |x| + (1 + i)e(−1−i)k0 |x| ) (43)
−∞ k + (s − ẇ(x, 0))/a + b
4 2 2 2 8 k03

Substitution of equation (43) into (42) yields the particular solution in the s, x domain for the
load, which also represents the total solution:

F̃ (s)
w̃(x, s) = − ((1 − i)e(−1+i)k0 |x| + (1 + i)e(−1−i)k0 |x| ) (44)
16EIk30

where


1 s 2 − ẇ(x, 0)
k0 = √ + b2 , s = σ + iω
4

2 a2
Now attention will be turned to the oscillator, the wheel mass. The equation of motion
(35), Laplace-transformed with respect to time, is given by (again assuming zero initial
 Modelling of wheels and rail discontinuities 779

displacement, but a non-zero initial velocity v0 , and indicating the transformed variable by
a tilde):
−mw u̇(0) + (mw s 2 + ksusp )ũ(s) = F̃ (s) (45)

From this equation it follows directly:

F̃ (s) + mw v0
ũ(s) = (46)
mw s 2 + ksusp

The coupling between equations (44) and (46) is established by means of the contact-force
(36). This expression, Laplace-transformed with respect to time, reads:

F̃ (s) = kH (−ũ(s) + w̃(0, s) + z̃(s)) (47)

Substitution of the expressions (44) for w̃(x, s) and (46) for ũ(s) yields, after some elaboration,
the final expression for the contact-force F̃ (s) in the Laplace-domain:

z̃(s) − (mw v0 )/(mw s 2 + ksusp )

F̃ (s) = (48)
1/kH + 1/(mw s 2 + ksusp ) + 1/(8EI k03 )

where



1 s 2 − ẇ(0, 0) EI kf
k0 = √ + b2 s = σ + iω; a= b=
4
and and .
2 a2 ρA EI

The input for v0 has been determined in section 4, equations (2) and (26). The input for ẇ(0, 0)
has been specified in the problem statement in section 6. The expressions for z(t) also have
been determined in section 4, but need to be transformed to the frequency domain. For the
Laplace-transformed functions z(t) according to equations (8) and (27) can be found:


u0 v  √ 
z̃(s) = · 2 1 − e− 2Ru0 · s/v (49)
2R s

and

1 e−tC s  tC s  1 1  −tC s 
z̃(s) = γ ·v· e − 1 − tC s + γ · v · e (tC s + 1) − e−tB s (tB s + 1)
s s 2 s
  
1 1  
+ u1 − γ 2 R e−tC s − e−tB s + u1 e−tB s (50)
2 4

respectively, with tC and tB as defined in equation (27).

In figure 14, the spectrum of the displacement according to equation (50) is given for the
same parameter values as have been used in figure 8. The symmetry of the real part and
antimetry of the imaginary part are obvious. The function rapidly decays with increasing
frequency (or s), as is necessary for convergence of the integration process in the inverse
transform.
780 M. J. M. M. Steenbergen

Figure 14. Real and imaginary part of the transformed excitation z̃(s) vs. ω (s = σ + iω; σ = 0.1).

8. Parametrical and numerical solutions in the time domain

The expression (48) must be inversely transformed to the time domain. This inverse transform
(which is basically an integration over the frequency) will be carried out numerically as the
analytical transform by means of contour integration implies the complex branch cutting.
The numerical procedure is explained in Appendix B. The result is that the inverse Laplace-
transform of F̃ (s) can be written as:

eσ t ∞  
F (t) = Re F̃ (s) eiωt dω (51)
π 0

This expression can be used for numerical integration of the solution in the Laplace domain.
The general expression, in integral form, for the dynamic contact-force wheel–rail as a function
of time, for an arbitrary excitation function z(t) with Laplace-image z̃(s) reads:

