Skytteb0l Et Al 2005
Skytteb0l Et Al 2005
Skytteb0l Et Al 2005
www.elsevier.com/locate/engfracmech
Abstract
The effect of welding residual stresses on fatigue crack growth in rail welds is studied. Finite element analysis is used
to calculate residual stresses in a flash-butt welded rail. The calculated residual stresses are found to be in good
agreement with experimentally determined residual stresses in a welded rail. The redistribution of residual stresses in the
welded rail is simulated for a straight track, during heavy-haul operation conditions, using a train–track model. Fatigue
crack growth of defects in the weld region is studied using fracture mechanics. In the investigation, a number of
parameters such as the axle load, crack location, crack size and rail temperature are varied.
2004 Elsevier Ltd. All rights reserved.
1. Introduction
Railway rails are manufactured in sections of 25–120 m length. The method of joining rails has changed
over the years. Earlier, the most common method was to use bolting. It was, however, found that bolted
joints were a large source of maintenance, and that they often broke in the plate, rail ends, or in the bolt as
reported by e.g., Sih [1] and Mayville [2]. As a consequence, new joining methods for rails were investigated,
and in the 1950’s, the first welded rails begun to appear in track. The utilisation of the welding method for
joining rails has since then been applied increasingly, and it is now standard practise all over the world. The
most important reasons to use continuous welded rails, in contrast to bolted rails, are the lower mainte-
nance cost and the improved dynamic behaviour of the train–track–rail system.
There are primarily two welding processes used today, flash-butt welding and alumino-thermic (ther-
mite) welding, but other methods are also employed such as gas pressure welding and enclosed arc welding
[3]. Flash-butt welding is used to weld short rail sections from the manufacturing together into longer
sections of about 400 m, which are then transported and welded in the track laying process. It is a resistance
*
Corresponding author. Tel.: +46-31-772-1498; fax: +46-31-772-3827.
E-mail address: anders.skyttebol@me.chalmers.se (A. Skyttebol).
0013-7944/$ - see front matter 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfracmech.2004.04.009
272 A. Skyttebol et al. / Engineering Fracture Mechanics 72 (2005) 271–285
welding method, often performed in stationary plants, which consists of electrical heating and hydraulic
forging of the rail ends. The thermite welding method on the other hand, is a casting method and it is the
most common method used in the field to weld sections together. In thermite welding, the rails are laid with
a gap and a ceramic mould is placed around the rail. The alumino-thermic mixture is ignited and molten
steel flows down the gap to make up the weld. Webster [4] has carried out residual stress field measurement
in thermite welds. However, both types of welding methods introduce a similar residual stress field in the
foot, the web and the head of the rail.
Small inclusions or cracks can appear in the weld cross-section [5]. The thermite weld process generates
cracks to a higher degree due to the casting nature, as compared with the flash-butt weld process. Defects in
thermite welds are often lack-of-fusion, shrinkage cavities, hot tears, gas pockets and inclusions from
entrapped slag or mould. Defects in flash-butt welds are almost always confined to iron oxide inclusions in
the fusion zone, internal imperfections in the weld, or microstructural alterations from post-weld surface
treatment [6]. Mutton et al. [7] have reported on surface cracks in the web of flash-butt welds and shrinkage
cracks in thermite welds. Obviously, future demands on train traffic are heavier axle loads, higher train
speeds and increased traffic density. To prevent accidents due to fracture of welds, and to lower mainte-
nance costs, detailed numerical and fatigue analysis of welds from different welding methods are of great
interest for the railway industry.
The stresses introduced in a rail by a passing train are the combination of several load mechanisms:
flexural action and local contact stresses from axle load and train speed, residual stresses and thermal
stresses from the ambient temperature or from the train wheel (e.g., braking). In order to balance the forces
that want to buckle the rail during high temperatures, called ‘‘sun-curves’’, and those that want to pull the
rail apart during cold weather, the rail is laid in track at a particular neutral temperature that varies with
the average temperature of the site. In Sweden, rails are normally laid so that they are stress-free at
an ambient temperature of 20 C. However, the annual rail temperature in Sweden varies from )40 to
40 C.
In this investigation, the train traffic situation at the Malmbanan (Iron-ore) line was studied. It is located
in the northern part of Sweden, and the train speed is low because of high axle loads, typically 50 km/h. The
axle load of a fully loaded freight car is 25–30 metric tonnes, and the annual load capacity on the line is 20–
22 MGT (in total 870,500 wheel passages); see Hiensch et al. [8] for more details of a section studied of the
Malmbanan line. Cracks have been found in welded joints, and the hypothesis was that the residual stress-
state from the welding process, together with pre-existing cracks in the rail/weld, has a great effect on the
time to fatigue failure.
