Engineering Failure Analysis 11 (2004) 151–166

www.elsevier.com/locate/engfailanal

Failure modes in aluminothermic rail welds under high axle

load conditions
P.J. Muttona,*, E.F. Alvarezb
a
Institute of Railway Technology, Department of Mechanical Engineering, PO Box 31, Monash University, Victoria 3800, Australia
b
BHP Billiton Iron Ore, PO Box 231, Port Hedland, WA 6721, Australia

Received 1 July 2002; accepted 21 May 2003

Abstract
Aluminothermic welding of rails is widely used within the railway industry for in-track welding during re-rail and
defect replacement. The process provides flexibility and low capital cost, but suffers from variable quality in finished
welds, due to the inherent limitations of the processes used, and their operator dependency. The service performance of
such welds can be considered in terms of the integrity of the joint, i.e. its ability to support the service loads without
fatigue damage or fracture, and the batter behaviour of the running surface, which in turn influences the extent of
impact loading. Under high axle load conditions, these performance requirements are unforgiving in terms of weld
quality and material characteristics. In response to recent failures in aluminothermic welds, and in recognition that such
welds represent one of the main risks for a catastrophic derailment, and a major limitation to further increases in axle
loads, a program was undertaken to both remedy the current problem with premature failures, and to develop improved
rail welding and track maintenance practices which would meet the performance demands of higher axle loads.
# 2003 Elsevier Ltd. All rights reserved.

1. Introduction

The heavy-haul railroad operations at BHP Billiton Iron Ore (BHPB-IO) comprise over 650 route kilo-
metres for the transport of iron ore from several mine-sites in the Pilbara region of Western Australia to
processing and ship-loading facilities at Port Hedland. Current haulage rates are approximately 75 million
net tonnes per annum, at nominal axle loads of up to 37.5 t. Train configurations comprise distributed
power (locotrol) consists of two or three 110-car rakes, i.e. train lengths of 220 or 330 cars.
The majority of the above track is laid on concrete sleepers, using 68 kg/m, high strength rails. These are
flashbutt-welded into 400 m strings, which are subsequently joined, in track, using aluminothermic welding.
Aluminothermic welding is also used for replacement of defective or broken rail (or welds), installation
of insulated rail joints, and track construction activities. A short preheat aluminothermic welding process
with a portion hardness of 340–360 HB is preferred, to provide hardness levels which are comparable with

* Corresponding author.
E-mail address: [email protected] (P.J. Mutton).

1350-6307/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfailanal.2003.05.003
152 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166

the parent rail. Of particular concern are the hardness levels in the fusion zone, and the width of the heat-
affected zones (HAZ’s). Welds that exhibit low fusion zone hardness levels or wide HAZ’s are susceptible
to batter (localized plastic deformation due variations in strength or hardness across the weld), which
results in increased impact loading and may contribute to premature failure of the weld(s). The extent of
weld batter can be limited by rail grinding, which is carried out as part of a preventative maintenance
strategy to control surface-initiated rolling contact fatigue (RCF) damage and maintain the desired rail
profiles. Rail grinding is not specifically carried out to control weld batter.
Two recent developments have prompted a re-appraisal of the performance of aluminothermic welds
under high axle load conditions; these were:

an increasing number of premature failures in relatively new welds, in both tangent and curved
track, and;

an assessment, using a failure mode and effects analysis (FMEA), of the consequences of further
increases in axle loads to 40 tonnes on component performance.

Notwithstanding the necessity to address the cause(s) of premature failures, the FMEA approach pro-
vided an opportunity to examine the long-term viability of aluminothermic welding procedures under
heavy haul conditions such as those at BHPB-IO. The program that was subsequently undertaken
involved:

examination of failed welds, to identify failure modes and likely contributory factors;

a review of alternative welding processes, and in-track welding procedures;

measurement of the loading conditions at, and the response of, aluminothermic welds under
representative track conditions;

a re-appraisal of the laboratory testing techniques used to assess the performance of aluminothermic
welds under high axle loads; and

implementation of improved inspection procedures for new welds.

Aluminothermic welding consumables are available in Australia from two main suppliers, Thermit Aus-
tralia Pty Ltd (part of the Elektro-Thermit Group of Germany), and Railtech Australia Ltd (part of the
French Railtech Group). The program covered welding processes available from these two manufacturers.

