AWC Derailment at Shap Summit 03/11/25

[M60lad]

Considering it finished and spent the night at Edge Hill when is it planned to go to Longsight for further tests?

 

[DaveyJones]

Considering it finished and spent the night at Edge Hill when is it planned to go to Longsight for further tests?

From a post I saw on Facebook, it's standing down at Edgehill for an exam and livery repairs

 

[Humftamuft]

Considering it finished and spent the night at Edge Hill when is it planned to go to Longsight for further tests?

No further tests required. Needs some livery repairs, so dropped into a planned maintenance slot at edgehill where this type of work is completed routinely.

 

[DaveyJones]

Looks like 390117 has been out and about in regular traffic since Sunday 7th.

 

[hexagon789]

Looks like 390117 has been out and about in regular traffic since Sunday 7th.

Been back in traffic since the 29th of November.

(Posts #419 and #420 for reference.)

 

[Mag_seven]

We will reopen this thread when the RAIB have reported.

thanks all.

 

[ainsworth74]

A brief update today from the RAIB in the form of an Urgent Safety Advice bulletin:

1. Safety issue​

Lineside monitoring equipment used on Network Rail managed infrastructure may not be able to detect the failure of slopes in some circumstances. As a result, this equipment may not provide data as expected to support safety decision-making, particularly during extreme weather conditions.

[...]

4. Background​

At around 06:10 on 3 November 2025, a passenger train operated by Avanti West Coast, the 04:28 Glasgow Central to London Euston service, derailed after striking a landslip near to Shap Summit, between Penrith North Lakes and Oxenholme Lake District stations.

The train was travelling at around 83 mph (134 km/h) when it struck landslip debris that had been washed onto the track. This material lifted the first bogie off the rails and to the right, where it ran derailed for around 560 metres. There were nine staff and 86 passengers on board the train at the time of the collision. Four people were treated for minor injuries as a result of the accident, and damage was caused to the train and to railway infrastructure.

The landslip was caused by a period of heavy and sustained rainfall. RAIB’s preliminary examination found that a drainage channel, which runs across the cutting slope above the washed-out material, was unable to accommodate the volume of water which was present. This led to the slope material below becoming saturated, initiating the landslip.

The cutting slope was fitted with remote monitoring equipment, which was designed to detect ground movement. At the time of the accident, the monitoring equipment at Shap was recording data and reporting to its online monitoring service. However, it had not been formally entered into operational use, so was not sending alerts to the Network Rail control centre. Similar equipment is operational on other parts of the railway infrastructure.

This type of equipment, when configured for Network Rail slope monitoring applications, is mounted on steel spikes every 2 metres along the base of the slope. The position of the sensors is recorded at intervals.

Movement of the sensors is recorded by the monitoring system as four colour-coded levels of alert, of which the highest two are considered to represent significant movement:

1. Green (information) – movement of between 10 and 30 mm
2. Amber (major) – movement of between 30 and 60 mm
3. Red (severe) – movement of between 60 and 90 mm
4. Black (critical) – movement of more than 90 mm

Around 4 hours before the accident, the sensors nearest to the landslip began to show minor movement of the earthwork, below the threshold needed to trigger a green alert. This movement continued for the next 2 hours, remaining below the green alert threshold.

At around 04:30, when the evidence available to RAIB suggests that the landslip occurred, the two sensors in the path of the debris were tipped over and subsumed by the material sliding down the slope. It would appear that this occurred too quickly for them to determine and transmit their movement and to generate an alert.

RAIB has further concluded that the sensors’ wireless signal was also unable to pass through the layer of material which covered them. This is based on them being able to re-establish a connection and report a variety of alert levels as the site was cleared.

Urgent Safety Advice 01/2025: Use of remote earthwork monitoring equipment

www.gov.uk

 

[Straus]

That is very significant

 

[CharlieSpotted]

Interesting! One would think that a system like this would have a failsafe mode - there would be a condition where if the signal from one (or more, perhaps adjacent) sensors was suddenly lost, that would trigger an alert, with high importance, too?

 

[endecotp]

So the sensors went from “less than 10mm of movement” to “buried under debris” too quickly to send out a message.

Which would be more useful, sending a message when there is a small movement (with likely false positives) or sending a message when buried (which might be too late)?