This investigation is a numerical study on fatigue crack growth and propagation of cracks in flash-butt
welded rails. It includes different sizes and locations of the cracks as well as different loading conditions.
The simulations incorporate several steps, and calculations, to make the fatigue assessment of the weld as
realistic as possible, see Fig. 1. The finite element (FE) method, using the commercial FE code ABAQUS/
Standard [9], was used to simulate the welding process, the dynamic global train–track response and the
elasto-plastic local stress–strain response in the rail. The fatigue crack growth and propagation analyses
were made in the commercial code SACC [10].
The sequence of steps in the fatigue evaluation is described in the Sections 3–6. Section 3 presents results
from residual stress measurements, materials testing, material characteristics and constitutive material
modelling. Section 4 presents the numerical simulation of the flash-butt welding process generating the
residual stress-state in the weld and the rail. The results from the simulation were then used in the analyses
presented in Section 5.
A. Skyttebol et al. / Engineering Fracture Mechanics 72 (2005) 271–285 273
∆σij,σij
σijres
Crack growth analyses
Assessment of surface and
interior cracks in welded rails
Section 5 introduces the FE tool, which consists of two coupled models, the track and rail models. In
short, the track model is used to simulate the dynamic global track response that causes bending and
rotation of the rail cross-section. The rail model features the possibility to include the following parameters
in the three-dimensional (3D) elasto-plastic analysis: residual stresses caused by the rail manufacturing
process, residual stresses in the rail due to warm or cold climate, residual stresses due to welding, influence
from global track behaviour and the local wheel–rail contact load. The influence from all the parameters
can either be separated or studied all together. As a result, the evaluation of fatigue crack growth in the rail
can easily be made for different combinations of parameters.
In Section 6, the results from the rail model analyses are used in the code SACC to investigate how the
size, the orientation and the position of an initial crack or weld defect can affect the fatigue life of a welded
rail.
The results from the residual stress measurements showed that compressive residual stresses appear in
the foot and the head of the rail, while large tensile residual stresses appear in the web. The measurements
were carried out on the surface of the rail in the weld, using the hole-drilling strain gauge method according
to ASTM E837-99 [11]. The strain gauges that were used were the HBM hole-drilling rosette 1-RY61-1.5/
120, and they were glued to the rail’s surface using the adhesive Z70. The drilling was carried out to a depth
of 250 lm. The released strains were measured with the HBM Spider8 and the Catman Express software
[12]. The hole-drilling method renders a measurement uncertainty of about 10%. The residual stress in the
weld cross-section is presented in Fig. 2 as rz ðyÞ; see Fig. 4 for definition of direction of stresses.
The behaviour of the pearlitic rail steel UIC grade 900A (R260) at room temperature was obtained from
materials testing at 20 C, while the behaviour at elevated temperatures was obtained from materials testing
at 600 C. The tests showed, as could be expected, that both the yield stress and the Young’s modulus drop
274 A. Skyttebol et al. / Engineering Fracture Mechanics 72 (2005) 271–285
180
160
140
100
80
60
40
FE calculation
20 Experiments
0
-600 -400 -200 0 200 400 600 800
σ z (MPa)
Fig. 2. Longitudinal residual stresses in the weld cross-section from experiments and FE calculation.
to about half when the temperature is increased from 20 to 600 C. For temperatures higher than 600 C, no
tests were performed and thus, the behaviour at these temperatures was estimated from Runesson et al. [14].
The material behaves in a rather complex manner at higher temperatures. In this investigation, it was
decided to represent the material with a constitutive material model which is elasto-plastic, with linear
kinematic hardening, and that includes an annealing behaviour. Note that if the objective with the analysis
was to study surface fatigue of the material on the rail head, a more elaborate nonlinear kinematic
hardening model would have been required. Such a model has the capacity to simulate cyclic plasticity and
ratchetting material response with decaying rate, see Ringsberg and Josefson [13], in contrast to a linear
kinematic hardening model which cannot describe the material behaviour on the rail head accurately for
repeated wheel–rail contacts. In the present investigation, only the part well below the surface of the rail
head was considered, and hence, a linear kinematic model was deemed sufficient.
