Class 345s and automated "auto reverse": Problems with it and whether it was a good idea to begin with

[Meerkat]

Mod Note: Posts #1 - #27 originally in this thread.

I remember seeing the JLE trial running north of Canning Town; at first it only ran from Stratford to North Greenwich, unconnected to the rest. Couple of weeks test I think.

One really does wonder what is the point of all this high-tech if it takes so much effort to test. Why not choose something known to work out of the box instead. As I understand it a key sticking point is "auto reverse" at Westbourne Park to handle 24tph, apparently it must be auto because nothing else would do ... for information, this is just what was achieved by Westinghouse in 1903 at Liverpool Central with the Mersey railway electrification, which coincidentally was also 24tph (every 5 minutes to Rock Ferry, every 5 minutes to the northern branch), and which equipment was STILL functioning over 70 years later. All automatic signals and point changing, single headshunt, compressed air points changed 5 seconds after the train cleared into the headshunt, and of course quite simply manually driven. Same frequency, same manoeuvre. Cost a bit less though, and doubtless will have lasted quite a bit longer.

How long were those trains? AIUI part of the reason for the automation is so that the driver can change ends?

 

[Mojo]

How long were those trains? AIUI part of the reason for the automation is so that the driver can change ends?

Could they not as an alternative have the trains be double-ended, ie. one driver gets on in the rear (east end) cab at Paddington, sits in the cab and waits for it to be fully berthed in the siding, and then when it is, they key in and take over the cab ready for its eastbound trip. When the train gets back to Paddington the original driver disembarks.

 

[Taunton]

Could they not as an alternative have the trains be double-ended, ie. one driver gets on in the rear (east end) cab at Paddington, sits in the cab and waits for it to be fully berthed in the siding, and then when it is, they key in and take over the cab ready for its eastbound trip. When the train gets back to Paddington the original driver disembarks.

That was exactly as it worked for much of a century in my example from Liverpool Central; there was a turnover driver there, who boarded each train in turn.

Time from driver at the south end stopping the train, points reversing, and driver at the north end restarting, was about 10 seconds.

 

[zwk500]

That was exactly as it worked for much of a century in my example from Liverpool Central; there was a turnover driver there, who boarded each train in turn.

Time from driver at the south end stopping the train, points reversing, and driver at the north end restarting, was about 10 seconds.

Did the Liverpool driver have to set up the GSMR and ETCS?

 

[Bald Rick]

Could they not as an alternative have the trains be double-ended, ie. one driver gets on in the rear (east end) cab at Paddington, sits in the cab and waits for it to be fully berthed in the siding, and then when it is, they key in and take over the cab ready for its eastbound trip. When the train gets back to Paddington the original driver disembarks.

Could do, but then that needs at least two extra drivers on shift at all times, plus it defeats the whole object of the exercise which is to automate the process.

 

[rebmcr]

"Why bother troubleshooting this windmill when three guys with a pestle and mortar have ground grain perfectly well for decades?"

 

[Taunton]

Did the Liverpool driver have to set up the GSMR and ETCS?

No, they put a notched key in, and everything worked straight off.

There's a moral for the whizzy 21st century GSMR/ETCS team in there somewhere ...

plus it defeats the whole object of the exercise which is to automate the process.

Sorry, but the whole object of the exercise is to transport passengers across London. Which it has been singularly failing to do for years now ...

 

[Roast Veg]

No, they put a notched key in, and everything worked straight off.

There's a moral for the whizzy 21st century GSMR/ETCS team in there somewhere ...

Which team is that then?

 

[306024]

Could they not as an alternative have the trains be double-ended, ie. one driver gets on in the rear (east end) cab at Paddington, sits in the cab and waits for it to be fully berthed in the siding, and then when it is, they key in and take over the cab ready for its eastbound trip. When the train gets back to Paddington the original driver disembarks.

That would lead to drivers swapping trains all over the place. One train missing or trains out of order and the diagrams could end up in a right mess. Keep it simple when doing drivers diagrams.

Could do, but then that needs at least two extra drivers on shift at all times, plus it defeats the whole object of the exercise which is to automate the process.

Except there is actually no need. The current timings allow sufficient time for changing ends at Westbourne Park. Auto reverse is a huge performance bonus but it isn’t mission critical yet unless the timetable gets tightened up in future. Planning to rely on auto reverse from day one would have been ’courageous’.

