Track information
Queensland Rail (QR) managed the railway where the derailment occurred, with the movement of rail traffic controlled from the QR Control Centre located at Townsville in Queensland.
The narrow gauge (1,067 mm) track at the derailment location consisted of 47 kg/m rail fastened to concrete sleepers by resilient clips with ballast to a nominal design depth of 200 mm. Heading west, approaching the derailment site, the track was largely tangent over undulating terrain.
Track condition
QR inspected the track regularly as part of its maintenance regime and no track anomalies were found that might have contributed to the broken axle.
QR applied a standard formula, known as track condition indices (TCI), to determine the overall condition of the track. The TCI was calculated by adding the condition index values of four track geometry parameters - top, twist, gauge and versine.[2] The resultant number represents the condition of one length of track. Weighted averages for all TCIs over a line section are known as overall track condition index (OTCI). Figure 3 shows the OTCI for the line between Charters Towers and Pentland. The lower the score, the better the general condition of the track. The figure shows a consistently low index, below the designated maintenance intervention level of 58 (red line).
Figure 3: Charters Towers to Pentland section OTCI

Source: Queensland Rail
The OTCI of the Mount Isa line, (Figure 4) from June 2011 until December 2017 similarly identified that the general condition of the track was below the intervention level.
Figure 4: Mt Isa line OTCI

Source: Queensland Rail
Train and crew information
Train 9T90 was a freight service operated by Australia Eastern Railroad (Aurizon) between Sun Metals Townsville and Phosphate Hill. It consisted of Aurizon locomotives (2832 leading) hauling 52 freight tanker wagons (GATX strings 6, 8, 11, 10). Incitec Pivot Limited (IPL) owned the GATX freight tanker wagons and the consignment. The train was about 730 m in total length and had a mass of about 4296 t. The consignment contained dangerous goods, including about 1.67 million litres of sulphuric acid.
The train was crewed by two drivers. Both drivers commenced work at Townsville at about 0245 on 28 September 2017. They were to take control of train 9T90 at Townsville and drive through to Hughenden (Figure 1), where they would finish their shift. After the incident, the train crew submitted to drug and alcohol testing and returned zero readings.
Rolling stock – locomotives
Aurizon was an accredited Queensland rolling stock operator. It owned and operated the diesel electric locomotives hauling train 9T90. Aurizon was providing a hook and pull service[3] to IPL for train 9T90 at the time of the derailment.
The locomotive was fitted with Ultra High Frequency (UHF) train control radio and a GPS. The GPS system enabled the monitoring of the locomotive location and speed by the NCO at the QR train control centre in Townsville. The following driver aids were also available:
- station protection device
- vigilance control system
- automatic train protection
- direct traffic control system.
Locomotive 2832 was also fitted with a data logger, which recorded various parameters including:
- time
- GPS position, speed and distance travelled
- throttle position
- driver vigilance
- motor current
- air reservoir/brake cylinder pressures.
There were no anomalies identified in the train speed, handling, or operational performance leading up to the derailment.
Rolling stock – tanker wagons
IPL was an accredited Queensland rolling stock operator and operated the United States‑designed, Australian‑made, GATX tanker wagons.
The tanker fleet comprised 145 wagons, classified as OSZY class wagons. The fleet operated as 11 13-wagon strings plus two spare wagons. A product hose interconnected each tanker wagon within a string. This configuration enabled the stabling of the strings at Phosphate Hill and decanting of product as required.
Each wagon, including the subject wagon OSZY44795, consisted of a tank mounted on two bogies with tare weight 22.26 t, gross weight of 80.66 t, and 12.9 m overall length. Each bogie consisted of two wheel sets. A wheel set was made up of two wheel discs, two roller bearings, and a solid axle shaft. The wheel discs were pressed onto the axle wheel seats and retained by an interference fit. Similarly, the bearings were pressed onto the axle bearing journals and retained by an interference fit.
Axles
The solid steel axles of the GATX fleet, designated as 840P1, are 1,879 mm long with a central barrel tapered from 154 mm in the centre to 170 mm towards the wheel seats, (Figure 5). Other operators also used rolling stock configurations with 840P1 axles.
The 840P1 axles are designed in accordance with Australian Standard AS1448, Carbon steel and carbon-manganese steels—Forgings (ruling section 300 mm maximum). The standard specifies minimum requirements in terms of chemical and mechanical properties of the steel.
Figure 5: 840P1 axle