The choice of constitutive material model was made on the basis that both the welding process, and the
subsequent fatigue analysis, should be carried out using the same material description. The yield stress, ry ,
was defined according to the von Mises yield criterion, where the yield function, /, is defined as
qffiffi
/ðr; X Þ ¼ 32jsdev j ry ð1Þ
with
sdev
ndev ¼ ð4Þ
jsdev j
where k_ is the plastic multiplier and H is the hardening modulus. The development of the plastic strain, e_ p ,
follows the associative flow rule as
A. Skyttebol et al. / Engineering Fracture Mechanics 72 (2005) 271–285 275
800
Dashed lines: Experiments
700
Solid lines: Material model
600
500
T = 20oC
σ (MPa)
400
300
200 T = 600oC
100
0
Tmelt = 1470oC
-100
-0.5 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Fig. 3. Mechanical material data from experiments and data used in the FE analysis with ABAQUS.
Table 1
Material constants as function of temperature used in the elasto-plastic analysis
T (C) E (GPa) m ry (MPa) a (1 · 106 /C)
20 210 0.3 430 13.3
600 110 0.3 242 15.7
>1000 10 0.4 20 17.4
qffiffi
e_ p ¼ k_ 2
n
3 dev
ð5Þ
Annealing material behaviour was accounted for in the analysis. In ABAQUS, the equivalent plastic
strain and the backstress tensor are set to zero at the annealing temperature. In the present investigation,
the annealing temperature was set to 1470 C which is the melting temperature for the current material. In
Fig. 3, the characteristics of UIC grade 900A material in a tensile test for different temperatures are shown
with dashed lines. Solid lines show the calculated elasto-plastic behaviour at different temperatures using
the linear kinematic hardening model. The temperature-dependent material data according to Table 1, and
the material behaviour according to Fig. 3, was used in the thermo-mechanical analysis and it was inter-
polated linearly between temperatures.
Flash-butt welding is a resistance welding method that is usually attributed to some advantages such as
high weld strength, no need of special preparation of weld surfaces, high production rate, small upset, good
heat concentration and the possibility to weld dissimilar materials. The technology consists of an electric
potential applied between two pieces of metal clamped adjacent to each other. This causes a current to flow
through the circuit sufficient to produce a flashing action. The metal is heated to the fusion point and the
weld is completed by applying an upset force.
The operation comprises three steps: preheating, burning-off and upsetting. Preheating is made by
pressing the two rails together and separation of them as an electric current passes across the interface. In
the second step, the edges are slightly parted and a low voltage is applied between the ends causing a flash
276 A. Skyttebol et al. / Engineering Fracture Mechanics 72 (2005) 271–285
arc. The surfaces are cleaned and uniformly heated during this step. In the third and last step, the upsetting,
a force is applied rapidly, which forge the two ends together and the molten metal between the ends are
expelled, see Ref. [16] for details.
The flash-butt weld simulation using ABAQUS was carried out in a sequence, starting with an electro-
thermal analysis that provides a temperature field history to the subsequent thermo-mechanical analysis.
The option for geometrical nonlinearities in ABAQUS was used for this analysis, as well as for the train–
track–rail interaction analysis in Section 5.
The electro-thermal analysis was fully coupled. Since there was symmetry between the two rails being
welded, and also in rail geometry cross-sections, only one half of a rail was modelled, see Fig. 4. The
electrical and thermal fields present during the welding were calculated using 8-node linear 3D solid ele-
ments with two degrees of freedom at each node, temperature and electric potential. In total, the FE model
contained 33,733 nodes.
Material data used in the analysis was taken from Madi [15], the rail manufacturer, or in some cases had
to be estimated. The thermal and mechanical properties as function of the temperature are shown in Fig. 5.
Boundary conditions, ambient temperature, convection and radiation were estimated from the environment
at the rail welding site [16]. Weld process parameters such as displacement, current, and electrical potential,
were measured or recorded by the welding equipment. Uncertain quantities were the contact resistance
between the electrode and the rail, the gap resistance between the two rails, and the power efficiency. In
order to account for these uncertainties, the temperature was measured in five locations along the rail. The
analysis was thereafter calibrated, with respect to the resistances, using these temperatures.
During welding, the electrode on the right rail was given a positive potential and the left rail had ground
potential. Thus, the weld surface was heated by the electric current that flows between the electrodes in
contact with the rail (shown as clips above and below the rail in Fig. 4). As the current passes through the
air gap between the rails, a high power was developed which caused melting of the material. As a result of
high temperatures, it was necessary to account for temperature dependence of the following material
constants: the specific heat, thermal and electrical conductivity was taken temperature dependent according
to Fig. 5, while a constant average value over the temperature range was used for the heat transfer coef-
ficient (convection) and the emmissivity (radiation). In addition, the material close to the rail end trans-
forms from austenite to pearlite during final cooling. Hardness measurements and microstructure
7
Specific heat/100 (m2s-2K-1)
thermal conductivity
. -1m-1)
5
0
0 200 400 600 800 1000 1200 1400
Temperature (oC)
evaluation of the heat affected zone (HAZ) reveal no major changes [16]. In this investigation, we have,
therefore, disregarded phase transformation effects both in the thermal and in the mechanical analyses.
