Professional Documents
Culture Documents
Rail welds
M. J. M. M. Steenbergen, Delft University of Technology,
The Netherlands; R.W. van Bezooijen, Id2 Consultancy,
The Netherlands
Key words: rail welds, weld geometry, dynamic wheel–rail interaction, weld
assessment, track deterioration.
12.1 Introduction
Continuously welded rail (CWR) was introduced on the railways in the
1930s, and nowadays is the standard in modern railway tracks. In many
cases, the traditional bolted or fish-plated rail connections are present in the
track only in the form of insulated rail joints for detection and signalling
purposes. The deflection of a bolted rail joint under static train axle loading
leads to a high dynamic impact component, which is a source of rapid
track deterioration and high noise levels (Kabo et al., 2006; Steenbergen,
2006; Cai et al., 2007). Therefore, the welded continuous connection was
a significant improvement.
In the present chapter, the main different rail welding technologies will be
discussed briefly in Section 12.2. After that, damage mechanisms that may
be initiated at or by rail welds are addressed in Section 12.3. An important
issue is the dynamic wheel–rail interaction at rail welds; this is the subject
of Section 12.4. In Section 12.5 the Dutch rail welding regulations will be
explained as a method for force-based weld assessment, as an alternative
to the conventional method purely based on geometry. Section 12.6 will
conclude with energy considerations in relation to welding irregularities and
track deterioration.
377
378 Wheel–rail interface handbook
(a)
(b)
12.1 Flash butt (a) and aluminothermic (b) rail welding on the Dutch
HSL-South.
Rail welds 379
different rail steel grades and alloys, such that austenite formed during rail
heating is re-transformed into pearlite and the formation of brittle martensite
is avoided. The manual thermite welding process in particular requires
craftsmanship, in order to control and optimize the steel phases and their
specific properties with pre-heating, controlled cooling down and post-weld
treatment under all circumstances and conditions.
Generally, the fatigue resistance of flash-butt welds is superior to that
of aluminothermic welds; estimations for heavy-haul traffic (Union Pacific
Railroad) are 365 MGT for thermite welds against 877 MGT for flash-
butt welds (Wimmer et al., 2002). In the case of severe welding geometry
irregularities, the fatigue life of welds may be significantly affected by the
effect of the dynamic wheel load and the occurrence of loose sleepers (Ishida
et al., 1999; Steenbergen, 2008b).
In order to obtain a good geometry, the final grinding of the rail surface
is in most cases done manually for both types of welding. In some cases,
however, the welds are treated using a grinding train, on newly built tracks,
yielding a high surface quality (Fig. 12.2).
(a)
(b)
12.2 Manual rail weld grinding (a) and grinding train (Speno) (b) on
the Dutch HSL-South.
and influence, for example, wheel seat fretting fatigue; furthermore, they
are responsible for non-uniform ballast settlement or local decompaction.
It is important to note that, especially on newly built tracks, the short-wave
contribution is almost entirely determined by the rail weld geometry.
Normally, new welds are being tested ultrasonically for internal damage,
cracks, voids or inclusions before acceptance. In general this is combined
with magnetic particle or eddy current inspection for the lop layer (5–7
mm). When a not-detected serious internal material defect, for example a
lack of fusion or porosity, is present in a weld, it often breaks at an early
stage (Shitara et al., 2003; Terashita and Tatsumi, 2003), especially in cold
winters with high tensile forces in the rail. Therefore, internal material or
fusion defects are in general not responsible for long-term deterioration at
the wheel–rail interface.
The effect of locally varying metallurgical properties along the rail surface
however is particularly relevant for rail welds. As has been mentioned, often
short indentations along the rail surface occur at the centre of rail welds, at
the transitions from the weld material to the parent material, the so-called
heat-affected zone (HAZ) (Fig. 12.3). The phenomenon is due to shrinkage
after grinding and further cooling down, especially when the steel temperature
was not low enough at the moment of final grinding. It can be observed
particularly for aluminothermic welds in rail repair works, executed under
time pressure and non-optimal conditions.
These geometrical indentations typically coincide with the local hardness
minima along the rail surface in the HAZ at both sides of the weld metal
(Mutton, 2000). This is shown qualitatively in the diagram in Fig. 12.4.
