Background
Hatfield is a city of 41265 people in southern England, located in Hertfordshire County 19km/12mi southeast of Luton and 28km/17mi north of London (both measurements in linear distance).
Hatfield has a station on the East Coast Main Line (ECML), a double- to quad-tracked electrified main line opening in 1850. The 633km/393mi connection from London (England) to Edinburgh (Scotland) is one of the United Kingdom’s main rail corridors alongside it’s west coast counterpart (West Coast Main Line, WCML), allowing trains to navigate most of the Island’s north-south expansion at speeds of up to 201kph/125mph as of 2024.
The Train Involved
1D38 was a midday express passenger service from London’s Kings Cross station to Leeds, provided by GNER (Great North Eastern Railway) with an IC 225 train set leased from HSBC, a British banking and financial service provider. The IC 225 is a high speed passenger train introduced in 1989 to modernize Intercity services. Each IC 225 train consists of a Class 91 locomotive, nine Mark 4 passenger cars and a so-called Driving Van Trailer (DVT). A full set measures 245m/804ft in length and reaches 201kph/125mph in regular service while being designed to reach 225kph/140mph, hence their name. They can carry up to 535 passengers if every single seat is occupied, including the onboard restaurant’s capacity (referred to as a Kitchen or Service Car).
The British Rail (BR) Class 91 is a four-axle high speed electric locomotive developed specifically for the IC 225-project, measuring 19.4m/63ft in length at a weight of 85 metric tons. Despite their rear end being designed to blend in with the Mark 4 passenger cars they do feature a driver’s cab at the “blunt” end, and can on rare occasions be seen running with that end leading. The locomotives were developed with bogie-mounted gearboxes and underslung transformers, leaving their interior between the cabs largely empty. This unusual choice was picked in order to reduce unsprung mass as well as the lower the center of gravity, reducing track wear and improving cornering capability at high speeds. The locomotives are combined with a Mark 4 DVT at the other end of the train, which is based off a near-identical body shell and looks just like a Class 91 at a glance while lacking any propulsion-equipment, only holding a driver’s cab to remotely control the locomotive on return trips. Their design is meant to make the train set appear symmetrical no matter which direction its going, similar to other countries’ high speed trains, while also offering good aerodynamics when the DVT is leading. The absence of propulsion equipment drops the weight of a DVT to 43.7 metric tons. 1D38 was led by Class 91 number 91023, christened “City of Durham”.
The majority of the IC 225 is made up of BR Mark 4 passenger cars, four-axle high speed passenger cars each measuring 23m/75ft in length at a weight of 40–44 metric tons empty depending on configuration. 1D38 was not even at 50% occupancy at the time of the accident, carrying just 170 passengers along with 12 crew members, excluding the locomotive crew.
The Accident
1D38 departs London’s Kings Cross station at 12:10pm on the 17th of October 2000 with a trainee driver at the controls of the locomotive while an experienced driver is supervising (and legally considered “in control”). A brake test is conducted in northern London before the train accelerates into the English countryside as passengers settle down for the journey.
The train is approaching Hatfield station at 185kph/115mph just 13 minutes later, heading into a long right hand curve when the left hand rail suddenly shatters beneath the first car’s wheels as if it were made of glass. The second car manages to follow the train over the growing gap as the fracture keeps growing, but the third car derails and is carried out of the turn. It runs into a signal (K563) at full speed, tearing it off the locomotive and the forward two cars as it drags cars 4–6 across the neighboring track towards the ditch.
Car 7, the service car, is forced towards what now was a considerable gap in the curved track, running off the end of the somewhat intact rails and overturning as the wheels dig into the gravel ballast, tearing the bogies off the train car. The car, landing on its side no more than 73m/240ft from the end of the rails, starts to rotate as it slides along, turning almost sideways to the direction of travel before slamming into a support post for the overhead wires roof-first at considerable speed. The impact uproots the post as it is stopped by the floor of the car, allowing it to carve out a large section from the car as it is dragged along, soon joined by the next support post to the north. Debris is thrown every-which way as the service car finally comes to a rest.