Source: Incitec Pivot Limited, annotated by ATSB.
Wayside equipment
Wayside equipment was installed at several locations on the occurrence line. The equipment included dragging equipment detectors, overload imbalance load detectors, hot wheel and hot bearing detectors.[4]
A review of data sourced from those detectors did not reveal a condition with wagon OSZY44795 that contributed to the axle failure, or subsequent derailment.
GATX fleet maintenance
The IPL maintenance program included three levels of periodic wagon inspection:
- Level one – a basic walk-around inspection that occurred at Mount Isa, Phosphate Hill and Townsville
- Level two – An annual visual inspection undertaken at the maintenance facility in Townsville
- Level three – a seven yearly inspection, including reline, undertaken at the maintenance facility in Townsville.
IPL engaged United Group Limited (UGL) to perform certain maintenance on the GATX fleet, which in turn engaged Aurizon for wheel set maintenance.[5] IPL provided the ATSB with records detailing the inspections undertaken for the previous 12 months. There were no defects noted by IPL that may have contributed to the axle failure.
Wheel sets
In addition to the periodic inspections, IPL had a planned maintenance program for the GATX fleet. This program included procedures for operation and maintenance of wheels sets, axles, and bearings.[6] In particular, the procedure detailed the operation and maintenance of axles, the relevant section being:
4 Operation and Maintenance - Axles
• Magnetic particle inspection of axle wheel seats and bearing journals including transition radii shall be performed whenever the wheels are removed from the axle.
• More frequent visual and ultrasonic inspection shall be carried out if necessary due to service conditions and axle designs.
• Axle wheel seats found to contain cracks less than 3mm long may be reclaimed by machining. Otherwise, cracked axles shall be scrapped.
• Visual inspection of the bearing journal fillets for corrosion, dents and cracking shall be performed whenever bearings are removed from the axle.
• Prior to inspection, axles must be cleaned and particular care should be taken when inspecting the critical zone, 300mm either side of the centreline. Axles grooved or gouged more than 3mm in depth must be condemned. All nicks, scratches or stampings, less than 3mm in depth must be machined or ground to a smooth contour.
Aurizon, the contracted wheelset maintainer, inspected and overhauled the GATX wheel sets at their wheel shop facility in Rockhampton, Queensland. Aurizon used their standards, procedures, and work instructions detailed in the Aurizon Incoming Inspection Work Instruction, in conjunction with IPL requirements, to inspect and overhaul wheel sets.
The Aurizon inspection process required wheel sets to be:
- cleaned
- inspected, measured, tested, and recorded
- components repaired / replaced (if required)
- machined (if required)
- reassembled
- condemned (if required).
Aurizon last inspected and overhauled the failed wheel set on 7 and 8 June 2017. The inspection included ultrasonic and magnetic particle inspection of the axle. During the inspection process, observations were noted on the Wheelset Incoming Inspection Form. The form noted that the axle passed both ultrasonic and magnetic particle inspections. The form also noted that the wheel set had new bearings fitted and both wheel treads machined. The wheel set was certified to re-enter service and subsequently installed under wagon OSZY44795. Wagon OSZY44795 travelled about 32,000 km prior to the axle failure.
Post incident examination
Material failure analysis
Failure analysis of both parts of the wheel set was conducted at the Queensland University of Technology (QUT) Central Analytical Research Facility in Brisbane under the supervision of the ATSB.
That examination identified that the axle fractured about 341 mm and 526 mm from the in-board side of each wheel disc, (Figure 6).
Figure 6: Fracture location on axle barrel

