Вы находитесь на странице: 1из 7

2002 WIT Press, Ashurst Lodge, Southampton, SO40 7AA, UK. All rights reserved.

Web: www.witpress.com Email witpress@witpress.com


Paper from: Computers in Railways VIII, J Allan, RJ Hill, CA Brebbia, G Sciutto and S Sone (Editors).
ISBN 1-85312-913-5

Cost-effective timetable and infrastructure


simulation
N. G. Harris & M. Cordon2
IThe Railway Consultancy Ltd, UK.
2Track Attack Software, UK.

Abstract
With increasing traffic levels on many of the worlds railways, the need to
understand the impact of one train service upon another has never been more
important. Similarly, the impact of different infrastructure options is also crucial,
especially where contractual payments between infrastructure providers and train
operators are partly dependent upon performance measured in terms of delays.
There are a number of computer packages already on the market for
examining these issues in the context of major investment decisions such as a
new line. However, with on-rail congestion spreading to secondary routes, and
occurring at more times of day, the need for simulation is increasing. For smaller
investment decisions, though, the correct identification of robust train paths,
whilst important, does not necessarily sustain major investment in the planning
process. Moreover, quick responses are required for many smaller problems,
such as the response of an infrastructure authority to an operators request for
paths.
The software developed here seeks to address these problems through an
ingenious design which enables the program to achieve full interlocking between
points, signals and trains. This means that simulations can be created and
modified quickly, so that more scenarios can be explored in a given time.
Moreover, the actual positions of trains are shown, rather than merely the blocks
which they occupy, thereby increasing the understanding of results. These two
features mean that less time can be spent in constructing simulations, and more
time in testing options and taking action on the basis of the results, which should
mean an improvement in the railway timetabling process.

2002 WIT Press, Ashurst Lodge, Southampton, SO40 7AA, UK. All rights reserved.
Web: www.witpress.com Email witpress@witpress.com
Paper from: Computers in Railways VIII, J Allan, RJ Hill, CA Brebbia, G Sciutto and S Sone (Editors).
ISBN 1-85312-913-5

390 Computers in Railways VIII

1 Background: A network nearing capacity


British Rail was very good at minimising costs by reducing infrastructure down
to the minimum level required to sustain the level of service operated at the time.
Sections of four track railway were reduced to two, and intermediate signalboxes removed. Over the last five years or so, however, rail traffic growth
induced largely through macro-economic conditions and increasing road
congestion has led to the number of train kms operated on the British network
rising by 25%. However, there has been virtually no change in capacity. As a
result, many lines (including secondary ones) operate well above the UICs
recommended level of 70% of capacity.
The identification of yet further paths is still required to satisfy new passenger
and freight demands, and this is an important task in maximising the overall
efficiency of the railway. However, increasing efficiency by definition reduces
the space on the train graph and the robustness of timetables. Delays caused by
one train lead to an increased number of delays to all trains, and knock-on
effects may continue until slack periods (e.g. at night). Railtrack have estimated
that each 1% of additional services adds 2.5% of delay to the system.
Moreover, further problems arise on a mixed-traffic railway, since trains with
different (e.g. speed and acceleration) characteristics do not mix very well.
Faster trains will catch up slower trains and then suffer delays. During times of
disruption, fast trains may fall behind slow trains, and be considerably delayed.
Although improvements can be achieved by segregating the trains onto different
tracks, this can be very expensive and traditionally has been limited to the busy
portions of routes.
However, it is more difficult to determine delays caused where these
additional tracks diverge, and later converge with the main lines, since junctions
have their own speed limitations. Indeed, the delay problem is more general in
that it will also apply to any complicated pointwork for route junctions and
pointwork at station throats. Trains are delayed simply because routes across
points have to be cleared by other trains, before signals can be set. It is becoming
clear that junction delays can negate the advantages expected by running faster
trains.
Simultaneously, as with other countries in Europe such as the Netherlands,
there has been increased public attention on punctuality. Passengers Charter
statistics have been collected for nearly ten years, and are now published by train
operating company (TOC). Each TOC has a performance regime with
infrastructure operators Railtrack, with all delays over one minute being
attributed to one company or another, a system designed to focus management
attention on this issue. As contractual payments between infrastructure provider
and train operators are partly dependent upon performance (measured in terms of
delays), this issue can affect companies financial performance by millions of 5
each year.
New timetables require the consideration of trackwork capacity. With the high
cost implications of any changes to timetables, it is necessary to test their
robustness to any delays that could occur, and any consequent delays to other

2002 WIT Press, Ashurst Lodge, Southampton, SO40 7AA, UK. All rights reserved.
Web: www.witpress.com Email witpress@witpress.com
Paper from: Computers in Railways VIII, J Allan, RJ Hill, CA Brebbia, G Sciutto and S Sone (Editors).
ISBN 1-85312-913-5

Computers in Railways VIII

39 1

trains. With penalty regimes being significant, it is likely that in the future it will
become mandatory to prove a timetable before it can be implemented.
This proof requires a simulation of the track, pointwork, interlocking, signals,
trains and timetable. As indicated above, a simulation on this small scale can
identify any inherent weaknesses that may be present in a proposed timetable.
As a bonus, any additional changes to the track can also be made and the
timetable can then be re-tested.