The thermo-mechanical analysis was sequentially coupled and executed using the temperature field hi-
story from the former analysis as input. In this analysis, the same FE mesh as in the electro-thermal analysis
was used, but the elements were now changed to the 8-node linear brick elements with 3 displacement
degrees of freedom in each node. The mechanical material properties and constitutive material behaviour is
described in Section 3.
Rollers supported the rail under the rail foot and a longitudinal symmetry condition was applied to the
weld surface. Initially, the rail was heated until the weld surface was molten. Thereafter, a force of 40 metric
tonnes was applied to the back end of the rail, which forged them together. The rail was cooled to room
180
160
140 σx
Rail height (mm)
σz
120
σy
100
80
60
40
20
0
-800 -600 -400 -200 0 200 400 600 800
σx, σy and σz (MPa)
Fig. 6. Residual stresses in the centre of the weld from the flash-butt weld FE analysis.
278 A. Skyttebol et al. / Engineering Fracture Mechanics 72 (2005) 271–285
temperature, air-cooling was assumed here, with a residual stress field as a result. In addition, to simulate
grinding of the weld, which always occurs, the expelled material was removed from the model. It had only
little effect on the resulting residual stress field. One reason may be that only removal of material was
accounted for in the analysis and not the increase in temperature and contact loads that also occur during
grinding. To conclude, there was good agreement in the results from the flash-butt welding simulation and
the residual stress measurements in the web and rail head, but less in the rail foot, see Figs. 2 and 6.
An FE tool was developed in previous work for the analysis of rolling contact fatigue (RCF) of railway
rails; see Ringsberg et al. [17] for details. It mimics wheel–rail rolling–sliding contact on any track, since it
incorporates both the dynamic global track response and the three-dimensional local elasto-plastic contact
conditions in the rail head. Section 5.1 gives a short description of the FE tool; Section 5.2 specifies the
contact loads of the heavy-haul Malmbanan line train traffic situation analysed; Section 5.3 presents four
analyses with different initial stress-states of the rail.
Two FE models, for track and rail, which are coupled by time-dependent boundary conditions, form the
FE tool. An elastic FE analysis using the track model calculates time-dependent displacements at two cross-
sections 12 cm apart; these are then used as boundary conditions in an elasto-plastic FE analysis using the
rail model. As a result, the influences of both the dynamic global track response and the three-dimensional
local elasto-plastic material response in the rail are incorporated in the rail fatigue or stress analysis. It was
shown in Ringsberg et al. [17] that the global track response predicted by the track model affects the
longitudinal residual stress-state in the rail head by some 10%, as compared with a situation where it is
disregarded.
The track model is an elastic beam element one which is used in dynamic analysis: 32 sleeper bays are
designed to represent an arbitrary track. The ballast material, the pads, the sleepers, and the rail beam are
all modelled by finite elements to enable realistic simulations of train traffic conditions. Distributed and
moving force combinations of normal, longitudinal, and lateral rail force components represent a train
vehicle travelling along the track.
The rail model is a three-dimensional FE model where 8-node linear brick elements form the mesh. Fig. 7
shows the rail model FE mesh used in the present investigation. The material volume near the wheel–rail
contact undergoes elasto-plastic material response. Hence, high mesh density was used to resolve all the
stress and strain gradients satisfactorily. The elasto-plastic material behaviour was modelled with the linear
kinematic hardening model presented in Section 3, and the material used was the UIC grade 900A. This
constitutive material model was deemed satisfactory for the present investigation where the material re-
sponse develops into elastic or elastic shakedown subsurface after some wheel–rail passages (load cycles).
Since train traffic on a straight track was investigated, the wheel–rail contact position was at the sym-
metry line in the rail longitudinal direction, see Ringsberg and Lindb€ack [18]. The wheel–rail contact loads
acting on the rail model, in the normal, longitudinal and lateral directions, were represented by moving
distributed normal pressure and shear stress distributions. The distributions were calculated according to
the Hertz theory of rolling contact between two elastic nonconforming solids with smooth and continu-
ous contact surfaces [19]. The normal load was represented by a Hertzian contact pressure distribution,
pðx; zÞ, where ðx; zÞ are the local coordinates that define the contact zone in the longitudinal and lateral
directions, respectively. The tangential loads, qðx; zÞ, were modelled as proportional to pðx; zÞ, that is
A. Skyttebol et al. / Engineering Fracture Mechanics 72 (2005) 271–285 279
qðx; zÞ ¼ lðlx ; lz Þ pðx; zÞ where ðlx ; lz Þ are the traction coefficients in the ðx; zÞ directions. Hence, regions of
stick and slip were disregarded, although they are known to occur.
