HomeRail NewsFinding the real causes

Finding the real causes

Listen to this article

After nearly ten years of operation, during which it has published 252 investigation reports, the Rail Accident Investigation Branch (RAIB) is well known within the industry. It was formed as a result of a recommendation by the Ladbroke Grove rail crash inquiry, which required the investigation of rail accidents to be undertaken by an independent body as is the case for the air and marine sectors, and became operational in October 2005.

The story of railway safety includes learning from accidents. This is one reason why the industry now has a good safety record with no passenger fatalities in the past eight years.

Paradoxically, as railways get safer there are also fewer accidents from which to learn. This makes it more difficult to assess both levels and areas of risk. Hence the need for a thorough investigation of the rare serious accidents or near misses that do occur from which there may be significant safety lessons to understand.

Different accidents, common themes

As an independent investigation body, RAIB is well placed to undertake such investigations and also ensure that it focuses on safety improvement rather than apportioning individual blame or company liability. By doing so, it can highlight areas of risk not fully addressed by the industry.

In a recent presentation to the IMechE Railway Division in Glasgow, Carolyn Griffiths, RAIB’s chief inspector of rail accidents, illustrated this point through two quite different derailment investigations. These were a broken axle on a Class 222 Meridian train at East Langton, Leicestershire, on 20 February 2010, and the flange-climbing of a wagon wheel on curved track at Camden Road, London, on 15 October 2013. Fortunately, neither accident resulted in any injuries. However, in different circumstances, both were potential multi-fatality events. As will be seen, Carolyn felt these accidents had common themes.

Twist faults at Camden

Prior to the October 2013 wagon derailment, the track at Camden Road on the Up North London line was in a poor condition. It had opposing twist faults and the section manager had recorded that “it is difficult to express on paper how poor and deteriorating the entire patrol (length) is”. The track-recording vehicle (TRV) had last measured track geometry 46 weeks prior to the derailment. A TRV should measure category 3 track, such as the North London Line, at normal and maximum intervals of respectively 16 and 36 weeks.

20 ft container after repacking [online]

Asymmetrically loaded scrap machines inside 20 ft container – repacked after the derailment to replicate original weight distribution.

Derailment by flange-climb occurs when the ratio of  the lateral force of the wheel flange on the rail (Y) to the vertical wheel load (Q) exceeds a critical limit value which depends on both the coefficient of friction and angle of contact between flange and rail head. This occurred at Camden Road due to the combination of twisted track and uneven wagon wheel loadings.

However this was not the only cause. Wear on the gauge face of the rail and the presence of metal particles indicates that the rail lubricator installed 75 metres before the point of derailment was not fully effective. The resultant increase in friction lowered the critical limit value and so increased the probability of derailment.

A further factor was that there was no check rail on the 187 metre radius curve at Camden Road although relevant standards require them for curves of less than 200 metres on passenger lines. A check rail would have prevented the derailment by preventing lateral force on the flange-climbing wheel.

Hidden hazard inside containers

Although various infrastructure defects contributed to this accident, the report makes it clear that the derailment resulted from a combination of these defects and an unevenly loaded container wagon. The RAIB report notes two similar previous incidents.

Although, as the duty holder, the operator is responsible for safe operation of its freight trains, it cannot check inside the containers which are often customs-bonded. It is therefore difficult to control the risk from unevenly loaded containers. This problem is compounded by the way containers are loaded on the 60 foot container wagons which can take combinations of 20ft, 30ft and 40ft containers which may be empty or fully loaded. As a result, the centre of gravity of the wagon and its containers can be a significant lateral and longitudinal distance from the wagon’s centre, as was the case at Camden Road in October 2013.

A significant part of the RAIB report is devoted to the effects of this asymmetric loading and how this is controlled. With the load in the 20ft container being disturbed as a result of the derailment, it was not possible to directly determine its original centre of gravity (CoG). To estimate the container’s load offset, it was repacked as closely as possible to its original configuration. An assessment was also made on the basis of measurements from a Wheelchex site over which it had passed prior to the derailment. As a result, the container’s CoG was estimated to be offset laterally by between 0.10 and 0.24 metres. Data from a British port indicated that 2% of containers have similar offset loading.

Class 222 Broken Axle at East Langton [online]

Derailed Class 222 bogie trailing wheelset with broken axle.