Source: QUT Central Analytical Research Facility.
QUT noted that the fracture surfaces were intact with minimal secondary damage. There was a 70-mm long fine circumferential white line present at the fracture initiation region at a depth of up to about 3 mm (Figure 7).
Figure 7: White paint on fracture surface near origin

Source: QUT Central Analytical Research Facility.
A scanning electron microscope (SEM) with an energy dispersive spectrometer (EDS) was used to determine the elemental composition of the white line. The SEM EDS analysis indicated the white line residue was rich in titanium, a pigment used in white paint. During the magnetic particle inspection process, white paint is applied to the axle barrel.
Both bearings, from the failed axle, appeared to be in a similar condition with their outer races having sustained secondary damage, most likely as the result of axle failure. The raceways and roller surfaces of the bearings displayed minimal operation-related damage. It was therefore considered likely that the condition of the bearings did not contribute to the axle failure.
Examination of the axle fracture surfaces also identified a single, distinct mechanical notch on the surface of the axle from which faint beach[7] marks originated (Figure 8).
Figure 8: Fatigue origin

Source: QUT Central Analytical Research Facility.
These beach marks indicated propagation of a fatigue crack. The fracture surface was flat and almost perpendicular to the axle barrel’s longitudinal axis. The fatigue fracture face displayed smooth texture indicating fatigue propagation. The fatigue area was relatively large, about 74 per cent of the fracture surface area, penetrating to a critical depth of about 94‑110 mm. Once the crack propagated that far, the remaining section could not support the load and failed rapidly due to overstress.
The mechanical notch measured about 1 mm in length and about 0.2 mm in depth (Figure 9). For this type of axle, the condemning limit is damage extending 3 mm deep or greater. IPL and Aurizon permitted repairs on notches up to 1.5 mm deep in the axle central area, and up to 2 mm deep elsewhere. There was no evidence that the coincident mechanical notch was detected or repaired during the last inspection on 7 June 2017. Although, given the shallow notch depth (0.2 mm), it may have been assessed as not requiring repair.
Figure 9: SEM image showing the initiating notch at the fatigue crack origin