The Malmbanan line between Kiruna (Sweden) and Narvik (Norway) is two-way-trafficked on one
track, i.e., loaded freight cars run from Kiruna to Narvik where the ore is unloaded before the cars return
(unloaded) to Kiruna on the same track. The loaded iron-ore cars have an average axle load of 30 metric
tonnes and they run at 50 km/h on straight track. Since these loaded iron-ore cars contribute the most to
RCF damage of the rail, their contact loads were used here.
A 30 metric tonne axle load corresponded to a rail normal force of 147 · 103 N, which, according to the
Hertz theory for the present wheel and rail geometry, equals a peak normal pressure of 1401 MPa. The
magnitude of the tangential loads was estimated from field measurements at track (using resultant contact
forces in the normal, longitudinal and lateral rail directions) and from material testing by twin-disc tests
(dry contact) [20]. The in-plane rail longitudinal force component due to traction was modelled using
lz ¼ 0:35 (estimated from twin-disc tests), and since straight track was modelled, only a small contribution
from the in-plane rail lateral force component due to traction was modelled using lx ¼ 0:10 (estimated from
the field measurements). Hence, in the rail model, there was a shear stress distribution for fully slipping
contact both in the longitudinal and lateral directions of the rail.
Railway rails in Sweden are normally laid at the reference temperature 20 C, where the rail is stress-free
when it is clamped to the sleepers. A small amount of the rail contraction and expansion can be taken up
across the rail through height and width expansion but the rail may be allowed to bulge slightly as long as it is
prevented to move sideways. The rail is held to the sleeper with strong clips that prevents the rail to move
along the track; hence, the sleepers take up most of the forces from temperature differences. The weight of the
280 A. Skyttebol et al. / Engineering Fracture Mechanics 72 (2005) 271–285
concrete sleeper and its friction to the ballast prevent any movement, and thus, the only expansion of the rail
that can occur is between the sleepers. The ambient temperature near the Malmbanan line varies annually
between )40 and 40 C, which gives rise to rather large thermal stresses in the rail. Therefore, the tem-
perature effect on the residual stress-state was also investigated for the case of the 30 metric tonnes axle load.
All together, four initial stress-state conditions of a rail were analysed with the FE tool:
A user-supplied subroutine was formulated in the FE code ABAQUS in order to transfer the results
from the flash-butt welding FE simulation (see Section 4) to the rail FE model calculations (see Section 5).
It was done by interpolation, between nodes, of node average stresses. These stresses became initial con-
ditions to the first step of the rail FE model calculation. No contact loads were applied for this step, and
equilibrium was reached after some iterations. The residual stress in the longitudinal direction of the rail
due to cold or warm climate were added using the subroutine as E a DT , where DT is the deviation in
temperature from the reference temperature, E is the Young’s modulus and a is the thermal expansion
coefficient.
The results from the welding residual stress fields which were calculated and measured in the present
investigation, for the flash-butt welding process, were found to be qualitatively the same as can be
experimentally found for thermite welding [4]. For the longitudinal stress field this means compressive
stresses in the rail head and rail foot, and tensile residual stresses in the web. It was, therefore, possible to
discuss growth of defects in the weld region for both welding methods. The calculations showed large stress
ranges due to global bending of the rail in the lower part of the rail head, whereas the stress ranges cal-
culated in the web and rail foot were much lower; see Fig. 8 for contour plots of the residual stress state in
the longitudinal direction of the rail. Furthermore, it was found that, apart from the contact zone in the rail
head, the weld material experienced elastic shakedown after very few load passages. With the weld material
subject to elastic conditions, linear-elastic fracture mechanics can be used to study growth of defects in the
weld region. Here, the code SACC [10] was employed for this purpose.
For the region of interest in the rail head, the calculated shear stress ranges were found to be consid-
erable less than the normal stress ranges. Hence, the stress-intensity factor was calculated based on the
normal stress component in the uncracked material normal to the plane containing the defect. Three dif-
ferent defects were considered, see Fig. 9:
(1) an embedded circular crack at mid level of the rail head subject to vertical stress ranges,
(2) an embedded circular crack in the rail head just above the web also subject to vertical stress ranges, and
(3) a half-circular surface crack at the lower part of the rail head subject to longitudinal stress ranges; crack
length/mouth on the surface divided by crack depth ¼ 2.
