Level crossing incident near Norwich new RAIB investigation

[moggie]

I believe that the reason was to reduce the chance of road users misusing level crossings which they believed had failed because the barriers had been down for "longer than normal" without a train passing.

A predictor system detects the speed of approaching trains so that the time interval between barriers being lowered and a train arriving is similar for all trains, irrespective of their speed.

As to organisation I'm unsure - probably Network Rail or its predecessors?

Yes indeed. My question was in part rhetorical. Predictor technology is as you say is to mitigate against misuse of the crossing by impatient road users. I'm pretty sure the other reason is to also to reduce road closure times where mixed rail traffic might otherwise result in longer strike in times for slower moving traffic , which anecdotally contributes to road user impatience. The agitating organisation pushing this technology on the rail industry - the ORR aided and abetted by the local MP's. The net result for the railway is the addition of complexity. Then they wonder why complexity in one system married with complexity at the wheel / rail interface and pressure to implement new systems pdq leads to unforeseen incidents.

I firmly believe that while RAIB can pinpoint the causal points of this incident, they ignore the underpinning culture of complexity and it's potential impact and the ability of the collective industry to consistently and effectively manage that complexity in a highly pressurised 'need it tomorrow' environment. Clearly that collective ability failed on this occassion.

 

[alxndr]

That default reset time of 16 seconds is the problem and perhaps the concept of a full reset entirely. Thinking about this further, I'm pretty sure traditional relay circuits for AHBCs DO NOT reset in this way at all, although I don't have any real-world example drawings to examine and confirm this, unfortunately. I believe failure to 'strike-out' should cause the barriers to stay down and red lights to continue flashing indefinitely. That's clearly undesirable from a road user safety perspective, but with no crossing clear check possible at this type of crossing it is the only failsafe option. The only realistic approach to minimising such events is to design very high reliability into the systems.

Yes and no. My understanding is the possibilities are:

• Track section occupies then clears - crossing operates and then after three minutes resets and stops.
• Track section occupies and doesn't clear - crossing operates, shows failed after three minutes, but crossing continues to operate.
• Correct strike-in but exit track section fails to clear - crossing operates as normal, after three minutes the crossing sequence restarts and shows failed.
• Strike-in track section occupies and clears without a train - crossing operates as normal and resets after three minutes.

Indeed, the report states:

102
At the time the level crossing control system was installed, the relevant standard in use by Railtrack was ‘Principles of Control for Automatic Half Barrier Crossings & Automatic Open Crossings – Remotely Monitored’ dated August 1985. At that time there was no mandated reset timer, but the standard states that where one is provided it should be set to 120 seconds. This timer, when provided, had the principal purpose of raising the barriers when a track circuit remained occupied after the passage of a train, although it would also allow the barriers to rise if a train stopped on the approach to a crossing.

103
The Network Rail signalling design handbook for automatic level crossings,10 introduced in September 2011, requires that ‘in the event that a strike-in track section becomes occupied (thus initiating the operating sequence) and subsequently clears with no train present, or the strike-in track section is not correctly reset following passage of a train, the crossing operating sequence shall be reset after a period of time, nominally three minutes’.

So, there are mechanisms for an AHB to reset in some circumstances, but they take far greater than 16 seconds, and indeed, also longer than the 99 seconds achievable with the predictor technology currently being employed.

 

[MarkyT]

And WHY was level Crossing Predictor technology ... thought necessary in the first place in the UK - and by what organisation?t

I believe that the reason was to reduce the chance of road users misusing level crossings which they believed had failed because the barriers had been down for "longer than normal" without a train passing.

A predictor system detects the speed of approaching trains so that the time interval between barriers being lowered and a train arriving is similar for all trains, irrespective of their speed.

As to organisation I'm unsure - probably Network Rail or its predecessors?

During Railtrack's attempts to broaden the UK supplier base, the predictor was a product available from the US catalogue of Vaughan-Harmon, who supplied the Whitlingham Junction - Cromer resignalling project that was completed in 2000, relying predominantly on that company's sister product, their US developed 'VHLC' vital logic controller interlockings, connected to a MCS (Modular Control System) signaller's workstation at Trowse Swing Bridge. Vaughan Harmon later became part of GE Transportation Systems (GETS) and then Alstom, who have recently supplied successor ElectroLogIXS equipment to the Wherry Lines scheme. The Predictor was an early (for UK) fully solid-state controller for such crossings which had previously relied on custom wired relay logic. Today there are a selection of approved solid-state controller products available for crossings that also save the complexity of relay logic wiring but do not rely on the predictor concept and can use conventional fixed track circuit (or axle counter) and treadle input. There are also equivalent non-contact inductive sensor based products that can be used in place of mechanically actuated treadle switches.

