2006 Q8 - Development of overlaps etc

[dorothy.pipet]

Another attempt for comments please.

[PJW]

Just had a skim read so far and it reads well. In general I don't think that continuous prose is the best way of gaining marks quickly in exam conditions, but actually the descriptive text book seems to suit this style of question quite well- I may have added more diagrams (e.g. perhaps re the SOD / overlap) and just wonder whether too much time spent at the begnning re absolute block. I'll evaluate in more detail and give further feedback later, but actually my gut feeling would be that this would have scored well on a question that few others would have attempted most of which would have been pretty weak and woffly, so your answer would have stood out amongst them.

Another attempt for comments please.

[PJW]

Yes I think this was a very good answer and ought to have obtained a Credit. Actually I don't think that it was far off a Distinction
I was tempted just to leave it at that- and you with a positive feeling.
However decided that it might be useful to discuss what I think you could have done to do that bit better and get the Distinction.

From an examiner's perspective they have a pack of 25 marks to award; they will need little persuasion to part with the first 5 or so, but conversely they will be very reluctant to part with their last 5. I think that it gets exponentially more diifficult to prise those last few marks out of them and suspect that this may be particularly the case in a question wheere there is one big allocation of potential marks and is generally pretty "open ended" comapred to the type where split into a group of smaller packets. I am not saying avoid such questions- actually they give you the frreedom to take the answer the way you want to and potentially avoid areas of weakness by concentrating attention elsewhere, so can be advantageous. I am saying that you have to be particulrly convincing in order to do extremely well. I think that this one is about as minimalist a question as you'll find in the past papers and thus gives a good example.

1. Definitely a good idea to start with a mechanically signalled layout that bears some relationship to the world of 1876 and the improved world the inspector was wanting to see instead; I think it is safet to assume that he was not envisioning ETCS L3 but something more like the railway of 1900.
The diagram was good, but it could have been better. The main deficiency, particularly in the context of this question, is that you did not show the BLOCK CLEARING POINT- broadly the equivalent of the MAS overlap. As you have drawn it, b couldn't accept a train from A unless the line was reerved for it, broadly as far as the singnalbox (which I am assuming to be 1/4 mile beyond the home signal on your diagram). Hence the junction cannot be used; indeed once having accepted a train from A then couldn't accept a train from the branch. Actually you shouldhave made the junction protecting signals as "inner homes" and provided a signal further out (and of course moved the distant out to be braking distance from it) to be an "outer (or acceptance) home".

You also seemed to suggest that there could be a train in station limits as well as a train in each of the block sections; true but actually you failed to convey the fact that whereas there can only (apart from exceptional conditions such as rescuing a failed train etc.) be one train per block section, there is no specific maximum tfor the number that can be within station limits. I think you probably know this, but you didn't SAY it and it is a key difference. A large station could have a significant number of stop signals and potentially there could be train stood at each. If there is a delay on the line to C preventing the clearance of the advance starter (section signal) the signalman (not signallers in those days, even if happened to be female!) would be dropping trains down signal by signal within station limits and it is in this scenario that the "thickness of the signal post" becomes particularly relevant.

Therefore you revealed some weakness in your understanding of semaphore signalling and in an element that was key to this question. At the overview level for the treatment of the whole question, this is not very damning but it is the sort of thing that would count significantly against you if I as the examiner were contemplating awarding a Distinction.

2. When discussing MAS overlaps, it would have been bebneficial to have been explicit re the two separate elements:
a) the locking of points, prevention of opposing moves and, to a certain extent sometimes, level crossings;
b) the proving of the wheeled path and train detection is the aspect level.