The first two cases correspond to inclusion defects or lack-of-fusion defects, while the third case cor-
responds to a crack starting from a defect caused by the final trimming of the web and rail head. For all
A. Skyttebol et al. / Engineering Fracture Mechanics 72 (2005) 271–285 281
Fig. 8. Contour plots of the residual stress in the longitudinal direction of the rail: (a) after the flash-butt weld analysis (no wheel
passage), (b) 30 metric tonnes axle load after eight wheel passages, (c) 35 metric tonnes axle load after eight wheel passages, and (d) 40
metric tonnes axle load after eight wheel passages.
cases, the code SACC was employed using two standard cases, an embedded crack (with zero eccentricity)
in a plate with finite thickness, and a circular surface crack in a plate with finite thickness. In both these
standard cases, a representative plate thickness t ¼ 40 mm was used and the size of the crack area, Acrack was
varied between 5 and 100 mm2 .
It was discussed in previous sections that heavy-haul traffic on the Malmbanan line was studied using
real traffic data for 30 metric tonnes axle load and a train speed 50 km/h. The present train–track inter-
action calculations were quasi-static. To include possible dynamic effects on contact loads, results are
presented for three different axle loads: 30, 35 and 40 metric tonnes, respectively.
Crack growth data for the rail steel UIC grade 900A was taken from the EU-project ICON [21]. A Paris
type crack growth law was used with C ¼ 2:47 109 (growth rate in mm/cycle and stress in MPa) and the
exponent n ¼ 3:33. Due to the high tensile stresses in the lower part of the rail head a very high stress ratio
R ¼ rmin =rmax was achieved, R ¼ 0:7–0:9. This means that existing cracks are fully open during the load
passage. For these high R-ratios no value of the threshold stress-intensity factor
pffiffiffiffi range, DKth , was found in
the literature for the current material. Therefore, the value DKth ¼ 2 MPa m was used which is repre-
sentative for steels at high R-values [22].
The results from the FE analysis of a residual stress-free rail, i.e., example (a) in Section 5.3, resulted in
stress ranges at elastic shakedown that were not sufficient for further crack growth of any of the three
defects. For the three examples (b)–(d) with residual stresses from the welding simulation, Table 2 shows
282 A. Skyttebol et al. / Engineering Fracture Mechanics 72 (2005) 271–285
d1
1. Embedded crack
d2
2. Embedded crack
3. Surface crack
Fig. 9. Analysed crack positions and configuration in the rail head: d1 ¼ 24 mm and d2 ¼ 43 mm.
Table 2
Analysed crack locations and stress characteristics for a temperature of 20 C
(1;2;3): Crack location (load direction) Acrack (mm2 ) rmax –rmin (MPa) Dr (MPa) Nf (106 cycles) Traffic years
Axle load: 30 metric tonnes
1: rail head mid (y-dir) 50 431–334 97 3.0 3.4
1: rail head mid (y-dir) 75 431–334 97 2.3 2.6
1: rail head mid (y-dir) 100 431–334 97 1.8 2.1
2: rail head lower (y-dir) 50 582–550 32 Infinity Kmax ¼ 30
2: rail head lower (y-dir) 75 582–550 32 Infinity Kmax ¼ 33
2: rail head lower (y-dir) 100 582–550 32 Infinity Kmax ¼ 35
3: surface lower rail head (z-dir) 5 735–697 38 Infinity Kmax ¼ 33
3: surface lower rail head (z-dir) 10 735–697 38 7.8 9.0
Table 3
Fatigue life of a rail with 30 metric tonnes axle load for different rail temperatures (thermal stresses), and without residual stresses from
welding
(1;2;3): Crack location Ambient rmax –rmin (MPa) Dr (MPa) Nf Traffic
(load direction); temperature (C) (106 cycles) years
embedded: Acrack ¼ 100 mm2
surface: Acrack ¼ 10 mm2
Axle load: 30 metric tonnes
1: rail head mid (y-dir) 20 (no residual stresses) )11–()108) 97 Infinity Kmax <0
1: rail head mid (y-dir) 40 356–259 97 2.3 2.6
1: rail head mid (y-dir) )40 656–559 97 0.1 0.1
2: rail head lower (y-dir) 20 (no residual stresses) )17–()50) 33 Infinity Kmax <0
2: rail head lower (y-dir) 40 508–475 33 Infinity Kmax ¼ 31
2: rail head lower (y-dir) )40 805–772 33 0 Kmax ¼ 49 > KIC
3: surface lower rail head (z-dir) 20 (no residual stresses) 47–8 39 Infinity Kmax ¼ 15
3: surface lower rail head (z-dir) 40 661–623 38 22.0 25.3
3: surface lower rail head (z-dir) )40 957–919 38 0 Kmax ¼ 47 > KIC
the calculated fatigue life for the three defects studied. The calculations were stopped when the maximum
stress-intensity factor reached the fracture toughness reported for welded UIC grade 900A material (in the
pffiffiffiffi
rail head region), KIC ¼ 40 MPa m (see Moller et al. [23]). Table 3 shows the influence of thermal stresses
from different rail temperatures for the 30 metric tonnes axle load.