2 Train service simulations


Major investment schemes (such as the West Coast Main Line upgrade, or the
cross-London Thameslink scheme) inevitably require a great deal of detailed
planning, which typically involves simulation of train operations, track layout
and power supply. A number of complex simulations exist, providing full
capability - but at a price of considerable time and effort in providing the input
data.
But, as we have seen, even minor schemes - such as a local authority wishing
to sponsor enhanced service frequencies on a regional line - are now pressing
against capacity. Railtrack managers are rightly increasingly concerned about
the performance implications of these potential additional services.
In the background, we have a trend of increasing use of simulations for other
purposes, including leisure. The Railway Consultancy Ltd came across one such
simulation in 1999, providing the user with an opportunity to run a key signalbox
on the network. We realised that the algorithm being used was extremely similar
to that required for a classic transport planning simulation. We therefore
developed a joint programme with Track Attack to provide an operational
simulation which was easy to use, cheap and quick to build up, and therefore
appropriate for examining problems on secondary routes.
2.1 Track Attack
The key inputs to any train service simulation are clearly:
0
the length, speed and acceleration/deceleration of trains;
the position and number of aspects of signals;
the track layout and associated routing and interlocking; and
0
the timetable under consideration.
All of these are contained within the Track Attack package, and the first three
dealt with at the infrastructure stage. A range of timetables can be built up
simultaneously using a separate timetabling program, whilst the use of random
delays enables the consideration of a range of operating conditions as might be
encountered across a few weeks of operation.
Development of the Track Attack software has sought to address the problem
of program usability through an ingenious design which enables the program to
achieve full interlocking between points, signals and trains. This means that
simulations can be created and modified quickly (completely new simulations

2002 WIT Press, Ashurst Lodge, Southampton, SO40 7AA, UK. All rights reserved.
Web: www.witpress.com Email witpress@witpress.com
Paper from: Computers in Railways VIII, J Allan, RJ Hill, CA Brebbia, G Sciutto and S Sone (Editors).
ISBN 1-85312-913-5

392

Computers in Railways VIII

with 50kms of route have been created in 10 days, with simple junctions created
in much shorter timescales), so that more scenarios can be explored in a given
time. Moreover, the actual positions of trains (and their lengths) are shown,
rather than merely the blocks which they occupy, thereby increasing the
understanding of results.
A useful feature of the timetabling element is the use of entry times to the
simulation and standard sectional running times (diagrams) for a given stopping
pattern. This avoids having to enter details separately for each train running on a
regular pattern. Once the timetable has been agreed, it is checked against the
infrastructure file, and interlocked auto-routes set up.
Scenarios should be run for a number of iterations, in order to reflect a range
of operating days. Graphical output shows a full track diagram, as per a
signalling centre. Four-character train destinations are shown in key sections, to
indicate which train is which; these descriptions move with the train, as per
normal signalling practice. However, unlike normal signalling practice, every
aspect of every signal is reproduced on the screen. This is extremely helpful in
understanding how and why succeeding trains brake behind a previous service.
The simulation can be run at various speeds between real-time and 8 times faster,
starting from any one of 63 pre-designated times through the traffic day. This
enables the user either to focus in on those times where problems occur, or to run
the simulation as fast as possible in order to get overall results. A batch facility is
available to leave the pc running (for instance) a whole month of simulations
without manual intervention.
Textual outputs include a summary of the total delay on each day, but it is
also possible to examine in detail which trains were delayed where, and for how
long. This type of analysis is essential in order to understand the fragility of some
timetables - in some circumstances, they may deteriorate severely based on a
relatively small input delay. Train graphs can also be viewed.
These features mean that less time can be spent in constructing simulations,
and more time in testing options and taking action on the basis of the results,
which should mean an improvement in the railway timetabling process.
The Track Attack software was developed using a simple subset of Turbo
Pascal programming language. The language was used in its simplest form to
minimise software bugs and their complexity. There are some assembler routines
for dealing with some interrupts.
The software was developed from a railway engineering viewpoint with
performance / simplicity being the prime requirements. There are no software
tricks used; each element of functionality has been created as a result of
performance needs. Despite this, the software is highly structured in order to
achieve reliability and understandability.
The software was developed from other leisure software that had been created
for existing locations such as London Euston, London Waterloo - Clapham
Junction, Birmingham New Street and ten other locations in the UK.
We have used Track Attack on a range of smaller projects, of which one of
particular interest is that on the GE main line.