Maybe the sensors were in the wrong place, i.e. they were nearer the track in the course of the slide, rather than higher up in the ground that actually gave way?

 

[edwin_m]

A sensor could also go out of contact for other reasons such as a loss of power or mechanical damage. That might be detected but treated by the system as a maintenance issue with the sensor rather than generating an alarm. If so then the software could possibly be re-configured to raise an alarm if a sensor goes offline after raising sub-threshold alerts, or if two or more go offline within a short time.

 

[Straus]

A sensor could also go out of contact for other reasons such as a loss of power or mechanical damage. That might be detected but treated by the system as a maintenance issue with the sensor rather than generating an alarm. If so then the software could possibly be re-configured to raise an alarm if a sensor goes offline after raising sub-threshold alerts, or if two or more go offline within a short time.

I suppose we'd need to know how many sub threshold alerts there are on a given day and how often sensors go offline for any reason. It could be that there are so many of both that the number of alerts would be unmanageable.

 

[edwin_m]

I suppose we'd need to know how many sub threshold alerts there are on a given day and how often sensors go offline for any reason. It could be that there are so many of both that the number of alerts would be unmanageable.

I was suggesting that a sub-threshold alert would not generate an alarm, but if a sensor went offline within a certain time after producing a sub-threshold alert then this would raise an alarm. So it would need both to happen to generate an alarm that needed immediate action.

 

[Straus]

I was suggesting that a sub-threshold alert would not generate an alarm, but if a sensor went offline within a certain time after producing a sub-threshold alert then this would raise an alarm. So it would need both to happen to generate an alarm that needed immediate action.

I understood - I mean we'd need to understand how often that occurs as it may not be an unusual set of circumstances.

 

[Aaron_TTL]

I'd have also though two sensors side by side going offline at the exact same time would trigger some sort of alert too.

 

[AndyPJG]

Thing is, this looks like more of a wash-out rather than a earth slip. Most of the cutting hasn;t moved, just a lot of earth washed down onto the track. (Drainage issue methinks)

It seems the derailment was caused by debris washed out from the cutting slope, but here's been no movement of the underlying ground to trigger the sensors (tilt sensors AIUI).
More likely to be a drainage issue of water running over the crest of the cutting?

 

[AndrewE]

It seems the derailment was caused by debris washed out from the cutting slope, but here's been no movement of the underlying ground to trigger the sensors (tilt sensors AIUI).
More likely to be a drainage issue of water running over the crest of the cutting?

That's what I thought, so I am surprised to see the term "landslip" used in the third paragraph.

 

[Annetts key]

Although the Network Rail consulted these earthworks movement systems with the unions, they were very light on any technical details of the system.

It's made and maintained by an outside company.

During the consultation, the unions did ask if any Network Rail engineering or technical staff would be installing or maintaining the systems, the answer from the company was no.

Only the supplier would install and maintain the equipment.

So it's unknown what the actual specification is and what if any failsafe provisions have been included.

 

[dgl]

A failsafe would be having the units linked by cables (could even be fibre) so that if a couple of units get buried by a landslip there is a chance of an undamaged unit, that might have not detected an issue, being able to report the fault.
I wonder if the manufacturer had thought that it might be the case that a landslip could happen so fast as to bury a sensor. I assume waterlogged soil/material (assuming wet soil/material was involved) is quite a good wireless attenuator.

 

[CharlieSpotted]

Fundamentally, a system designed to detect earth movements that can be rendered inoperative by... earth movements... doesn't sound like a terribly useful system.

 

[Straus]

Fundamentally, a system designed to detect earth movements that can be rendered inoperative by... earth movements... doesn't sound like a terribly useful system.

"we never saw this coming"

 

[edwin_m]

It seems the derailment was caused by debris washed out from the cutting slope, but here's been no movement of the underlying ground to trigger the sensors (tilt sensors AIUI).
More likely to be a drainage issue of water running over the crest of the cutting?

The report confirms it was water from the crest drain, but this appears to have brought down enough earth to simultaneously topple a couple of sensors but also cover them over before they could communicate this back to base.

 

[Horizon22]

I'd have also though two sensors side by side going offline at the exact same time would trigger some sort of alert too.

This - along with the suggestion in post 434 combined should be enough to satisfy the specific conditions for further attention without producing loads of spurious alerts around the country.