The rear three vehicles (cars 8 and 9 as well as the DVT) had separated from the rest of the train as the service car overturned, but have no choice but to enter the gap in the rails as well. Car 8 partially falls over before grounding to a halt, detached from the train behind it, leaving it hanging into the ditch at an angle of 70° to the left. The ninth passenger car remains upright, coming to a stop deeply embedded in the gravel ballast and thus also stopping the DVT behind it. 17 seconds have passed between the train reaching the site and silence falling on the wreckage. 4 people die, all in the remains of the service car, another 70 people survive with injuries.
Aftermath
The locomotive crew doesn’t initially know what happened, being blindsided by the train suddenly triggering an emergency stop on its own. The locomotive covered another 910m/1000 yards with full brakes applied, despite the resistance from the derailing train, before coming to a stop beyond the right hand curve. The derailing train had destroyed overhead wires and trackside wiring for communications and signaling, causing a myriad of warnings and errors to show up at the local signal box. Emergency calls from the driver and head conductor reached the emergency services, notifying them that a catastrophic derailment had occurred. The first responders reached the site within 10 minutes of the derailment, being met by the first survivors evacuating the train on their own. They had to be careful not to get too close to torn overhead wires, as the confirmation about them being switched off wouldn’t be issued until 12:45pm.
Most of the train had largely retained structural integrity during the derailment, something the Mark 4 cars would later be complimented for, with the service car being a drastic exemption. While it hadn’t suffered significant flexing or compression there was still extensive damage as the two steel posts from the overhead wires, torn off their concrete footers by the train, had carved out the body of the car floor to ceiling for several feet, leaving the car without more than half of its roof and making look like something had taken a bite out of it. Seats, tables, pieces of the wall and roof along with various furnishings were found strewn between the foundation of the first support post and the final position of the car.
Investigators were met with the rather obvious cause of “several feet of rail have ceased to exist”, the mandatory examination of the locomotive and analysis of the locomotive crew’s action brought, as expected, no indication that a technical defect on the train and/or the locomotive crew’s actions had played any role in the accident. It appeared the trainee had been driving a perfectly intact train just as they were meant to, enabling an uneventful journey until the track disappeared below the train. The fracture of the outside rail had also caused tension forces to displace the left rail at the far end of the gap to the left, so even if the train would, by some physically impossible miracle, have stayed on course, it wouldn’t have been able to “slot” into the rails and proceed. The investigators found severe cracking on several hundred meters of the left hand rail around the gaps as well as on the remaining right hand rail. This cracking was traced to rolling fatigue, a form of wear caused by the rail bending and straightening ever so slightly as trains run over it, every time a train wheel runs over it. A common example to demonstrate the forming and consequence of fatigue cracking is to take a paperclip and bend it back and forth repeatedly. Eventually, it will break off. This happens because microscopic cracks start to form in and spread through the metal, until they reduce the solid diameter of the object enough for the forces (which are about the same each time) to cause a complete fracture.
A painstaking search effort turned up 300 pieces of metal that used to make up the initial 35m/115ft gap in the rail (the “main gap”) strewn in the wreckage (if you’re calculating along, that brings the average size of a fragment to 11.66cm/4.5in), with the possibility that further pieces weren’t found. Another fragmented section of 54m/177ft was located further along the wreckage, at which point the train was already beyond saving.
The investigators took all the pieces of rail that they traced to the main gap to an off-site facility and, in what must have been the world’s most frustrating puzzle, managed to reconstruct large sections of the rail from the various bits they had recovered. They discovered even more severe cracking throughout the reconstructed sections, indicating a lack of maintenance over a prolonged time. Damage to the recovered pieces included spalling-damage, which refers to cracks forming somewhat parallel to a surface, breaking pieces out of the material instead of simply running through it as a crack. Some of the spall from Hatfield measured as big as 10cm/3.9in long at a thickness of 5 millimeters (0.2 inches).