Source: QUT Central Analytical Research Facility.
The microstructure of the axle material appeared banded and consisted predominantly of equiaxed pearlite and ferrite, indicating that it was likely in a normalised condition (correctly manufactured). Manganese sulphide inclusion stringers were observed in the sample. These stringers can contribute to corrosion fatigue, but would have a negligible effect on normal fatigue crack growth.
Non-destructive testing, consistent with the relevant Australian Standards, was conducted on the axle sections. Ultrasonic examination did not detect any major discontinuities within the axle body. Fluorescent magnetic particle inspection found small, isolated inconsequential surface flaws.
Samples taken from the axle were chemically analysed and found to be largely consistent with Australian Standard AS1448/K5.[8] These results were compared to the manufacturer’s test certificate for axle steel batch used during its manufacture in July 1999 (Table 1).
Table 1: Chemical analysis
|
C
|
Si
|
Mn
|
P
|
S
|
Ni
|
Cr
|
Mo
|
Cu
|
AS1448/K5
|
.35-.45
|
.1-.35
|
.5-1.0
|
<.05
|
<.05
|
<.35
|
<.3
|
<.1
|
.35
|
Batch records
|
.37
|
.23
|
.69
|
.013
|
.02
|
.09
|
.1
|
.02
|
.26
|
Failed axle
|
.34
|
.23
|
.38
|
.01
|
.02
|
.08
|
.1
|
.03
|
.25
|
(C) Carbon, (Si) Silicon, (Mn) Manganese, (P) Phosphorus, (S) Sulphur, (Ni) Nickel, (Cr) Chromium, (Mo) Molybdenum, (Cu) Copper
The results showed that carbon was slightly low (0.34 per cent rather than 0.37 per cent) and manganese was low (0.38 per cent rather than 0.69 per cent) in the axle sample. The low carbon reading was negligible and likely did not affect the overall mechanical strength of the axle or notch resistance. Similarly, there is no evidence that the low manganese reading affected the mechanical strength of the axle, (Table 2).
Samples from the failed axle were also mechanically tested and found to meet, or exceed, the minimum mechanical property requirements of the Australian Standard AS1448/K5 (Table 2).
Table 2: Mechanical analysis
|
Australian Standard AS1448/K5
|
Failed axle
|
Yield (MPa)
|
270
|
370
|
Tensile (MPa)
|
540
|
574
|
Elongation (%)
|
16
|
23
|
Magnetic particle inspection
Magnetic particle inspection (MPI) is a non-destructive testing process widely used to inspect ferromagnetic materials for surface cracks. The rail industry commonly uses MPI to inspect axles for cracks. IPL specified the use of applicable standards, including the Aurizon standards and work instructions. The Aurizon document suite also referred to the relevant Australian standards for axles and MPI, which included rail‑related standards.
The Rail Industry Safety and Standards Board is responsible for the development and management of rail‑related Australian Standards, rules, codes of practice and guidelines, all of which have national application. The Australian Standard AS7515:2014 Axles described the requirements for the design, manufacture and maintenance of rolling stock axles to prevent derailments caused by axle failures.
Part 7 of AS7515:2014 described the use of MPI during axle inspection and referred to Australian Standard AS1171:1998 Non-destructive testing – Magnetic particle testing of ferromagnetic products, components and structures. AS7515 stated that a risk‑assessed consideration of:
- previous failures
- service conditions
- high failure consequences
- axle designer recommendations.
may indicate that more frequent inspections were required.
AS1171 specified the requirements of magnetic particle testing for the detection of surface and near-surface discontinuities in ferromagnetic products, components and structures. The standard provided detail on:
- testing personnel requirements
- equipment and materials
- methods of test
- process control procedure and requirements
- test records and reports.
Qualifications
The effectiveness of magnetic particle testing depends on the technical competence of the personnel performing the tests and on their ability to interpret indications, as specified in AS 1171. The testing personnel at Rockhampton were appropriately qualified and medically fit, including meeting the visual acuity requirements, for the task.
Equipment testing
Before use, the alternating current electromagnetic yoke (AC yoke) used for the magnetic particle testing was required to be performance checked using Aurizon Work Instruction – Inspection and Reconditioning of Wheelsets WI/2016017. The work instruction reflected the testing requirements contained in AS1171, which stated that a dead weight and standard test piece were to be used. The dead weight test involved the use of not less than a 4.5 kg sample of mild steel with the AC yoke pole spacing between 75 and 300 mm. A standard test piece with known discontinuities was also specified.
Equipment testing in practice
During a site visit to the Aurizon Rockhampton wheel shop, the ATSB observed that only the dead weight test was conducted. A standard test piece with known discontinuities was not used. Additionally, the local work instructions did not reflect the Aurizon and AS1711 requirements to performance check the AC yoke using the standard test piece.
Figure 10: AC yoke

Source: AS1171 Figure 3.2 (b).
Process
Maintenance personnel at the Rockhampton wheel shop had access to the following documents when conducting MPI:
- Aurizon Heavy Maintenance – Rockhampton, In coming Inspection Work Instruction WI‑RO‑WAB-10-001
- Aurizon Rockhampton Work Instruction – Wheel Shop work instruction for MPI of axles WI‑RO-WAB-10-014
- Aurizon Inspection and Reconditioning of Wheelsets WI/2016017
- Australian Standard AS1171:1998 Non-destructive testing – Magnetic particle testing of ferromagnetic products, components and structures
- Australian Standard AS7515:2014 Axles.
Process in practice
During a site visit to the Rockhampton wheel shop, the ATSB noted that during the inspection/testing process, the magnetic ink media (Ardrox 800/3) was not reapplied in between AC yoke placements, nor was there the positional overlap between each test. Both of these processes were required and detailed in Australian Standard AS1171.
Figure 11: AC yoke measurement