 

[35B]

Reading the RAIB comments, I'm unclear what movement is measured against. It's not clear whether it's movement from a datum point, or from the current location.

 

[themiller]

Reading the RAIB comments, I'm unclear what movement is measured against. It's not clear whether it's movement from a datum point, or from the current location.

I would expect that the white ‘mushrooms’ on the posts contain accelerometers which detect movement of the posts. Each post should be isolated from all the others to avoid interference between them. There’ll also be a battery powered transmitter which communicates with a central relay (possibly via a mobile phone signal). The relay will report to a central monitoring hub which would be the bit that triggers alarms. The signal from the ‘mushrooms’ won’t be continuous to save battery power but will fire a signal periodically hence a period which leaves that post unmonitored - in much the same way that smoke detectors in fire detection systems work

 

[Russel]

Fundamentally, a system designed to detect earth movements that can be rendered inoperative by... earth movements... doesn't sound like a terribly useful system.

Yes, it does seem like quite a flaw if a couple of these devices loose contact at the same time while others, presumably unaffected devices still function doesn't trigger some kind of alert.

 

[35B]

I would expect that the white ‘mushrooms’ on the posts contain accelerometers which detect movement of the posts. Each post should be isolated from all the others to avoid interference between them. There’ll also be a battery powered transmitter which communicates with a central relay (possibly via a mobile phone signal). The relay will report to a central monitoring hub which would be the bit that triggers alarms. The signal from the ‘mushrooms’ won’t be continuous to save battery power but will fire a signal periodically hence a period which leaves that post unmonitored - in much the same way that smoke detectors in fire detection systems work

If that's correct, then the system is weaker than I'd thought - a lot of holes for movement to not be detectable in, and solutions leaning towards false positives to offset significant weaknesses.

 

[edwin_m]

I would expect that the white ‘mushrooms’ on the posts contain accelerometers which detect movement of the posts. Each post should be isolated from all the others to avoid interference between them. There’ll also be a battery powered transmitter which communicates with a central relay (possibly via a mobile phone signal). The relay will report to a central monitoring hub which would be the bit that triggers alarms. The signal from the ‘mushrooms’ won’t be continuous to save battery power but will fire a signal periodically hence a period which leaves that post unmonitored - in much the same way that smoke detectors in fire detection systems work

I think some of them measure tilt - if the ground slips they will almost certainly tip to some extent and that's easier to detect than trying to fix position by GPS to an accuracy of millimetres. The tilt may be expressed as a movement of the top in millimetres relative to the bottom.

If they only ever transmit at intervals then it makes it more likely they could get covered over between one transmission and the next. It would be better if they transmitted immediately on detecting a movement above certain limits. But maybe that causes problems for battery life because the sensor part then has to be active all the time.

 

[MarkyT]

I think some of them measure tilt - if the ground slips they will almost certainly tip to some extent and that's easier to detect than trying to fix position by GPS to an accuracy of millimetres. The tilt may be expressed as a movement of the top in millimetres relative to the bottom.

If they only ever transmit at intervals then it makes it more likely they could get covered over between one transmission and the next. It would be better if they transmitted immediately on detecting a movement above certain limits. But maybe that causes problems for battery life because the sensor part then has to be active all the time.

The control device could require a maximum interval between reports from each sensor. Maybe the first such missed message might be forgiven for the sake of reliability, but at the absence of the next expected communication after that from the same sensor, an operational alarm would be triggered. I favour cabling between the devices as suggested above. This could provide a back-up communications channel for the tilt sensor when its radio system was disabled. The cabling need not connect to the control module, as a sensor could use the interconnections to exploit adjacent sensors' antennae to communicate by radio instead. Furthermore, a 'weak' cable laid along the slope base might itself be monitored for continuity as another sensor for major movements.

 

[Annetts key]

I think some of them measure tilt - if the ground slips they will almost certainly tip to some extent and that's easier to detect than trying to fix position by GPS to an accuracy of millimetres. The tilt may be expressed as a movement of the top in millimetres relative to the bottom.

If they only ever transmit at intervals then it makes it more likely they could get covered over between one transmission and the next. It would be better if they transmitted immediately on detecting a movement above certain limits. But maybe that causes problems for battery life because the sensor part then has to be active all the time.

I gave all the details of the system that I know of in post #236 earlier