 

[Taunton]

That would lead to drivers swapping trains all over the place. One train missing or trains out of order and the diagrams could end up in a right mess. Keep it simple when doing drivers diagrams.

Oh guys, everyone sorted out such issues long ago.

At Liverpool your great-grandfathers got out of the train, walked back down the island platform while it was shunting, and reboarded their own train at the front when it was ready for departure again. Meanwhile the two turnover men, one each end, just took the two cabs and spent the shift doing that.

 

[306024]

Oh guys, everyone sorted out such issues long ago.

At Liverpool your great-grandfathers got out of the train, walked back down the island platform while it was shunting, and reboarded their own train at the front when it was ready for departure again. Meanwhile the two turnover men, one each end, just took the two cabs and spent the shift doing that.

Exactly, that is keeping it simple, as the main line drivers effectively keep with the same train. As I read it that wasn’t what Mojo was suggesting.

 

[stuu]

No, they put a notched key in, and everything worked straight off.

There's a moral for the whizzy 21st century GSMR/ETCS team in there somewhere ...

The basic point of digital signalling is to be automated and therefore not reliant on lots of people in signal boxes. In the past, labour was cheap so schemes like the Mersey example you used earlier would need many more people to run. Now labour is expensive it needs to be replaced with cheap to run computers, which are more of a pain to get working reliably.

 

[Bald Rick]

Sorry, but the whole object of the exercise is to transport passengers across London. Which it has been singularly failing to do for years now ...

Sorry, I meant the process for turning trains round. It’s also a decent precedent to set.

 

[Nicholas Lewis]

Sorry, but the whole object of the exercise is to transport passengers across London. Which it has been singularly failing to do for years now ...

Spot on.

 

[Taunton]

The basic point of digital signalling is to be automated and therefore not reliant on lots of people in signal boxes. In the past, labour was cheap so schemes like the Mersey example you used earlier would need many more people to run. Now labour is expensive it needs to be replaced with cheap to run computers, which are more of a pain to get working reliably.

This is an unfortunate comparison, as the reversing/signal installation of the Mersey Railway at Liverpool in 1903 was an absolute pioneer in automation, and operated without any signalmen in signalboxes. It was also installed by Westinghouse, along with electrification and automatic signalling along much of the rest of the line, as a showpiece of how to do it, the signalboxes being reduced to emergency use only. The points were operated from a compressed air main at a time when power points were almost unknown; the hiss-bang-thump-hiss of the two sets at the headshunt could be heard throughout the station every minute or two. They ended up with far less operational staff than before. And it was still all in use up to about 1977.

 

[Snow1964]

The basic point of digital signalling is to be automated and therefore not reliant on lots of people in signal boxes. In the past, labour was cheap so schemes like the Mersey example you used earlier would need many more people to run. Now labour is expensive it needs to be replaced with cheap to run computers, which are more of a pain to get working reliably.

The computer might be cheap to run, but if developing it takes hundreds of people years, then not really gaining anything over having an extra person.

How many people have been involved in the computers and software. 10, 50, 100, 200, or whatever, and for at least 5 years. How is that cheaper than employing extra people.

 

[zwk500]

This is an unfortunate comparison, as the reversing/signal installation of the Mersey Railway at Liverpool in 1903 was an absolute pioneer in automation, and operated without any signalmen in signalboxes. It was also installed by Westinghouse, along with electrification and automatic signalling along much of the rest of the line, as a showpiece of how to do it, the signalboxes being reduced to emergency use only. The points were operated from a compressed air main at a time when power points were almost unknown; the hiss-bang-thump-hiss of the two sets at the headshunt could be heard throughout the station every minute or two. They ended up with far less operational staff than before. And it was still all in use up to about 1977.

Your plan still requires 2 people to be available all the time a Paddington for more than 18 hours a day. They're going to need to eat and take breaks between that, so that's 4 to 6 extra driver shifts per day required to be covered, and more drivers than that will need to be hired to cover for leave and sickness.
There's also the question of what would happen to the service if a driver became unable to continue during their shift. Would you have a cover shift just ready and waiting or would you start cancelling trains?
And also to point out that not every crossrail train will be turning round at Paddington in the final service. Its not the same situation as Liverpool Central because the shunting trains need to fit in between trains to/from Ealing Broadway.

The computer might be cheap to run, but if developing it takes hundreds of people years, then not really gaining anything over having an extra person.

But the software can be applied many times over once is working, a person can only be in one cab at one time.