Even though the sizes of the crack areas for the surface cracks were less than the areas of the embedded
cracks, it was observed that the weld was more sensitive (less fatigue-resistant) to the surface cracks than
the embedded cracks. The reason was that the stress ranges in the location for the surface crack were larger.
In addition, internal cracks can be found in thermite welds, and due to the casting process, such shrinkage
cracks are normally found in the foot or lower part of the web. For the rail weld in the present investigation
in a straight track, the stress ranges caused by service loading were much higher in the rail head than in the
foot. Experimental results from Mutton et al. [7], for a similar axle load, show that the stress range in the
web may be higher for a weld in a curved track due to the torsion of the web. Depending on possible
horizontal misalignment of the rail weld additional stress ranges may also appear in the weld region. Hence,
the situation studied here, a weld in a straight track, may not represent a severe loading case in terms of
fatigue crack propagation. However, it gives a qualitative picture of fatigue growth of different defects in a
rail weld.
7. Conclusions
In this investigation, the flash-butt welding process and the fatigue crack growth in the rail weld under
train passages on a rail was simulated. The objective was to investigate the influence of the residual stress
field in the rail from welding, ambient temperature and train loading, on the fatigue crack growth of pre-
existent flaws in the weld.
The investigation includes analyses of the residual stresses from welding, the redistribution of these
residual stresses when the rail is subject to service loading, stresses from train–track interaction, thermal
stresses, and fatigue crack growth analyses. The crack growth analyses are approximate in the sense that
only defects in planes perpendicular to the lateral or longitudinal directions are considered, and crack
growth at high R-ratios are not fully known. However, the analyses show that typical crack sizes that can be
found in a weld may grow to failure in very short time if the residual stress field interact with the axle load.
284 A. Skyttebol et al. / Engineering Fracture Mechanics 72 (2005) 271–285
The effect of thermal (residual) stresses on the time to failure is important to include since it has a large
influence on the fatigue life.
Three different crack locations were chosen from observations in the field. The weld analysis seems to be
able to predict the residual stress field with acceptable agreement to the results from the measurements.
Except for the rail head, the rail experiences elastic shakedown after already a few wheel passages. For the
30 metric tonnes axle load used today at the Malmbanan line, the time to fatigue fracture is between 2 and 3
years but this time period may be lower, in particular for welds in cold climate/low temperatures. The
results and conclusions can be summarised as follows:
• The residual stress field from welding causes defects in a weld to grow, thus continuous inspection of
welds is important.
• The time to fatigue failure is strongly dependent on the ambient temperature.
• The time to failure depends on the axle load, and it was found that the largest decrease in fatigue life was
between 30 and 35 metric tonnes.
• Elastic shakedown occurs in the weld region already after a few wheel passages, but the residual stress
field is still high and is only slightly relaxed, i.e. the mean stress in the weld region is high.
• Surface cracks are more dangerous than embedded cracks in the web-rail head area.
• The size and direction of the crack is of importance, transverse cracks seem to be most likely in the area
where the web meets the rail head.
• Small crack areas of the order of 5–10 mm2 may lead to fatigue failure in a few years if the axle load is
increased to 35 metric tonnes.
Acknowledgements
The work was funded by the Swedish National Centre of Excellence in Railway Mechanics Research
CHARMEC (CHAlmers Railway MECanics) in collaboration with Swedish Industry and Chalmers
University of Technology, and by the Swedish Agency for Innovation Systems (VINNOVA). Banverket
(the Swedish National Rail Administration) is acknowledged for collaboration with measurements at the
welding plant.
References
[1] Sih GC, Tzou DY. Rail-end bolt hole fatigue crack in three dimensions. Theor Appl Fract Mech 1985;3:97–111.
[2] Mayville RA, Hilton PD. Fracture mechanics analysis of rail-end bolt hole crack. Theor Appl Fract Mech 1984;1:51–60.
[3] Fukada Y, Yamamoto R, Harasawa H, Nakonowatari H. Experience in maintaining railtrack in Japan. Weld World 2003;47:123–
37.
[4] Webster PJ, Mills G, Wang XD, Kang WP, Holden TM. Residual stresses in alumino-thermic welded rails. J Strain Anal
1997;32:389–400.