2. Failure characteristics

During the 18 month period to June 2001 failures in Thermit SkV-F aluminothermic welds comprised
approximately 75% of all broken rail reports for the Newman mainline. Of these, the majority occurred in
newer welds, i.e. welds that had been in track less than 6 months. The tendency for early failures in alu-
minothermic welds is consistent with historical data [Fig. 1(a)]. Monitoring of weld failure rates as a
function of time in service therefore provides a clear indication of the quality of new welds, and hence any
variation in welding procedures or the quality of welding consumables.
A review of weld defect and failure statistics over a three-year period to October 2000 also indicated an
increase in the number of ultrasonic defects in aluminothermic welds, coinciding with periods during which
the number of new welds decreased [Fig. 1(b)].
The two main failure types were straight-break or vertical fractures either at the weld centreline or at the
edge of the weld collar or reinforcement, and horizontal split-web or ‘‘S’’ fractures starting in the web [1].
Most of the failures were of the horizontal split-web type, which present a greater derailment risk than
straight-break failures.
 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166 153

Fig. 1. Failure statistics for aluminothermic welds—September 1997–October 2000.

2.1. Straight-break failures

The majority of straight-break failures were found to be associated with centreline shrinkage defects in
the foot [Fig. 2(a)], which initiated a vertical fracture at the centreline of the weld. Examination of failed
welds and an audit of welding procedures indicated that the most likely cause of the defect was insufficient
preheat. However test welds manufactured in accordance with the consumable supplier’s recommended
procedures were also found to exhibit some centreline shrinkage, albeit of smaller size than in welds that
had failed in track. This indicated that the process was inherently prone to centreline shrinkage when used
with the 68 kg/m rail section.
Straight-break failures also occurred as a vertical fracture at the edge of the weld collar, initiating in the
upper web region (100–110 mm above the rail foot), at the surface of the web [Fig. 2(b)]. The initiation
region coincided with a widening of the fusion zone, in which weld metal extended some 10–15 mm beyond
154 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166

Fig. 2. Straight break failure modes.

the weld collar. Inclusions present at the initiation site in some, but not all, failed welds were identified as
slag arising from the portion reaction.
Straight break failures were more likely to occur in the colder months, when rail temperatures fell below
the stress-free temperature. An additional factor was the distribution of residual stresses in the web. A
heat-blocking procedure was used at one stage to reduce the probability of horizontal split-web failures, by
reducing the magnitude of residual stresses in the vertical direction. However incorrect heat-blocking pro-
cedures were found to increase the magnitude of longitudinal residual stresses in the vicinity of the weld
collar, increasing the risk of tensile failures in this region. The use of the heat-blocking procedure was
subsequently discontinued.
 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166 155

2.2. Horizontal split-web fractures

Horizontal split-web (HSW) or ‘‘big dipper’’ failures are a relatively common failure mode in both
flashbutt and aluminothermic welds in the Pilbara heavy haul railroads [2,3]. The factors that contribute
to this failure mode are well understood [4], with most failures occurring in curves of 600–900 m radius,
or in tangent track sections prone to vehicle hunting, due to the cyclic nature of torsional stresses in the
rail web.
Most recent horizontal split-web fractures showed a weld defect such as large inclusions on the fracture
surface, at the edge of the weld collar, and no evidence of fatigue crack growth (Fig. 3). This indicated that
the fractures started at the defect and progressed rapidly, possibly under loading from one or two wheels.
Examination of the inclusions in several failed welds showed that it was mainly alumina (aluminium
oxide) slag arising from the portion reaction.
The examination of failed welds therefore indicated that:

the short-preheat welding process was prone to centreline shrinkage, even in welds produced using
recommended procedures;

Fig. 3. Horizontal split web failure mode.

156 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166

optimising weld characteristics in terms of the hardness distribution (to minimise weld batter) could,
under some circumstances, compromise the structural performance of the weld and increase the
probability of weld failure;

weld quality was variable, with inclusions in the web (at the edge of the weld collar) of particular
concern; and

the loading conditions to which welds were subjected were more severe than anticipated, when
compared with previous data for high axle load operations [2,3].