The derailed wagon was carrying an empty 40ft container of 3.9 tonnes and the previously-mentioned 20ft container which was loaded with scrap machines and had a gross weight of 28.8 tonnes. As a result, the wagon’s CoG was approximately 3.2 metres towards the front from the wagon’s centre line.

The estimated result of all this uneven loading was that the wagon’s leading bogie was carrying 2.7 times the weight of the trailing bogie and the left-hand wheels were carrying between 1.2 and 1.4 times the load on the right hand wheels.

This asymmetric loading, combined with the poor condition of the track and the lack of a check rail, were thought to be the main causes of this accident. Incidentally, the damage to the track, a viaduct wall and adjacent overhead line equipment that were the result of this derailment, closed the Up and Down North London lines for six days.

Broken axle at 94 mph

In Carolyn’s second example, the broken axle on a class 222 at East Langton, in February 2010, occurred on a train travelling at 94 mph when a hollow axle failed inside a final drive. This was the first such known accident anywhere in Europe. This axle had been in service for 920,000 miles.

Subsequent metallurgical analysis showed that it had been subject to a maximum temperature of around 1,100 to 1,200°C at the seat of the gear end (GE) output bearing, causing a loss of axle bending strength and resulting in its fracture.

As this heat destroyed much of the evidence, the investigation had to consider five possible causes. After a process of elimination, it was concluded that the initial cause of the failure was the GE output bearing stiffening up so that it could no longer rotate normally, resulting in the axle spinning inside the bearing inner race. Computer modelling indicated that this could generate a heat input of around 30kW. After a further process of elimination, it was concluded that the most likely cause of the bearing failure was a loose fit between the GE bearing inner ring and axle.

These conclusions were reached with the assistance of a technical group set up specifically for the investigation comprising of RAIB, ORR, East Midlands Trains, Bombardier Transportation, Eversholt Rail Group, Serco (consultant for failed gearbox strip down and axle investigation) and ESR Technology (consultant for bearing investigation). This group also oversaw various metallurgical examinations and rig testing of final drives with various low GE bearing interference fits. It also investigated six other powered axles with between 0.67 and 1.2 million miles service that were found to have poor GE bearing fits and so could have been potential failures.

Loss of interference

Class 222 FD and Gearbox [online]As this was an unprecedented failure of a normally highly-reliable safety-critical component, it was important to understand how a bearing could have lost its interference fit. The design intent was that the inner rings of both GE and non-GE bearings should have an interference fit on the axle of 75 -120μm (1μm = 0.001 mm). However, on this axle, the GE bearing is immediately adjacent to a gear wheel with a heavier fit of 243 – 309μm. Mathematical modelling indicated that this heavier interference fit would shrink the adjacent seat of the GE bearing by between 24 and 30μm on diameter for a hollow axle. The corresponding amount of shrinkage for a solid axle would be between 5 and 10μm.

When the axle was made in 2004, the manufacturer did not have a company procedure for dimensional checking as it does now. The final drive manufacturer also did not check axle bearing seat dimensions. On the failed axle, the undamaged non-GE seat was found to be undersize and tapered. This indicated that possibly the GE bearing seat, which had been damaged in the failure, may also have been slightly under size.

Although the GE output bearing’s interference fit on the axle was probably less than that specified, the axle had run almost a million miles over six years before its failure. This is indicative of a progressive loss of interference fit. The report explains that this could have been caused by a combination of two mechanisms. The first is inner ring growth, a gradual and normal increase of the bearing bore occurring over time due to the bearing’s metallurgy, its operating temperature and loading. The second mechanism is fretting – wear and corrosion that occurs when there are small relative movements between two surfaces in loaded contact.

Evidence of fretting was found on several other GE  output bearings in the class 222 fleet. Fretting wear combined with inner ring growth could have reduced the interference fit to a level at which there would be circumferential slip, which then accelerates the loss of fit.

Common themes – whole system

Although a broken axle and an unevenly loaded wagon on twisted track are two quite different derailments, they have common aspects. One of these is the requirement for a whole system approach. The Camden Road derailment was caused by a combination of track and train issues for which the infrastructure controller and freight operating company had separate responsibility.