Damage to the rail at the site was too severe and expansive to identify an exact starting point of the fracturing, with different pieces of rail debris showing impact-marks indicating that they had been struck by train wheels running on ballast after the rail had shattered below them.
Severe cracking was also found up to 1km/0.6mi south of the site on different sections of rail, indicating that the cause of the severe damage lay with insufficient maintenance rather than the manufacturing process of the fractured rail, which was only 5 years old.
Ultrasound testing of both the detached fragments and intact rails surrounding the site was conducted following the U3-standard which was the common way to conduct those tests at time time as well as the improved U14-standards in selected spots. A specific focus was placed on the upper section of the rail (referred to as the rail head), especially its outer sections (referred to as gauge corners) as centrifugal forces would press the flange of the train’s wheel against the side of the rail as the train is forced into a curve, shifting pressure from acting straight down onto the rail to coming from the side, pushing on the gauge corner.
The investigation found that neither method managed to reliably produce accurate readings of the entire rail. Especially U14, which was designed to be better at detecting damage in gauge corners (referred to as “Gauge Corner Cracking”, GCC), had issues detecting just that as soon as cracks were oriented slightly outside a closely specified range of angles to the testing device. Testing also required the probe to sit flush atop the rail being tested. If the rail showed surface cracking or even spalling there would be no satisfactory contact-patch, leading to invalid results (U3) or to the spot being simply declared untestable (U14), both of which meant there was no data recorded on the condition of the rail beyond that of its surface, creating a “Blind Spot” where the condition beyond what the naked eye could see was unknown.
This tendency for damage to hide further damage was compounded by a problem with the procedure’s paper trail. There were requirements to record the settings of the ultrasonic devices and the discovered damages, but there were no instructions on how the data should be recorded, leaving room for confusion as the same crack could be entered into the logbook in a multitude of ways. Furthermore, there was also no requirement for how to report inspection results to superiors. Notably, if ultrasonic testing failed to produce any result those Blind Spots were not required to have their locations recorded meaning a localized retesting was just about impossible and one had to hope the equipment would work better next time. Only found damage was reported, and a blind spot wasn’t considered damage, it was just nothing.
Railtrack, one of the companies responsible for England’s rail infrastructure at the time, had already been aware of the ultrasonic testing’s shortcomings, penning a letter in December 1999 where they called for reform and expansion of the maintenance guidelines as the current procedures and documentation guidelines provided insufficient protection from fatigue cracking.
The track at the site of the accident was maintained by Balfour Beatty Rail Maintenance Ltd (BBRML), who subcontracted the work from Railtrack. Looking into the maintenance records provided by BBRML brought the investigators finally on track to figure out why 1D38 had been left without a rail ever so suddenly. Mistakes made in the maintenance of the area surrounding the site of the accident included:
- A January 1999 U3-test which failed to show the rail foot (“Loss of Rail Bottom”, LORB) did not trigger a U14-test to try and get complete results.
- A November 1999 U3-test of the rail which ended up shattering found severe gauge corner cracking along with LORB-spots. A note “Curve needs re-railing” was attached but no further action was taken (or wasn’t recorded).
- An April 2000 U3-test was apparently abandoned due to “total LORB from heavy GCC”, seeing the rail labelled untestable. The same section was listed as both LORB and untestable, presumably by different individuals, but this doubled listing wasn’t picked up by superiors.
- Severe GCC was discovered at a different location in October 2000, 10 days after the accident at Hatfield, and was rectified by re-railing within a week. Why the site at Hatfield, despite similar/worse damage, wasn’t given the same urgency remains unknown.
In summary, the investigation found that U3-tests at Hatfield were made at the required frequencies, but additional U14-tests were not conducted when required. The rail which fractured beneath 1D38 was in such bad shape by January 1999 that testing couldn’t be carried out in a proper and complete fashion, escalating to the point were urgent retracking would be required in December of that year. The worn out rail should have been replaced by April 2000 at the latest.