3. Review of welding processes and procedures

Auditing of in-track welding procedures is carried out on a regular basis. In addition, there is an ongoing
program to monitor and assess developments in rail welding procedures that may be applicable to the
railroad operations.
One important aspect of the aluminothermic welding process is the crucible design. Historically a long-life
(re-usable) crucible has been used; however single-use crucibles have been developed by both consumable
manufacturers. The single-use crucible eliminates the need for cleaning of the crucible between consecutive
welds, and hence should improve weld quality.
A single-use crucible welding procedure had previously been approved for in-track use. Over 200 of these
welds had been installed, with relatively few defects and no failures up to the present time. However the
process was no longer available.
Recent developments in aluminothermic welding procedures included an improved short preheat, Triple
Riser (TR) process from Thermit Australia Pty Ltd, and the single-use crucible Boutet PLK process from
Railtech Australia Ltd. The Thermit TR process was initially available only with the long-life crucible,
although a single-use crucible version has since been developed. Both of these processes were subjected to
an initial (laboratory) assessment program, with emphasis on the following characteristics:

the hardness distribution as measured by a longitudinal traverse immediately below the running
surface of the rail;

quality and cleanliness of the weld metal, in particular the presence or otherwise of inclusions in the
web, shrinkage defects in web and foot, and surface imperfections such as hot tears in the web;

tensile properties of the weld metal, in particular those relating to catastrophic failure in an overload
situation;

the distribution of residual stresses in the web, and the magnitude of vertical stresses developed
under simulated service loading conditions; and

horizontal split-web fatigue behaviour.

The hardness distributions obtained in Thermit TR and Boutet PLK welds were similar, with an increase
in HAZ width compared with the SkV-F process (Fig. 4). This indicated that weld batter would be more
severe than with the SkV-F process, although this could be offset by using a slightly harder portion (subject
to availability from the manufacturer(s)). Previous in-track test programs had identified a minimum fusion
zone hardness of 340 HB in order to control weld batter to acceptable levels. The width of the HAZ could
be reduced slightly by using one of the low-alloy, heat treated rail grades currently available.
Fig. 5 compares the various weld types in terms of the level of residual stress levels in the web, and the
combined effect of residual stresses and those induced by application of an eccentric vertical load on the rail
head (to simulate horizontal split-web behaviour). The data are for vertical stresses measured at the weld
centreline. Residual stress levels are lower in the Boutet PLK process, although the shape of the weld collar
results in higher stress levels in the lower web region under eccentric vertical loading.
 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166 157

Fig. 4. Longitudinal hardness distribution in aluminothermic welds.

Fig. 5. Distribution of vertical stresses at weld centerline.

Previous studies of the horizontal split-web behaviour of aluminothermic welds [2,3] have made use of
fatigue crack propagation data such as the threshold stress intensity factor range (
Kth) or fracture
toughness (KIC) to determine the stresses required for fatigue crack initiation from a pre-existing defect, or
weld fracture, respectively.
pffiffiffiffi For an overload failure, the relevant parameter is the fracture toughness, with a
range of 25–35 MPa m typical of aluminothermic weld metal [5–7]. For the current program, the sensi-
tivity of the weld metal to surface defects such as inclusions or shrinkage cracks was examined by means of
notched tensile tests, using a procedure based on the ASTM standard [8]. Specimens were prepared from
the web of the weld, and a small notch machined into the gauge length to simulate an inclusion at the edge
of the weld collar. A comparison of the tensile strength with corresponding data for un-notched specimens
provided an indication of notch sensitivity. Table 1 summarises typical data for the SkV-F, TR and PLK
processes. For the SkV-F and TR processes in the un-notched condition the mean UTS was 805 MPa,
reducing by an average of 14% in the notched condition. PLK welds exhibited lower un-notched UTS, but
did not break at the notch when tested in the notched condition.
158 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166

Table 1
Tensile properties of weld metal

Weld type 0.2% P.S. (MPa) El. (%) UTS (MPa)

Un-notched Notched

SkV-F 787 1 809 667

TR 745 2 802 723
PLK – 0.5 642 650a
a
Did not fail at notch.

Test welds were also subjected to horizontal split-web fatigue tests [9], during which a vertical load of 200
kN was applied at an eccentricity of 22.5 mm from the centreline of the head to simulate torsional loading
on the web. Fig. 6 shows the arrangement of the test rig. Welds were required to complete 5 million cycles
without failure.
Welding procedures that were approved on the basis of the laboratory test program were subsequently
used for in-track welding. This initially involved a limited number of trial welds, during which welding
gangs were trained in the new procedures. Use of the newer procedures was subsequently expanded, subject
to the availability of welding consumables. The Thermit TR process (using long-life crucibles) was initially
introduced for all in-track welding in the Newman mainline, with the exception of step- or junction-welds
involving rails of differing head losses. However there were some ongoing concerns with weld quality, with
several failures associated with inclusions or shrinkage cracks at the edge of the weld collar.
Following initial (successful) trials of Boutet PLK welds, this process was subsequently adopted for all
in-track welding requirements. An additional test program is currently underway to examine the fatigue
behaviour of PLK stepped welds with rail height differences in the range 5 to 14 mm.