 
∞
eσ t z̃(s) − (mw v0 )/(mw s 2 + ksusp )
F (t) = Re   eiωt dω (52)
π 0 1/kH + 1/(mw s 2 + ksusp ) + 1/ 8EIk03

where

1 s 2 + v0 EI kf
k0 = √ + b2 , a= b= s = σ + iω.
4
, and
2 a2 ρA EI

The value of v0 is zero for continuous single-point-contact; for two special cases of double-
point-contact, the values have been determined in equations (2) and (26).
In the following, some numerical evaluations of expression (52) will be made. First, numer-
ical values will be adopted for the parameters playing a role. In table 1, some material and
geometrical properties for both track and vehicle are specified. Data have been taken from ref.
[2] and mainly from ref. [13].
The considered rail irregularities are defined by γ = 0.01 rad, u1 = 0.0025 m and u0 =
0.001 m, respectively, see figure 15.
The wheel–rail contact stiffness due to elastic deformations of the wheel-body and the rail
surface can be derived from the Hertz contact theory. Here, the simplified expression from
 Modelling of wheels and rail discontinuities 781

Table 1. Physical properties of vehicle and track components.

Young’s modulus of steel (E) 2.1 × 1011 N/m2

Poisson’s ratio of steel (v) 0.3
Steel density (ρ) 7850 kg/m3
Railhead radius (Rrail ) (UIC) 0.3 m
Rail cross-sectional area (A) (UIC54) 6930 × 10−6 m2
Rail moment of inertia (I ) (UIC54) 2346 × 10−8 m4
Rail foundation stiffness (kf ) 48 × 106 N/m2
Train wheel radius (R) 0.46 m
Train wheel mass (mw ) (locomotion) 2000 kg
Stiffness primary suspension (Ksusp ) (locomotion) 2.25 × 106 N/m
Vertical static wheel-load (Q) (locomotion) 112.5 × 103 N
Train velocity (ν) 22.2 m/s (80 km/h)

ref. [2] will be used, which read as follows:


 2 √
 3E Q R · Rrail
kH =  3
 2 (53)
2 1 − υ2

Substituting material and geometrical properties from table 1, the following expression ensues:

kH = 3 3 × 1022 Q (54)

Calculation results for the step (figure 15) are shown in the graphs of figure 16, at the left
zoomed in for short time duration and at the right for longer time duration (steady-state
response).
It is remarked that the first peak in the contact-force (which is the impact force itself) is
not calculated, as it is used as a discrete initial condition; only the response after impact is
shown. The contact-force starts at zero, with certain slope, due to the initial velocity of the
wheel-mass. The oscillation is disturbed by the further system excitation given by equation (8),
which ends at t = 0.0014 s (equation (7)), when the wheel has ‘climbed’ the step. The highest
frequencies travel away in the rail from the contact-point with the highest velocity. After very
short time, the system comes in a sort of ‘overall’ resonance frequency of the total system,
corresponding to the so-called trapped mode, where also the equivalent track mass, foundation
stiffness and primary suspension play a role. The trapped mode of a mixed lumped-continuous
system corresponds to the lowest energy state of free vibration of the system. The frequency
corresponding to the trapped mode (which is estimated as 190 rad/s) should be smaller than
the cut-off frequency of the beam, which can be verified easily. The cut-off frequency of the
rail can be found from the stiffness of the foundation and the distributed mass of the rail per

Figure 15. Considered rail irregularities and their definition.

782 M. J. M. M. Steenbergen

Figure 16. Model response (wheel–rail contact-force) after impact for short (left) and longer (right) time duration;
step excitation.