Grinding/resurfacing of the rail at the site of the accident had been planned for 1998, but didn’t occur until the 5th of October 2000. Grinding had no effect on the cracks within the rail, yet still there was no replacement performed, not even a low speed zone was placed on the section. Cracking so severe that the examiners couldn’t see the center of the rail in the ultrasound should have led to urgent emergency action as it was clear proof of extensive damage.
The investigation concluded their examination of BBRML’s maintenance performance with a several pages long list of issues and shortcomings, including employees across the hierarchy failing to follow instructions and uphold standards, engineers not being trained on track defects and the importance of urgent management of those defects, failure to keep two separate damage-databases (one of damages found and one of damages known) in synchronization and failure to correctly asses the risk posed by findings, especially if those findings included blind spots or contradicting inputs.
To put it bluntly, it appeared that Railtrack had contracted a company unfit for the task at hand, likely going for a low cost offer. BBRML’s manager for the section of track were the accident occurred flatly admitted that he held no knowledge of railway engineering or safety, despite this being among the requirements when the company applied for the maintenance contract. The only explanation the investigation could find for the sudden drop in maintenance quality was continued cost-cutting as part of privatization, hiring lower-skilled workers who could, after more or less thorough training, perform the tasks well enough that it didn’t immediately raise suspicion without having to pay them more than absolutely necessary. Especially since a lot of their work, as lacking signatures on documentation showed, went without review. This had allowed BBRML to provide an attractive offer to Railtrack who, being a private company listed on the stock exchange, were very interested in not spending more than necessary on anything. Internal issues and oversights had also created a backlog of untreated damages so long that over 1800 emergency speed restrictions were issued after the Hatfield derailment while they tried to catch up on repairs and re-railing of various spots up and down the ECML before another train would find itself without a rail beneath its wheels. The speed restrictions and track closures wrought havoc on both passenger and freight rail services for over a year, with some rail service providers having to cancel up to 400 trains per week. It’s estimated that the disruption cost the UK economy 6 Million British Pounds (12.2 Million £/ 14.3 Million Euros/15.6 Million USD in 2024) per day.
Railtrack itself didn’t survive the scandalous and costly aftermath of the investigation, finally folding in 2002 with their operations being taken over by the government-controlled non-profit Network Rail. The investigation had concluded that they had hired a company who did sub-standard maintenance work including a failure to properly process discovered defects, allowing rails to deteriorate to the point where they couldn’t withstand the physical stress of regular operation. Charges surrounding manslaughter and breach of health and safety standards were subsequently filed against Network Rail (as the successor of Railtrack) and BBRML’s division responsible for the area where the accident occurred as well as against five managing employees from Railtrack and BBRML. The trial started in January 2005 and saw Network Rail and BBRML found guilty on the breach of health and safety standards while the judge dismissed the manslaughter-charges. Network Rail had to pay 3.5 Million British Pounds while BBRML was fined 10 Million British Pounds (6.5 Million £/7.7 Million Euros/8.4 Million USD and 18.7 Million £/21.9 Million Euros/23.9 Million USD in 2024, respectively).
BR Class 91 91023, having escaped the derailment undamaged, gained another bit of tragic fame just four months after the accident, when the 2001 Selby train collision saw a train it was pushing derail after striking a car stuck on the tracks only to be sent into the path of an oncoming freight train. 91023 only suffered minor damage, but 10 people aboard its train lost their lives. The accident, also referred to as the “Great Heck Train Crash”, was the subject of an early article on this blog.It was repaired and returned to service once more, renumbered 91132, while fans gave it the unofficial nickname “Lucky”. It served for another 20 years before meeting an unspectacular end, being scrapped as the first of its type in summer 2021. The IC 225 is seeing increasingly reduced usage with trains going to private rail service providers or being retired outright as the newer BR Class 800 and Class 801 (both nicknamed “Azuma”) replace it on its original services.
_______________________________________________________________
I have chosen to stay off Medium’s monetization offers to keep these stories as accessible as possible, but you can support me with a small tip via “Buy me a coffee” if you feel like it.
_______________________________________________________________
A kind reader is posting the installments on reddit for me, I cannot interact with you there but I will read the feedback and corrections. You can find the post right here.