 

[Dr Hoo]

During Railtrack's attempts to broaden the UK supplier base, the predictor was a product available from the US catalogue of Vaughan-Harmon, who supplied the Whitlingham Junction - Cromer resignalling project that was completed in 2000, relying predominantly on that company's sister product, their US developed 'VHLC' vital logic controller interlockings, connected to a MCS (Modular Control System) signaller's workstation at Trowse Swing Bridge. Vaughan Harmon later became part of GE Transportation Systems (GETS) and then Alstom, who have recently supplied successor ElectroLogIXS equipment to the Wherry Lines scheme. The Predictor was an early (for UK) fully solid-state controller for such crossings which had previously relied on custom wired relay logic. Today there are a selection of approved solid-state controller products available for crossings that also save the complexity of relay logic wiring but do not rely on the predictor concept and can use conventional fixed track circuit (or axle counter) and treadle input. There are also equivalent non-contact inductive sensor based products that can be used in place of mechanically actuated treadle switches.

Thanks for that informative post. So the set-up predated 'Network Rail' (and the ORR's involvement in safety issues). I presume that was the era of DfT-issued Level Crossing Orders and HSE (then incorporating HMRI) oversight?

 

[Bald Rick]

Thanks for that informative post. So the set-up predated 'Network Rail' (and the ORR's involvement in safety issues). I presume that was the era of DfT-issued Level Crossing Orders and HSE (then incorporating HMRI) oversight?

The project was definitely delivered by Railtrack (close out just crept into the NR era) and it was definitely in the era of HMRI oversight.

 

[MarkyT]

Yes and no. My understanding is the possibilities are:

• Track section occupies then clears - crossing operates and then after three minutes resets and stops.
• Track section occupies and doesn't clear - crossing operates, shows failed after three minutes, but crossing continues to operate.
• Correct strike-in but exit track section fails to clear - crossing operates as normal, after three minutes the crossing sequence restarts and shows failed.
• Strike-in track section occupies and clears without a train - crossing operates as normal and resets after three minutes.

Indeed, the report states:

So, there are mechanisms for an AHB to reset in some circumstances, but they take far greater than 16 seconds, and indeed, also longer than the 99 seconds achievable with the predictor technology currently being employed.

Thanks for that. Thinking further, for a conventional strike-in that remains occupied for an unusually long time before the strike-out sequence has started, the assumption must be that, if it really is a train rather than an equipment failure, then to have taken so long to arrive means it has either come to a complete stand or is moving so slowly that a crossing collision would be very low risk. With a predictor, such a slow approach wouldn't trigger the warning sequence until the train was much closer to the crossing in the first place (I don't think the reinforcement treadle would change that behaviour). If a train had come to a stand for any length of time, the warning sequence would start in the appropriate place soon after the train started moving again as the loop resistance started changing at a sufficient rate (assuming there was an effective shunt of the track circuit at all!). That can't happen with the conventional strike-in case, where, once timed out for the movement on the track concerned, the barriers would stay up. For the case where the strike-out track circuit remains occupied in a conventional installation (where bi-directional controls are provided) the crossing is effectively registering a new strike-in from the opposite direction after the timeout.

I'd just like to say in closing that of all the common level crossing types still in use on public roads, none frighten me to the bone more than the AHBC (apart from their short-lived cousin the AOCR)! That was the case when I worked for a time on their relay control circuit designs over three decades ago, and nothing has occurred since to change my mind. I'm very pleased to see many of them being replaced by bridges or being converted to safer full barrier or locally monitored types.

 

[Tio Terry]

At a guess - and it is very much a guess - this form of predictor was included in the Product Acceptance Trial for the Vaughan-Harmon signalling system installed on the Cromer line. It looks very much like the approval process fell foul of the Railtrack/NetworkRail changes combined with the Vaughan-Harmon/GETS/Alstom amalgamations.

Perhaps that's also something RAIB should be looking at?