Vickers hardness levels (at 5 mm below the surface) typically fluctuate
between 250 and 400 HV. For comparison, the hardness of the parent rail
generally starts around 260 HV for new common rail grades (increasing to
300 HV after grinding and initial cold work-hardening due to train operation)
up to 400 HV for new head-hardened rails.
The unevenness in the rail surface, resulting in a local fluctuation of contact
stresses in the rolling contact, may induce differences in shakedown, work-
hardening and plastic deformation of the rail surface along the rail (Böhmer
and Klimpel, 2002), leading to non-homogeneous wear, surface cracking and
a progressively deteriorating geometry. This process is also called weld batter
(Mutton and Alvarez, 2004). Examples of this phenomenon are shown in Fig.
12.5. In in-situ welding it is almost impossible to avoid these indentations,
except for new track construction works, where the grinding can be done
separately from the welding.
Due to the effects that have been discussed previously, rail welds are often
considered, and also observed in practice, as corrugation or squat initiators
(Hiensch et al., 2002; Li et al., 2006, 2007). In Figs 12.6 and 12.7 examples
are shown of squats on aluminothermic and flash-butt welds, respectively.
Rail welds 383
The residual stress field which is induced over the cross-section of the
rail as a result of welding and heat transfer processes may be a further cause
of internal damage propagation or failure (Chen et al., 2006a,b). Webster
et al. (1997) showed that in aluminothermic welds the longitudinal residual
stresses are compressive at the top and the bottom of the rail, which is
(a)
(b)
12.3 (a) Weld and parent material; (b) the HAZ after thermite welding
and grinding.
384 Wheel–rail interface handbook
beneficial for the fatigue lifetime. In the web region, however, residual stresses
are strongly tensile, which makes that region very susceptible to damage
initiation and growth. Tawfik et al. (2006) performed a similar analysis for
flash-butt welds, and showed again a strong residual tensile stress field in
the web of the rail, leading to specific failure modes originating from the
rail web. Since the rail web mostly accounts for the shear stresses in the
cross-section, these are failure modes induced by shearing under high axle
loads, such as the horizontal split web mode. Tawfik et al. (2008) showed
that the tensile residual stress field may be alleviated by localised post-weld
heat treatment at the base of the foot directly after welding. Without heat
treatment, stress values in longitudinal direction may be as high as 300
MPa. Skyttebol et al. (Skyttebol, 2004; Skyttebol et al., 2005) showed that
the residual stress field in combination with stresses resulting from ambient
temperature loading and train axle loading may lead to very rapid crack or
damage growth and a significantly reduced fatigue life. Mutton and Alvarez
(2004) observed that the fatigue process leading to the horizontal split web
failure mode on heavy-haul lines (especially on curves and tangent track)
is in many cases low-cycle fatigue with few high axle overloads and rapid
fracture growing from weld collar slag inclusions. Ilic et al. (1999) showed
experimentally the effect of post-weld heat treatment also on the crystalline
structure of thermite welds (grain size reduction and homogenizing) leading
to higher ductility.
(a)
(b)
(a)
(b)
(a)
(b)
12.7 Squat initiations on flash-butt rail welds or in the HAZ.
The results in Fig. 12.8 have been obtained from FE simulation with the
DARTS-NL package. Parameter values were adopted as follows: a wheel
mass (half unsprung mass) of 970 kg; a sleeper mass of 300 kg; the 54E1
rail profile is used; the primary suspension stiffness equals 1.8.106 N/m (per
wheel); the railpad stiffness 1.2.109 N/m and the ballast stiffness 30.106 N/m
per sleeper. The latter value is rather small, to account for the generally bad
compaction of the ballast underneath the sleepers close to the irregularity;
388 Wheel–rail interface handbook
due to ‘ballast pumping’ loose sleepers often occur (Ishida et al., 1999). The
package uses a non-linear Hertz contact model.