Fig. 6. Test configuration for horizontal split web fatigue.

 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166 159

4. Weld instrumentation

The initial examination of failed welds indicated a rapid fracture mode, with little or no evidence of
fatigue crack growth. This aspect, plus the relatively high number of horizontal split-web failures in tan-
gent track, indicated that the loading conditions at welds were more severe than anticipated. An instru-
mentation program was therefore undertaken to quantify these loading conditions and the resultant
response of the welds, taking into consideration the following aspects:

track location (curved and tangent track);

wheel and rail profiles;

the longitudinal profile at the weld(s), including peaking or dipping associated with alignment of the
rail ends during the welding procedure, and localised dipping attributable to the hardness differential
across the weld(s);

vehicle speed;

wheel tread irregularities (wheel flats).

The instrumentation layout included strain–gauge load bridges at the weld(s) and parent rail, in order to
measure vertical loads at both locations and derive impact factors for the weld, and single gauges for the
measurement of vertical and longitudinal stresses in the web and foot of the weld(s). Measurement of
longitudinal profiles at the welds was by means of a laser-based dip gauge.
Fig. 7 shows the relationship between dynamic impact factor and train speed for a dipped and peaked
weld. Dynamic impact factors were generally 41.6, and decreased with reducing train speed below 50 kph.
Fig. 8 shows the relationship between the depth (or height) of the irregularity at the weld and the
dynamic impact factor at mainline speed (70–75 kph). Poor support under dipped welds (ballast pumping)
contributed to increased impact loading, compared with equivalent fully-supported welds.
The instrumentation program enabled the response of the different weld collar shapes to be compared
under in-track loading conditions. Use was made of vertical stresses measured at the weld centreline, 100
mm above the rail foot; this position coincided with the initiation site for horizontal split-web failures.
Fig. 9 shows the web stress response of SkV-F and TR welds in tangent track, compared with the
equivalent position in parent rail. The SkV-F weld collar shape resulted in an apparent increased sensitivity
to vertical load eccentricity compared with the TR weld.

Fig. 7. Dynamic impact factor vs train speed (tangent track).

160 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166

Fig. 8. Dynamic impact factor at mainline speed vs weld irregularity size.

Fig. 9. Comparison of web stress response in SkV-F and TR welds (tangent track).

The magnitude of vertical stress levels in the rail web, at the weld centreline, did not correlate with the
magnitude of the vertical load (Fig. 10). Vertical stresses in the region of the rail (or weld) are influenced to
a large extent by the loading position on the rail head, with contact towards gauge or field sides of the rail
head promoting increased stress levels. By comparison, longitudinal stress levels in the rail foot show a
linear relationship to the magnitude of the applied vertical load (Fig. 11).
The vertical load data shown in Figs. 9 and 10 also indicated that peak loads approaching 400 kN (i.e. twice
the nominal static load) occur when wheels with tread irregularities (wheel flats) traverse the dipped welds.
Vertical stresses of > 150 MPa tension were measured in the rail web, at the weld centreline, in both
tangent and curved track locations. Maximum tensile stresses of up to 250 MPa occurred in some instances
in curved track. At these stress levels the probability of overload failure (horizontal split-web mode) was
significantly higher, as indicated by a linear-elastic fracture mechanics analysis for this failure mode
(Fig. 12).
The tendency for eccentric loading of the rail head was influenced by wheel profile condition (and hence
service interval since re-profiling), bogie condition, and rail profile. Revisions to the low rail profile in the
 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166 161

Fig. 10. Net torsional stress in rail web, at weld centreline vs vertical load.

Fig. 11. Mean longitudinal stress in rail foot vs vertical load.

Fig. 12. LEFM relationship between defect size and fracture stress for horizontal split web failure mode.
162 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166

sharper (600–800 m radius) curves, to shift the contact band away from the field side of the head, resulted in a
decrease in torsional loading on the rail web, and hence reduced tendency for horizontal split web failures.
Apart from residual stresses and those resulting from wheel loading, welds are subjected to longitudinal
stresses resulting from two sources, viz:

temperature fluctuations relative to the stress-free or neutral temperature (at which the rail was
initially welded); and

locomotive traction and braking forces.

