|
|
Post by zbang on May 4, 2020 16:57:42 GMT
Out of the the GCR discussion elsewhere-
At least in the context of non-computerized schemes, how is an lever frame/interlocking plant "proved" to be correct? That is, the correct locks are applied and sequence enforced? Also that one lever doesn't affect things it shouldn't.
On the face of it, most locking should be fairly simple combinatorial logic with the occasional timer, however when moving from a worksheet/book to a physical manifestation, it's almost easy to put a peg in the wrong place or a wire on the wrong terminal. At first glance, testing is an n^2 problem trying every combination of levers, however that can't directly work because of enforced sequences and it becomes a huge number rather quickly. Likewise, it's easy to try the expected sequences but are seemingly unrelated actions checked at the same time? How is this documented?
Or is a frame only visually checked that the locks match the book?
(Normally this would send me into a pile of books, but I have neither the time nor brain power at the moment.)
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on May 4, 2020 17:04:24 GMT
The locking of a frame would not just rely on someone looking at it.
Every possible scenario would be tested and tried amd double checked by someone independent to make sure the mechanical locking held up with any conflicts
The designers also make mistakes we are all human after all
|
|
|
|
Post by zbang on May 4, 2020 17:47:52 GMT
Having spent time working with industrial controls and designing/building test systems, I understand the need for proving a system and the basic methods involved. Also for monitoring unintended consequences ("Hey, flow at pump #5 and motor not running, is that right?", "Um, no, close the valves and raise an alarm").
What I'm interested in here is how those principles are applied to a lever frame/interlocking plant. "Every possible scenario" on a 50 lever frame is quite a lot. It's easy to check simple things, such as if a FPL prevents the associated points being moved, but does that procedure also include whether it affects an genuinely-unrelated signal at the far end of the station?
There's also the concept of "implied trust" where simple sequences are proved correct and then become a single component in a more complex sequence. This can limit the amount of direct action to prove a sequence by avoiding what could be come a recursive problem.
Or one let's be it be an n-squared problem and gets on with the work.
When Theory meets Practice, what actually happens?
FPL- Facing Points Lock
|
|
Tom
Administrator
Signalfel?
Posts: 4,196 
|
Post by Tom on May 4, 2020 18:30:18 GMT
It depends a lot on whether or not you have a new interlocking, or it's an alteration to an existing one. Generally on an existing interlocking you spend more time proving the presence of correct functionality associated with the change, rather than absence of unintended functionality (though I have been known on occasion to do some negative testing, as time permits). Prior to the commissioning the Principles Tester will undertake a thorough review of the design and the signalling being introduced. Does it comply with all the standards? Has the designer omitted something? Are there unusual conditions which need to be accounted for? Are there any logic conditions 'masked' by the operation of other conditions, and how will I separate them out to prove each one? This review period often includes a period of to and fro discussion between the tester and the designer. When testing a new installation we would undertake what's known as 'flood' testing as an initial test. Flooding means drop every track circuit, break the detection on every set of points, etc, so that none of the interlocking's inputs are present. This is ideally done without the interlocking connected to the railway, but instead to a test desk, but beware, this is what they used when the train derailed at Waterloo, so you have to be careful. Here's an example of a test desk, used at King's Cross back in 2015:  This was a relatively small interlocking, and there were still 66 function inputs to be simulated. Imagine doing that with a site like Acton Town! With the test desk connected, try and clear a signal. Nothing should happen. Introduce each condition you need in turn, until you have all the conditions you need to clear that signal. If the signal clears when it should with all other conditions broken, you have proved the absence of unintended functionality for that signal to clear. You then repeat it for all other signals, then do the same with their approach and back locking, etc. Usually at this point, you'd take the test desk away and connect the interlocking to the railway. Does each track circuit/point/signal/trainstop correspond correctly to the interlocking? If it does, read on... Once you get more and more proof of the correct functionality on a signal by signal basis, you move to more dynamic testing. Can I do parallel moves? Does a second train approaching a signal before a junction prevent the previous route from correctly unlocking? Can I lock a junction up with a single fault? Does a route unlock prematurely? From then, it's more akin to 'playing trains' - you simulate train moves, try combinations of things, how close you can get trains etc. After that - connect up the control system (which may or may not have been tested as an integral part of the vital signalling), and check there are no race conditions or response delays. Finally, bring in test trains. These are your final tests, and are used to double check that real-life operational response times don't have any untoward effect. |
|
Tom
Administrator
Signalfel?