In Fig. 12.8 the following quantities are shown as a function of time:
• the geometry of the irregularity z;
• the dynamic component of the vertical wheel displacement of the first
wheel of a passing bogie, defined positive in upward direction and
calculated in a convective reference frame moving along with the
wheel;
1.2
0.8 Artificial weld irregularity 1 m
z [mm]
1
uwheel/utrack,dyn [mm]
0.6
0.2
–0.2
0.04 0.08 0.12
–0.6 t [s]
Wheel displacement
–1 Track displacement
120
P1
Fcontact [kN]
100 P2
80
60
10
0
–10 0.04 0.08 0.12
–20 t [s]
–30
6
Mrail,dyn [kNm]
4
2
0
–2 0.04 0.08 0.12
–4 t [s]
160
100
Max. dyn. contact force [kN]
60
y = 0.28x
R2 = 0.97
40
20
0
0 40 80 120 160 200 240 280 320
Train speed [km/h]
which is responsible for the scatter in the relationship. The linear relationship
between the maximum dynamic contact force (or the dynamic amplification
factor) and the train speed is confirmed experimentally by the results obtained
by Mutton and Alvarez (2004). A very similar result was found, both from
simulations and experiments, by Jenkins et al. (1974) at British Rail for the
peak forces occurring at dipped rail joints. This result was reflected in the
well-known Jenkins formula for the calculation of P1 and P2 forces. They
found these peak forces to be governed by the dip angle and the train speed;
furthermore the relationship between maximum dynamic contact forces and
both variables proved to be linear in a good approximation.
Jenkins’ results for rail joints, together with the results obtained for rail
welds, are shown in the graph in Fig. 12.11. In this figure, the dynamic peak
forces for rail joints and welds are depicted as a function of the product of
the maximum absolute inclination and the train speed; a ≈ tan a is defined
as the angle of a dipped rail end with the horizontal for rail joints, and as
the maximum absolute inclination (on 25 mm basis) of the geometry for rail
welds. It is clear that the force level for rail welds is significantly lower than
392 Wheel–rail interface handbook
350
250
200
P2 for rail joints
150
100
0
0 100 200 300 400 500
Tan a · v [mrad · m/s]
12.11 Maximum dynamic wheel–rail forces (Fdyn) for rail welds and
rail joints, as a function of the product of the maximum absolute
inclination (on 25 mm basis) or the dip angle and the speed.
for rail joints; the reduction factor is approximately three. This is partly due
to the quasi-static rail deflection under axle loading, which is much larger
for a jointed rail (behaving as a hinged beam) than for a welded rail. This
quasi-static deflection increases the dip angle at the moment of wheel passage
(Steenbergen, 2006).
Steenbergen (2008a,b) also investigated the dynamic wheel–rail interaction
at rail welds with a linear numerical–analytical frequency–domain model,
comprising a rail on elastic foundation and a train axle up to the primary
suspension level. This study was performed for a better understanding of
the backgrounds of wheel–rail interaction at short irregularities and trend
behaviour. Analytical results show a perfectly linear relationship between
the slope of a ramp in the rail geometry and the magnitude of the dynamic
contact force in the time domain. The linear trend of this contact force as
a function also of the train velocity is shown in Fig. 12.12, for the case of
a ramp of 2 mrad and a length of 25 mm. Thus, for the elementary case
of a ramp in the rail surface, the first peak force (P1) may be considered
as directly proportional to the speed and the ramp angle, according to the
expression:
Fdyn,max = x · V · qmax: x = 0.18 [12.2]
Rail welds 393
35
25
20
Fit ---:
15 y = 0.36 x;
R2 = 1
10
0
0 15 30 45 60 75 90
Velocity [m/s]
1.2
Measured geometry
z [mm]
–0.4 –0.4 –0.3 –0.2 –0.1 0 0.1 0.2 0.3 0.4 0.5
x[m]
–1.2
350
250
150
Fdyn [kN]
50
–50 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Vt[m]
–150 Static wheel load
(a)
0.04
z [mm]
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
–0.04 x[m]
Measured geometry
20 5-point smoothing
10
Fdyn [kN]
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Vt[m]
–10
–20
(b)
12.13 Examples of measured welds, their geometry after 5-point
smoothing and calculated time histories of the dynamic wheel–rail
contact force at 140 km/h.
200
y = 152.56x
v = 300 km/h
R2 = 0.18
120
80
y = 100.45x + 28.12
R2 = 0.35
40
0
0 0.3 0.6 0.9 1.2 1.5
Vertical abs. peak deviation [mm] (unsmoothed)
(a)
200
v = 300 km/h y = 15.82x
R2 = 0.86
160
120
80
40
0
0 4 8 129 16
Max. abs. inclination [mrad) (5-point smoothed)
(b)
12.14 Peak value of the dynamic wheel–rail force as a function of the
vertical peak deviation (a) and the maximum absolute inclination (b)
respectively, at 300 km/h.
slightly higher value can be explained from the fact that inclinations on a 25
mm basis have been used in FEM analysis. The inclination decreases when
the sample interval is extended from 5 to 25 mm. It can be concluded that
x ≈ 0.2 in practice (for 54 E1 rail).