In its report on the Camden Road derailment, RAIB found that neither company had quantified the risk from unevenly loaded wagons. As a result, it suggested that the industry’s approach to the understanding of risk for asymmetrically loaded wagons had not been effectively co-ordinated and, consequently, the level of risk and required mitigation measures were unknown.

The Class 222 final drive involved in the East Langton derailment was supplied to a specification by the train manufacturer which had overall design responsibility for the train. Its quantified risk assessment did consider the risk of axle failure from an overheating final drive. However, it concluded that having the final drive produced by a reputable transmission manufacturer to industry best practice, a record of satisfactory service on similar fleets, and specified maintenance procedures provided sufficient mitigation.

In fact, while the Class 222 axles were hollow and relatively thin-walled (45 mm), the interference fits for solid axles had been applied. The bearing manufacturer was unaware that its products were to be fitted to a hollow axle and no analysis had been done to assess the effect of the gear wheel interference fit on the adjacent bearing seat.

Common themes – traditional standards

Standards are often derived from lessons learnt from previous accidents. However, such lessons do not always remain relevant.

Bearing Cutaway [online]

In common with other railways, track twist is measured over three metres. This may have been an appropriate derailment mitigation when there were numerous three- metre wheelbase four-wheeled wagons. However such wagons were discontinued many years ago. The derailed wagon at Camden Road had a 14-metre bogie spacing with two metres between the axles on each bogie. When the track was measured after the derailment its twist, measured over three metres, was found to exceed the seven-day action limit but was within the 36 hour action limit. There are no corresponding limits for twist measured over 14 metres.

For many years, oil sampling has been used to determine potential failures yet, three days before the East Langton derailment, the failed final drive’s oil sample was well within the then-specified iron content caution limits. However, when the oil samples were retrospectively analysed to determine the cumulative iron over the whole life of the final drive, it was found to be significantly greater than other final drives which had loose GE output bearings.

Furthermore, during the investigation it was found that oil samples taken from final drives with loose GE output bearing fits had a significantly darker colour than normal oil samples, which indicated that colour may be a useful additional parameter to monitor in oil sampling. Prior to the derailment there was no requirement to assess the colour of final drive oil samples.

The real lessons

For those with a technical interest in the issues that led to these derailments, the 69 and 93 page RAIB reports into the Camden Road and East Langton derailments are a fascinating read. They can be found at www.raib.gov.uk and they include appropriate recommendations to prevent recurrence of the causal and underlying factors.

As a result of the work done by RAIB to investigate these derailments and publish reports that clearly show their causes, no doubt much has been and will be done to prevent the recurrence of similar accidents. However the real challenge is preventing apparently-unpredictable rare accidents that could have serious consequences.

In part, this requires acceptance of the common themes of these derailments. One of these is that the required mitigation measures would have been evident beforehand had a robust analysis been undertaken of the interaction between all associated technical factors. The other is the need to challenge standards that may no longer be appropriate or provide all required mitigation.

In the RAIB 2014 Annual Report, Carolyn Griffiths considered that its investigations may bring new focus to areas of risk that are not always evident to the industry. These accident reports support this view and so demonstrate the value of an independent investigation body.

David Shirres BSc CEng MIMechE DEM
David Shirres BSc CEng MIMechE DEMhttp://therailengineer.com

Rolling stock, depots, Scottish and Russian railways

David Shirres joined British Rail in 1968 as a scholarship student and graduated in Mechanical Engineering from Sussex University. He has also been awarded a Diploma in Engineering Management by the Institution of Mechanical Engineers.

His roles in British Rail included Maintenance Assistant at Slade Green, Depot Engineer at Haymarket, Scottish DM&EE Training Engineer and ScotRail Safety Systems Manager.

In 1975, he took a three-year break as a volunteer to manage an irrigation project in Bangladesh.

He retired from Network Rail in 2009 after a 37-year railway career. At that time, he was working on the Airdrie to Bathgate project in a role that included the management of utilities and consents. Prior to that, his roles in the privatised railway included various quality, safety and environmental management posts.

David was appointed Editor of Rail Engineer in January 2017 and, since 2010, has written many articles for the magazine on a wide variety of topics including events in Scotland, rail innovation and Russian Railways. In 2013, the latter gave him an award for being its international journalist of the year.

He is also an active member of the IMechE’s Railway Division, having been Chair and Secretary of its Scottish Centre.


Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.