Often the limits for these two separate elements are the same, but not always- one is primarily worrying about a junction collision whereas the other primarily a rear-end collision.

i) An example is when considering the SOD= Safe Overrun Distance for TPWS which can well be beyond the limit of the "overlap".

ii) In some cases we define separately an EOL = Extended overlap, locking up additional railway beyond the "overlap" which for clarity may be referred to as the AOL= Aspect overlap.

iii) You could also have mentioned POL = Phantom Overlap where the defined locking limit is at a place that is not the limit of a train detection section and thus the proving of tracks clear extends beyond it (actually often the case where wish to permit simultaneous moves up to opposite direction signals that are separated only by one track section which is then included as the overlap control for the outer signals in each direction- the locking overlaps however are not shared, it is just that the two separate lengths fit within the length of that track section.

iv) You did cover ROL= Restricted overlap well- it was good that you discussed the likelihood and consequences of providing, but perhaps you should have been a little more explicit that there is a trade off between a slightly increased risk of collision following a SPAD and the significant improvement to operational flexibility achieved; indeed you could go further and argue that the overall risk of the layout is actually reduced because of the ability to clear that other signal (a signal that is at red can be SPADed, one that is shoing a proceed aspect cannot!). Slightly unfortunate that you introduced the concept utilising the word "reduced" rather than say, shortened; it may have been sensible to have explained when a reduced overlap is utilised before rather than after describing the ROL, but at least by including you did salvage the situation preventing the examiner thinking that you were confused.

Don't tell a Great Western man that AWS was the earliest system; it used updated technology to implement what had become the standard ATC installation at distant signals on its network some 50 years prior.

Indeed there were purely mechanical train stops even earlier and a fair percentage of London Underground still utilises trainstops for its train protection that would not have been unfamiliar to an engineer of 1900.

Actually GW ATC and indeed AWS are not really "to reduce hazards at overlaps" (whatever that means- I think you meant: "mitigate the SPAD hazard by containing any overrun within the overlap") because they are really only good for encouraging the driver to brake when encountering a cautionary signal and doing so automatically unless the driver intervenes. They do mitigate against the SPAD occuring in the first place by reducing the likelihood that the driver will fail to brake down to a speed low enough that they can brake to a stand within the sighting distance of the stop signal.

Conversely the LU trainstops are primarily (ignoring the "blind train stops") there to mitigate the consequences once the SPAD has occurred; the overlap length is always calculated according to site conditions and train characteristics rather than being a nominal length.

When TPWS was first retro fitted to existing railway it was intended to address the consequences of SPAD similarly, but not to the same degree; LU design "guarantees" the overlap whereas TOWS was just trying to reduce the risk significantly.
Nowadays TPWS for new projects and major alterations has moved rather closer to the LU approach, TI22 etc. but the nature of the beast, and particularly the range of different rolling stock, means that cannot actually give that guarantee. Of course the OSS loops, although primarily for the same reason of containing the overrun within the SOD, can in some cases reduce the likelihood of a SPAD occuring. Perhaps one could argue that the LU blind trainstops (on the approach to some signals that are operated by track timers so that they drop just prior to the train reaching them if being driven on a suitable braking curve to stop at the signal) might also have that effect, yet that would be by the drivers watching them day after day and reinforcing their driving technique through familiarity.

I think that you could have done well to expand on a discussion in this area, talking about multiple OSS loops and even Conditional Double Red control where there is a high speed approaching a junction conflict [see next post reply]. If only you then related this back to the outer and inner home signals for mechanical signalling that we discussed at the beginning, you would now be firmly on your way to that Distinction!


Another area that could have given content for this question would have been ATP, ETCS etc. The mainline approach is to give the river the information they need to stop at the signal, whereas the train itself is guaranteeing to stop by the end of the overlap. There is always the issue of odometry uncertainty to deal with so the train has to permit itself to be driven as far as the signal ebven in circumstances in which the train lacks certainty regarding how far ahead it still is.
This gives rise to the concept of "release speed", effectively the train accepting that it has to allow a driver to continue even when it thinks it might be about to SPAD, but only at a relatively slow speed. It calculates the speed limit to be low enough that it can afford to wait until it absolutely knows a SPAD must have been committed before intervening and yet still have enough remaining distance for the emergency braking to be able to stop safely within the overlap.

With ETCS, whether or not there are lineside signals, the position at which the driver should stop is the EOA = End of Authority which is separate to the SvL = Supervised Location which is the place by which the train must stop- hence effectively an "overlap".