250
300 km/h
200 km/h
150
140 km/h
100
80 km/h
50
40 km/h
0
0 4 8 12 16
Max. abs. inclination [mrad] (5-point smoothed)
(a)
100
Max. abs. dyn. contact force [kN]
Inclination : 5 mrad
y = 0.94x
75 R2 = 0.99
50
25
0
0 10 20 30 40 50 60 70 80 90
Train speed [m/s]
(b)
of tolerances. These tolerances are defined for a given basis length, which
is commonly 1 m. A common value for the vertical tolerance is 0–0.3
mm (Esveld, 2001). Two drawbacks of this traditional method are that no
restrictions are being posed to the ‘smoothness’ of the rail surface, whereas
this shape has a direct relation to the magnitude and the spectrum of the
dynamic wheel–rail interaction force, and the fact that the train speed has
no influence, whereas its effect on the contact forces is far from negligible.
The spectrum of the wheel–rail forces in its turn is related to the rate of track
degradation. A typical result of the current rail welding assessment methods
(especially aluminothermic in-situ welding) is shown in Fig. 12.16. After the
Rail welds 397
0.6
y [mm] ‘Set-up’ Grinding area ‘Set-up’
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
x [m]
welding process, the ‘top’ resulting from the aligning of both rail ends, as far
as it does not fit within the tolerance, is ground off. The result is often far
from smooth, and introduces large dynamic variations in contact stresses.
the development of digital straightedges that sample the vertical rail
geometry enables more advanced assessment methods, that do account
for the ‘smoothness’ of the rail geometry. it is not feasible to compute the
wheel–rail contact forces in each step of weld grinding, in order to limit the
dynamic contact stresses. A compromise between theoretical efficiency and
practical feasibility, at the cost of exactitude, can be looked for by correlating
a parameter describing the geometry of the weld and the maximum contact
force that would occur for a predefined reference track and wheelset, and
the train speed. the relationship derived earlier (eq. 12.1) could be used
for this purpose, accounting for both the ‘smoothness’ of the rail surface
through its maximum first derivative and for the train speed. For assessment
purposes, the dimensionless rail weld quality index (QI) is proposed. It is
defined as the actually measured maximum absolute first derivative (on 25
mm basis) normalized with a norm value, which can be defined differently
for different values of the line section speed:
dz i (x )
dx max, actual
QI = (≤ 1: acceptance; >1: rejection) [12.3]
dz
dx norm
nor
0.2
y [mm]
0
–0.2
0 0.2 0.4 0.6 0.8 1
x [m]
0.3
y [mm]
0.1
–0.1
0 0.2 0.4 0.6 0.8 1
x [m]
12.18 Example of a weld with a pronounced step due to bad rail end
alignment.
0.2
y [mm]
0
–0.2
0 0.2 0.4 0.6 0.8 1
x [m]
12.19 Example of a weld with a ‘smooth’ surface but with negative
height coordinates.
1
95%
0.8
Cumulative frequency
0.6
carried out during the installation of the Dutch HSL-South (Winter et al.,
2007). In Fig. 12.20, the cumulative distribution of the maximum absolute
first derivatives (25 mm basis) of 296 test welds (a mix of flash-butt and
thermite welds) is also shown before and after grinding with a GWM 550
grinding train. Before grinding with the GWM the 95 percentile value of the
maximum first derivative is 5.3; this reduces to 1.1 afterwards. Finally, Fig.
12.20 displays the cumulative distributions of the maximum absolute first
derivative of the rail welds on a double track test section (East and West)
of 6 km on the HSL-South. These welds were manually pre-ground, and
finally ground with a Speno grinding train (types RR 24 and 48). Based on
the results in Fig. 12.20, the 0.7 mrad limit seems too strict a value, and 1
mrad is proposed as a limit value in the QI determination for welds in new
high-speed tracks (with norm speed 300 km/h).
In order to consistently determine speed-dependent norm values for the
maximum inclination, a pre-defined dynamic force level should be adopted.