 

[Taunton]

But applying the same software many times over just doesn't happen, does it? Every single project we hear of whole teams working on it. It's seemingly impossible to just buy off the shelf.

Incidentally, the Liverpool installation was installed over a weekend, and worked flawlessly the following Monday morning, and thereafter. George Westinghouse himself apparently came over from the USA personally to oversee it. Now did we say "Crossrail"?

 

[stuu]

But applying the same software many times over just doesn't happen, does it? Every single project we hear of whole teams working on it. It's seemingly impossible to just buy off the shelf.

Incidentally, the Liverpool installation was installed over a weekend, and worked flawlessly the following Monday morning, and thereafter. George Westinghouse himself apparently came over from the USA personally to oversee it. Now did we say "Crossrail"?

Unlikely - this site says it was installed in 1921. Westinghouse died in 1914, so that would have shown an admirable work ethic. The automatic signalling still needed lots of maintenance to keep it working, all those mechanical train stops and pneumatic pipework would need regular inspections. Again labour was cheap then, and isn't now.

The ETCS installation on the GWML is supposedly off the shelf, but Crossrail's decision to use a different signalling system through the tunnel and then use the standard lights on sticks system on the GEML means lots of interfaces between signals, which is the main cause of the headache... although I do agree there does need to be a step back from the complexity of the system

 

[Roast Veg]

But applying the same software many times over just doesn't happen, does it? Every single project we hear of whole teams working on it. It's seemingly impossible to just buy off the shelf.

New specification somes in, codebase forked from the last project, client specification met with new code. The problem is that the old codebase wasn't designed for the requirements of the new project, so you have to dig pretty deep to refactor it into what you want. Combine that with a very fast moving economy of software engineers who don't stay in the same job for more than a few months and all knowledge of the old code goes away...

 

[Taunton]

Unlikely - this site says it was installed in 1921. Westinghouse died in 1914, so that would have shown an admirable work ethic. The automatic signalling still needed lots of maintenance to keep it working, all those mechanical train stops and pneumatic pipework would need regular inspections. Again labour was cheap then, and isn't now.

The ETCS installation on the GWML is supposedly off the shelf, but Crossrail's decision to use a different signalling system through the tunnel and then use the standard lights on sticks system on the GEML means lots of interfaces between signals, which is the main cause of the headache... although I do agree there does need to be a step back from the complexity of the system

It went in with the electrification in 1903. Don't believe Wikipedia. George W came over because it was a total Westinghouse project, the trains, the signals, the power supply.

I think we are in absolute agreement on Crossrail unwisely using multiple signal systems (is it three or four? Is the Heathrow branch different again?), and changing from one to the other every few minutes. I really hope it doesn't turn out to be the Achilles Heel of the system.

I have had some involvement with automatic software-controlled traffic signals "off the shelf", which don't need a competent traffic engineer to program them. Mainstream supplier (they also do trains). Sure, they work as such, but the traffic backs up all around because off the shelf has no optimisation to the individual circumstances. At the end of the day it takes a (senior) engineer longer to configure and get them working optimally, reading multiple manuals, than in the old days, despite the substantial price. The supplier will do this, for a further substantial price, but all they do is come to the office and ask what do you want to do.

 

[rebmcr]

I think we are in absolute agreement on Crossrail unwisely using multiple signal systems (is it three or four? Is the Heathrow branch different again?), and changing from one to the other every few minutes. I really hope it doesn't turn out to be the Achilles Heel of the system.

I'm expecting the GWML's ETCS installation to be extended into the tunnels before too much time passes (something like 5 to 15 years). CBTC is only there in the first place because (ironically) they didn't want unproven ETCS to delay the tunnel opening.

 

[Roast Veg]

I think we are in absolute agreement on Crossrail unwisely using multiple signal systems (is it three or four? Is the Heathrow branch different again?), and changing from one to the other every few minutes. I really hope it doesn't turn out to be the Achilles Heel of the system.

I don't know if there's a difference between the ECTS levels on the GWML vs the Airport branch, but this image, suggests that there wouldn't be a transition from GWML to the Airport branch once complete.

I wouldn't be so sure about CBTC getting replaced even in 15 years, as there will be a full maintenance contract for some time to come.[/url]

 

[stuu]

I think we are in absolute agreement on Crossrail unwisely using multiple signal systems (is it three or four? Is the Heathrow branch different again?), and changing from one to the other every few minutes. I really hope it doesn't turn out to be the Achilles Heel of the system.