Posts: 4,196
|
Post by Tom on May 4, 2020 19:12:36 GMT
So, what about mechanical locking? Well, conveniently I'm pretty well placed to answer that one, having tested all the mechanical locking alterations undertaken by LU in the last five years.
One frame was completely relocked from scratch and I did the majority of the testing for that off site, on a frame used only for testing purposes.
Beforehand, in a similar manner to circuit logic, you sit and review the mechanical locking arrangements. Usually this involves working out in your head what the locking should be, and comparing that to the version in the design, then trying to understand and justify the differences between your solution and theirs.
When you go to the test frame, you check the mechanical locking is correct to the dog chart (the drawing which shows the positions of all the locking pieces) and then try to move each lever in turn. It should only move with the pre-conditions (known as the lever leads) required for it to be able to move met.
For instance, let's say that No. 19 lever can only move if no. 18 lever is reversed. Can 19's move with 18's normal? Can 19's move if 18's is mid-way between normal and reverse? It should only move when 18's lever is fully reversed.
(Repeat for all over levers)
Once 19's lever is reversed, it should lock 18's reverse. Does it? When is the locking removed? It should only unlock 18's lever when 19's is fully normal.
Another example is that 16's lever reverse locks 18's lever normal. Does it? If 16's is reverse, Does it lock other levers unnecessarily - can I still move levers 13, 14, 16, and 20 freely?
Once I was satisfied, the locking is disassembled and taken to site. When it was refitted, I went through a simplified testing routine to make sure everything still locked correctly according to the locking table - this took about 20 minutes on a small frame.
The majority of alterations I have been involved with have been mods to existing locking. It's a bit different as you review the locking arrangements but don't test the lever leads too widely - you take those as given, and if they're wrong they tend to be fairly self-announcing.
When it's an alteration, you're there watching over the locking technicians as they alter the locking in many cases, so you know exactly what locking pieces are being removed and installed.
Next, I take out all the fuses to the outside world (so I can't accidentally throw points) and disconnect the electric locks, so that I know I'm not having any mechanical locking masked by electrical locking. (This doesn't stop you from being masked by other pieces of mechanical locking though - more on that later.)
When it's all in, I tend to test the alteration first to make sure it works, then get the locking tech to secure the screws holding the locking pieces in place. Then, check it all corresponds to the locking table - which on a bigger frame can need two people and take several hours - Uxbridge takes about three hours and there is a lot of masking where you can't see a lock because of another lock doing the same thing. The only way to get around that is to take the other pieces of locking away - which isn't advised, so I have to use another alternative method, which is the comparison against the dog chart to confirm that the locking is there, even though I can't prove the locks are physically interacting.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on May 4, 2020 19:35:34 GMT
Was Kings Cross your handy work then? I seem to recall some people wasn't happy with it
|
|
Tom
Administrator
Signalfel?
Posts: 4,196
|
Post by Tom on May 4, 2020 21:05:15 GMT
I was heavily involved, yes.
|
|
|
|
Post by PiccNT on May 4, 2020 22:49:56 GMT
Which brings me to my next question. I have asked a few people and not received an adequate response. When I was growing up on the Northern line as a kid and used to bum the odd cab ride, I remember some of the motormen used to pluralise the signal number such as Y2 becoming Y2's. A signal that was often spoken about.
Now working for LU, I hear this from the more experienced T/Op's and on occasion over the train radio by the Controller/Signaller. Tom above used this way of describing points and just wondered why it's done?
|
|
Tom
Administrator
Signalfel?
