Tornado

You really think so? I'm afraid that I'm of the school of thought that regards ETCS as one big white elephant and a complete waste of money. It may be a good idea on a dedicated high speed railway like the once planned HS2 but it is totally unnecessary on a conventional multi purpose railway which is where, if anywhere, that steam belongs. Fortunately I don't see it happening in my life time - and I intend on being around for a few years yet, God willing.

Peter

 

I understand that ETCS potentially allows closer headways and hence more paths. That would be helpful on the WCML.

 

My comments were surrounding the technological achievement of fitting 21st century technology to a 19th century analogue design - from an engineering perspective its phenomenal. Too often we hear criticism of projects when they go wrong, often overshadowing great engineering (Crossrail being one) - this is a genuine good news story.

As for the merits of in cab signalling, well I guess I have a pretty good perspective being a regular user of the Cambrian line. Here it enabled an hourly service, which has been of huge benefit, which was impossible before. The technology, after a lot of teething problems, now seems fairly reliable. It is now regarded as a bit long in the tooth (much like the 158s which ply the route) and the new equipment is smaller, smarter and less power hungry.
Like anything on the railway, its hugely expensive, but once installed, there's much less maintenance and not so much equipment to be vandalised, stolen or swallowed up by Climate Change events - so I can see the attraction.

 

Exactly this.

 

But hence the consequences when "human error" entered the fray as at Clapham - a reason why some control has been taken out of human hands.

 

I think that is a misreading - changing signalling systems doesn’t take control out of human hands, but changes which hands become critical. As a general principle, automation of signalling systems removes the consequences of a driver and - more recently - a signalman error; but it shifts the consequences onto those designing, implementing and maintaining the systems. The lesson from Clapham was about the need for robust processes around wiring changes and maintenance.

Any system is only as good as its weakest link. When there is a significant accident on a line controlled by ERTMS - as there inevitably will be - the scrutiny will fall on the software development and management process; and rail enthusiasts will learn a whole new vocabulary of system architects, software design, code review, regression testing and so on.

Tom

 

And, as the collision at Waterloo and near miss at Cardiff Central a few years ago demonstrated, the learnings can't be one-offs, but will require continued vigilance in those disciplines.

 

ETCS isn't just about capacity, it is or should be about reducing the amount of lineside equipment and hence lineside working required for maintenance (with associated risks to staff safety etc.), and actual maintenance costs.
Capacity can justify it on the ECML, the other factors more so on secondary lines.
How many steam railtours will be possible once it's rolled out across not just bits of the ECML, but bits of other main lines as well? Not many, if Tornado remains the only steam loco fitted. Whatever the other apparent failings of the A1SLT, they should be congratulated for allowing their loco to be the testbed for this technology.
I saw a figure of upward of £500k per steam loco quoted the other day. If NR don't foot the bill, I'm not sure that the loco owners will want to.

 

Compare railway signalling developments with air traffic control systems. ERTMS has similar technologies to ADS-B and associated automatic piloting technologies. Both are also incorporating AI capabilities. I hope the faith in these systems is not misplaced.

 

Define "AI" in this context. If it's LLMs, I'll go and hide somewhere well away from aircraft. But it has many strengths for pattern recognition that may be highly appropriate.

 

Stuff such as XGBoost and Random Forest.

 

So forms of machine learning. I'll go with that, subject (of course) to the usual pre-requisites of genuinely thorough testing.

 

TCAS works most of the time in the air, but not always. I expect ERTMS to have similar results.

 

Do you think these systems will be developed to the point that pilots, or train drivers, just get bored with paying attention and miss an impending collision due to a system failure of some sort?

 

Today's Times has an interesting obituary of Sir David "Den" Davies, a railway engineer heavily involved after Ladbroke Grove. He stated “drivers are already operating very near to the likely limits of human performance for a highly repetitive safety function”; a point that we easily forget. Similar has also been observed of some recent aviation incidents, where underdeveloped or degraded reflexes and responses due to automation have been suggested as factors.

However, and by the same token, such automation has taken significant variability out of the system - and mitigated against deaths and injuries.

 

I do think there is a risk of this, I guess it will depend on what sort of monitoring is introduced of driver and pilot awareness. Trouble with that of course it is the sort of thing you need to get through ASLEF or BALPA.
I am not conversant with the level off computer control that drives ERTMS, but we all know computer issues are not exactly uncommon.

 

Don't forget we have already had a serious incident (with sadly a death) on the Cambrian only last year.
Not wanting to jump ahead of the official report, but early indications are that ERTMS applied the brakes and slowed down the train heading towards Shrewsbury that day, perhaps preventing an even more catastrophic accident by reducing the closing speed of the impending head on, however, it could not overcome the railhead conditions that were apparent as the 158 heading the other way slid helplessly down Talerdigg.