Track locations that are subjected to repetitive changes in the level of tractive effort from locomotives
may develop higher levels of rail longitudinal stress, as will rapid changes in rail temperature during the
onset of colder weather. These stresses increase the probability of straight-break failures at welds. A pro-
gram was undertaken to assess the distribution of stress-free temperatures in mainline, in conjunction with
routine monitoring of rail temperature and rail end gap data whenever track is cut for the installation of
new welds. Results from the latter measurements indicate that in some track sections the stress-free tem-
perature may be considerably higher than the recommended value. Procedures for the management of rail
longitudinal stresses have since been upgraded.

5. Inspection procedures

5.1. Weld quality

Non-destructive inspection of aluminothermic welds has historically involved a one-off manual ultra-
sonic testing of new welds following installation, and automated testing as part of the regular rail flaw
detection program. The equipment used for such inspections has recently been upgraded [10], providing
improved capability to detect discontinuities such as inclusions in aluminothermic welds. However ultra-
sonic inspection has not proved capable of detecting defects such as centreline shrinkage in the foot of
aluminothermic welds.
A program of radiographic inspection was therefore initiated, with emphasis on all recently-installed
welds. This program has subsequently continued to include all new welds immediately after installation,
and selected testing of older welds for comparison purposes. The radiographic inspection procedure uses a
gamma-ray source, and covers the web and both sides of the rail foot. The technique is capable of dis-
criminating between shrinkage defects and slag inclusions, both of which can be rated in terms of size (length
of shrinkage defect or diameter of inclusions) and number. Calibration of the radiographic data has been
carried out, where possible, by sectioning and examination of welds that have been removed from track.
Initial results for the SkV-F process indicated approximately 50% of the newer welds (less than 12
months old) exhibited centreline shrinkage (CLS) in the foot of the weld. The corresponding proportion in
older welds was somewhat lower, reflecting the fact that welds with unacceptable levels of CLS may have
since failed and been removed from track. By comparison, the TR and PLK process welds exhibit negli-
gible CLS defects.
Both straight-break and split-web failure types have been associated with the presence of inclusions at
the edge of the weld collar. A significant advantage of the radiographic testing has been the ability to detect
and size these inclusions, although it was not possible to locate these in the transverse direction (through
the web thickness). Inclusions located at the edge of the weld collar are considered more detrimental in
terms of potential crack initiators. Approximately 75% of the short-preheat welds exhibited significant
inclusion levels in the web, while the TR and PLK weld types have lower inclusion numbers. The single-use
crucible procedures, in particular, are capable of providing much cleaner welds.
 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166 163

The use of both ultrasonic and radiographic inspection for new aluminothermic welds increases the cost
of new welds, although this can be offset against the reduced risk of weld failures and disruption to rail
traffic. Nevertheless, developments in NDT techniques are being monitored in order to identify procedures
that offer improved defect detection capabilities and reduced inspection costs.

5.2. Rail impact loads

One outcome of the in-track weld monitoring program is evidence of variability in the longitudinal
profile of new welds, due to the combined effects of incorrect alignment of the rail ends prior to welding,

Fig. 13. Spalled aluminothermic weld.

164 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166

Fig. 14. Accelerometer response for spalled aluminothermic weld.

and subsequent weld batter. This has immediate consequences in terms of increased impact loading at
welds, and associated deterioration (ballast breakdown) in the underlying track structure. The severity of
impact loading at welds is currently being assessed, on a continuous basis, using an ore car fitted with an
instrumentation package that includes accelerometers mounted above the bearings.
Fig. 13(a) shows an example of an aluminothermic weld that was slightly peaked during installation, and
subsequently developed localised batter and spalling in the fusion zone. The corresponding longitudinal
profile along the centreline of the rail head is shown in Fig. 13(b), with spalling to a depth of approximately
2 mm.
Fig. 14 shows acceleration data for the above weld, obtained from the instrumented ore car. The data are
for loaded and empty traffic, with slightly higher acceleration levels occurring in the loaded condition, due
to slight differences in the unsprung mass of the wheelset. The effectiveness of rail grinding is also shown,
with peak acceleration levels in the loaded condition decreasing from 25 g before grinding, to 20 g.
Measurement of vertical loads at this weld, after grinding, averaged 300 kN for the loaded traffic, with
peaks of up to 380 kN (i.e. of similar magnitude to that obtained when a wheel flat impacts a dipped weld).