2002 WIT Press, Ashurst Lodge, Southampton, SO40 7AA, UK. All rights reserved.
Web: www.witpress.com Email witpress@witpress.com
Paper from: Computers in Railways VIII, J Allan, RJ Hill, CA Brebbia, G Sciutto and S Sone (Editors).
ISBN 1-85312-913-5

Computers in Railways VIII

393

3 Case Study: The Great Eastern Main Line


In 2000, The Railway Consultancy Ltd was approached by a property developer
in Essex planning a new business park and residential development adjacent to
the main railway line between London and Norwich, on the outskirts of
Colchester. This line is only double-track beyond Shenfield (on the edge of
London), and has little if any usable overtaking capacity, since the loops at
Ingatestone (Northbound), Chelmsford (Northbound) Marks Tey (Southbound)
and Witham (both directions) are either too short and/or have limited access and
egress speeds. However, the loops are used for reversing trains and for parking
failed trains.
Although at the last COMPRAIL conference a number of delegates
commented on line capacity problems with 6tph, the peak frequency on this line
is 13tph, and the standard offpeak train frequency is Stph, comprising:
e
2tph Intercity London - Ipswich - Norwich (operated by Anglia);
e
2tph fast London - Ipswich (operated by FGE, to provide a joint 4tph
service to Ipswich);
e
ltph semi-fast London - Clacton (operated by FGE);
e
ltph slow London - Braintree (as far as Witham) (operated by FGE);
e
ltph slow London - Colchester Town (operated by FGE);
e
1 train every two hours Basingstoke - around London - Chelmsford
(operated by Anglia);
e
1 container train every two hours to/from Felixstowe Docks (operated
by Freightliner).
Operating problems arise because the fastest trains take 24 minutes to cover
the 51 kms between Shenfield and Colchester, but the slowest 35 minutes. This
1 1-minute difference in running time has to be accommodated between trains
running 15 minutes apart. Already, it has proved impossible for all the stations to
be served by the slowest train (to Colchester Town), because that would
increase the journey time to a level which would necessitate overtaking.
Our demand forecasts quickly showed that the proposed station, at Stanway,
would be worthwhile in commercial terms. However, it is immediately apparent
that there is no capacity for additional stops in slow trains without causing delays
to faster ones. We therefore investigated a number of infrastructure options
providing either short passing loops, or longer dynamic passing loops.
Simulation clearly shows the level to which the line is being used at capacity, and
the relative ease at which one train can land up delaying several others. Delays
occur frequently on this section of line where fast trains catch up previous ones
calling at local stations. This occurs particularly often outside Chelmsford, which
is served by all passenger trains except those toifrom Norwich.
Short passing loops provide an increase in capacity and a further option for
traffic regulation in cases of severe disruption, but they do not provide for
increased day-to-day reliability. This is because they require fast and slow trains

2002 WIT Press, Ashurst Lodge, Southampton, SO40 7AA, UK. All rights reserved.
Web: www.witpress.com Email witpress@witpress.com
Paper from: Computers in Railways VIII, J Allan, RJ Hill, CA Brebbia, G Sciutto and S Sone (Editors).
ISBN 1-85312-913-5

394

Computers in Railways VIII

2002 WIT Press, Ashurst Lodge, Southampton, SO40 7AA, UK. All rights reserved.
Web: www.witpress.com Email witpress@witpress.com
Paper from: Computers in Railways VIII, J Allan, RJ Hill, CA Brebbia, G Sciutto and S Sone (Editors).
ISBN 1-85312-913-5

Computers in Railways VIII

395

to be running exactly in their slots, if one or other of them is not to be delayed.


Dynamic passing loops are much longer and enable a greater degree of
robustness to be incorporated into timetables. One option analysed as part of this
study is to serve another new station at Beaulieu Park, on the outskirts of
Chelmsford, where further residential development is taking place. A 5-km-long
loop could be constructed, including platforms at this station, which would
therefore provide a significant variation in running time, and hence the possibility
of a robust timetable serving the two new stations. However, the cost
implications of this are well beyond what either property developer might
reasonably contribute to a local station, whilst the benefits of it accrue to all
operators on the line. With Railtrack (the infrastructure operator) currently in
financial administration, investment by them looks unlikely, and Government
funding through the SRA (Strategic Rail Authority) will be necessary if these
network enhancements are to be made.
At present, the SRA are looking at refranchising rail services on this part of
the network from 2004, and are looking at the timetable in detail. A Line
Development Group is being formed to look at the options, which include the
infrastructure alternatives for this key route, and the ability to carry even more
traffic on the existing network. The costs and benefits of a range of options will
need to be examined, but the line capacity problems are recognised, and the
beginning of a new franchise period brings about an opportunity which both
passengers and freight customers will hope is taken.

4 Conclusions
On a busy network, reliability concerns require the proper testing of options,
even for relatively minor changes. Such testing for a rail system requires
simulation, although good (and, importantly, fast) results can be obtained from
relatively-unsophisticated systems. This paper has reported one such system,
designed to understand the key operating problems with a minimum amount of
input.

Вам также может понравиться