The measured width of the AC yoke on the day of the site visit. Source: ATSB
The AC yoke poles are adjustable and the yoke used at the wheel shop had been set to a distance of about one third the length of the axle shaft. The AC yoke was placed on one end of the axle (perpendicular to the centre line), energy applied to create magnetic flux, checked for potential indications, and then moved to the next third and the process repeated. There was no overlap between consecutive tests observed during the ATSB site visit. Once one side of the axle was inspected, the work instruction specified that axle be rotated through 90° and the process repeated until the entire axle was tested (Figure 12). During the site visit, the ATSB noted that the axle was rotated 120° rather than 90°. This equated to nine tests for the axle instead of the required 12.
Figure 12: AC yoke placement

Source: Aurizon Heavy Maintenance – Rockhampton
Axle failures on the Mount Isa line
Between 2008 and 2013, 41 main line derailments occurred on the Mount Isa line. Due to the significant number of derailments, the then-rail regulator, Queensland Department of Transport and Main Roads (TMR) rail regulation unit, undertook a study titled Mount Isa Derailment Analysis 2008 -2013.[9] The study examined the causes with the aim of reducing their frequency and increasing the availability and capacity of the corridor. In October 2015, the rail regulation unit published the report.
The report focused on the operating practices of the:
- rail infrastructure manager in terms of incident prevention, maintenance, repair and upgrade
- rolling stock operators who used the Mount Isa line with respect to the age and suitability of rolling stock, inspection, maintenance practices and incident investigation.
The examination of the 41 derailments relevant to the scope of the report found 35 of these occurred on the main line, five when travelling through passing loops, and one when traversing a yard.
The rail regulation unit assessed each derailment to identify the principal causal factor, identifying rolling stock as the most frequent contributor, followed by track‑related defects (Figure 13).
Figure 13: Principal causal factors

Source: Queensland Department of Transport and Main Roads. Mount Isa Derailment Analysis 2008 -2013 report attachment B
The number of rolling stock-related causal factors was broken down into occurrence and train types (Figure 14).
Figure 14: Defect by train type