This force level should be related to the damage level but, because little
is known of this relationship and furthermore this is strongly frequency-
dependent, any choice of this level is rather arbitrary. Two facts may be
taken as a point of departure:
1. the empirical vertical tolerance of 0.3 mm has been used worldwide for
several decades and has not led to systematic problems; the maximum
peak deviation resulting from the standardized inclination should, for
140 km/h (which is a conventional passenger train speed), not have a
different order of magnitude;
400 Wheel–rail interface handbook
2. the norm value for high-speed lines (300 km/h) has been established at
the feasibility limit of 1 mrad.
The norm value of 1 mrad at 300 km/h leads, according to Eq. (12.1)
(with x = 0.2), to a dynamic force level of 17 kN. The value of 0.7 mrad,
obtained for new straightened longrail, leads to a force level of 12 kN. The
curves according to Eq. (12.1) corresponding with both force levels are
depicted in Fig. 12.21. The following speed intervals have been used: 0–40
km/h, 40–80 km/h, 80–140 km/h, 140–200 km/h (conventional lines) and
200–300 km/h (high-speed lines).
The norm values should ideally be situated in the hatched area between
the two curves in Fig. 12.21, with a cut-off by the feasibility limit. However,
due to the hyperbolical behaviour, the values for low speeds grow to infinity,
whereas it is easy to grind, even manually, to a much better quality. Therefore,
the smooth function depicted in Fig. 12.21, leaving the hatched area at 40
and 80 km/h, has been adopted by the Dutch Rail Infra Manager ProRail.
In 2005, the geometrical standards as shown in Table 12.1 were introduced
for metallurgical rail welds in the Netherlands.
Table 12.2 gives the amplitudes of a sinusoidally shaped weld with a half
wavelength of 1 m, and corresponding to the inclinations in Table 12.1. These
amplitudes give an impression of the maximum vertical peak deviations
corresponding to the inclination values. Using the 1.8 mrad criterion for 140
km/h, the maximum amplitude equals 0.57 mm, which is larger than – but
in the same order of magnitude as – the 0.3 mm tolerance (1).
8
Inclination norm value (25 mm basis) [mrad]
2
Feasibility limit
1
0
0 40 80 120 160 200 240 280 320
Train speed [km/h]
0–40 3.2
40–80 2.4
80–140 1.8
140–200 1.3
200–300 1
NB: 60, 100 (special tracks) 2.8, 2.2
Table 12.2 Inclination norm values (25 mm) and the resulting 2 m wave
amplitude
40 3.2 1.02
80 2.4 0.76
140 1.8 0.57
200 1.3 0.41
300 1 0.32
Norm values
40 km/h
300
200
140
80
0.8
73 %
Cumulative frequency
0.6
58 %
46 %
0.4 CDF of 239 weld measurements
27 %
0.2 16 % satisfies tolerance 0–0.3 mm
17 %
0
0 1 2 3 4 5 6 7 8 9 10
Max. abs. weld inclination (25 mm basis) [mrad]
12.22 Comparison of acceptance levels of rail welds according to the
traditional method based on tolerances and the QI-method.
been used for the computations earlier in this chapter. It is observed that
the acceptance level increases drastically (except for 300 km/h). At the
conventional 140 km/h the acceptance increases by 30 %. This relaxation,
which, according to the investigation in this chapter should not lead to an
increasing rate of deterioration, is due to the fact that peak deviations larger
than 0.3 mm are accepted, provided that the contact geometry is smooth.
The weld geometry assessment method elaborated in this section can be
simply implemented in practice. This is briefly illustrated in the following.
After welding and cooling down of the rails, the quality of the geometry
can be measured using an electronic digital straightedge. In the processor
the routines for the calculation of the quality index can be programmed. The
device samples the rail geometry and plots the normalized first derivative
(dependent on the train speed) and the QI on a screen. With the help of
this output, the geometry can be optimised such that the standards are met.
An example of such a screen output is given in Fig. 12.23. At the location
where the longitudinal rail geometry shows a relatively large inclination,
the requirement is not met (the QI is 1.7), and the weld should be ground
before acceptance. In Fig. 12.24, an example is shown of rail weld geometry
assessment using an electronical straightedge.
mm 2 2
QI: 1.7
v: 40 km/h
1.5
1 1
0.5
0 0
0.2 0.4 0.6 0.8 1
–1
Weld geometry Normalised 1st derivative Norm – per definition 1
(a)
(b)
12.24 Ground rail weld (a) and geometry assessment (b) using a
digital straightedge.