[5] Cheesewright PR. The improvement of thermit weld reliability. In: Proceedings of the seminar ‘‘Rail Technology’’, Nottingham,
UK, September 21–26, 1981. Organized by the British Rail Research and Development Division and the Association of American
Railroads. p. 305–20.
[6] Feddersen CE, Broek D. Fatigue crack propagation in rail steels. In: Stone DH, Knupp GG, editors. Rail steels––developments,
processing and use, ASTM STP 644. West Conshohocken, PA, USA: American Society for Testing and Materials (ASTM)
International; 1976. p. 414–29.
[7] Mutton PJ, Epp KJ, Alvarez E, Lynch M. A review of wheel–rail interaction and component performance under high axle load
conditions. In: Ekberg A, Kabo E, Ringsberg JW, editors. Proceedings of the Sixth International Conference on Contact
Mechanics and Wear of Rail/Wheel Systems (CM2003). Sweden: Vasastadens Bokbinderi AB; 2003. p. 273–8.
[8] Hiensch M, Larsson P-O, Nilsson O, Didier L, Kapoor A, Franklin F, et al. Two-material rail development: field test results
regarding rolling contact fatigue and squeal noise behaviour. In: Ekberg A, Kabo E, Ringsberg JW, editors. Proceedings of the
A. Skyttebol et al. / Engineering Fracture Mechanics 72 (2005) 271–285 285
Sixth International Conference on Contact Mechanics and Wear of Rail/Wheel Systems (CM2003). Sweden: Vasastadens
Bokbinderi AB; 2003. p. 39–47.
[9] Hibbitt, Karlsson & Sorensen. ABAQUS/Analysis user’s manual version 6.4. Pawtucket, RI, USA: ABAQUS, Inc; 2003.
[10] SACC Version 4.1, revision 9. Det norske Veritas. Stockholm, Sweden, 2003.
[11] ASTM E837-99: standard test method for determining residual stresses by the hole-drilling strain-gage method. In: Annual Books
of ASTM Standards, vol. 03.01. West Conshohocken, PA, USA: American Society for Testing and Materials (ASTM)
International; 2000. p. 675–84.
[12] Hottinger Baldwin Messtechnik GMBH, Spider 8 and Catman Express version 3.1 operating manual. Darmstadt, Germany. 2001.
[13] Ringsberg JW, Josefson BL. Finite element analyses of rolling contact fatigue crack initiation in railheads. In: Proceedings of
IMechE, Part F. J Rail Rapid Transit 2001;215:243–59.
[14] Runesson K, Skyttebol A, Lindgren LE. Nonlinear finite element analysis and applications to welded structures. In: de Borst R,
Mang HA, editors. Comprehensive structural integrity.Ritchie RO, Karihaloo M, editors. Numerical and computational methods,
3. UK: Elsevier; 2003. p. 255–320.
[15] Madi VN. Transformation behaviour and weldability of alloy rail steels, Doctoral thesis. Illinois Institute of Technology, Chicago,
Illinois, USA. 1988. p. 253.
[16] Skyttebol A, Josefson BL. On the influence of residual stresses after flash-butt welding on the fatigue of railway rail. Research
Report 2003:1, Department of Applied Mechanics, Chalmers University of Technology, G€ oteborg, Sweden, 2003 [to appear in
Proceedings of Numerical Analyses of Weldability in Seggau-Graz, Austria, 2003].
[17] Ringsberg JW, Bjarnehed H, Johansson A, Josefson BL. Rolling contact fatigue of railway rails––finite element modelling of
residual stresses, strains and crack initiation. In: Proceedings of IMechE, Part F. J Rail Rapid Transit 2000;214:7–19.
[18] Ringsberg JW, Lindb€ack T. Rolling contact fatigue analysis of rails including numerical simulations of the rail manufacturing
process and repeated wheel–rail contact loads. Int J Fatigue 2003;25:547–58.
[19] Johnson KL. Contact mechanics. Cambridge, UK: Cambridge University Press; 1985.
[20] Ringsberg JW. Life prediction of rolling contact fatigue crack initiation. Int J Fatigue 2001;23:575–86.
[21] ICON (Integrated Study of Rolling Contact Fatigue). Final Technical Report. European Rail Research Institute, Utrecht, The
Netherlands. 2000.
[22] Dowling NE. Mechanical behaviour of materials. 2nd ed. Upper Saddle River, NJ, USA: Prentice Hall; 1999.
[23] Moller R, Mutton PJ, Steinhorst M. Improving the performance of aluminothermic rail welding technology, through selective
alloying of the rail head. In: Proceedings of the Seventh International Heavy-Haul Conference, Brisbane, Australia, June 10–14,
2001. p. 8.