Source: Queensland Department of Transport and Main Roads. Mount Isa Derailment Analysis 2008 -2013 report attachment B
The rail regulation unit cited legacy issues associated with rail infrastructure and rolling stock, together with loading irregularities, and identified a series of recommendations to the rolling stock and rail infrastructure managers for consideration.
In response to the identified rolling stock defects/failures, it was recommended that rolling stock operators should:
- improve record keeping and reporting of wheel set and bearing faults, including wayside alarms and derailment history
- review the process for disseminating information regarding the service history of wheel sets to the staff that undertake non-destructive testing of the wheel sets
- review their investigation practices and reporting to capture the history and types of wheel sets involved in catastrophic failures and derailments.
Following the release of the report, TMR established the Mount Isa Line Safety Working Group (SWG) in March 2016. A key objective of the SWG was to provide a platform to rail transport operators for jointly addressing specific recommendations. The SWG incorporated representatives from TMR and the rail transport operators that conducted rail safety work on the Mount Isa railway.
GATX axle failure analysis
In addition to this occurrence, IPL experienced four axle failures on the GATX fleet. Following each axle failure, IPL engaged a specialist consultant to determine the reason for each failure (Table 3).
Table 3: Axle failures, findings, and recommendations
Date
|
Finding of investigation
|
Recommendation from investigation
|
July 2012
|
Axle contained large fatigue cracks and failed near the centre. Although the fracture surface was damaged obscuring the exact location, the fatigue crack originated from a single point fatigue crack initiation site. (similar to this incident)
|
Investigate the cause of impact damage and corrosion damage on the outer surface of the axle/shaft especially adjacent to the fracture initiation area.
|
November 2014
|
Axle contained large fatigue cracks and failed near the centre. Although the fracture surface was damaged obscuring the exact location, the fatigue crack originated from a single point fatigue crack initiation site. (similar to this incident)
|
Conduct root cause analysis covering handling, storage, design, maintenance, and in-service damage.
|
July 2016
|
Axle contained large fatigue cracks and failed near the centre. Although the fracture surface was damaged obscuring the exact location, the fatigue crack originated from a single point fatigue crack initiation site. (similar to this incident)
|
Improving visual inspection of wagons, axle design, wayside detection, auditing wheel set maintenance procedures. Finite element analysis on axles.
|
August 2018 (Hughenden)
|
No report produced.
The ATSB obtained evidence from this derailment to analyse it as part of the investigation. The failed axles were the same type (840P1), same operation, and similar failure location.
|
No report produced.
|
Following the recommendations of each report, IPL implemented actions to address the failing axles. Notably, IPL commissioned a finite element analysis (FEA) of the 840P1 axle design.[10] The organisation that conducted the FEA detailed the fatigue assessment in the 840P1 axle and proposed a new axle design ‑ 840P2. The new axle design had a central barrel tapered from 165 mm in the centre to 174 mm towards the wheel seats. The FEA, suggested the maximum von-Mises[11] stresses at the centre of the axles were 28.26 and 15.45 MPa for the existing and proposed axles respectively. [12]
The stresses experienced by the axles were significantly lower than the 370 MPa yield strength of the material and the tensile strength of 574 MPa shown in Table 2. The endurance limit of steels is generally accepted to be 40 per cent of the tensile strength of the material. In this case, the calculated stresses were well below the endurance level for the material. Consequently, by that measure the axles had been designed appropriately and should have infinite life.
Fatigue crack size and growth rate
Based on the FEA report, the computed critical crack sizes[13] at the centre of both axles was:
- existing 840P1 - 110 mm (consistent with the failed axles)
- proposed 840P2 - 130 mm.
The critical crack size for the existing axle design correlated with the observed crack sizes in failed axles. Additionally, the FEA report performed fatigue analysis using software based on fracture mechanics idealisation and Paris’ Law.[14] The resulting fatigue lives are shown in Table 4.
Table 4: Computed fatigue lives
|
Life to grow crack to critical depth (cycles)
|
Life to grow crack to critical depth (hours)#
|
Existing 840P1 axle ^
|
19,660,018
|
612
|
Proposed 840P2 axle *
|
78,755,984
|
2,454
|
Notes: # Based on the assumption that the tanker wagon is travelling at 80 km/h and the wheel diameter is 800 mm.
^ Initial crack depth was 2 mm. Below this value resulted in no crack growth.
* Initial crack depth was 5 mm. Below this value resulted in no crack growth.
Based on the FEA report, and the average of two return trips per week, the existing axle design (840P1) with a 2 mm deep notch would be theoretically at critical risk of failure at 612 hours (26 weeks) of loaded operation. Additionally, the proposed axle design (840P2) with a 5 mm deep notch would be theoretically at critical risk of failure at 2,454 hours (100 weeks) of loaded operation.
The FEA report commented:
It has been shown that under normal operating loads the stress-levels in the axles (both the existing and proposed designs) are relatively low and indicate the axles are adequately designed. Therefore in an ideal environment these axles would not be expected to initiate defects, and therefore would not be expected to fail in the manner in which they have in the field.
Post-incident fleet-wide inspection
Following the derailment of train 9T92 at Hughenden on 15 August 2018, IPL performed in-situ MPIs on the entire GATX wagon fleet. By the end of September 2018, it had detected 17 cracked axles and removed them from service. IPL observed that the initiating notch in those instances was minimal (less than 2 mm) in depth.
Summary
In an operational environment, ballast and other objects can strike an axle possibly forming a notch. If the notch is more than 2 mm deep, it will likely initiate a fatigue crack and propagate to the critical size and fail, unless detected earlier. Conversely, based on the FEA report, notches less than 2 mm deep may not propagate. However, based on the results of fleet‑wide inspections, and given the depth of the coincident notch on the failed axle being 0.2 mm, it appears that in practice other stressors allow a notch less than 2 mm to initiate a crack and propagate until failure.
__________