unit of length and equals 940 rad/s. However, at this lower frequency also the bogie mass
starts to play a role, which was not incorporated in the model. The static wheel-load was
taken as 112.5 kN, where in figure 16 negative contact-forces larger than 500 kN occur. This
means nonlinearities (loss of contact) will play an important role: in reality the wheel will
start bouncing on the rail after impact.
In figure 17, the system response is shown again for the step excitation of figure 15, but
now assuming continuous single-point-contact. Thus, the Hertzian spring base passes the step
and no impact occurs, v0 = 0.
The response shows a behaviour in time, which is completely different from the behaviour
as found with two-point-contact. The contact-force starts with a large positive value, due to the
instantaneous compression of the Hertz spring. Afterwards, the system again ends up in the
trapped mode, which is even clearer than in figure 16. Comparison of the results of figures 16
and 17 with the results from figure 12, for rigid track, shows that the results are basically
similar.
In figure 18, calculation results are given for the kinked rail geometry, which is shown in
figure 15.
Basically, the same happens as for the step. However, a strange effect occurs due to on the
one hand the response to the initial impact, and at the other hand the response to the system
excitation as the wheel ‘climbs’ the ramp: they mutually cancel to a large extent. The system
excitation stops at tC = 0.011 s (equation (27)), which is clearly visible in figure 18. Again,
the system ends in the trapped mode. As the static preload equals 112.5 kN, no loss of contact
will occur after impact.

Figure 17. Wheel–rail contact-force for a step excitation, assuming single-point-contact, for short (left) and longer
(right) time duration.
 Modelling of wheels and rail discontinuities 783

Figure 18. Model response (wheel–rail contact-force) after impact for a kinked rail geometry, for short (left) and
longer (right) time duration.

9. Concluding remarks

The applicability of lumped wheel models has been studied for situations where two- or multi-
point-contact between wheel and rail can occur, like dipped rail joints. It has been shown that,
in principle, occurring mechanisms cannot be described in a fully correct way by application
of these models. The assumption of continuous single-point-contact (which is implicitly made
using a model of a lumped mass on a Hertz spring) automatically excludes the possibility
of impact, as this is caused by a discontinuity in the single-point-contact. The problem can
be avoided by applying an instantaneous vertical velocity change to the wheel mass at the
moment of two-point-contact, with magnitude following from the vertical rail geometry and
the wheel radius. However, in this way, a singularity in the contact-force occurs, which grows
to infinity in a zero-time duration. In reality, this effect is smoothed by different effects, but the
best possible approximation of what occurs in reality is obtained for a lumped wheel model.
In practice, it is important to avoid short-length rail irregularities where two-point-contact
can occur, as the resulting impact forces are detrimental to both track and vehicle. As an
example, one may refer to the sensitivity of concrete sleepers to high frequency and impact
loading, with respect to crack initiation and propagation. In conventional CWR-tracks, IRJs are
the most critical elements with respect to the vertical rail geometry. In figure 19, measurements
are shown of the vertical geometry of two different IRJs, both with the same duration of use,
in unloaded condition.
In both measurements, the effect of wear and plastification of the end of one of both rail
sections is clearly visible. One of the welds, however, shows a relatively large dip angle.
When the joint is loaded by a passing axle, this dip angle will further increase, causing large

Figure 19. Measurements of the vertical geometry along the rail of two IRJs.
784 M. J. M. M. Steenbergen

Figure 20. Bi-linear relation between dip angle (α  90◦ ) of an IRJ and the corresponding impact force.

impact loads. Because of these impact loads, the ballast under the sleepers below the joint
will be compacted, raising the problem of ‘fluttering’ sleepers with voids underneath. Owing
to this phenomenon, the dip angle will further increase, which on its turn will again increase
the impact loads. The result of this process is a very rapid track deterioration. Therefore, it
is important to realise and to maintain joint-geometries without negative vertical tolerances,
thus avoiding impact (figure 20).
High impact loads at the wheel–rail interface are not only caused by geometrical (kine-
matical, spatial) discontinuities, as has been discussed in this article, but also by short-length
irregularities where, although the kinematical contact is single-point in time, the nonlinear
dynamic response causes discontinuity in time (‘hammering’). Therefore, in a more general
sense, short wavelengths or high frequencies may be stated to be most detrimental to the track.
By eliminating the impact load components at the wheel–rail interface, the level of dynamic
loading on the track can be significantly reduced, which on its turn will reduce numerous
well-known and cost-intensive problems like rolling contact fatigue, corrugation and rapid
wear [14].