Three in theory, I'm not sure where they have got to with ETCS on the GWML, although I bet it's not ahead of schedule, so a Heathrow train might have to transition from CBTC to TPWS then to ETCS

Edited to add: I wonder if this move to more complicated systems which are (in theory), less labour intensive and thus cheaper to run is down to governments everywhere liking new shiny things and therefore are happy to pay Capex but hate paying out Opex?

 

[coppercapped]

A couple of points, m'Lud...

• All trains currently serving Heathrow, 345s and 387s, make the TPWS to ETCS 9 (and vice versa) transition all day every day. It seems to work — but is, of course, a possible failure mode.
• Two of the published reasons for the choice of CBTC were that (i) TfL already used it and (ii) it made auto-reverse possible. ETCS can't do auto-reverse and there are no published plans to add this function.

From what one can gather it is not the transitions which have been causing all the issues in the Crossrail testing up to now but the general lack of maturity in the train control software and the interfaces between it and the CBTC signalling system and between the CBTC and the other systems such as the ventilation and platform edge doors, especially under fault conditions. Which is not to say that this version of CBTC doesn't have its own internal hiccups.

 

[Bald Rick]

ETCS can't do auto-reverse and there are no published plans to add this function.

Objection your honour, it is being proposed , albeit I don’t know the status of the development. At least that’s what a very knowledgeable Signalling / Control engineer told me. (Who happens to work on Crossrail).

 

[Meerkat]

It went in with the electrification in 1903. Don't believe Wikipedia. George W came over because it was a total Westinghouse project, the trains, the signals, the power supply.

I think we are in absolute agreement on Crossrail unwisely using multiple signal systems (is it three or four? Is the Heathrow branch different again?), and changing from one to the other every few minutes. I really hope it doesn't turn out to be the Achilles Heel of the system.

I have had some involvement with automatic software-controlled traffic signals "off the shelf", which don't need a competent traffic engineer to program them. Mainstream supplier (they also do trains). Sure, they work as such, but the traffic backs up all around because off the shelf has no optimisation to the individual circumstances. At the end of the day it takes a (senior) engineer longer to configure and get them working optimally, reading multiple manuals, than in the old days, despite the substantial price. The supplier will do this, for a further substantial price, but all they do is come to the office and ask what do you want to do.

That sounds a terrible product - if it can be programmable but doesnt have a suitable user interface.

 

[Horizon22]

A couple of points, m'Lud...

• All trains currently serving Heathrow, 345s and 387s, make the TPWS to ETCS 9 (and vice versa) transition all day every day. It seems to work — but is, of course, a possible failure mode.
• Two of the published reasons for the choice of CBTC were that (i) TfL already used it and (ii) it made auto-reverse possible. ETCS can't do auto-reverse and there are no published plans to add this function.

From what one can gather it is not the transitions which have been causing all the issues in the Crossrail testing up to now but the general lack of maturity in the train control software and the interfaces between it and the CBTC signalling system and between the CBTC and the other systems such as the ventilation and platform edge doors, especially under fault conditions. Which is not to say that this version of CBTC doesn't have its own internal hiccups.

There's probably somewhere between 5-10 "failure to transitions" every day (including the HeX 387s). Normally this only adds delays of 5-10 minutes, but when you're talking about a central section of 24tph (albeit CBTC), it has the potnential to be a big risk.

Ultimately having so many competing systems was bound to be an issue and these potential / actual problems have been severely understated / underestimated.

 

[Nicholas Lewis]

There's probably somewhere between 5-10 "failure to transitions" every day (including the HeX 387s). Normally this only adds delays of 5-10 minutes, but when you're talking about a central section of 24tph (albeit CBTC), it has the potnential to be a big risk.

Ultimately having so many competing systems was bound to be an issue and these potential / actual problems have been severely understated / underestimated.

Not sure i would find this acceptable but presumably the transitions are on the airport branch so don't have a significant impact on main line operations.

 

[Horizon22]

Not sure i would find this acceptable but presumably the transitions are on the airport branch so don't have a significant impact on main line operations.

Apparently recent software updates on both the 387s and 345s means an ETCS Reset (what a failure to transition oftern requires) can be as low as 2-3 minutes, but this sounds like an "on paper" calculation as opposed to reality. Not sure if TPWS <> CTBC has had many issues at Westbourne Park, but haven't heard of many.