For metro railways using CBTC (generally implemented as "moving block" and without lineside signals) apparently there is no need for overlaps. However I said apparently, because I don't actually believe it.
Certainly positioning accuracy is extremely good (being in tunnel -as many are- helps since rail adhesion conditions are more predictable, putting down are large number of absolute position references (e.g. RFID tags) also helps) and so the story is that no overlaps are required; the train will always stop a short safety distance prior to a defined conflict point.
That sounds great and you nearly believe it- until you try to apply it to a railway where there is pointwork just beyond a platform end and that you need to get the train right up to the end of the platform since the train only just fits so don't want to stop short. Then there is a problem and you find that you need to place the confict point within the pointwork and thus have to lock the points that are beyond the nominal stopping position. Obviously you have to invent a name for this (it simply can't be called "overlap locking" because the system is a priori defined not to need overlaps so it just wouldn't be politic to suggest otherwise). This is the time at which the contractor reminds the client of the project soundbite along the lines of: "don't change the product, change the client" and if this means some tunnelling to move the junction away from the station or the platforms further from the junction, then that is not their problem; indeed there are many "Londons" in the USA and it actually would be rather more convenient to implement the project in one of them instead of the UK version.

Obviously there are rather more ideas here than could ever be fitted into 30 minutes at this level of detail; I think that often the best technique is to deliberately make your scope as wide as you can. In this question if you can cover the whole range from mechanical witha bsolute block through generations of MAS with increasing degrees of train protection and then end with transmission based signalling with moving block, that is ideal. Also the wider your coverage the less time you will understandably have available to go into any level of detail- hence less chance of you revealing any imperfections of your understanding to the examiner.

Just had a skim read so far and it reads well. In general I don't think that continuous prose is the best way of gaining marks quickly in exam conditions, but actually the descriptive text book seems to suit this style of question quite well- I may have added more diagrams (e.g. perhaps re the SOD / overlap) and just wonder whether too much time spent at the begnning re absolute block. I'll evaluate in more detail and give further feedback later, but actually my gut feeling would be that this would have scored well on a question that few others would have attempted most of which would have been pretty weak and woffly, so your answer would have stood out amongst them.

Another attempt for comments please.

[dorothy.pipet]

Thank-you, that is very encouraging at this stage.
You are right that my knowledge of semaphore signalling is sketchy, but also I'm not sure of how SODs are calculated and used - in my role I only a scheme plan with the TPWS already placed, and Design Logs rarely show the detail.
When I looked at the 2006 paper, I found few appealing questions and this looked like a "least worst" option, so I have surprised myself.

[PJW]

The original TPWS fitment approach was broadly (look at 10037 for the detail)-

1. Look at every signal in turn and dertmine whether any exclusion code applied (not a stop signal, no junction conflict, protection by virtue of trapping etc.); if no approved "get out", then fit TSS within defined position (very close to signal).

2. Look at the entries in a generic table and find one for the applicable approach speed for that signal; this dictates where the optimum position for the OSS is defined to be, and the spacing of the arm/trigger loop that iself defines the "set speed" (note that this translation is different for passenger or freight as the on board timer is deliberately different so that the speed at which intervention would occur is quite a bit lower for a freight cmpred to passenger).
This is inevitably a compromise, just the best position to maximise the average protection. Appropriate for the retrofit project needing to fit the country very quickly and using pretty inexperienced design resource. The intention initially was just to address the majority of the risk associated with SPAD; it was always known that fast trains would SPAD considerably beyond the overlap, but actually these represented only a very small percentage of all SPADs and even if one of these should occur then at least the consequences of collision would have been reduced significantly because some braking would have occurred.

HOWEVER, that didn't last for long as the "worry brigade" decided that needed to do something more since there would still significant residual risk (albeit much reduced than that with which we had previously been reasonably content). I think it was partially because the national press got the story that TPWS didn't work above 60mph (and interpreted incorrectly that this meant "not at all" rather than "wouldn't completely stop train within a 180m overlap").