References

[1] Jenkins, H.H., Stephenson, J., Clayton, G.A., Morland, G.W. and Lyon, D., 1974, The Effect of track and vehicle
parameters on wheel/rail vertical dynamic forces. Railway Engineering Journal, January, 2–16.
[2] Esveld, C., 2001, Modern Railway Track, MRT productions, Zaltbommel, The Netherlands.
[3] Andersson, C. and Dahlberg, T., 1998, Wheel/rail impacts at a railway turnout crossing. Proceedings of the
Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 212, 123–134.
[4] Dukkipati, R.V. and Dong, R., 1999, The dynamic effects of conventional freight car running over a dipped-joint.
Vehicle System Dynamics, 31, 95–111.
[5] Chen, Y.C. and Kuang, J.H., 2002, Contact stress variations near the insulated rail joints. Proceedings of the
Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 216, 265–273.
[6] Chen, Y.C., 2003, The effect of proximity of a rail end in elastic-plastic contact between a wheel and a rail.
Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 217, 189–201.
[7] Wu, T.X. and Thompson, D.J., 2003, On the impact noise generation due to a wheel passing over rail joints.
Journal of Sound and Vibration, 267, 485–496.
[8] Wu, T.X. and Thompson, D.J., 2004, The effects of track non-linearity on wheel-rail impact. Proceedings of the
Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 218, 1–10.
[9] Schupp, G., Weidemann, C. and Mauer, L., 2004, Modelling the contact between wheel and rail within multibody
system simulation. Vehicle System Dynamics, 41, 349–364.
[10] Koro, K., Abe, K., Ishida, M. and Suzuki, T., Timoshenko beam finite element for vehicle-track vibration
analysis and its application to jointed railway track. Proceedings of the Institution of Mechanical Engineers,
Part F: Journal of Rail and Rapid Transit, 218, 159–172.
[11] Kataoka, H., Abe N., Wakatsuki, O. and Oikawa, Y., 1997, A dynamic stress analysis of joint rails using finite
beam element model (in Japanese). Proceedings of the 14th Japan National Symposium on Boundary Element
Methods, 93–98.
[12] ADAMS/Rail, Help Guide, 2002, Available online at: www.mscsoftware.com/products (accessed 23 February
2005).
[13] De Man, A.P., 2002, Dynatrack, a survey of dynamic railway track properties and their quality. Thesis, Delft
University of Technology, Delft, The Netherlands.
 Modelling of wheels and rail discontinuities 785

[14] Popp, K. and Schiehlen, W., 2003, Rolling-contact-fatigue and wear of rails: economic and technical aspects,
In: P. Pointner (Ed.) System Dynamics and Long-Term Behaviour of Railway Vehicles, Track and Subgrade
(Springer Verlag: Berlin, Heidelberg, New York), pp. 51–62.

Appendix A

The poles of the integrand (42) are equal to the solutions of the equation


s 2 − ẇ(x, 0) 1 s 2 − ẇ(x, 0)
k + + b2 = 0 ⇒ = √ (±1 ± i) + b2
4 4
k1,2,3,4 (A1)
a2 2 a2
or

k1 = (1 + i) k0 ; k2 = (−1 − i) k0 ; k3 = (1 − i) k0 ; k4 = (−1 + i) k0 (A2)

where


1 s 2 − ẇ(x, 0)
k0 = √ + b2
4
(A3)
2 a2
The behaviour of the imaginary part of the poles is shown in figure A1, where the definition
of the Laplace parameter s has been used, s = σ + iω with σ > 0.
It can be concluded: Im(k1 ) > 0, Im(k2 ) < 0, Im(k3 ) < 0, Im(k4 ) > 0, from which it is
known, that which poles lay in the upper half-plane (k1 and k4 ) and which ones in the lower
half-plane (k2 and k3 ) of the complex plane. According to Jordan’s lemma, the contour should
be closed in the upper half-plane in the case of x < 0 and in the lower half-plane in the case
of x > 0. Both closed contours and the poles are shown in figure A2.
As the integrand is single-valued, Cauchy’s residue theorem can be applied now. First, it
will be applied for the first contour, x < 0. Inside the contour are two poles. The residue
theorem states:
∞
e−ikx
dk
−∞ k + (s − ẇ(x, 0))/a + b
4 2 2 2