0.08
0.04
z [mm]
0
0.2 0.4 0.6 0.8 1
–0.04
x = Vt [m]
–0.08
20
Dyn. force [kN]
10
0
0.2 0.4 0.6 0.8 1
–10 x [m]
–20
3
2
Power [kW]
Einput
1
0
0.2 0.4 0.6 0.8 1
–1 x [m]
(a)
12.25 Power input into the track per train wheel for the welds
depicted in Fig. 12.13 at 140 km/h.
Rail welds 405
1.2
0.4
z [mm]
250
150
50
–50 0.2 0.4 0.6 0.8 1
–150 x [m]
800
Power [kW]
600
400 Einput
200
0
0.2 0.4 0.6 0.8 1
x [m]
(b)
12.25 Cont’d
between 700 and 800 kW, for each passing wheel (using a linear calculation).
Given that modern locomotives generally have a traction power between
2000 and 8000 kW (the latter for high-speed trains), a power loss of 800
kW per wheel into the track due to a bad weld is very severe.
Figure 12.26 shows computational results for the elementary case of a
small ramp in the rail surface, which was considered in Fig. 12.12. In Fig.
12.26a, maxima of the power input into the track due to P1 are shown as
a function of the ramp inclination, for a given train speed of 140 km/h.
In Fig. 12.26b this is repeated as a function of the velocity, for a given
inclination of 2 mrad. The basis length of the ramp equals 25 mm in both
cases. In both cases, the relationship is exactly second-order polynomial.
Both relationships with the maximum dynamic force were linear. This indicates
the severity of rail surface irregularities for long-term track behaviour, which
is governed by energy dissipation mechanisms. These mechanisms need
future investigation.
406 Wheel–rail interface handbook
50
v = 140 km/h
30
P1 [kW]
20
10
0
0 2 4 6 8 10 12
Inclination [mrad]
7
q = 2 mrad
Maximum power input due to
6
y = 7.86e–05x2 –1.59E–03x
5 R2 = 1.00E + 00
P1 [kW]
0
0 40 80 120 160 200 240 280 320
Velocity [km/h]
12.26 Maxima of the power input into the track due to P1 for a
ramp with a basis of 25 mm. (a) Values as a function of the ramp
inclination, for 140 km/h; (b) values as a function of the velocity, for
an inclination of 2 mrad.
12.7 References
Böhmer A and Klimpel T (2002), Plastic deformation of corrugated rails – a numerical
approach using material data of rail steel, Wear, 253, 150–61.
Cai W, Wen Z, Jin X and Zhai W (2007), Dynamic stress analysis of rail joint with
height difference defect using finite element method, Engineering Failure Analysis,
14, 1488–499.
Chen Y, Lawrence F V, Barkan C P L and Dantzig J A (2006a), Heat transfer modelling
of rail thermite welding, Proc IMechE, Part F: Journal of Rail and Rapid Transit,
220, 207–17.
Chen Y, Lawrence F V, Barkan C P L and Dantzig J A (2006b), Weld defect formation
in rail thermite welds, Proc IMechE, Part F: Journal of Rail and Rapid Transit, 220,
373–84.
Rail welds 407
Esveld C (2001), Modern Railway Track (2nd edn), Zaltbommel, the Netherlands, MRT-
productions.
Esveld C (2005), Measurements of Rail and Weld Geometry on HSL-South, ECS report,
Zaltbommel, the Netherlands.
Hiensch M, Nielsen J C O and Verheijen E (2002), Rail corrugation in the Netherlands
– measurements and simulations, Wear, 253, 140–49.
Ilic N, Jovanovic M T, Todorovic M, Trtanj M and Saponjic P (1999), Microstructural and
mechanical characterization of postweld heat-treated thermite weld in rails, Materials
Characterisation, 43, 243–50.
Ishida M, Moto T, Kono A and Jin Y (1999), Influence of loose sleeper on track dynamics
and bending fatigue of rail welds, Quarterly Report of RTRI, 40, 80–85.
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.
Kabo E, Nielsen J C O and Ekberg A (2006), Prediction of dynamic train-track interaction
and subsequent material deterioration in the presence of insulated rail joints, Vehicle
System Dynamics, 44, 718–29.