3. Hence the OSS+ was invented as an outer protection; generally it was positioned as far as it reasonably could from the location at the signal- (750m) and used for situations where the approach speed such that it was needed.

That was effectively the basis on which the retrofit was initially performed in the very early 2000s.

4. Of course even the OSS+ couldn't cope with the very highest speeds and so people then started worrying about this scenario on the approach to a junction.
Hence the invention of the "double red" control so that unless the junction signal is cleared then the signal in rear would be also held to red; hence could fit this outer signal with a TSS & OSS (not to protect its own route, but to protect the junction beyond the next signal), This was a great idea in that it, via aspect sequence, effectively extended the TPWS protection beyond the technological limit for the junction signal; BUT the flip side is that it completely ruined capacity and junction reoccupation times with predictable effects on the train service.

Hence it had to be further complicated by implementing instead as "Conditional Double Red", which means that the outer signal is approach released once the approaching train has passed its OSS (i.e. it either is coming slow enough that it escaped the speed trap and thus well undercontrol, or it is comng fast but with the emergency brakes already applied). [In some way it has analogies with M / W class routes and their different lengths of overlaps].
Then we started worrying about whether it was right to be releasing the aspect of the outer signal by the route being set across the junction from the inner signal without imposing its approach locking (so this had to be changed to "once route set" so that effective even if the aspect not yet cleared) so here we added another dollop of complexity that many of us think isn't really justified by the risk it is supposedly addressing.

It was only once all this had been completed and new schemes were being developed with TPWS as an integral component, that people started thinking that if we could tolerate slightly longer overlaps then could get good protection at higher speeds without the worst of these complications; conversely if short overlaps really needed then could justify this to be acceptable by providing enough TPWS.

A few of us (including one of the current mod 3 examiners) tried for a long time to rationalise the various standards 10137, 10038 and 00028 because there was a lot of inconsistency and no single point of truth. We very nearly succeeded but fell at the last fence (one contoversial sentence in the wording of the briefing note for the new standard aboout to be published) due to what perhaps are best stated as "personalities and politics". However out of this work did emerge the current methodolgy for determining OSS positions......

A. Having decided to fit a TSS and knowing what the SOD length is beyond the signal, calculate the speed at which a train[#] will stop within that length.

B.If the speed just calculated is less that the potential approach speed then that speed is to be the "set speed" of OSS1.
Position the OSS1 such that it is placed suitably between
i) the braking cuve for a sensible braking rate to the signal (so that won't get invalid interventions on trains being suitably controlled by their drivers)
ii) the braking curve for emergency braking to the end of the SOD (so have a very good expectation that overrun will be contained safely).

C. Now calculate the speed at which this OSS1 will be effective at stopping a train within the SOD. If this is equal to the maximum approach speed then good there is protection; if not then repeat the activity with this speed as the set speed for OSS2 and position this in a similar manner.

D. Repeat working away from the signal adding further OSS with each having a set speed equal to that which is the limit of effectiveness of the one closer to the signal, until effective for the highest possibl;e approach speed.

E. Now that know the minimum number of loops that gives protection, "juggle" positions a bit if this has advantages whilst still giving that protection.

You can probably see why it is all formulae in Excel now! Hence I guess that very few actually understand the methodology utilised but just put in the entries needed and then push the "calculate" button and check that no error messages and write down what it says.

# I have glossed over a bit re "train will stop within that length". Where there are different types of trains with different emergency braking rates and brake build up time (more or less everywhere), then there can be "robust discussions" relating what rolling stock is to be considered. The whole thing is very much a compromise between giving good confidence of protection on the one hand and avoiding false activations on the other; a mix of diverse rolling stock types can make it totally impossible to satisfy all scenarios.

Bet that is more than you wanted to know!

Thank-you, that is very encouraging at this stage.
You are right that my knowledge of semaphore signalling is sketchy, but also I'm not sure of how SODs are calculated and used - in my role I only a scheme plan with the TPWS already placed, and Design Logs rarely show the detail.
When I looked at the 2006 paper, I found few appealing questions and this looked like a "least worst" option, so I have surprised myself.