 

[ex-probationer]

It is LOCALLY MONITORED crossings of various types, with drivers' red/white flashing lights (for approaching rail traffic) that normally reopen to road traffic after failure timeout; I think in these cases the barrier machines are biassed to rise under total power failure, so the barriers are effectively forced down by the controllers and actuators, unlike AHBCs where they are designed to fall under gravity.

Mark,

It was my understanding that the principle design difference between an AHB and ABCL was the barrier power pack hydraulic valve release. An AHB barrier is 'failsafe', where the absence of power to the hydraulic valve allows the barrier to fall, whereas an ABCL requires a feed to the hydraulic valve to allow the barrier to fall.

The 'locally monitored' barrier power packs are painted blue to avoid confusion with a standard barrier power pack used for an AHB. My understanding is that both require the hydraulic pump to be powered to raise the barriers and use a standard BR 843 Barrier Pedestal.

In the event of a power failure, an ABCL may remain open to road traffic and any train approaching will be warned to stop by the flashing red 'Driver Crossing Indicator'.

Kevin



ABCL Barrier Pedestal

Minor edits and photo added

 

[Taunton]

For all that there is implied criticism above of the crossing controls, they do seem to have worked quite satisfactorily for the railway as installed, and for some 20 years afterwards, including through 20 leaf fall seasons, without apparent issue. It was the introduction of a new train type which caused the issue to arise in their very first leaf fall season. So we need to be looking more at new train design and testing rather than fully established trackside kit.

 

[trebor79]

For all that there is implied criticism above of the crossing controls, they do seem to have worked quite satisfactorily for the railway as installed, and for some 20 years afterwards, including through 20 leaf fall seasons, without apparent issue. It was the introduction of a new train type which caused the issue to arise in their very first leaf fall season. So we need to be looking more at new train design and testing rather than fully established trackside kit.

The report states specifically it was nothing to do with the introduction of the new trains per se, and that a similar issue could have arisen if several trains with freshly turned tyres had traversed the route. Whilst that might not seem very likely, it isn't uncommon for the same one or two units to run up and down the same line all day - eg Sudbury branch.
Fundamentally, the crossing controls had a wrong side failure mechanism which should have been spotted and designed out before they were installed 20 years ago. It's really only chance that has prevented a serious accident.

 

[ex-probationer]

I thought I read something at the time about the new trains being fitted with lubricators that may have left a thin film of oil on the rail. I understood these lubricators may have been isolated following the problems on the Cromer Line.

 

[trebor79]

I think that was one of the theories in the time immediately following the incident, before the actual cause had been determined. They were also terminating the Cambridges short at Ely, then shunting into the sidings to have the wheelsets manually cleaned before returning to Norwich. I guess naturally the assumption was that t must be some issue related to the new trains, and therefore a brainstorm on what the issue might be led to various actions to mitigate until the actual problem was discovered.

 

[Philip Phlopp]

I think that was one of the theories in the time immediately following the incident, before the actual cause had been determined. They were also terminating the Cambridges short at Ely, then shunting into the sidings to have the wheelsets manually cleaned before returning to Norwich. I guess naturally the assumption was that t must be some issue related to the new trains, and therefore a brainstorm on what the issue might be led to various actions to mitigate until the actual problem was discovered.

It was good thinking by those involved - think what the possible causes could be and mitigate them as quickly as possible.

I'm pleased there wasn't some of the more recent "we'll cancel everything on safety grounds and wait for RAIB to tell us" thinking, or the really old BR-era thinking of never ever stopping the job "it's only happened the once, it can't be too likely to happen again".

 

[trebor79]

It was good thinking by those involved - think what the possible causes could be and mitigate them as quickly as possible.

I'm pleased there wasn't some of the more recent "we'll cancel everything on safety grounds and wait for RAIB to tell us" thinking, or the really old BR-era thinking of never ever stopping the job "it's only happened the once, it can't be too likely to happen again".

Yes I absolutely agree. Obviously there was a lot of service disruption at the time but equally it was obvious that a lot of effort was being put into keeping as much running as possible.
I recognised the outputs of the thinking from some of my own experiences dealing with crisis management in a different industry.