Posts: 4,196
|
Post by Tom on May 4, 2020 23:07:46 GMT
And it was ever thus!
It's easier than saying 'number 18 lever' - though I would never write it like that in anything official - it would always be No. 18 lever or Lever no. 18.
|
|
|
|
Post by brigham on May 5, 2020 7:43:43 GMT
And it was ever thus! It's easier than saying 'number 18 lever' - though I would never write it like that in anything official - it would always be No. 18 lever or Lever no. 18. So it's a possessive, rather than a plural. Just everyday English, in fact. Like 9's letterbox, or 17's front door. (Bit of Post Office jargon there!) |
|
Tom
Administrator
Signalfel?
Posts: 4,196
|
Post by Tom on May 5, 2020 19:15:12 GMT
Yes, though technically it shouldn't be possessive.
|
|
|
|
Post by brigham on May 6, 2020 7:37:12 GMT
Yes, though technically it shouldn't be possessive. I'm not sure why it shouldn't be possessive. It's signal 18's lever, and Number 9's letterbox, after all. (I can personally vouch for the latter). |
|
|
|
Post by PiccNT on May 6, 2020 11:11:05 GMT
I think I'm beginning to understand this now. When my Trains Manager tells me my train is at PM2's, what he means therefore is my train is at PM2's signal. That right?
|
|
North End
Beneath Newington Causeway
Posts: 1,769
|
Post by North End on May 6, 2020 11:16:20 GMT
I think I'm beginning to understand this now. When my Trains Manager tells me my train is at PM2's, what he means therefore is my train is at PM2's signal. That right? Yes. He’s probably also unknowingly telling you that he’s been on the job for a while, or at least has picked it up from such a person, as it’s a form of expression which has tended to fall out of use over time - on the operating side at least. How quaint that you still have signals to refer to! ;-) |
|
|
|
Post by PiccNT on May 6, 2020 12:22:18 GMT
I think I'm beginning to understand this now. When my Trains Manager tells me my train is at PM2's, what he means therefore is my train is at PM2's signal. That right? Yes. He’s probably also unknowingly telling you that he’s been on the job for a while, or at least has picked it up from such a person, as it’s a form of expression which has tended to fall out of use over time - on the operating side at least. How quaint that you still have signals to refer to! ;-) And long may they last :-) |
|
|
|
Post by nig on May 6, 2020 13:50:12 GMT
I think I'm beginning to understand this now. When my Trains Manager tells me my train is at PM2's, what he means therefore is my train is at PM2's signal. That right? newer train managers would say approaching oakwood which would mean oakwood station |
|
|
|
Post by PiccNT on May 6, 2020 14:55:06 GMT
I think I'm beginning to understand this now. When my Trains Manager tells me my train is at PM2's, what he means therefore is my train is at PM2's signal. That right? newer train managers would say approaching oakwood which would mean oakwood station They should then be ignored as that's PK2's ;-) |
|
|
|
Post by nig on May 6, 2020 15:15:55 GMT
newer train managers would say approaching oakwood which would mean oakwood station They should then be ignored as that's PK2's ;-) and the not so new train managers saying train approaching cockfosters meaning cockfosters station .. sorry going off topic |
|
|
|
Post by rheostar on May 11, 2020 7:58:03 GMT
They should then be ignored as that's PK2's ;-) and the not so new train managers saying train approaching cockfosters meaning cockfosters station .. sorry going off topic The phrase used is, "...he's just hit 1's up at Cock." |
|
|
|
Post by PiccNT on May 11, 2020 10:16:08 GMT
Or in the case of my mate last Friday, he's just hit 90's down at Acton. Oops.
|
|
|
|
Post by rheostar on May 11, 2020 10:30:22 GMT
Or in the case of my mate last Friday, he's just hit 90's down at Acton. Oops. That's never good! |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on May 11, 2020 14:39:36 GMT
Or in the case of my mate last Friday, he's just hit 90's down at Acton. Oops. Won’t be the first or last |
|