Li Z, Zhao X, Esveld C and Dollevoet R (2006), Causes of squats: correlation analysis
and numerical modeling, Proceedings 7th International Conference on Contact
Mechanics and Wear of Rail/Wheel Systems, Brisbane, Qld Australia, 24–27 September,
439–46.
Li Z, Zhao X, Esveld C and Dollevoet R (2007), Rail stresses, strain and fatigue under
dynamic wheel-rail interaction, Proceedings International Heavy Haul Association
Specialist Technical Session, Kiruna, Sweden, 11–13 June, 389–96.
Mutton P J (2000), Material aspects of weld behaviour in wheel-rail contact, Proceedings
Fifth International Conference on Contact Mechanics and Wear of Rail/Wheel Systems,
Tokyo, Japan, 25–27 July, 131–5.
Mutton P J and Alvarez E F (2004), Failure modes in aluminothermic rail welds under
high axle load conditions, Engineering Failure Analysis, 11, 151–66.
ProRail (2007), Directives RLN00127–1&2, Part 1 – Operationele eisen voor metallurgische
lassen in bovenbouwconstructies; Part 2 – Kadereisen voor metallurgische lassen in
bovenbouwconstructies, ProRail, Utrecht, the Netherlands.
Shitara H, Terashita Y, Tatsumi M and Fukada Y (2003), Nondestructive testing and
evaluation methods for rail welds in Japan, Quarterly Report of RTRI, 44, 53–8.
Skyttebol A (2004), Continuous Welded Railway Rails: Residual Stress Analyses, Fatigue
Assessments and Experiments, PhD Thesis, Chalmers University of Technology,
Gothemburg, Sweden.
Skyttebol A, Josefson B L and Ringsberg J W (2005), Fatigue crack growth in a welded
rail under the influence of residual stresses, Engineering Fracture Mechanics, 72,
271–85.
Steenbergen M J M M (2006), Modelling of wheels and rail discontinuities in dynamic
wheel-rail contact analysis, Vehicle System Dynamics, 44, 763–87.
Steenbergen M J M M (2008a), Quantification of dynamic wheel-rail contact forces at
short rail irregularities and application to measured rail welds, Journal of Sound and
Vibration, 312, 606–29.
Steenbergen M J M M (2008b), Wheel–Rail Interaction at Short-Wave Irregularities,
PhD Thesis, Delft University of Technology, Delft, the Netherlands.
Steenbergen M J M M and Esveld C (2006a), ‘Rail weld geometry and assessment
concepts’, Proc. IMechE, Part F: Journal of Rail and Rapid Transit, 220, 257–71.
408 Wheel–rail interface handbook
Steenbergen M J M M and Esveld C (2006b), Relation between the geometry of rail welds
and the dynamic wheel–rail response; numerical simulations for measured welds, Proc
IMechE, Part F: Journal of Rail and Rapid Transit, 220, 409–24.
Tawfik D, Kirstein O, Mutton P J and Chiu W K (2006), Verification of residual stresses
in flash-butt-weld rails using neutron diffraction, Physica B, 385–6, 894–6.
Tawfik D, Mutton P J and Chiu W K (2008), Experimental and numerical investigations:
alleviating tensile residual stresses in flash-butt welds by localised rapid post-weld
heat treatment, Journal of Materials Processing Technology, 196, 279–91.
Terashita Y and Tatsumi M (2003), Analysis of damaged rail weld, Quarterly Report of
Railway Technical Research Institute, 44, 59–64.
Webster P J, Mills G, Wang X D, Kang W P and Holden T M (1997), Residual stresses
in alumino-thermic welded rails, Proc. IMechE, Journal of Strain Analysis, 32,
389–400.
Wimmer W E, Connell D A and Boos M J (2002), Joint elimination through mobile flash-
butt welding on Union Pacific Railroad, Proc. of Innotrans 2002, Berlin, Germany,
24–27 September.
Winter T, Meijvis P A J, Paans W J M, Steenbergen M J M M and Esveld C (2007),
Track quality achieved on HSL-South – reduction of short-wave irregularities cuts
life cycle cost, European Railway Review, 13(3), 48–53.
Wöhnhart A and Wenty R (2002), Mobile flash-butt welding: three decades of experience,
Rail Engineering International, 3, 11–16.