 

[MarkyT]

Modern suspension, wheel-tread and rail-head profiles attempt to reduce the so-called 'hunting' behaviour of wheelsets in traditional rolling stock. Hunting, a side-to-side oscillation, particularly prevalent on straight track, leads to poor ride quality at high speed and increased wear of both rail and wheel tread. It has a side effect of keeping a wider area of the railhead clean however, which was beneficial in maintaining good electrical contact for track circuits, even ensuring reliable operation of the very low voltage DC TC types used in traditional mechanical signalling. The lack of hunting, together with the transition from tread brakes (keeping the wheel surface clean) was a major problem with the second generation DMUs introduced in the 1980s. That led to widespread modifications of track circuits to higher rail voltages, and development of the on-board TCA device. The TCA induces a high frequency wetting current circulating around a pair of wheelsets, helping to break down any high resistance due to contamination and allowing the track current to also flow more easily. Pacers even had problems in this area, even though their individual axle weight was often higher than some of the 1st gen DMUs they replaced. The Anglian FLIRTS are typically LONGER trains and also have a higher axle weight (with articulation) than the 2nd gen DMUs they're replacing, yet this incident demonstrates the industry still experiences problems in this area. To me it confirms my long-held opinion that track circuits of all kinds are becoming increasingly anachronistic and need to be superseded in most applications by more reliable alternative methods of train detection. That, broadly, is NRs current policy on this matter for new schemes.

 

[Roast Veg]

Perhaps a couple of lineside sprinklers offer a short term solution?

 

[moggie]

During Railtrack's attempts to broaden the UK supplier base, the predictor was a product available from the US catalogue of Vaughan-Harmon, who supplied the Whitlingham Junction - Cromer resignalling project that was completed in 2000, relying predominantly on that company's sister product, their US developed 'VHLC' vital logic controller interlockings, connected to a MCS (Modular Control System) signaller's workstation at Trowse Swing Bridge. Vaughan Harmon later became part of GE Transportation Systems (GETS) and then Alstom, who have recently supplied successor ElectroLogIXS equipment to the Wherry Lines scheme. The Predictor was an early (for UK) fully solid-state controller for such crossings which had previously relied on custom wired relay logic. Today there are a selection of approved solid-state controller products available for crossings that also save the complexity of relay logic wiring but do not rely on the predictor concept and can use conventional fixed track circuit (or axle counter) and treadle input. There are also equivalent non-contact inductive sensor based products that can be used in place of mechanically actuated treadle switches.

Which once again shines a light on the state of the signalling industry in those Railtrack days. I wasn't sure how far the boundaries of the Norwich - Cromer resignalling project extended in relation to the crossing involved in this incident but what I am certain of is that Vaughan's (as they were known in the industry) were principally a Train Describer system supplier in the UK market at the time, not interlocking or level crossing technology. And while their partnership with Harmon at the time provided Vaughan's access to the VHLC (PLC technology) the US and UK signalling systems are very different beasts, especially when it comes to UK Signalling Principles.

As you state this was another example of Railtrack's procurement adventure into pastures unknown* to expand their signalling supplier base (a very much unsuccessful expansion). One wonders quite how much a new entrant (with limited technical experience) expanding their presence into the signalling arena of lineside signalling / interlocking from a niche TD offer might have contributed to the failure to identify the nature of the risk introduced by their imported Predictor technology? As you state it's reassuring that the technology has evolved as its matured. But one of the drawbacks of early adoption of novel technology is continued support to NR when it needs alteration. It's either a lack of suitably competent and available people and / or bespoke software design tools to undertake the necessary change years after the system has reached obsolescence - thus exploding the cost for such alteration beyond reasonable sums.

So to summarise. Novel operational features (the Predictor), new entrant (inexperienced) into the supplier field, erosion of corporate system knowledge, new rolling stock introduction, fragmented rail industry, ineffective communication between parties, unclear responsibilities for the interface between track and train parties. Bound to be a success!

PS: I think I'm correct in saying that there has been no more VHLC introduced by the successors of Vaughan's (GETS) since. MCS (a good product) yes. Where VHLC and it's successor technology have been implemented subsequently this has been carried out under sole UK license by SNC Lavalin (Atkins) with Alstom the equipment supplier.

 

[eman_resu]

Predictors are installed down here also to cut down on lineside enclosures, power reticulation and cable routes and cabling that axle counter or traditional track circuited approaches require. Some of my recent projects with these have saved over 4000 m in new routes and cabling per crossing, and replaced with a tuned shunt at the start of the approach.