6. Summary

The increased incidence of defects and failures in aluminothermic welds in the Newman mainline and
associated tracks at BHPB-IO has highlighted the inherent limitations of this welding process as axle loads
approach 40 t. The current approach to testing and approval of aluminothermic welding procedures
therefore addresses the following aspects:

is the process technically sound, i.e. is it capable of producing consistent and high quality welds with
the desired performance characteristics if the recommended procedures are followed?

is the process inherently robust, i.e. is the quality of finished welds dependent on the operator(s)
attention to following recommended procedures such as setting up of the rail ends, gap widths and
preheat conditions?

Welding processes which incorporate a single-use crucible offer significant advantages, not only from the
perspective of improved weld quality, but also because the process is easier to use and hence more attrac-
tive to the welding crews.
 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166 165

Since the change-over to the TR process, and subsequently to the PLK process, the incidence of
failures in newer welds has decreased to negligible levels. No failures have occurred to date in PLK
welds.
The recent improvement in defect and failure rates in newer aluminothermic welds has provided an
indication of the success of the current approach to assessing and approving welding procedures for in-
track use. However the question still remains as to the long-term viability of aluminothermic welding of
rails as axle loads approach or exceed 40 t. The laboratory test program, which covers mechanical
properties, distribution of residual stresses and stresses developed under simulated service loads, and
fatigue performance, provides an indication of the maximum loading conditions or stress levels that the
welds can sustain without increasing the risk of failure. The in-track instrumentation program has
demonstrated that the loading conditions at aluminothermic welds are severe, with the geometry of the
weld, the longitudinal and transverse profiles at the running surface, and rail and/or sleeper support
conditions influential in determining the level of impact loading and the resultant stress levels in the
weld(s). In particular, wheel-rail contact conditions that promote eccentric loading of the rail head, in
conjunction with lateral loads arising from bogie steering forces or vehicle response(s) to track geometry
errors increase web vertical stresses to unacceptable levels. These aspects require attention if alumi-
nothermic welding of rails is to be continued as the primary in-track welding procedure for heavy haul
conditions

Acknowledgements

This paper is based on a similar paper presented at the 13th International Rail Track Conference (G.A.
Offereins and P.J. Mutton (2001), Recent Experiences with the Performance of Aluminothermic Rail Welds
under High Axle Loads, Rail Track Association Australia), which has been revised to include further
developments in aluminothermic welding procedures and weld performance at BHP Billiton Iron Ore.
The support and/or contribution of the following organisations is gratefully acknowledged:

Thermit Australia Pty Ltd;

Railtech Australia Ltd;

Track Maintenance, BHP Billiton Iron Ore;

Rail Technology International Pty Ltd;

Institute of Railway Technology, Monash University.

References

[1] Offereins G. BHP reacts quickly to a spate of weld failures. International Railway Journal 2001;41:15–16.
[2] Dudley N, Oswald S, Vines MJ. Rail welding for heavy axle load unit train operations, Proc. Joint Australasian Welding and
Testing Conference (Perth), 1994, Australian Welding Institute/AINDT.
[3] Mutton PJ, Soeleiman S. Performance of aluminothermic welds under high axle loads, Proc. 4th International Heavy Haul
Railway Conference, Brisbane, 1989, IEAust, pp. 180–7.
[4] Mutton PJ, Horizontal split web fatigue failure of rail welds, Proc. Conf. Failure Analysis in Materials Engineering, Institute of
Metals and Materials Australasia (IMMA) Melbourne, 1994.
[5] Mutton PJ, Moller RH. Improved aluminothermic rail welding technology, Proc. 10th International Rail Track Conference,
Brisbane, 1994, Rail Track Association, pp. 145–51.
[6] Moller R, Mutton P, Steinhorst M. Improving the performance of aluminothermic rail welding technology, through selective
alloying of the rail head, Proc. 7th International Heavy Haul Conference, Brisbane, 2001, IHHA, pp.331–8.
[7] Lonsdale CP, Lewandowski JJ. Fracture toughness of thermite rail weld steel, Proc. 39th Metal Working and Steel Processing
Conference, 1998, Vol 35, pp. 1083–90, Iron & Steel Society.
166 P.J. Mutton, E.F. Alvarez / Engineering Failure Analysis 11 (2004) 151–166

[8] American Society for Testing and Materials. Standard test method for sharp-notch testing with cylindrical specimens, ASTM
E602-91, 1997.
[9] Standards Australia, AS1085.15: Railway permanent way material, Part 15: Aluminothermic rail welding, 1995.
[10] Griffiths D, Ivachev A. Rail reliability improvement by two orders of magnitude reduction in critical defect detectability, Proc.
7th International Heavy Haul Conference, Brisbane, 2001, IHHA, pp. 403–9.