 e−ikx  e−ikx 
= 2π i Residue of 4 at k1,4 = 2π i 
k ,k
k + (s 2 − ẇ(x, 0))/a 2 + b2 k ,k
4 k 3 k1,4
1 4 1 4

(A4)

Figure A1. Behaviour of the imaginary parts of the poles vs. ω (for small σ > 0).
786 M. J. M. M. Steenbergen

Figure A2. Closed integration contours for x < 0 and x > 0, respectively.

Evaluation of this expression, with the poles defined by equation (A3), yields:

∞
e−ikx  e−ikx) 
dk = 2π i 
−∞ k 4 + (s 2 − ẇ(x, 0))/a 2 + b2 k1 ,k4
4k 3 k1,4
 −i(1+i)k0 x 
e e−i(−1+i)k0 x
= 2π i +
4 ((1 + i)k0 )3 4 ((−1 + i)k0 )3
π k0 x  
= 3
e (1 − i) e−ik0 x + (1 + i) eik0 x
8 k0

Since x < 0, this result can be written as:

π  
3
(1 − i) e(−1+i)k0 |x| + (1 + i) e(−1−i)k0 |x| (A5)
8 k0

In an analogous way follows for the 2nd contour (which is in the negative sense; a minus
should be added), x > 0:

∞
e−ikx  e−ikx 
dk = −2π i 
−∞ k 4 + (s 2 − ẇ(x, 0))/a 2 + b2 k2 ,k3
4 k 3 k2,3
 −i(−1−i)k0 x 
e e−i(1−i)k0 x
= −2π i +
4 ((−1 − i)k0 )3 4 ((1 − i)k0 )3
π  
= 3
(1 − i) e(−1+i)k0 x + (1 + i) e(−1−i)k0 x
8 k0

This may be written as (x > 0):

π  
3
(1 − i) e(−1+i)k0 |x| + (1 + i) e(−1−i)k0 |x| (A6)
8 k0
 Modelling of wheels and rail discontinuities 787

which is exactly the same result as found above (expression (A5)) for the first contour. Thus,
the integral can be evaluated for all x as (for x = 0 the contour may be closed anyhow):
∞
e−ikx π  
dk = (1 − i) e(−1+i)k0 |x| + (1 + i) e(−1−i)k0 |x|
−∞ k + (s − ẇ(x, 0))/a + b
4 2 2 2 3
8 k0
(A7)

Appendix B

The inverse Laplace-transform of F̃ (s) is defined as:

σ +i∞
1
F (t) = F̃ (s)est ds (B1)
2πi σ −i∞

Substituting s = σ + iω with σ some small positive value, or ds = idω, and −∞ < ω < ∞,
yields:
eσ t ∞
F (t) = F̃ (σ + iω)eiωt dω
2π −∞
The integrand in this expression can be elaborated as follows:

F̃ · eiωt = (Re(F̃ ) + iIm(F̃ )) · (cos ωt + i sin ωt)

= Re(F̃ ) cos ωt − Im(F̃ ) sin ωt +i (Re(F̃ ) sin ωt + Im(F̃ ) cos ωt)
 
Real part Imaginary part

where the imaginary part of the integrand should add up to 0, as the force F (t) is real. This
means, sin ωt being anti-symmetric and cos ωt symmetric with respect to ω, that Re(F̃ ) is a
symmetric and Im(F̃ ) an anti-symmetric function of ω. Thus, for the integral may be written
as follows:
∞ ∞
eσ t eσ t
F (t) = Re(F̃ (s))eiωt dω = Re(F̃ (s))eiωt dω (B2)
2π −∞ π 0